Search

In the evolution of the Internet from Web 1.0 to Web 3.0, algorithms have played an irreplaceable role, enabling information technology to functionally realize the transmission of one-way content to decentralized interaction. Algorithms have a subtle impact on people's lives, work, and learning, and this impact is so great that it has produced the idea of algorithmic society and algorithmic era. Algorithms belong to the basic category of digital law and are an important tool for digital rule of law. In existing research, scholars are more concerned about how to regulate algorithms, and less concerned about the legal nature of algorithms. There has been a brief discussion on this issue. However, how the legal nature of algorithms should be defined is still inconclusive. This article aims to answer this question. To this end, this paper sorts out the existing theoretical schemes of the legal nature of algorithms and their interrelationships, including speech, due process, patents, and trade secrets, and analyzes the sources and arguments of these theoretical schemes. There are internal and external reasons why the legal nature of algorithms has formed many disputes. These reasons are also difficult to define the legal nature of algorithms, including: the algorithm ontology and carrier are easy to confuse, the algorithm and the program are indiscriminate, the algorithm and data, the result mix, and the limitation of governance rules. When seeking justification for various theoretical claims, one logical premise is easily overlooked, that is, the criteria for determining the legal nature of algorithms. This paper puts forward these standards, namely the principle of legality, the principle of respecting the fruits of labor, the principle of maintaining public safety, the principle of promoting innovation, and the principle of maximizing social benefits. On this basis, this paper examines various theoretical schemes, and finally proves that the legal nature of the algorithm is a trade secret, not something else. In terms of methodology, this paper uses the method of identification analysis to discuss various scenarios one by one. Compared with previous literature, this paper makes contributions in the following aspects: first, this paper sorts out the existing debates on the legal nature of algorithms in detail, grasps the arguments of these theoretical schemes, and specifically examines them. Second, this paper analyzes the reasons for many disputes over the legal nature of algorithms. The widespread use of algorithms has also caused heated debates about algorithms. Algorithms have been in the field of legal science for a long time, but there are frequent misunderstandings about how their legal nature should be defined, and the reasons behind them deserve careful analysis. Third, this paper clarifies the criteria for defining the legal nature of algorithms, and argues that algorithms are trade secrets on this basis. Algorithms cannot be called speech, nor are they due process. The argument that algorithms are patents is also untenable, and although algorithms have the possibility of patentability in the current legal system, algorithm patenting is not a good solution. [ABSTRACT FROM AUTHOR]

Abstract: Accurate identification of cotton top bud is important for cotton topping operation to detect cotton terminal bud accurately in field, a recognition method using Faster R-CNN (Faster Region Convolutional Neural Networks, Faster R-CNN) by integrating dynamic mechanism was proposed to solve the recognition difficulties caused by the small size of cotton terminal bud when it is topped in densely planted fields. The RegNetX-6.4GF model was used as the backbone network to improve its image feature extractor capabilities. Due to number of proposals under a higher IOU(Intersection Over Union, IOU) as well as the matching degree between anchor and the target shape affect the performance of the detector, the method proposed in this paper changed the original anchor generation mechanism by combining FPN (Feature Pyramid Network, FPN) and Guided Anchoring in RPN (Region Proposal Network, RPN), which will cause the distribution of the proposals generated by the RPN of the algorithm under different IOUs dynamically change during the training process. To adapt the dynamic change of proposals distribution, we integrated Dynamic Region Convolutional Neural Networks ( Dynamic R-CNN) in Faster R-CNN, which can dynamically adjust the value of IOU to obtain high quality proposals. And the GROIE (Generic ROI Extractor, GROIE) mechanism was inducted to extract ROI (Region of Interest, ROI) to improve the feature fusion capability. In this paper, 4 819 images of gossypium hirsutum population which contain seven leaf types were taken from the top of cotton plant at distance of 30-50 cm (medium distance) and 50-100 cm (long distance) under uniform light, oblique strong light, direct strong light, and shadows. Those images were processed as MS COCO 2017 format dataset and assigned to the training set, validation set, and test set, which contained 2 815, 704, and 1 300 pictures respectively. The experimental results demonstrated that FPS (Frames Per Second, FPS) of proposed model was up to 10.3 frames/s and the Mean Average Precision (MAP) of bud identification reached to 98.1% which was 3.2 percentage points higher than original Faster R-CNN model. The validation set were used to compare performance of mainstream recognition algorithm and proposed method. Results showed that the improved Faster R-CNN's MAP was 7.3% higher than original Faster R-CNN, which was also higher than RetinaNet, Cascade R-CNN (Cascade Region Convolutional Neural Networks, Cascade R-CNN) and RepPoints by 78.9%, 10.1% and 8.3% when IOU was set to 0.5. The improved Faster R-CNN proposed in this paper meets the accuracy and real-time requirements of cotton topping operation. [ABSTRACT FROM AUTHOR]

The existing pig counting system based on image processing are seriously affected by light conditions, and the counting accuracy is poor when pigs are crowded and obscured.To realize the intelligent pig counting, the paper proposed a counting scheme based on the improved instance segmentation algorithm was proposed in this paper. Aiming at the problems of image illumination and target edge blur, Laplace operator was used to preprocess the images. The feature extraction network of MASK R-CNN network was improved by using Resnet-152 as the Mask R-CNN feature extraction network, and the original Feature Pyramid Network(FPN) was followed by a bottom-up enhancement path, which directly fused the low-level edge features with the high-level features to improve the recognition ability of the target edge bluring. The non maximum suppression process and loss function were optimized and improved to improve the segmentation accuracy. The experiments were carried out at three different real pig farms to verify the counting accuracy, respectively in Hebei province, Jilin province and Inner Mongolia. The size of the cage in Hebei pig farm was 5.5 m × 1.8 m, with an average of 12 pigs in a single pen, with a feeding density of 1.21 pigs/m²; the size of the cage in Jilin pig farm was 5.5 m × 3.9 m, with an average of 22 pigs in a single pen, and the average rearing density was 1.03 pigs/m²; the size of the cage in Inner Mongolia pig farm was 11.4 m × 5.28 m, with an average of 80 pigs in a single pen, the average feeding density was 1.32 pigs/m². The RGB camera is was installed on the top of the pen and acquired the image in daytime. 2 400 images were collected in total, and 2000 images were selected after image preprocessing, and 1 250 images of three pig farms were selected as the original data set according to the ratio of 2:2:1. The training set and verification set were enhanced to 1 500 and 150 images and 250 images for the test set, The experimental results showed that 98 images could realize exact counting and 2 images missed 1 pig in the Jilin pig farm, the accuracy of pig counting was 98%. In Hebei pig farm, 99 images could realize the exact counting and the accuracy of pig counting was 99%. For Inner Mongolia pig farms with high feeding density, the accuracy of pig counting was 86%, among the 50 test images, 7 images missed detection, 4 images missed 1 target, 3 images missed 2 targets, The results can provide the application of the artificial intelligent in agriculture field. [ABSTRACT FROM AUTHOR]

Pomegranates are one of the economic fruits in China. The timely detection of pomegranate diseases and the corresponding preventive measures are important to increase crop yields and reduce economic losses. To tackle the issues that traditional edge detection operators usually resulted in low accuracy of the detection of pomegranate diseases and its contour, low anti-noise capability, and created false edges, this study presented an improved Prewitt operator with an adaptive threshold. Firstly, in the image pre-processing stage, the pending image was enhanced by high-frequency emphasize filter, and the high-frequency component, which represented to the details of the pomegranate sample in the image, was increased by the filter. On the contrary, the low-frequency component, which represented to the background of the pomegranate sample, was attenuated to facilitate the following image processing. Since the most original images contained the Gaussian noise, bilateral filtering was used to process the noise present in the image. The weighted average of the brightness values of adjacent pixels was used to represent the intensity of a pixel. The weighted average method was used based on Gaussian distribution. The weight calculation took into account both the Euclidean distance between the pixels and the radiation difference in the pixel range domain to better maintain edge feature information. Secondly, a fifth-order convolutional mask was proposed. The weights of the elements in the mask were set according to the properties of the Gaussian noise probability distribution to reduce the effect of noise on the algorithm. The weight of the central element of the convolutional mask was increased so that the edge information of the image had higher contrast. In terms of gradient calculation, eight direction templates were used to perform the convolutional operation of the image, then the gradient values of each direction were calculated. After that, the corresponding gradient values were obtained, and the gradient values of eight directions were combined into eight groups according to the rule of orthogonality. As a result, there was 90 degrees difference between each group of gradients. Then calculated the L2 norm of each set of gradients, and used the largest value of L2 norm as the gradient of the current pixel. It was shown that it was only necessary to calculate the convolutions in half the direction due to the convolutional results in opposite directions were inverse to each other. The gradients of the rest of the direction could be obtained by its symmetric property. The adaptive threshold was implemented through the minimum error method to decrease the probability of false detection and prune irrelevant details. The mean and variance of the target and background were calculated, meanwhile, the minimum error objective function was obtained according to the principle of the minimum error classification. The value that minimizes the objective function was taken as the optimal threshold. To demonstrate the effectiveness of the method in this paper, 607 test images were manually collected as samples in the natural environment from different angles and different illumination conditions. To further reduce the interference of light changes and suppress background noise, the pending images were converted into grayscale images. Therefore, the algorithm was tested using the converted grayscale images. The experimental results showed that the algorithm proposed in this study achieved a better definition at the edge of the lesion. Compared with the traditional edge detection operator, it achieved a higher recognition accuracy and lower running time. The peak signal to noise ratio of the obtained image was higher, and the obtained edge was complete and accurate. The algorithm proposed in this paper could quickly and accurately detect the area of pomegranate lesions, providing a fundamental reference for the future prevention of pomegranate diseases. [ABSTRACT FROM AUTHOR]

Depth estimation is critical to 3D reconstruction and object location in intelligent agricultural machinery vision system, and a common method in it is stereo matching. Traditional stereo matching method used low-quality image extracted manually. Because the color and texture in the image of field plant is nonuniform, the artificial features in the image are poorly distinguishable and mismatching could occur as a result. This would compromise the accuracy of the depth of the map. While the supervised learning-based convolution neural network (CNN) is able to estimate the depth of each pixel in plant image directly, it is expensive to annotate the depth data. In this paper, we present a depth estimation model based on the self-supervised learning to phenotype tomato canopy. The tasks of the depth estimation method were to reconstruct the image. The dense disparity maps were estimated indirectly using the rectified stereo pair of images as the network input, from which a bilinear interpolation was used to sample the input images to reconstruct the warping images. We developed three channel wise group convolutional (CWGC) modules, including the dimension invariable convolution module, the down-sampling convolution module and the up-sampling convolution module, and used them to construct the convolutional auto-encoder - a key infrastructure in the depth estimation method. Considering the shortage of manual features for comparing image similarity, we used the loss in image convolutional feature similarity as one objective of the network training. A CWGC-based CNN classification network (CWGCNet) was developed to extract the low-level features automatically. In addition to the loss in image convolutional feature similarity, we also considered the whole training loss, which include the image appearance matching loss, disparity smoothness loss and left-right disparity consistency loss. A stereo pair of images of tomato was sampled using a binocular camera in a greenhouse. After epipolar rectification, the pair of images was constructed for training and testing of the depth estimation model. Using the Microsoft Cognitive Toolkit (CNTK), the CWGCNet and the depth estimation network of the tomato images were calculated using Python. Both training and testing experiments were conducted in a computer with a Tesla K40c GPU (graphics processing unit). The results showed that the shallow convolutional layer of the CWGCNet successfully extracted the low-level multiformity image features to calculate the loss in image convolutional feature similarity. The convolutional auto-encoder developed in this paper was able to significantly improve the disparity map estimated by the depth estimation model. The loss function in image convolutional feature similarity had a remarkable effect on accuracy of the image depth. The accuracy of the disparity map estimated by the model increased with the number of convolution modules for calculating the loss in convolutional feature similarity. When sampled within 9.0 m, the root means square error (RMSE) and the mean absolute error (MAE) of the corner distance estimated by the model were less than 2.5 cm and 1.8 cm, respectively, while when sampled within 3.0m, the associated errors were less than 0.7cm and 0.5cm, respectively. The coefficient of determination (R²) of the proposed model was 0.8081, and the test speed was 28 fps (frames per second). Compared with the existing models, the proposed model reduced the RMSE and MAE by 33.1% and 35.6% respectively, while increased calculation speed by 52.2%. [ABSTRACT FROM AUTHOR]

In comparison with the former ZUC, this paper adopted a different way of design in initialization and the generation of MAC in ZUC-256 algorithm, and significantly promoted the cryptographic performance. In order to implement the algorithm by software in optimized and efficient way, this paper introduced the general structure and workflow of the algorithm in the first, and then implemented the basic functions of the algorithm by programming referring to the original algorithm. Furthermore, taking into consideration the transfer execution process of the array parameter in programming function together with the computational characteristics of the algorithm itself, this paper analyzed the factors affecting the execution efficiency of the program, and then promoted basic function program previously completed. After the actual programming test, the effectiveness of the two optimization methods is verified. In general, in two different software running reference machine environments, the optimized code execution efficiency combined with the two optimization schemes is improved 8% ~ 9% and 18% ~ 36% higher than that of the basic function implementation code respectively. It has practical significance for the future application of ZUC-256 cryptography in software engineering. [ABSTRACT FROM AUTHOR]

The symptoms of the same tomato disease in different stages are obviously different, and different diseases show some similarities. Traditional pattern recognition methods can not reflect the dynamic changes of the pathological characterization, and the practicability is poor. To solve this problem, this paper proposed a Multi-Scale AlexNet recognition model for mobile platform based on convolutional neural network (CNN), and implemented a tomato leaf disease image recognition system for agricultural workers based on Android. Many parameters and large memory utilization of traditional AlexNet model are unfit for mobile platform, this paper adjusted the network structure of the traditional model by removing the local response normalization(LRN) layer, modifying the full connection layer, setting up different convolution kernel extraction features, designed a multi- scale recognition model based on the AlexNet. The model consists of 6 layers. It can optimize the training time and memory utilization and achieve high precision. After removing the LRN layer, there was a 30% decrease in running time. Extending the single convolution kernel into multi-scale (1 × 1,3 × 3,5 × 5,7 × 7) convolution kernel then fused at the first layer, removing full connection layer 6 and 7, and taking the place of global average pooling layer, then the model size was only 30.2 M. The forward propagation rate (FRP) and backward propagation rate (BPR) were reduced, and the global average pooling is better than the global maximum pooling on recognition accuracy. So the Multi-Scale AlexNet model used global average pooling in the 5th layer. In image preprocessing phase, in order to avoiding over fitting of the trained model caused by the unbalanced distribution of sample numbers, we had zoomed, flipped, color jittering, add noise and rotated the original pictures of dataset randomly to get the augmented dataset, and used 70% of the pictures as the train dataset and the rest as the validation dataset(20%) and test dataset(10%). These pictures were quantized to 224 × 224 dpi for Multi-Scale AlexNet training, and the original dataset and augmented dataset were used to train models. In order to validate the performance of the proposed model, comparative tests were done between Multi-Scale AlexNet and traditional pattern recognition method. It repeated 600 tests. The results showed that the CNN model achieved 92.7%, the high average recognition accuracy of each disease and each disease in the early, middle and late stages. Compared with the other CNN Net model(MobileNet, SequeezeNet, LeNet-5), the Multi-Scale AlexNet achieved the highest recognition accuracy, and reached 95.8% on the late stage disease dataset, but the SequeezeNet model used less memory. The MobileNet and SequeezeNet model reached lower recognition accuracy on the middle and late stages dataset, that because their convolution size was small. The recognition system was implemented on Android platform, and then test was done on field dataset. The results showed that the average recognition accuracy was 89.2%, its less value was due to the complex background of image. Then the system can meet the requirements of disease image recognition on mobile platform in production practice. The research results provide a method for disease image recognition based on convolution neural network, and provide a reference for automatic identification of crop diseases and engineering applications. [ABSTRACT FROM AUTHOR]

With the advancement of remote sensing technology, the high spatial resolution remote sensing image contains rich special information with a great detail. At the same time, the complexity of high spatial resolution remote sensing images also requires higher the classification technology of remote sensing images. However, in the face of high spatial resolution remote sensing image more obvious geometrical structure and the more rich texture characteristics, how to design rational system of characteristics, select the appropriate sorting algorithms to accurately and quickly grasp the number of rural land of building and its distribution status, are of great significance to balance urban and rural areas, save land, and realize sustainable development. This will help in exploring the application of deep learning model in high spatial resolution remote sensing image building extraction, and have research significance for improving the classification accuracy of high resolution remote sensing image. In this paper, the semantic segmentation model (SegNet) was used for extracting buildings. SegNet is mainly composed of encoder network, decoder network and pixel-wise classification layer. The encoder network transforms high-dimensional vectors into low-dimensional vectors, enabling low-dimensional extraction of high-dimensional features. The decoder network maps low-resolution feature maps to high spatial resolution feature maps, realizing the reconstruction of low-dimensional vectors to high-dimensional vectors. The softmax classifier separately classifies each pixel, which outputs the probability that each pixel belongs to each class. In this paper, a 3000 pixel × 3000 pixel and two 2000 pixel × 2000 pixel slices were taken from the global remote sensing image of Bazhou City, Hebei Province as training samples, and a 3000 pixel × 3000 pixel slice was taken as the verification sample. In this paper, five comparative experiments were used to extract the buildings, including PSPNet, support vector machine, random forest, ISO clustering and maximum likelihood method. The confusion matrix of each classification method was obtained by calculating the difference between the classification results of the comparison experiment and the real value. From the traditional classification algorithm to the shallow learning algorithm to the deep learning algorithm, the Kappa coefficient and overall accuracy of classification kept constantly increasing, among which SegNet semantic segmentation algorithm based on the deep convolutional network performed better than the other five algorithms in extracting buildings from high spatial resolution remote sensing image. The Kappa coefficient and the overall accuracy of SegNet semantic segmentation algorithm were 0.90 and 96.61%, respectively, and the ground truth value was basically the same as the classification result. The F1Score of building extraction of SegNet semantic segmentation algorithm based on deep convolution network was 0.91, but the other five algorithms were below 0.87. SegNet had the lowest error rate of 9.71% for buildings, indicating that the ability to identify buildings of semantic segmentation algorithm from high spatial resolution remote sensing was superior to traditional classification algorithms, shallow layer learning algorithms based on machine learning, and PSPNet semantic segmentation algorithm based on deep convolution network. The Kappa coefficient and overall accuracy of the remaining five classification algorithms were respectively below 0.83 and 94.68%, and the difference between the ground truth value and the classification result was relatively large. SegNet can not only make use of spectral information but also make full use of abundant spatial information. During SegNet training, more essential features can be learned, and more ideal features suitable for pattern classification were finally formed, which can enhance the ability of convergence and generalization of the model and improve the classification accuracy. Traditional classification algorithms, such as ISO clustering and maximum likelihood method, failed to make use of the rich spatial information of the high-resolution remote sensing image, so the accuracy was relatively low. Due to limited computing units and large amount of high spatial resolution remote sensing image data, shallow layer learning algorithms based on machine learning such as support vector machines and random forest cannot effectively express complex features of ground objects, so their advantages are not obvious in building extraction from the high spatial resolution remote sensing images. The experimental results showed that the SegNet based on deep learning has the best performance, and it has important theoretical significance to explore the application of deep learning model to remote sensing image classification methods. At the same time, the research results also provide a reference for improving the classification accuracy of high resolution remote sensing images. [ABSTRACT FROM AUTHOR]

Identification of plastic-mulched farmland using UAV image is quite few. This paper proposes a method of combining with texture features, image morphology algorithm and threshold segmentation algorithm to extract plastic-mulched farmland using UAV Red-Green-Blue (RGB) images. In order to test the performance of this method, this paper took Ludian County of Zhaotong City, Yunnan Province as the research area, and obtained 2 images in the research area as experimental data. The complexity of land cover type in the 2 images was different. In complex area, the main land cover types included vegetation, impervious layer (building and road), plastic-mulched farmland (mainly black plastic mulch with a small amount of white plastic mulch), and bare soil (containing the plastic residues of a previous year). In simple area, the land cover types were similar with those in complex area; however, all plastic-mulched farmland was covered by black plastic mulch and there were no plastic residues in bare soil. Firstly, we calculated the gray level co-occurrence matrix of 2 images in different window sizes (3×3, 5×5, 7×7, 9×9, 11×11, 13×13, 15×15), directions (0, 45°, 90° and 135°) and steps (1, 2 and 3) and extracted 8 texture features from each band of RGB images including mean, variance, synergy, contrast, dissimilarity, information entropy, second moment and correlation. Secondly, we combined the original RGB image with different texture features to make maximum likelihood classification and determined the best extraction parameters of the texture features by comparing the overall pixel accuracy, user accuracy and product accuracy of the plastic mulch in complex area. The best extraction parameters of texture features were the window size of 15×15, the direction of 0, and the step of 2, which were also used to extract texture features of the image in simple area. Thirdly, we selected the optimal texture combination based on importance evaluation of texture features using Random Forest Algorithm and combined them with original UAV RGB image to make maximum likelihood and get preliminary classification maps in both complex area and simple area. Fourthly, we recoded the preliminary classification maps into binary maps (1 refers to plastic mulch and 0 refers to the other land cover types) and made majority filtering to remove noises (such as the plastic residues of a previous year). Then, we used image morphology algorithms to convert the strip plastic mulch into the plastic-mulched farmland and set area threshold to extract plastic-mulched farmland distribution. The area thresholds were 35 m2 in complex area and 500 m2 in simple area. Finally, taking the digitized mulched farmland as references (ground truth data), the accuracy of the recognition results of mulched farmland was assessed by error matrix and area error. The results showed that the texture features extracted by the optimal parameters could greatly improve the classification accuracy. The image morphology algorithm and the threshold segmentation method could effectively extract the block-shaped plastic-mulched farmland. The overall accuracy, Kappa coefficient, product accuracy, user accuracy and area error were 94.84%, 0.89, 92.48%, 93.39%, 0.38% in complex area, and 96.74%, 0.93, 97.39%, 94.63%, 1.95% in simple area, respectively. Compared with step and direction, the size of window had greater influence on plastic mulch classification accuracy. Among 8 texture features, mean contributed most to extracting plastic mulch. The method of extracting plastic-mulched farmland based on the fusion of supervised classification and image morphology algorithm proposed in this paper can provide reference for the development of identification algorithm about plastic-mulched farmland. [ABSTRACT FROM AUTHOR]

An important goal of hydrobiology is the simulation, reconstruction and restoration of important fish habitats, in which fish species form aquatic ecosystems’ climax communities, enabling structural and functional restoration of river ecosystems. Traditional methods for identifying key fish habitat locations such as spawning grounds, include fish resource surveys, observation of fish spawning behavior and interviews with fishermen. But these are subject to problems including poor accuracy and large error. Precise positioning of fish can accurately locate key habitats (such as spawning grounds) based on key life cycle phases (such as spawning periods), permitting observation of corresponding habitat parameters. Fish movement trajectory data can also deepen understanding of fish habits and habitats, permitting suitable habitat indicators to be scientifically determined, and providing theoretical and technical support for fish protection and habitat restoration efforts. Ultrasonic tag tracking technology is widely used in fish behavior research due to its long underwater propagation distance and broad applicability. But most existing researches derived fish movement trajectories from hydrophone data using equipment manufacturers’ software or services, and few articles concerning fish positioning principles and methods optimized for natural aquatic environments have been published. The Chan’s algorithm (1994) in literature[24] and robust least squares estimation were combined to get the location of ultrasonically-tagged fish in this paper, leveraging the strengths of these methods to overcome their disadvantages when used singly. Chan’s algorithm was first used to obtain approximate coordinates of fish, which were used as initial position estimates from which the final position estimates were obtained with robust least squares. Prior information such as water depth and fish swimming speed could also be taken into account, making the proposed positioning method well-suited for dealing with ultrasonically-tagged fish in natural aquatic environments. The proposed method was suitable for existing ultrasound hydrophones, and effectively solved problems with large observation errors. Based on these research results, the UWP (under water positioning) software package was developed. To verify the effectiveness of the proposed method, an observation network was constructed which consisting of 16 hydrophones uniformly distributed over a area of 120 m×120 m in Huangbai River, Yichang. 4 ultrasonic signal tags were used to evaluate the positioning results, 2 was co-located with hydrophones for static simulation, and the other 2 affixed to a boat hull for dynamic simulation. Comparisons with Beidou/GNSS RTK with centimeter-level accuracy positioning estimates over 115 groups of test results, using millisecond-level accuracy observation data from existing hydrophones, swimming trajectories of ultrasonically-tagged fish could be obtained to an accuracy of about 2.15 m. While complex water environments degraded this accuracy, where single observations contained gross errors exceeding 10 m, 100% of these errors could be identified. The success rate for identification of observations with gross error was a gradually declining function of gross errors, dropping to 84.3% for 3 such observations. With over 3 gross error-bearing observations, the success rate declined significantly. With over 5 gross error-bearing observations where gross error-bearing observations accounted for over 31.25% of all observations, application of the proposed method could detect the error’ existence, but was unable to identify the error-bearing observations effectively. The ultrasonic tag precise positioning method of fish proposed in this paper provide an effective method for determining the accurate swimming trajectory of fish in rivers, lakes and seas with low visibility. In addition, by modifying the data communication interface, this method can be effectively applied to ultrasonicall-taggeds fish and hydrophones of different manufacturers. In the future, it can play a more important role in promoting ecological environmental protection, and human beings’ understanding of ecological and behavioral evolution in the aquatic environment at the population level. [ABSTRACT FROM AUTHOR]

It is the key to prevent and control dairy cow diseases by real-time monitoring and detection of abnormal behavior early. In this field, it is the most common for the detection and recognition of lame behavior especially. However, the technology of the livestock abnormal behavior recognition is still facing the problem of poor real-time performance. Regarding the above problems, 2 improvement strategies were proposed in this paper. Firstly, a sparse overcomplete dictionary learning algorithm based on conjugate gradient pursuit-KSVD (CGP-KSVD) algorithm was proposed which was used to description and representation of lame behavioral semantic features. The proposed idea was to optimize search direction in the stage of sparse coding construction though introducing a conjugate gradient pursuit algorithm. Meanwhile, the proposed method could avoid the computational load caused by storing and calculating large-scale Hessen matrix effectively. By this way, it achieved the high convergence speed of the sparse overcomplete dictionary learning. Secondly, the basic features of representing the videos were extracted by the fusion of spatio-temporal interest points and spatio-temporal interest points extracted in the dense trajectory map. With the proposed method, it is possible to reduce the redundancy features and the computational load while retaining the rich details. To verify the proposed algorithm, we designed 3 experiments. Experiment 1 was used to achieve the recognition accuracy by comparing the proposed algorithm with the classical sparse dictionary learning algorithms. Experiment 2 and Experiment 3 were used to study the efficiency of the algorithm. In the experiments, 500 black and white adult cows were selected randomly and a total of 1200 video samples were achieved for training and testing by the manual marks. In Experiment 1, the average accuracy of the algorithm proposed in this paper is 5.5 percentage points higher than that of LLC (locality-constrained linear coding) and 3.9 percentage points higher than that of K-SVD. The fusion feature proposed in this paper was used as the basic features of LLC and K-SVD and the above methods were recorded as LLC + fusion feature and K-SVD + fusion feature. The experiment results show that the average accuracy of the proposed algorithm is 3.8 and 1.4 percentage points higher than that of LLC + fusion feature and K-SVD + fusion feature, respectively. In Experiment 2 and 3, the theoretical analysis results show that the algorithm proposed in this paper has the lowest computational time complexity, and the average response time of the CGP-KSVD algorithm is 0.043 s, faster than the 2 other algorithms, BP-KSVD (basis pursuit-KSVD) and OMP-KSVD (orthogonal matching pursuit-KSVD). And the 90-hour video test results show that the CGP-KSVD algorithm has the highest recognition rates of 93.3% and 92.7% respectively with the playback video and online test video, which are increased by 33.3 and 35.6 percentage points compared with the BP-KSVD algorithm, and 13.3 and 14.1 percentage points compared with the OMP-KSVD algorithm. [ABSTRACT FROM AUTHOR]

The existing pig counting system based on image processing are seriously affected by light conditions, and the counting accuracy is poor when pigs are crowded and obscured.To realize the intelligent pig counting, the paper proposed a counting scheme based on the improved instance segmentation algorithm was proposed in this study. Aiming at the problems of image illumination and target edge blur, Laplace operator was used to preprocess the images. The feature extraction network of MASK R-CNN network was improved by using Resnet-152 as the Mask R-CNN feature extraction network, and the original Feature Pyramid Network(FPN) was followed by a bottom-up enhancement path, which directly fused the low-level edge features with the high-level features to improve the recognition ability of the target edge bluring. The non maximum suppression process and loss function were optimized and improved to improve the segmentation accuracy. The experiments were carried out at three different real pig farms to verify the counting accuracy, respectively in Hebei province, Jilin province and Inner Mongolia. The size of the cage in Hebei pig farm was 5.5 m × 1.8 m, with an average of 12 pigs in a single pen, with a feeding density of 1.21 pigs/m²; the size of the cage in Jilin pig farm was 5.5 m × 3.9 m, with an average of 22 pigs in a single pen, and the average rearing density was 1.03 pigs/m²; the size of the cage in Inner Mongolia pig farm was 11.4 m × 5.28 m, with an average of 80 pigs in a single pen, the average feeding density was 1.32 pigs/m². The RGB camera is was installed on the top of the pen and acquired the image in daytime. 2 400 images were collected in total, and 2000 images were selected after image preprocessing, and 1 250 images of three pig farms were selected as the original data set according to the ratio of 2:2:1. The training set and verification set were enhanced to 1 500 and 150 images and 250 images for the test set, The experimental results showed that 98 images could realize exact counting and 2 images missed 1 pig in the Jilin pig farm, the accuracy of pig counting was 98%. In Hebei pig farm, 99 images could realize the exact counting and the accuracy of pig counting was 99%. For Inner Mongolia pig farms with high feeding density, the accuracy of pig counting was 86%, among the 50 test images, 7 images missed detection, 4 images missed 1 target, 3 images missed 2 targets, The results can provide the application of the artificial intelligent in agriculture field. [ABSTRACT FROM AUTHOR]

Ochotona curzoniae is an endemic species and key species in the alpine meadow of the Tibetan Plateau and also it is a main kind of organism that destroys the grassland ecology of the plateau. In order to prevent the dangers of the ochotona curzoniae, we need to study the living habits of ochotona curzoniae and investigate the degree of harm of ochotona curzoniae, and then we can control the number of ochotona curzoniae through the effective preventive measures. With the development of sensing technology and image processing, we can provide an objective basis through intelligent monitoring system to control the damage of ochotona curzoniae. The object detection of ochotona curzoniae is a key technology in the intelligent monitoring equipment because it can provide the object contour feature for behavior analysis of ochotona curzoniae. The object detection of ochotona curzoniae is very difficult, because the ochotona curzoniae video possesses the characteristics of complex background, low contrast, the object color with intensity inhomogeneity, diversity and mutability. The traditional object detection method cannot extract the object contours accurately. This paper presents a fast object detection method based on space-time domain. Firstly, the centroid position of the object in the current frame image is determined by the background subtraction, and then the rough segmented image and the initial contour are obtained based on the centroid position. The rough segmented image is segmented by the improved Chan-Vese model, and then we can obtain the object contours. In view of the fact that the level set function needs to be initialized in the process of improved Chan-Vese model, and the initialized computation is enhanced with the increase of the image scale, the centroid of the object is taken as the center to intercept the image containing the object as the roughly segmented image. Then, the improved Chan-Vese model is used to segment the roughly segmented image, so as to reduce the time consumption of Chan-Vese model segmentation. In addition, as Chan-Vese model can’t fully segment the image of object whose color is diverse and mutable, we use the improved Chan-Vese model to segment the roughly segmented image. The internal pixels of image evolution contours were processed by K-means clustering, and the clustering center point values were obtained. The internal fitting values of Chan-Vese model were constructed by the clustering center point values and the image mean filtered intensity information, thereby improving the adaptability of Chan-Vese model for complex object image segmentation．In addition, rectangular Dirac function was used to replace regularized Dirac function in the energy function of Chan-Vese model, and the calculation of level set evolution equation could be limited to the zero level set so as to avoid the influence of the image background disturbance on the segmentation result. In this paper, the video processing with 50 frames of images shows that the time consumption of this method is only 15.25 s, the average value of Dice similarity coefficient is 0.852 929, and the average value of Jaccard index is 0.744 57. In summary, the object detection method proposed in this paper can accurately extract the object contour and has a high real-time performance. [ABSTRACT FROM AUTHOR]

The maternal behaviors reflect the health and welfare of the sows, which directly affect the economic benefit of the pig farm. Computer vision provides an effective, low-cost and non-contact method for monitoring the behavior of animal for precision farming. Under the scene of piggery, it is a challenge for 24-hour automatic recognition of lactating sow postures due to the daily illumination variations, influence of heat lamp, and adhesion between piglets and sows. This paper proposed an automatic recognition algorithm of lactating sow postures based on improved Faster R-CNN (convolutional neural network) using depth video images. To improve the recognition accuracy and satisfy the real-time need, we designed a ZF-D2R (ZF with deeper layers and 2 residual learning frameworks) network by introducing residual learning frameworks into ZF network. First, 3 convolutional layers were added in the ZF network to design ZF-D (ZF with deeper layers). Then, in ZF-D network, shortcut connections were used to form 2 residual learning frameworks. The whole network made up the ZF-D2R network. Moreover, the Center Loss was introduced to Fast R-CNN detector to construct a joint classification loss function. With the joint supervision signals of F-SoftmaxLoss and Center Loss in Fast R-CNN detector, a robust model was trained to obtain the deep feature representations with the 2 key learning objectives, which led to intra-class compactness and inter-class dispersion as much as possible. So, the joint supervision of F-SoftmaxLoss and Center Loss could reduce recognition errors caused by the similar features between different postures. By taking ZF-D2R as basic net and adding the Center Loss to Fast R-CNN detector, the improved Faster R-CNN was built. Experiments to obtain the actual data set of lactating sow posture from the depth video of sows in the 28 pens were performed. The data set included 2 451 standing images, 2 461 sitting images, 2 488 sternal recumbency images, 2 519 ventral recumbency images and 2658 lateral recumbency images. And 5 000 images were randomly chosen as the testing set. The rest of the images were used as training set. To enhance the diversity of training data, dataset augmentation including rotating and mirroring was employed. Based on the Caffe deep learning framework, our improved Faster R-CNN was trained with end-to-end approximate joint methods. By adding 2 residual learning frameworks to ZF-D, the ZF-D2R model improved the MAP (mean of average precision) by 1.28 percentage points. After introducing the Center Loss supervision signal, the MAP of the optimal model reached 93.25%, obtaining an increase of 1.3 percentage points, and the MAP of the method proposed achieved 93.25%. And APs (average precisions) of the 5 classes of postures i.e. standing, sitting, sternal recumbency, ventral recumbency and lateral recumbency were 96.73%, 94.62%, 86.28%, 89.57% and 99.04%, respectively. The MAP of our approach was 3.86 and 1.24 percentage points higher than that of Faster R-CNN based on ZF basic net and Faster RCNN based on the deeper VGG16 basic net, respectively. Our method processed images at a speed of 0.058 s per frame, 0.034 s faster than Faster R-CNN based on VGG16. Our proposed method could improve the recognition accuracy and simultaneously ensure the real-time performance. Compared with DPM (deformable part model) detector plus CNN posture classifier, the MAP of the end-to-end recognition method proposed in this paper was increased by 37.87 percentage points, and the speed was raised by 0.855 s per frame. Our method can be used for the 24-hour recognition of sow behaviors and lays the foundation for the analysis of sow dynamic behavior by video. [ABSTRACT FROM AUTHOR]

White Agaricus bisporus is an excellent source of the B vitamins, riboflavin, niacin, and pantothenic acid and also a good source of the dietary mineral phosphorus. It is one of the most commonly and widely consumed mushrooms in the world. The production of white Agaricus bisporus has been industrialized in China. However, during the last production chain, it needs a lot of labors to sort and grade the white Agaricus bisporus. The manual sorting and grading mode has many disadvantages such as larger error, low productivity, non-uniform standard, and so on. With the development of machine vision technology, it has been successfully used for automatic inspection and sorting, especially in agricultural industry due to its nondestructive characteristic. An automatic sorting and grading system based on machine vision was designed in this paper. Firstly, an automatic sorting and grading hardware system was designed. It included conveying mechanism, image acquiring system, control module and actuator. The conveying mechanism consists of the fixed support, conveyor, roller, driving unit, tension unit, cleaner and guide plate. The image acquiring system consists of a line scan camera, lens, light source and its controller. The control module consists of photoelectric sensor controlling the camera, electromagnetic valve and relay starter controlling the actuator. The actuator consists of the air compressor, duplex pieces, muffler, air cylinder, joints, gas channel, guide rod, slider, connector, flap, baffle, and so on. Secondly, it put forward an image algorithm based on the watershed method, Canny operator, OR operation and closed operation to determine the diameter of white Agaricus bisporus. The first watershed algorithm combining the global threshold segmentation method and maximum entropy threshold segmentation method is used to remove the shadow of image. The second watershed algorithm based on Canny operator, OR operator and closed operation is used to remove the disturbance of petiole. The minimum enclosing rectangle method is used to get the diameter of white button mushroom. Thirdly, a precise control strategy based on the conveyor speed, distance between trigger and flap piece, trigger time and algorithm processing time was designed. Finally, the software based on OpenCV 2.4.10 and Visual Studio 2010 was developed in this paper to acquire, analyze and process the image and output the control instruction to control the valve by USB-4761 module. It also has the visual parameter configuration function for camera, control module and grading standard. When the whole system starts at the first time, the light source of image acquiring system is adjusted by the light controller to make the image clear and stable. The air compressor is powered on and pressurized to the rated pressure of 0.3 MPa. The motor of conveyor is started and the speed of the conveyor belt is adjusted according to the actual production requirement. The white Agaricus bisporus goes into the guide bar with the conveyor belt. When the white Agaricus bisporus goes into the region of image acquiring, it triggers the industrial camera to scan. The image data are transmitted to the industrial computer by image capture card. The software analyzes the image on line. The analyzed result is sent to the related solenoid valve through the digital control module. When the white Agaricus bisporus arrives near the container of related grade, it is sorted into the container by the related flap piece driven by the solenoid valve. In order to validate the applicability and reliability, the test of grading performance and effect was carried out with the prototype of sorting and grading system at 12.7 m/min conveying speed and 1900 Hz line frequency. The results showed that the average maximum grading speed was 102.41 pieces/min, the accuracy of grading was 97.42%, the damage rate was 0.05% and the undetected rate was 0.96%. The grading speed improved by 38. 86%, the accuracy improved by 6.84% and the damage rate reduced by 0.13% compared to the manual grading. The system can stably and continuously operate. For long time grading, the advantage of intelligent system is more obvious due to the fatigue of labor. The whole system realizes the on-line automatically sorting and grading for fresh white Agaricus bisporus with the minimum destruction. [ABSTRACT FROM AUTHOR]

Body shape parts acquisition and recognition from moving cattle are difficult to realize automatically especially with the similar color and texture frames or images. Usually, the cattle’s abnormal behavior is hardly detected by human because abnormal behavior detection needs continuous observation, but we can solve this problem with our method. In this paper, we use the Microsoft Company's smart vision sensor (named Kinect) to collect the information of 2 modals from cattle movement (the depth information and RGB information, DRGB). The depth information can be acquired by the infrared sensor and the depth value means the distance between the object and the sensor. The RGB information can be acquired by the normal camera. Based on the color modal information, we propose a randomized nearest neighbor pixel comparison (RNNPC) method to snatch at continuous action frames without static frames. By comparing the distinction of pixels between 2 adjacent frames, we can judge whether there is obvious or micro actions appearing in these 2 neighboring images. And if the distinction value is more than the threshold, the camera would record the cattle’s continuous movements or actions automatically. Based on the depth modal information, we can calculate the mean value of depth which is obtained in the dynamic region by the RNNPC method. And with the mean value of depth, we can filter and transform the invalid pixels into dark pixels in the continuous frames. Meanwhile, the intact shape of the cattle in the original picture is preserved by our method (save the pixels with no change). From the results of filtration, we can see that the original image background area is reduced, and the method (we use SelectiveSearch in this paper) also directly reduces the number of candidate regions compared with the conventional recognition algorithm. According to the training samples scale, the network performance and the characteristic of the key parts of cattle from continuous frames, we adjust the parameters of the deep convolutional neural network, and we set the iterating upper limit and finish the network training when the number of iterating reaches this limit. Finally, with the trained network, we can realize the recognition of key parts from moving cattle in the processed continuous frames. The experimental results show that RNNPC method can save 72% of storage space, and the ratio of valid data for the other 38% of continuous frames can reach 94%. By filtering the invalid pixels, the invalid background information or objects of the original images can be almost removed, and the number of candidate objects generated by the conventional object recognition algorithm can be reduced by an order of magnitude. Compared to the original image, 90% candidate regions are reduced with SelectiveSearch algorithm on DRGB image. By adjusting network parameters, we can improve the convergence speed of deep convolutional neural network in training the samples with similar color and texture, and the single training iteration time does not significantly increase. The average classification accuracy of the net can reach 75.88%, and the image processing rate is 4.32 FPS, and under the same condition, the effect is better than that by the original Fast RCNN. By using the method mentioned above, we can realize body shape parts acquisition and recognition from moving cattle. [ABSTRACT FROM AUTHOR]

Virtual rice technology has played an important role in modern agricultural production decision-making, yield prediction, crop breeding and growth conditions optimization. In the visualization of virtual rice, the phenomenon of interpenetration between organs is often found. There is need to use collision detection technology to avoid this phenomenon. However, when the number and size of the population increase, there will be a lower collision detection efficiency. To improve the efficiency of leaf collision detection during the dynamic growth simulation of rice population, methods for rapid construction of mixed leaf tree (MLT) and fast detection of CPU/GPU are proposed in this paper. The main ideas of these methods are based on the morphological structure characteristics of leaf parabola and the accelerating characteristics of CPU/GPU. On the single leaf scale, a new calculation of OBB bounding box direction axis method is proposed to reduce the construction complexity of OBB bounding box, and the MLT is constructed based on the upper axis aligned bounding box (AABB) and the lower oriented bounding box (OBB). The AABB bounding box is built to quickly exclude the disjointed leaf pairs and the OBB bounding box is built to ensure the accuracy of the collision detection. According to the morphological characteristics of rice leaves, the calculation method of the new OBB bounding box direction axis is proposed to replace the traditional calculation method based on covariance matrix and mean value, which reduces the complexity of OBB bounding box construction. The new calculation method is applicable to the cases in which the leaves of rice are not twisted, curled, broken, and so on. In these cases, the mid-vein curve of the rice leaves can be considered as the first order derivable and the second order continuous. The connection line between the starting point and the end point in the mid-vein curve is a direction axis of the OBB bounding box, in the initial structure of rice leaves, the z-axis is the other axis of the leaf OBB bounding box, and the third direction axis can be determined by the 2 determined direction axes. On the group scale, firstly, according to the regularity of rice population cultivation and the law of rice growth, the rule of collision detection between rice is proposed: No collision detection is calculated between rows or columns of non-adjacent plants, and if row or column distance is greater than the sum of the length of the 2 longest leaves of rice, there is no collision detection between them. Use these rules to reduce the number of rice leaves for collision detection. Then the CPU/GPU acceleration scheme was designed by using the dependence of the collision detection between the leaves of the individual plants and rice population: The construction of the rice population MLT on the CPU side is carried out, and the intersection of the calculation on the GPU side is calculated. Each thread block represents the result of the intersection detection of a leaf, and for each thread in the thread block a crossing detection is calculated between a pair of leaves; when the thread block in the calculation of all the results has no collision, it is determined that thread block represents no collision, otherwise it is determined to have the collision. After all the calculation is completed, the intersection test result is returned to the CPU side for processing. In order to verify the effectiveness of this study, the experiment of collision detection efficiency was carried out with the leaves of large-scale rice population at tillering stage. The results show that the time consumed by collision detection method proposed in this paper is 50% less than the traditional AABB and OBB method, which effectively improves the collision detection speed between leaves. When the rice population is large, the running time of CPU/GPU parallel acceleration is 98% less than that of the CPU, and the collision detection efficiency is greatly improved. [ABSTRACT FROM AUTHOR]

Agricultural implements tilt angle measurement is one of the key technologies to achieve agricultural implements and equipment precision operations. For example, the precision navigation control and the leveling control of agricultural implements are all dependent on the accurate measurement of tilt angle. What’s more, agricultural implements of tilt angle are one of the key parameters of agricultural mechanics modeling and agricultural implements safety warning learning. In order to further improve the quality of agricultural implements operation, we developed a new agricultural implement tilt angle measurement system in this paper and verified by tests on triaxial turntable platform and field. Modern micro-electromechanical systems (MEMS) technologies provide the moderate-cost and miniaturized solutions for the development of attitude reference system. Using highly-integrated inertial measurement units (IMUs) ADIS16445 provided by ADI company and micro ARM processor STM32F446 provided by ST company, we built the hardware platform. ADIS16445 ISensor® includes tri-axial gyroscopes and tri-axial accelerometers, the raw sensors data was sampled by STM32F446RC processor through SPI interface. The attitude calculation was carried out based on the Euler angle algorithm. The Kalman filter model with four state vectors and two observations was established to fuse the accelerometer and gyroscope information to achieve the accurate measurement of the tilt angle of agricultural implements. Considering the zero bias and drift characteristics of the gyroscope and the motion characteristics of the MEMS micro sensor, adaptive error covariance matrix Q and R rules were established to achieve precise tilt angle measurement of agricultural implements under different working conditions. Tests were conducted on SGT320E triaxial turntable platform and ZP9500 high level sprayer provided by LOVOL company dual in the field with the assistance of antenna positioning and attitude module BD982 provided by Trimble company. The SGT320E triaxial turntable platform was the standard equipment for testing the angular motion parameters and inertial systems. By setting the triaxial motion parameters to simulate a variety of motion states, it had speed, position and sine swing modes on all triaxial with a rate resolution of 0.0001°/s. In this paper, we used six position accelerometer calibration method and gyroscope error model to verify the performance of accelerometers and gyroscopes. Three-axis multi-function turntable test results showed that ADIS16445 built-in gyroscopes’ and accelerometers’ zero bias were under 0.15°/s and 0.075 mg, qualified to meet the system design hardware requirements. Kalman fusion algorithm were more accuracy and effective compare to simple integral by gyroscope and can solve the problem of zero bias and drift characteristics of the gyroscope with tilt static measurement error accuracy was 0.15°, typical dynamic measurement accuracy was 0.3°, maximum measurement error was less than 0.5°. The BD982 supports high precision positioning, attitude and heading output with high stability and fast dynamic response, which is widely used in construction implements, automobiles, agricultural implements and other fields, making it to be the leader of the industry. In this paper, the baseline length was 1.4 m with the measurement accuracy of 0.1°. Test results from high level sprayer showed that the average error of the attitude inclination was less than 0.55°, maximum measurement error was less than 0.91°, which satisfied the precise operation requirement of the agricultural equipment. Test results also verified that self-adaptive Kalman filter algorithm was more accuracy and stable than normal Kalman filter algorithm, which made the system development by this paper have more applicability. The agricultural implements tilt angle measurement system developed in this paper not only can reducing costs but also can improving the quality of agricultural implements operations. [ABSTRACT FROM AUTHOR]

Accurate crop-load estimation is very important for efficient management of nutrients and harvest operations. Current machine vision techniques for crop-load estimation have achieved only limited success mostly due to partial occlusion, shape irregularity, varying illumination and multiple sizes. Detecting immature green fruit is a more challenging task for similar color of fruit and background. The key starting point of this paper for detecting immature citrus fruit was the observation that the light distribution on citrus fruit follows a general pattern in which the light intensity decreases with the distance from a local maximum due to specular reflection. Immature citrus fruit detection was achieved by detecting this pattern with concentric circles or parts of circles. This pattern was proposed with the maximally stable extremal region (MSER) method and validated by hierarchical contour analysis (HCA), and was the first proposed in this paper. The images were captured by a color camera under low natural light conditions with a flashlight, and the green component of the color images was used for further analysis. After smoothing the whole image by Weiner filter, the regions of interest (ROIs) in the image were extracted by the method of MSER. The ROIs detected by MSER were those whose support was nearly the same over a range of thresholds, so the regions on citrus fruit were detected by MSER for the pattern that the light intensity decreases stably and gently with the distance from a local maximum. However, many regions on leaves and background were also detected as ROIs and should be excluded in the next step. A novel algorithmic technique was proposed to remove these regions on background, and this method was named as the HCA. Firstly, shape analysis was used for each ROI and only those ROIs were considered as valid if the shape was nearly circular. Secondly, multiple levels of contours around each valid ROI were extracted and fitted with the circular Hough transform (CHT). Lastly, multiple fitted circles would be merged into one if their most parts were overlapped together, this step was called circle merging and the merged circles were considered as the last detected citrus fruits. The algorithm was tested on a testing dataset with 20 images and achieved the recall rate of 81.2% and the precision rate of 83.5%. The harmonic average of them was 82.3%. The result showed that 24.8 fruits were correctly detected, 5.7 fruits were not detected and 4.9 ROIs were wrong detected as fruit in each image on average. The processing time of the proposed method was 3.70 s totally on each image, on average, in which 0.57 s was used for MSER detection and 3.13 s was used for HCA. The result showed that the proposed method can detect green citrus fruit in a very difficult and challenging scene with so many fruits in one image and extensive partial occlusion. The good performance of partial occlusion tolerance of the proposed method in this paper is mainly due to that the proposed HCA doesn't use the shape of outer contour of fruit, but uses multiple concentric contours which come from the pattern of light intensity distribution on fruit surface. The research framework in this paper can give a novel thought on other green fruit detection besides citrus fruit. [ABSTRACT FROM AUTHOR]

Harris algorithm is a classical corner detection algorithm. It can extract corners of image quickly and has a certain degree of anti-noise ability, but it has corner location error to some extent. It needs to artificially set 2 threshold parameters, and it can not easily eliminate false corners such as edge points, so it has somewhat lower accuracy of corner detection. For above-mentioned reasons, a modified Harris corner detection algorithm based on auto-correlation matrix of image pixel was proposed in this paper, and the purpose was not only to solve the problem of the variability and randomness of setting thresholds for corner response function (CRF) and non-maximum suppression in Harris algorithm, but also to improve the accuracy of corner location. In our paper, the most important innovation is embodied in 2 aspects: One is avoiding to set 2 thresholds of traditional Harris corner detection algorithm artificially, the other is locating corner more accurately by modified non-maximum suppression method. Firstly, original image was filtered by directional filtering and Gaussian low-pass filtering, and feature corner image was constructed by calculating determinant of every pixel's auto-correlation matrix. Potential corners of image could be heightened effectively, which had more significant intensity than other surrounding pixels, and could be recognized easily in feature corner image. Secondly, in order to improve intelligent level of the modified algorithm, we selected adaptive OTSU algorithm to determine segmentation threshold. The segmentation threshold of feature corner image could be calculated by OTSU algorithm, and the pre-selected regions were obtained. So the search range of corner detection was significantly decreased. On the basis, an optimized non-maximum suppression method was adopted in our research, which could divide each pre-selected region into several 3×3 square subranges, and correct corners were extracted from potential corners of each square subrange, false corners were eliminated effectively. Finally, in order to validate the efficiency and reliability of the modified algorithm, 5 groups of comparison experiments were performed in our research. Five images, including generic image format (jpg, bmp), and multi-band remote sensing image format (GF-2 data), were selected to test performance of the modified algorithm and Harris algorithm, which contained the total of detection corners, the number of correct corners, the number of false corners, the number of omissive corners, and the detection rate of correct corners. According 5 groups of comparison experiments, the accuracy of corner detection in different types of images is improved, for crop vegetation remote sensing image, the accuracy of corner detection can reach 27.06 percentage points. We can draw a conclusion that the improved algorithm can not only calculate the optimal threshold automatically, but also locate the corners more accurately. Therefore, our modified algorithm can greatly improve the precision of corner detection. The proposed algorithm is more accurate and efficient than traditional algorithm, its adaptive characteristic makes it easy to be integrated in an image processing system or image registration module, and it has higher feasibility and application value. Experiments also show that there is some insufficiency to be improved in our research, for example, some corners in picture of cubes could not be detected correctly with either our modified algorithm or Harris algorithm. In our future research, we propose to partition an image into several sub-image blocks, and independently determine each sub-image block's segmentation threshold by OTSU algorithm, so that the corners not prominent in full image can be significantly strengthened in sub-image blocks, and can be detected correctly. The research could provide reference for agricultural remote sensing image data detection. [ABSTRACT FROM AUTHOR]

The breeding environment plays an important role in healthy growth and development of the pigs, and it is also a guarantee of the excellent traits. The growth of pig is the integrated result of their own traits and external environmental factors. For different environments, pigs will show different behaviors. It is found that the appropriate growth environment can be reflected by their degree of dispersion in the pens. When the environment is suitable, the pig group will lie down together. When the temperature is low, they will huddle together; and when the environmental temperature is high, the pigs will scatter around. Image segmentation is the important link to analyze pig behavior status based on computer vision technology, and pig image segmentation is an image representation that provides easy-to-understand and analysis for the behavior feature extraction, parameter measurement, image analysis, pattern recognition of individual pig. The accurate and effective image segmentation algorithm is for pig behavioral intelligent analysis, understanding and environmental intelligent decision. The traditional Graph Cut algorithm uses the energy function to compute all the pixels in the image, it takes a lot of time. It has the low segmentation efficiency and it cannot accurately segment specific pig. Aiming at the limitation of traditional Graph Cut algorithm and the characteristic of image itself, on the basis of the interactive watershed algorithm, in this paper, we proposed an interactive image segmentation method based on improved Graph Cut algorithm aiming at better use of interaction information provided by the user and effectively control of the number of super pixels. To some extent, the algorithm could also avoid the problems of over segmentation or under segmentation. The image was divided into regions based on the interactive watershed algorithm, and the foreground and background were specified. Each region block was regarded as the super pixel, the gray scale of super pixel was used as the vertex. Establish edges between adjacent vertices and the new network map was constructed instead of the traditional weighted graph. In order to reconstruct the energy function to complete the effective segmentation of the foreground and background, the maximum flow, and minimization cut algorithm were used to cut the image. Results of experiments showed that the method can effectively segment pig target, it had high segmentation quality and high efficiency. In this paper, the peak signal to noise ratio (PSNR) and structural similarity (SSIM) were used as objective evaluation standard. The range of the PSNR was [30, 40] and the range of the SSIM was [0.9, 1] in this method, which was the most accurate segmentation algorithm in the shown segmentation algorithm. It indicated that two evaluation criterions were consistent with the subjective evaluation, and the stability and the reliability of this method were improved. The average time consumed in this algorithm was only 33.7% of in the traditional Graph Cut algorithm. It greatly improved the efficiency of segmentation, and it had highly robust for the segmentation of specific target pigs. The results of this paper can provide technical support for subsequent target identification, detection, tracking and monitoring, evaluation of external environments, such as alarms. [ABSTRACT FROM AUTHOR]

As the wireless sensor network node's energy is limited, how to effectively use the limited resources and the effective transmission of data becomes a hot topic. In order to extend the life cycle of farmland wireless sensor network and improve the packet delivery rate of sensor network, a wireless sensor network architecture suitable for farmland information collection was constructed, and a hybrid type clustering routing algorithm (HCRA) was proposed in this paper. Two classical algorithms, LEACH (low-energy adaptive clustering hierarchy) and HEED (hybrid energy-efficient distributed clustering), were evaluated for their merit to extend life cycle of wireless sensors. In the LEACH algorithm, all cluster head nodes communicate directly with the sink nodes, so that the cluster head node far from the convergence node energy consumption was serious, and cannot guarantee the cluster head evenly distributed, which could make some cluster head nodes' energy consumption too large, affecting the network life cycle. The HEED algorithm, on the other hand, used the cluster head to communicate directly with the sink node, which consumed a lot of energy. Because the farmland information collection range was large, those two algorithms were not fully applicable to the wireless sensor network information gathering in farmland environment. Based on this evaluation and the principle of the HEED algorithm, we proposed a hybrid clustering routing algorithm (HCRA) for farmland information collection. In the proposed algorithm, the network model and the energy consumption models were described separately. In the network model, the monitoring area of the sensor network was abstracted as a circle area. Each node in the network had its own unique ID, and the location was not changed after the deployment was completed. All nodes were distributed evenly in the area. In the energy consumption model, the energy consumption of the wireless sensor network increased with the increase of the distance between the transmitting node and the receiving node, regardless of the power dissipation loss using the free space model or the multi-channel attenuation model. Besides, we also descried the HCRA algorithm in detail in this paper, including formation of cluster, cluster bidding and inter-cluster routing. The formation of the cluster was as follows: the center node sink was placed in the center position, and then from the sink node, taking r (the distance between layers and layers), 2r, 3r…as the radius, the entire circular area was divided into m flows. The entire area was divided into six parts, each of which was subdivided into small areas with equal area according to the number of layers. Each small area was a cluster in the wireless sensor network. The cluster head bidding process was as follows: the cluster head node was selected according to the heed algorithm, i.e., the cluster head was generated according to the initial probability of each node. When the cluster head was selected, other non-cluster head nodes joined the cluster selectively according to the collected information from the competition stage. The inter-cluster routing was the information transfer process between cluster heads, cluster head selected the next cluster head based on energy consumption. Finally, the proposed HCRA algorithm along with LEACH and HEED algorithm were simulated experimentally under the same conditions. The experimental results showed that the HCRA algorithm had 28% longer network lifetime than the LEACH algorithm, about 12% faster than the HEED algorithm, under the iteration period of 1000 times. The HCRA algorithm had a packet delivery rate, which was about 30 percentage points higher than the LEACH, and about 13 percentage points higher than HEED. [ABSTRACT FROM AUTHOR]

The extensive application of residual current protection device in rural low-voltage power grid plays an important role to avoid electric shock casualties and fire accident caused by the leakage. Malfunction and failure action often occur in online residual current protection device due to the irrelevant between the setting value of action current and electric shock current of organism. Many researchers conducted a number of breakthrough research on detection of leakage current and hardware architectures of residual current protection technology, which improved the technology performance of residual current operated protective device to some extent, but it could not overcome the low efficiency of correct delivery rate. There were no mature technology and products at home and abroad on detection and characteristics of the law for biological shock signal when the organism was in electrical shock, which could not meet the need of reliable power system under many complicated factors. In this paper, detection model of biological electric shock current was researched based on Hilbert-Huang transformation. Therefore, aiming at how to detect electric shock time and recognize current signal of the biological electric shock branch in residual current, residual current and electric current signal of organism electric shock were set for example, Hilbert-Huang transformation method was used to determine local amplitude of the IMF component with the largest correlation coefficient in the natural modal function of residual current when biological shock occurred, this local amplitude was 34.02 mA, which reached 0.99 correlation coefficient with the original signal, meanwhile, there was a similar law of changes of spectral characteristics between residual current and electric shock current transient process. Biological current signal were consisted of 5 IMF components and one residual component, which accounted for 60.64% of total samples. The IMF component with the biggest correlation coefficient has much bigger variation range of amplitude. In actual signal processing, mutations of high frequency IMF could be used to determine the biological shock time, and IMF component with high amplitude share and correlation coefficient could be used to extract current amplitude of electric shock branch. Hence, in this study, based on those results above, firstly, mutation characteristics of high frequency IMF component amplitude in biological current signal could be used as a criterion and judgment method for electric shock time, which could automatically identify the moment of failure and locate the calculation. Simulation of the actual data processing accuracy was 94.17%. Moreover, low frequency natural modal IMF component was extracted from residual current decomposition, which should be higher relevance and limited quantity. At last, method was established for detecting current amplitude of biological shock branch based on natural mode component of residual current through application of stepwise multiple linear regression method. Simulation result shows that the average relative error is 5.46%，which indicates that the method proposed in this paper has good potential rapid technique for developing a new generation-residual current protection device based on biological electric shock current and plays an important role to avoid personal electric shock casualties and electrical fire in as well as safe operation low voltage power grid. [ABSTRACT FROM AUTHOR]

The application of multi-rotor unmanned aerial vehicle monitoring system for forest pest information collecting has many advantages, such as low running cost, operating flexibility, easy access to data, high image resolution etc. It has been regarded as a quick access to forest insect pest information collecting. By use of unmanned aerial vehicle system, valid segmentation and extraction of pest images acquired with the help of multi-rotor unmanned aerial vehicle can be used to calculate the insect pest proportion in monitored sample field. It can provide forest conservation experts with evidence for assessing the insect pest damage. To conduct forest monitoring work and calculate the proportion of pest infested area in monitored sample field with more preciseness and fast turnaround, in this paper, we aimed to solve poor time response circle and limited monitoring range problems that exist in current forestry information monitoring method. Firstly, in this paper, we built both hardware and software systems of multi-rotor unmanned aerial vehicle. Aerial vehicle equipped with image collecting devices was used to monitor in forestry pest insect infested area and collect data in the Liaoning testing forest. In order to obtain proper resolution images, aerial vehicle took off the center of the chosen monitoring area vertically to collect photo resources. By considering needed resolution requirements on image segmentation comprehensively, the height of about 50 m was chosen for image acquisition. On the analytical basis of monitoring images, an image segmentation method based on composite gradient watershed algorithm was proposed. This method introduced global histogram equalization to eliminate the influence of dark texture and adopted the morphological hybrid open-closing reconstruction filter to complete the denoising work of the image samples, eliminate the image interference to the segmentation effect, and suppress the over-segmentation phenomenon in image segmentation process. The gray-scale image was obtained by gray-scale transformation of the pre-processed image. The non-correlation regions (road and bare ground) were extracted by calculating the composite gradient of each pixel point in the gray image. Interference to the segmentation result may arise in segmenting process due to the similar color of non-correlation region and pest insect infested area. In this paper, the mentioned region was removed from the original image, which greatly avoided the interference of the non-related region to the pest area and ensured the accuracy of the result. Finally, the watershed algorithm was applied to realize the segmentation and extraction of insect pest area in images. In order to verify the effectiveness of the proposed method, the traditional watershed algorithm and K-means clustering algorithm were used for comparing experiment methods in the segmentation of eight images with different levels of insect pest. With the help of mage segmentation device, the accurate pest insect infested area was labeled manually, and it was taken as reference value in pest insect proportion calculating step. The experiment result showed that the segmentation effect was much more similar to the manual operation result. Specifically, the relative error rate decreased by 6.56% and 3.17% and the relative limit measuring accuracy was improved by 7.19% and 2.41% in this proposed method when traditional watershed algorithm was compared with K-means cluster algorithm. Our result showed that multi-rotor unmanned aerial vehicle was helpful in real time and effective monitoring of forest pest insect. The algorithm proposed in this paper was able to accurately segment and extract pest insect area in monitoring images and the proportion of pest area in whole sample fields was acquired, thus providing valid data support for forest pest monitoring and preventing work in the future. [ABSTRACT FROM AUTHOR]

Timely and accurate identification of oestrus behavior of sows can effectively increase the conception rate and litter size, which is of great significance to improve the breeding level and economic benefits of breeding enterprises.The traditional identification methods of oestrus behavior of sows have the problems of high working intensity, strong subjectivity and low automation level. Some identification methods of oestrus behavior of sows based on internet of things also have some problems, such as high false alarm times, high error rate and long recognition time. In order to solve the problems of low identification accuracy and efficiency, a oestrus behavior recognition method of sows based on Moth-Flame Optimization optimized Long Short Term Memory(MFO-LSTM) was proposed in this paper, and the verification test was carried out at WuFeng breeding farm in Xinghualing district, Taiyuan city, Shanxi Province. The posture data of sows were obtained by posture sensors installed on the neck of the sows. The posture data describing the sow's motion state contained three-axis acceleration, three-axis angular velocity, three-axis angle, three-axis magnetic field and quaternion. The collected posture data was manually marked according to the videos, and preprocessed to obtain a posture classification data set. Then, the MFO algorithm was used to optimize the numbers of first and second hidden layer neurons, maximum training period, block size and learning rate, thus, the LSTM network model was built. Postures of sows were divided into three categories, i.e. standing, lying and mounting. Through the statistics of the duration of each complete mounting behavior of the sows, the range of the recognition time of the oestrus behavior was obtained. The sows posture classification results were counted with different oestrus behavior recognition duration, and two characteristics of mounting behavior and activity were then extracted, so as to obtain the recognition feature matrix of oestrus behavior of sows. Finally, the feature matrix was input into the MFO-LSTM classification model to judge whether the sow was oestrus. The experimental results showed that the classification method of sows posture proposed could effectively distinguish the three postures of standing, lying and mounting. The classification effects of proposed method were better than that of the Support Vector Machine (SVM), Probabilistic Neural Network(PNN), Learning Vector Quantization (LVQ) and Extreme Learning Machine(ELM). The average accuracy, recall rate and F1 in the attitude data set were 98.02%, 96.26% and 96.18, respectively. On the basis of accurately identifying the sows’ posture, the effect of oestrus behavior recognition based on MFO-LSTM algorithm was verified. The test results showed that the recognition effect was best when the oestrus recognition duration was 30 min at this time, the error rate, recall rate and specificity of oestrus recognition were 13.43%, 90.63% and 81.63%, respectively. The oestrus behavior of sows recognition method proposed effectively reduced the error rate while maintaining a high recall rate and specificity. Compared with other methods, the error rate was reduced by more than 80%, the oestrus behavior of sow could be recognized after 30 min of oestrus. [ABSTRACT FROM AUTHOR]

Agricultural machinery field automatic navigation technology is widely used in farming, sowing, weeding, fertilizing, spraying, harvesting and other agricultural production process. This technology can improve the efficiency of the mechanical efficiency and reduce the missing areas of operation, labor intensity and the complexity of the operation. Because machine vision can be used to obtain and perceive the relative position information of crop rows, current crop growth status and field environment in real time, it is widely applied in online crop detection and identification. In this paper, a method based on automatic accumulation threshold of Hough Transformation was presented in order to improve the adaptability of the crop row recognition algorithm for different kinds and growth periods of vegetables with machine vision. The method was composed of image preprocessing, feature point detection, optimal accumulation threshold acquisition and crop row extraction. Firstly, to reduce the adverse effects of light change and restrain the background noise, a* component of Lab color model was selected for transforming RGB image to grayscale image. Optimal adaptive threshold and morphology close-open operation was applied for minimizing error segmentation probability and eliminating irrelevant detail. Secondly, the feature points of crop rows were extracted by sectionalized vertical projection method. The original image was divided into several horizontal segments and target pixel ratio and vertical projection width were used as double threshold in the luminance projection view of each segment to determine the location of feature points and distinguish noise. Thirdly, the Hough transformation method with different accumulative thresholds was performed to fit straight lines for all feature points in the image coordinate system, then they were all converted to Hough space accumulator as points. These points were clustered into the same number as crop rows by K-means clustering method. According to the camera projection, the optimal accumulator threshold was acquired by the position relation of clustering centroid and minimum inter-class variance. Finally, the fitting line parameters of real crop rows were the clustering centroid parameters of the accumulation space under the optimal accumulation threshold, then the parameters were converted into the crop lines in the image coordinate system. The crop row identification tests of lettuce and cabbage were carried out in the greenhouse and filed according to the conditions of crops in different growing periods, different weed densities, and different light conditions in the field. The greenhouse experiment showed that the algorithm can effectively identify crop rows with an average recognition accuracy of 97.1% for two crops of different growth periods under different weed densities. The outdoor experiment showed that the algorithm can also identify crop rows with 94.6% recognition accuracy under different row numbers and light conditions. Time consumption for optimal accumulator threshold algorithm and crop rows extraction algorithm were no more than 1.5 and 0.2 s, and the average accuracy rate of crop row detection was achieved 95.8%. In view of the practical application of field operations, as the environmental parameters basically do not change significantly in a short time, the optimal accumulation threshold was only needed to be obtained once, which can ensure the time consumption of algorithm was about 0.2 s. [ABSTRACT FROM AUTHOR]

Traditional methods for classifying crop pests are based on the appearance of the pests such as their color, morphology and texture using algorithms s back propagation neural network and support vector machine. These methods are sensitive to the environments where the pests appear. Also, the imbalance between training sample numbers, difference in scales at which the training samples are taken, along with that the pests might seasonally change their color and shape, make it difficult for these algorithms to detect and recognize the pests. In order to improve the accuracy of detecting crop pests appearing in diverse environments, this paper proposed a deep convolutional neural network method by combining the spatial pyramid pooling with the improved YOLOv3 deep convolutional neural network algorithm. The proposed method first located the pests on a test image and then identified the species they belongs to. Using the improved network structure in YOLOv3, the proposed method used up-sample and convolution operations to calculate the de-convolution. Bilinear interpolation was used to enhance the output of the network, and the depth characteristics of the pests were extracted by the depth residual neural network. As a result, the improved network can effectively detect and recognize small pests in the images. By fusing with the spatial pyramid pooling, the Yolov3-SPP network can map the extracted eigenvectors to different spatial dimensions. The network can be used at various scales to detect pests of different sizes. Results from identifying and testing 20 types of pests collected by traps showed that the average accuracy of the proposed method was 88.07%, with a detection speed of 26 frames/s. Compared with the typical YOLOv3 algorithm, the proposed methods improved accuracy by 2.8 percentage points. We also reconstructed the maps of the network at each stage and compared them with the typical YOLOv3 algorithm in attempts to demonstrate that the features extracted by the proposed method was more recognizable. The results revealed that the typical YOLOv3 algorithm could miss the targeted objects, while the proposed method is not only able to detect pests of different sizes, it also has high recognition accuracy. Using the same dataset, we compared the proposed method with other detection algorithms such as HOG+SVM, Faster R-CNN, YOLOv3. The comparison showed that the proposed method was 19.61 percentage points higher than that of HOG + SVM, and 9.64 percentage points higher than that of Faster R-CNN, Faster R-CNN was unable to extract small pests, and its recognition accuracy was only 78.43%. [ABSTRACT FROM AUTHOR]

Cows’health is the foundation of large scale dairy farm’development. How to monitor cows accurately and management efficiently is important to the development of scale dairy farm. In the spring of 2018, 17 high-yielding Holstein lactation cows were selected from the same group of Anxing dairy farm in Hulin city for experimental study. The sample cows’average weight (500±50) kg and lactation age (203±83) days. The average daily milk yield of each cow was (30±2) kg. Materials selected for this experiment include lactating cows, computers, cow collars (MooMonitor+, Dairymaster, Ireland), data base stations, and amazon cloud storage terminal. The cows and computer were provided by Anxing dairy farm. The cow collar and data base station were provided by Dairy Master company in Ireland, and the cow collar is equipped with the MEMS (micro electro-mechanical system) accelerometer, according to the principle of accelerometer sensor technology, big data clustering analysis was carried out for the time of acceleration changes in different directions, and the position of cows was tracked through RFID tags to monitor the activity behavior of cows. Refer to the number of samples selected by domestic and foreign scholars related to cow feeding behavior, ruminant behavior, reclining and resting behavior and estrus behavior. From April 1, 2018 to May 31, 2018, Dairymaster’s Moonmonitor + information collection system was used to monitor and collect test data of 17 cows’behavior time every day (24h) in Anxing dairy farm, and physical health conditions of the cows were recorded the worker at the same time, such as mastitis, lameness, diarrhea and trauma. During the test period, test data were downloaded from the cloud data storage server through the monitor system at 08:00 am every day for 61 days. The data collection interval was 24 h, the downloaded data includes the rest time (min/24 h), rumination time (min/24 h) and feeding time (min/24 h) of the cows, and 1 037 data records were obtained. And 937 data were randomly selected as training data set, one hundred data were randomly selected as validation data to verify the model prediction. Then, binary logistic regression analysis method and statistical analysis software SPSS23.0 were used to study the collected data. In order to meet the binary logistic regression analysis conditions, the different behaviors time were converted into classification variables. The independent variables of binomial Logistic regression model were entered by force method. The entry criteria of variables was α<0.05, and the exclusion standard was α>0.1. The results showed that: 1) The behavior time changed differently in different cows, and the average rest time changed greatly. The average rest time during abnormal behavior period increased by 25.7% compared with that during normal behavior period. Compared with the normal behavior period, the ruminant time and feeding time during abnormal behavior period decreased by 12.7% and 2.3%, respectively. 2) The behavior time of healthy Holstein lactation cows was normally distributed, with an average rest time of 300-600 min, ruminant time of 400-700 min and feeding time of 200-400 min per day (24 h). The behavior time of abnormal cows was distributed discreetly without obvious distribution rule. 3) Rest time and feeding time are the main influencing factors of the prediction model, among which feeding time had a greater influence on the prediction probability of the model than rest time. When other conditions remain unchanged, the prediction probability change of abnormal behavior of cows increased by 4.2 times for one additional unit of feeding time. Compared with the results of human visual observation, the prediction accuracy of the model was 91%. Therefore, the paper can provide a reference for scientific and accurate management of modern large-scale dairy farms. [ABSTRACT FROM AUTHOR]

Tomato has become the most cultivated and consumed vegetable crop in the world, and China has become one of the largest producers and consumers of tomatoes in the world. The pectin content in cell wall of tomato during ripening is closely related to fruit hardness, which is one of the important indicators to determine the maturity and reflect the quality of tomato. The requirement of tomato maturity classification and evaluation promotes the development of non-destructive, fast and accurate detection methods of tomato hardness.Hyperspectral imaging integrates spectroscopy and imaging technology in an analysis system, which transfers tomato maturity assessment from subjective, manual classification and measurement methods. Hyperspectral imaging has been widely used in the rapid acquisition of information to classify, detect or identify the quality of various fruits. A novel method for tomato hardness detection based on hyperspectral imaging and modified interval Random Frog was proposed in this paper. Firstly, hyperspectral images of 120 tomato samples in different mature periods were captured by hyperspectral imaging system covering near-infrared region (865.11nm-1 711.71nm). And the hardness data of tomato was obtained by texture analyzer. Secondly, the spectral data were pretreated by multiplicative scatter correction (MSC) and normalized preprocessing to eliminate noise and improve signal-to-noise ratio. The validity of the characteristic wavelength plays a crucial role in the prediction performance of the model. Therefore, we need an effective method to extract the effective wavelength to improve the accuracy of the model. Interval random frog (iRF) algorithm considers all possible spectral wavelengths and ranks all the wavelengths based on selected probability. But one of the disadvantages of this method is large number of iterations and slow convergence. In view of above disadvantages, the traditional iRF algorithm was optimized in terms of constructing initial variable subset method. A modified interval Random Frog (miRF)was proposed to extract the characteristic wavelength effectively. Finally, a prediction model was developed based on partial least squares regression (PLSR) method to detect tomato hardness. The results indicated that the convergence efficiency and accuracy of miRF has a significantly improvement compared with the iRF method. The iRF has selected 100 feature bands, accounting for 40.65% of the full band, and its runtime was 32.1min. miRF has selected 47 feature bands, accounting for 19.1% of the full band, and its runtime was 1.6 min. It can be seen that miRF greatly reduces the running time of the algorithm. The characteristic wavelengths selected by iRF and miRF methods were mainly distributed in 1 582 nm-1 655 nm, followed by 1 160 nm-1 190 nm and 1 353 nm-1 383 nm, indicating that above regions were sensitive bands to tomato hardness. In order to prove the effectiveness of the proposed algorithm, the results of miRF-PLSR were compared with those of iRF-PLSR and SPA-PLSR. The prediction set correlation coefficients (RP) of the SPA-PLSR model and the iRF-PLSR model were 0.803 9 and 0.936 6 respectively. And the RP of miRF-PLSR model was 0.968 5. The root mean square error (RMSEP) of the SPA-PLSR model and the iRF-PLSR model were 0.007 7 kg/mm2 and 0.004 4 kg/mm2respectively. And the RMSEP of miRF-PLSR model was 0.004 0 kg/mm2. The experiments results show that the miRF-PLSR model has the best prediction results in all models. [ABSTRACT FROM AUTHOR]

There is poor infrastructure and weak power grid in Chinese western rural areas. Photovoltaic (PV) and wind power pro-poor investments do not consider supporting transmission and distribution facilities. The economy of biogas from biomass waste is not good, due to that it is affected by seasonal variations in temperature. Utilizing PV and wind power to supply energy for biogas can improve biomass energy utilization and solve the problem of environmental pollution, while the absorptive capacity of the PV and wind power is increased, and the comprehensive utilization of biomass and renewable energy in place can be achieved. It has important significance for development of new countryside. Based on national PV and wind power poverty relief policy, this paper proposed rural micro energy grid architecture that combines PV system, wind power system, micro turbine, biogas fired boilers, heat recovery boiler, lithium-bromide absorption-type refrigerator, battery storage, heat and cooling storage, air-source heat pumps for cooling exchange, air-source heat pumps for heating exchange, and so on. Mathematical models of micro turbine CCHP (combined cooling heating and power) system, air-source heat pumps system, heat and cooling storage system and battery storage system were built up. With micro energy grid cost in a single day as an objective function, considering electric power balance, heating power balance, cooling power balance, power exchange with electricity grid and the other constraints, the micro energy grid optimal model was established. Because of premature and local optimization problem for particle swarm algorithm, this paper uses dynamic inertia weight crossbreeding particle swarm optimization algorithm for solving. Taking Chinese west village as an example, according to the actual situation, electric and heating power were supplied in the winter, but electric and cooling power were supplied in the summer. Electricity price applied the time of use price issued by the National Development and Reform Commission. Parameters of energy supply equipment and energy storage equipment, time of use price, and equipment maintenance cost per unit power were determined. Forecasted data were given, which combine PV and wind power outputs, electricity heating and cooling load for typical day. Simulation platform was built in MATLAB 2014a. Electric heating and cooling balance curve of typical day was acquired. System running cost comparison of typical day based on improved and basic algorithm was performed. In addition, according to forecasted curve referred to above, parameters of various devices, time of use price and equipment maintenance cost, the un-optimized system running cost was calculated. Results showed that, through the dispatch of each device in the system, the outputs of energy supplying devices were more reasonable, and energy storage devices played a role of load shifting. The daily running cost based on dynamic inertia weight crossbreeding particle swarm optimization algorithm was less than that based on basic particle swarm and un-optimized cost. To sum up, the proposed algorithm is adopted to dispatch various devices in micro energy grid, it can reduce system running cost effectively, and micro energy grid can be operated economically; the correctness of the models and algorithms can be proved. [ABSTRACT FROM AUTHOR]

In order to extract the crop rows of wheat quickly and accurately, a new method of wheat crop row detection was proposed based on particle swarm optimization (PSO) - clustering. The first step is image segmentation. The purpose of image segmentation was to separate the green plants from background, and it required the following 2 steps: Firstly, gray-level transformation, which could be done in RGB color space. In this paper, the color excess green index 2G-R-B was used; Secondly, image binarization was conducted. Among the global thresholding techniques for image binarization, Otsu method is one of the best threshold ways. So, this paper used Otsu algorithm to binarize the above obtained gray-level image. In order to reduce the burden of the next work, it was essential to extract a number of feature points indicating the crop rows. The specific algorithm can be divided into 2 steps: Firstly, get the left and right boundary points of each crop row. Secondly, extract the midpoint between left and right boundary points. After the original crop image was processed by the above steps, we got the feature points of the crop rows. According to the characteristics that the distances from the feature points around the crop row centreline to this straight line were all smaller than a certain distance threshold, we used the clustering method based on PSO to determine the real center points indicating crop rows. In the crop rows detection algorithm based on the PSO-clustering, the line in the data space composed of the feature points was considered as a particle. Finally, the centrelines were detected by fitting a straight line with the least square method. In order to prove the superiority of the algorithm, we compared the algorithm with standard Hough transform and the algorithm proposed in another literature. We tested the performance from the aspects of the detection accuracy and processing time for different images. Here, a total of 350 images have been tested. The number of the wheat images in overwintering stage was 197 and the number of the wheat images in green stage was 153. For the algorithm proposed in this paper, the number of the wheat images in overwintering stage successfully detected was 190 and that in green stage successfully detected was 143. Comparatively speaking, for the algorithm with standard Hough transform, the numbers of the wheat images in overwintering and green stage that were successfully detected were 180 and 100, respectively. For the algorithm proposed in another literature, the numbers were 168 and 93, respectively. Three representative pictures were selected in the experiment, which included the different environment i.e. the lack of crops, soil block, and high density weed. For the 3 images, the identification errors of the proposed algorithm were 0.631 0°, 0.773 5° and 1.065 7°, respectively. The identification errors of the standard Hough were 0.968 3°, 2.715 8° and 4.463 3°, respectively. The identification errors of the algorithm proposed in another literature were 2.260 5°, 1.531 9° and 5.829 1°, respectively. Therefore, compared with the other 2 algorithms, the proposed algorithm has the advantages of high real time and high accuracy, which can meet the practical requirements of field operation of agricultural robots. [ABSTRACT FROM AUTHOR]

For the problem of obstacle avoidance fruit harvesting in unstructured environment of harvesting robot manipulator, a kind of obstacle avoidance path planning algorithm of hybrid harvesting manipulator based on joint configuration space was proposed in this paper. This paper simplified the structure of the 2P4R hybrid camellia oleifera fruit harvesting robot manipulator which is composed of a palletizing manipulator together with a spherical wrist serial manipulator, and named the link which connecting palletizing manipulator and serial manipulator as moving platform. The manipulator could accomplish six kinds of movements including waist revolution, translational motion of vertical slider and horizontal slider, and three kinds of revolute motions of the wrist part. In this paper, only five joint motions were taken into consideration, it means the last link was considered as a fix part of its previous link when collision free path planning operation was conducting, cause the motion of the last link is not related with the position of the end effector which only affect the posture of it. With the proposed algorithm, firstly select a goal point for the moving platform of the manipulator in Cartesian space which the goal point should located in the workspace of parallel manipulator and not inside the obstacles. The initial and goal configuration of the serial manipulator were determined by the goal point of moving platform, goal position of end-effector and initial posture of serial manipulator. Secondly, the obstacle mapping model in serial manipulator joint configuration space was built by using traversal method, and then attempt to search a collision-free route from initial point to goal point in this space where each point uniquely corresponding to a configuration of the serial manipulator by using rapid-exploring random tree (RRT) algorithm. Thirdly, if algorithm failed to found such a route in previous step, then repeat goal point selecting and collision-free route searching operation, otherwise built the joint configuration space of the parallel manipulator. Fourthly, determined the unique posture of parallel manipulator with the mapping relationship between driving joint value and the position coordinates of moving platform. The goal posture of parallel manipulator corresponded with a goal point in its joint space. Obstacle model was rebuilt in joint configuration space of parallel manipulator, and then reselect goal point of moving platform if the point in configuration space which corresponding to goal configuration is located in mapping model of obstacle, otherwise search a collision-free route from the start point to the goal point in configuration space which corresponding with the initial posture and goal configuration of parallel manipulator respectively by using RRT algorithm. The obstacle avoidance path of hybrid manipulator was obtained from the synthesized results of the serial manipulator and parallel manipulator collision free path. In order to verify the feasibility and effectiveness of the proposed algorithm, the path planning simulation of a hybrid manipulator in Matlab was carried out. The proposed algorithm was also applied in the obstacle avoidance path planning experiment of camellia oleifera fruit harvesting manipulator. The simulation and experiment results showed that the path which planned by proposed algorithm could successfully driving the end effector form initial position to the goal position without any collision, the results validated the feasibility and effectiveness of the proposed algorithm in this paper. [ABSTRACT FROM AUTHOR]

In order to detect cutting surface roughness of grape grafting hard branch, this paper built a detection system of cutting surface roughness and designed an image detection algorithm of feature extraction based on measurement principle of light-section method. The detection system was made up of 9J light-section microscope, industrial camera and computer. Sampling length was an important role in the calculation of roughness. In order to make up for the shortage of single sampling length (1 420 μm), this paper applied image mosaic technology to images acquired by multiple sampling, and thus got a longer sampling length. Meanwhile, a method was proposed to product matching template automatically, which could avoid the manual operation and improve the efficiency of image mosaic when matching the reference image and target image based on gray level information. Then the image mosaic algorithm was tested and the results showed that: The average running time was increased by 1.104 s and the average sampling length was increased by 1131.77 μm when the number of mosaic image increased by a roughness feature image. Moreover, fuzzy set theory was applied to the process of gray-scale transformation and can effectively guarantee the integrity of the single edge of the image after segmentation. In this paper, Otsu algorithm was used to segment image. In order to filter the defeat profile due to catheter lumen and tracheid cavity contained by cutting surface itself, local pixels of the roughness feature binary image were operated by way of human-computer interaction. The pixels of the defect position in the foreground image were set to zero and not involved in the subsequent roughness calculation. A method was also proposed to extract the single side edge by scanning pixels of per column one by one, the first pixel whose value was not zero at each column belonged to edge pixels set. According to the measurement principle of light-section method, the corresponding position relation could be established between the single edge extracted from roughness feature image and roughness profile in one place of grape cutting surface. So the value of roughness height parameters Ra and Rz could be obtained by calculating the single edge. In order to verify the feasibility that the design of image detection algorithm could measure cutting surface roughness of grape hard branch, the comparative experiment was conducted between the detection system built and KEYENCE VK-200 laser microscope. The experimental results showed that the relative error of Ra was 6.73%, which was measured by the roughness image detection algorithm. Therefore, within the allowable range of measurement error, the image detection algorithm based on light-section method and designed to measure cutting surface roughness of grape hard branch has good feasibility, which can help provide technical support for the further research of the impact of cutting parameters on cutting surface roughness and grafting survival rate of grape hard branch. [ABSTRACT FROM AUTHOR]

The research on crop phenotype is one of the important measures in crop breeding. Crop breeding is a technology on selecting the good seeds by the crop phenotype. The farm remote sensing image analysis is a simple and effective method for rapid analysis of crop phenotype. However, the farm remote sensing image acquired by UAV (unmanned aerial vehicle) will be affected by the noise. In order to process and analyze the remote sensing image accurately, the remote sensing image should be first denoised. Typical image denoising algorithm of frequency domain is wavelet threshold algorithm. Wavelet transform can identify the singular signal accurately; however, the traditional wavelet can't effectively deal with multidimensional signal. For example, two-dimensional wavelet obtained from one-dimensional wavelet tensor only has horizontal and vertical directions, and the wavelet filter is isotropic, which is not sensitive to the image with more directions of the texture; while using the traditional wavelet threshold denoising algorithm to handle noise image, it is easy to make texture region blurry compared with the common wavelets. It is well known that the shear transformation has been introduced into the definition of the shearlet, which makes shearlet can represent more directions. In order to capture the directions of local geometric features of image, shearlet must be compactly suppressed. Anisotropic expansion scale transform can be applied in multi-resolution analysis. From coarse scale to fine scale approximation signal, namely in the approximation image in smooth regions using coarse scale, and in the approximation image with directions of texture using fine scale, the optimal approximation of singular curves can be achieved through the application of multi-resolution analysis method. Shearlet filter is anisotropic, and it is very sensitive to the direction of the texture; however, with the increase of noise standard deviation, after denoising, using shearlet algorithm generates artifacts easily. In brief, the denoising algorithm based on the total variation model has the advantage of edge preserving and noise reduction. In the application of shearlet for image denoising, we find that the shearlet with noise in the texture region can detect the texture direction, and can effectively remove the noise; but when the noise standard deviation is large, the application of shearlet algorithm in the image can cause the denoised image to tend to produce some artifacts, which will be identified as noise texture perhaps due to the shear wave in the smooth region. In order to solve the problem of discrete shearlet algorithm, this paper proposes a method based on the combination of shearlet and total variation model to eliminate the artifacts. First, select the generated Symmlet quadrature mirror filter, and ascertain anisotropic scale parameters and shear parameters. Perform discrete shearlet transformation to farmland remote sensing image with multi-resolution analysis. Wavelet coefficients can be obtained using the conventional method by anisotropic wavelet transformation. Wavelet coefficients are projected into wavelet of image. Hard threshold algorithm is used to handle noise coefficients. Wavelet coefficients are adopted to reconstruct farm remote sensing image. After using total variation model to smooth artifacts, the mirror extension is used to preprocess the image. In part of the experiment, the algorithm proposed in this paper is compared with the shearlet algorithm and total variation algorithm, and this algorithm gets a PSNR (peak signal to noise ratio) 1dB more than that from total variation model, and an iteration number of total variation less than that from direct total variation. In addition, not only can this algorithm effectively denoise farm remote sensing image noise and visual effects, but also can remove artifact better than the discrete shearlet transform algorithm. [ABSTRACT FROM AUTHOR]

The basic behavioral characteristics of live pigs are mainly shown through daily food intake frequency, water intake frequency, and excretion frequency. These factors indicate the health states of pig growth. Monitoring and analyzing the behavioral characteristics of pigs are important basis to understand their health situations. Currently, we mainly use artificial way to monitor livestock behavior in China. This method consumes large amounts of human labor and energy, and the observed data obtained in this way is subjective. It is difficult to ensure the accuracy and the continuity of the records. We take good advantage of pig detection and tracking technology based on machine vision to monitor the behavior of pigs to evaluate the health status of pigs in time, and to reduce the morbidity and mortality of pigs and increase the slaughtering rate of pigs. It has important practical significance and application value in improving people’s confidence in pork quality and increasing the income of farmers. Target tracking technology is the basis of the moving target identification and abnormal behavior tracking, recording and analysis. We research the real-time monitoring of the target pigs foraging based on the particle filter target tracking technology. Particle filter algorithm closely approximates Bayesian filtering algorithm based on Monte Carlo simulation, and it is used in target tracking widely. Conceptually, a particle filter tracker maintains a probability distribution over the state (location, scale, and so on) of the object being tracked. Particle filters represent this distribution as a set of weighted samples, or particles. Each particle represents a possible instantiation of the state of the object. In other words, each particle is a guess representing one possible location of the object being tracked. The set of particles contain more weight at locations where the object being tracked is more likely to be. This weighted distribution is propagated through time using a set of equations known as the Bayesian filtering equations, and we can determine the trajectory of the tracked object by the particle with the highest weight or the weighted mean of the particle set at each time step. In view of the pig behavior characteristics and the actual situation of the farms' video image acquisition, this paper takes a group of pigs raised as detection tracking target. On the basis of analyzing and summarizing in particle filter tracking algorithm, we carried out particle filter target tracking technology for pigs which is based on the color characteristics to achieve the goal of tracking pigs. In order to solve the problems in the color characteristics of particle filter target tracking for pigs, we fused the color characteristics and the target contour centroid feature. The specific methods were as follows: First of all, according to the particle filter tracking algorithm based on single color feature of target tracking on the position of the rectangle coordinates, and the height and width of the target tracking rectangular box, we calculated the center of the target tracking rectangle coordinates. Secondly, we determined the centroid position of moving pigs on the basis of the comparison and analysis of moving target centroid position and the minimum circumscribed rectangle length-width ratio. Finally, according to the target contour centroid location and the center of the tracking target rectangle coordinates, we calculated the amount of deviation between them. When the deviation of target contour centroid and tracking rectangular box was too large, we took a second correction for tracking the target coordinates based on the particle filter algorithm with multi-feature fusion. The improved algorithm presented in this paper updated the tracking rectangular coordinates through the target contour centroid coordinates, and gave the new tracking rectangular box. This paper constructs the target pig tracking system based on particle filter algorithm, achieves a multi-feature fusion particle filter tracking algorithm through area real-time monitoring, and completes the statistics of the target pig's feeding time and food intake frequency. Experiment results prove that this algorithm can automatically accurately track, record and analyze the feeding behaviour of the target pigs, and effectively deal with the problems such as target short-time missing. The feeding frequency and time of the target pigs are almost the same as the manual statistics. [ABSTRACT FROM AUTHOR]

Recognition of larval shrimp is the key to realize the quantification of the behavior of larval shrimp. Recently, the studies on aquatic animal behavior recognition mainly use the typical features method or frame differential method. It is hard to accurately identify larval shrimp due to the lack of key profile information caused by the strong motion distortion and background noise. To solve this problem, we took the Penaeus vannamei as research object, and proposed a recognition method for moving larval shrimp based on the improved PCA (principal component analysis) and Adaboost algorithm in this paper. Firstly, we extracted and optimized the effective features. The improved PCA algorithm was applied to the larval shrimp images with the size of 100 × 100 pixels to reduce the image dimension and extract the external features. We then analyzed and determined the number of principal components according to the distribution of the contribution rate. After that we expounded the theory of the Adaboost algorithm, and used these principal components to build different weak classifiers. A total of 100 images of larval shrimp and background samples were used for training. During the training every weak classifier was selected automatically, and a stage classifier was generated after 20 iterations. To verify the effectiveness of the proposed algorithm, the recognition experiment was performed with 150 sample images containing 100 images of larval shrimp and 50 sample images of background. The images were captured in the Key Laboratory of Fishery Equipment and Engineering Technology of Ministry of Agriculture in Shanghai, China. The experiment results showed that, the recognition rate for stationary and slowly-moving larval shrimp and fast-moving larval shrimp was 100% and 96% respectively, and the overall recognition rate reached 98%. Among all the test samples, the recognition time for each image ranged from 25 to 35 ms, and the average time was 27.898 ms. Compared with the conventional typical feature method and the frame differential method, the proposed method can be used to recognize the moving larval shrimp. Furthermore, the proposed method has better performance, because it effectively restrains the influence of complex situation such as the velocity and direction of motion and the sample's size. This paper provides a technical basis for the larval shrimp's behavior quantification. [ABSTRACT FROM AUTHOR]

Soil moisture (SM) is an important parameter for drought monitoring as well as soil erosion, crop production and surface temperature. Microwave remote sensing especially at L-band is one of the most promising approaches to monitor the variable of SM at the global scale with frequent revisiting times. We carried out a large number of experimental and theoretical studies to assess the potential of monitoring soil moisture by means of both active and passive microwave satellite observations. Improving the inversion accuracy of soil moisture by combining active and passive microwave remote sensing data is an important task for the development of quantitative remote sensing. In this paper, the main objective was to estimate soil volumetric water content (SVWC) of bare soil collaborative using the active and passive observations. A novel approach, which was based on the simulation database of the advanced integral equation model (AIEM), was presented. In the method, we took the advantages of the active and passive microwave remote sensing data to obtain SVWC. Based on the analysis of the simulation database, some significant patterns had been found that the characteristics of the vertical polarization microwave emissivity (EV) had stronger sensitivity to the soil moisture than the roughness while the active microwave backscatter coefficient was more sensitive to the soil roughness than passive microwave. Among the vertical polarization microwave emissivity (EV), horizontal polarization microwave emissivity (EH) and VV backscatter coefficient (σvv), the effect of roughness on the σvv was greater than on passive microwave surface emissivity (EV and EH) and the effect of roughness on EV was minimized. The EV and the σvv were used to estimate the soil surface roughness, and on the premise of knowing soil surface roughness, soil moisture can be accurately calculated by passive microwave data. This algorithm had following two steps, the first step was to estimate the soil surface roughness using the vertical polarization emissivity and VV polarization backscatter coefficient; and the second step was to estimate SVWC using a combination of the EV and EH at L band under the condition of soil roughness was known. At last, based on the dataset of the SMAP experimental campaigns carried out in 2012 (SMAPVEX12), we evaluated the superiority of the algorithm. The validation data were obtained through the online Data Pool at the National Snow and Ice Data Center of the SMAPVEX12, the good inversion results were achieved with R2 equal to 0.6637 and the RMSE equal to 0.0607 cm3/cm3. One of the biggest advantages of the algorithm developed in this paper was that the coefficient of algorithm was calculated by the simulated data set, which different from the conventional experience algorithm that depended on the field data, which reduced greatly the limitations of the algorithm in practical application. The threshold value 0.25 of NDVI was used to select the appropriate sampling points for validation and the error of the result partly came from the effect of the little vegetation covering the surface, so the algorithm proposed in this paper should be revalidated using the observed data of the bare soil in the future. [ABSTRACT FROM AUTHOR]

Automatic classification of insect species in field crops such as corn, soybean, wheat, and canola is more difficult than the generic object classification because of complex background in filed and high appearance similarity among insect species. In this paper, we propose an insect recognition system on the basis of advanced sparse coding and spatial pyramid model. We firstly learn features from a large amount of unlabeled insect image patches to construct an over-complete dictionary. The sparse coding of insect image patches is obtained by encoding over the dictionary. To enhance discriminative ability of the sparse coding, we then apply multiple scales of filters coupled with different spaces. Finally, the multiple space features of sparse coding are seamlessly embedded into a multi-kernel framework for robust classification. Traditionally, insect recognition has mainly relied on manual identification by expert entomologists. However, for laymen without a thorough understanding of the terminology of insect taxonomy and morphological characteristics, it is hard to discriminate insect categories at the species level. Therefore, effective identification of insects is a key issue that needs to be well addressed. To improve the recognition accuracy, we develop an insect recognition system using advanced sparse coding, spatial pyramid model and multiple-kernel learning techniques. Different from traditional feature representation, a novel feature representation that is multiple-space sparse coding of insect objects is proposed by this work. The work flow of our method can be decomposed into 2 stages. The first stage focuses on image or insect object representation. At this stage, the features of insect images are extracted using advanced sparse coding and spatial pyramid model. The second stage, which deals with effective fusion of multiple insect-categorization features, constructs a kernel-level fusion classifier using all the sparse coding features. At the first stage, for an insect image given, the features of insect images are extracted firstly. The features are then represented as a linear combination of the corresponding training feature dictionary. Then, a multiple-space sparse coding with spatial pyramid model is used to represent insect image in a joint sparse way over all the features. In this process, for the object image, an over-complete dictionary with unlabeled insect images is learned first. Then, the local image patches of the insect object are represented by their sparse codes with the training dictionary. Despite the fact that appearance is modeled using local patches, the global structure information is necessary for accurate insect identification. Consequently, insect appearance is represented by concatenating the location and orientation sparse-coding of all image patches. To obtain the more compact representations of insect images, we use spatial pyramid model at multi-scale levels, which achieves better robustness to noise and clutter, and thus better copes with severe variations in the pose, scale or rotation. In this paper, we use the 3-scale level pyramid to represent insect image. At the top level of the pyramid, there are 4 image patches which represent the whole image. Each image block size is 50×50. The middle level contains 16 equal size non-overlapping image patches, for which each image block size is 25×25; and the bottom level has 64 image patches for which each image block size is 12×12. Then, the local image features of the 3-level pyramid are combined to represent the insect appearance. The larger scale level provides better geometric features when the classifying insects undergo large appearance variations, while smaller scale level obtains finer features. Finally，the features from fine to coarse levels across different scales are concatenated together to generate the final feature representation of insect images for insect classification. At the second stage, the multiple-kernel learning approach is adopted to combine multiple-space sparse coding. As different features of insect images contribute differently to the classification of insect species, the multiple-space sparse coding technique can combine multiple features of insect species to enhance the recognition performance. Given positive and negative insect samples, the features are extracted. Local image patches of the samples are then represented by multiple-space sparse coding using the corresponding training dictionary. Finally, the multiple-kernel learning classifier is constructed by learning the multiple-space sparse coding of the negative and positive samples for insect categorization and recognition. To meet the need of practical insect image identification, we collected insect images covering various species across several common field crops including corn, soybean, wheat, and canola. Samples of 35 common pest species found in field crops were collected, such as Pieris rapae (Linnaeus), and Leptocorisa acuta (Thunberg). Experimental results showed that our proposed method performed well on the classification of insect species, and outperformed the state-of-the-art methods of the generic insect recognition. Our method improved the recognition rate by more than 9% compared to other methods for the same data sets. In addition, the proposed method had also a good performance and enhanced the average recognition accuracy by 14.1% for the different data sets. [ABSTRACT FROM AUTHOR]

Ontology represents a structured view of the domain containing rich semantic meanings, thus plays an important role for various knowledge-intensive applications. Domain ontology contains the concept of complete information and extensive links between concepts. Construction of agricultural domain ontology can provide the foundation of knowledge organization for vertical agricultural research engine and promoting agricultural informationization and realizing cooperative service of agricultural information. Semantic similarity measure plays an important role in information retrieval and information integration based on ontology. While, traditional semantic similarity algorithms of domain ontology only focus on single influencing factor, which leads to poor convergence performances, lower accuracy, strong subjectivity and other defects. In this paper, a weighted semantic similarity algorithm based on agricultural domain ontology was proposed. According to the characteristics of the different structure of ontology, the major factors that influence the similarity are the structure factor, property of the concept, information, etc. But the structural factors are impacted by relationship type, density of node, depth of node and other factors. First, according to the structural characteristics of ontology model, a new method for calculating the node density was proposed in this study. At the same time, an integrated structure similarity model based on relationship type, node density, depth integrated structure similarity, semantic distance was given, which was called the structure factors. Second, according to the literature and empirical knowledge, the property grid of ontology concept pairs was accessed to gain the attributes of the concept. Third, according to ontology hierarchial network, B-U probability based on root and leaf nodes and semantic information was calculated, which did not rely on the expertise and was objective. Fourth, combining semantic structure, information and property factors, an integrated semantic similarity algorithm was proposed, which considered that different impact factors had a different important degree in the calculation of semantic similarity and were given different weights to agricultural ontology relations. Finally, taking semantic similarity computation of part of agriculture ontology for example, the calculation process of semantic similarity on “sweet corn” and “waxy maize” was enumerated in detail. According to the semantic similarity algorithm proposed in this paper, comparing the calculation results of semantic similarity (0.8206) and standard deviation (0.0565) with other algorithms, it was closer to the intuitive cognition and expert advice, which can effectively improve the accuracy and validity of semantic similarity computation. In this paper, we presented the effort of computing the semantic similarity values via studying the relationship between concept pairs of agricultural ontologies at different depth of ontology hierarchical structure. We evaluated the accuracy of semantic similarity calculation between four different algorithms (algorithm in this paper, algorithm based on information content, algorithm based on distance, and algorithm using standard deviation commonly applied in statistics). The results of this study demonstrated that with proper selection of parameters and comprehensive similarity computation measures, we can significantly reduce difficulty of distinguishing the concept of weak correlation. This study provides a deeper understanding of the application of semantic similarity to agricultural ontologies, and shows how to choose appropriate semantic similarity measures for agricultural information retrieval. [ABSTRACT FROM AUTHOR]

The temperature distribution of pig skin is an important indicator to characterize its physiological state and disease. However, due to the surface hair coat, the skin temperature accuracy which detected by infrared thermography (IRT) is affected and its ability to diagnosis of fever and disease is reduced. The purpose of this paper is to explore the influence patterns of the coat on the skin temperature distribution and propose the thermal image processing method to eliminate the influence of the coat on temperature accuracy. The animals for experimental data were 12 sows in empty pregnant period with the average ambient temperature of 27.4 and humidity in the piggery of 80.3% respectively. The body surface °C temperature was measured by hand-held infrared thermal imager (Fluke, Ti 300) with a resolution of 240 pixels×180 pixels and sensitivity of 50 mK. And it also carried a laser distance measuring sensor with a resolution of 0.01 m to measure the distance between the measured object and the thermal imager. The statistics of the temperature distribution detected by IRT from the region of interest (ROI) under normal coat (NC) was compared to that under shed coat (SC) state. The statistical data indicated that the hair coat produced a large number of “canyon”-like low temperature noise in temperature distribution in NC state, which reduced the minimum temperature and average temperature of the ROI, but had no significant effect on the maximum temperature with diagnostic ability. According to the noise distribution characteristics and the influence pattern, an image noise filtering algorithm named the grid maximum value bicubic interpolation filter (GMBI) was proposed. The GMBI algorithm consisted of three steps including image mesh segmentation, filtering with maximum value and image bicubic interpolation. The key problem of GMBI was how to select the appropriate neighborhood size to ensure that each block contained at least one skin temperature value and the resolution was as high as possible. In this study, mathematical statistics was employed and it was found out that the optimal neighborhood size was 4.25 mm. In order to evaluate the validity of the algorithm quantitatively, the mean square error (MSE), peak signal-to-noise ratio (PSNR) and the difference of maximum, minimum and mean between the processed images by GMBI and the SC thermal images were calculated. The experimental data showed that the differences of minimum and average were greatly reduced from the original 1.59 and 0.47 to 0.13 and 0.07 °C, which both were within the maximum allowable error range(±0.3 °C) for disease diagnosis. Moreover, the MSE decreased from 0.38 to 0.05, while PSNR increased from 45.14 dB to 53.66 dB. In conclusion, the GMBI purposed in this study can filter the majority of noise caused by hair in temperature distribution and significantly improve skin temperature detection accuracy. [ABSTRACT FROM AUTHOR]

The amount of wheat ears per unit area is an important parameter for assessing wheat yield and wheat planting density. Generally, phenotype parameters should be obtained by manual count technique which is time-consuming and needs great effort. Aiming at the traditional image segmentation method based on color feature, texture feature and Haar feature, in this paper, we proposed an improved K-means algorithm for estimating the number of wheat ears. This method established a direct mapping relationship from the low-level features of the image to the number of wheat ears in the image through the color feature so that the target did not need to be segmented or detected. This type of method was more suitable for complex lighting and dense wheat ears, and had higher computational efficiency than counting methods based on image segmentation. This method made full use of the color feature of wheat ear image, took the area feature of local region extracted from local region as the basis of wheat ear judgment, and used the number of sub-regions in clustering region as the estimation value of wheat ear number, thus avoiding the task of target detection and location, and greatly improving the accuracy of Wheat ear counting. According to classification results of K-means method, the wheat ear image was divided into three regions. The green area pixel value representing the wheat eat region was set to 255, the pixel value of other areas was set to 0, so the binary image of wheat ears was obtained. Most of the binarized areas in the image were wheat ears, and a small part of the binarized area that was too small or too large was caused by the leaves. Following the steps were used to count wheat ears: 1) extracting the connected regions of the binary image and labeling them; 2) calculating the area of each connected region, the area was represented by the number of pixels in the region; 3) using the following method to filter out the connected region where the area was too small or too large; 4) finally, the number of wheat ears was obtained by counting the retained binary regions. This method established a direct classification relation from the image color feature of the lower layer to the image containing wheat ears so that the image did not need to be segmented or detected. Based on color feature clustering, this method could estimate the number of wheat ears in local area of space, and it was more scalable without training. The average prediction precision of the total wheatear number in a wheatear image for 12 wheatear images was 94.69%. And the Statistical error of the total wheatear number was between 2.17% and 7.41%, the average statistical error was 5.31%. The statistical experiment results showed that the accuracy of this algorithm to wheat ear counting was better than the traditional method based on image color feature and texture feature. [ABSTRACT FROM AUTHOR]

It is important for crop growth to maintain suitable temperature in plant factory, however large heating energy consumption has been proved to be an obstacle that restricts its development. In Europe, the cost of heating accounts for about 30% of the total operation cost during the winter, but in the north of latitude 43° of China, the proportion reaches 60% to 70%. The traditional heating equipment such as coal-fired boilers has an energy utilization rate of only 40% to 50%. So it is very necessary to apply renewable energy to plant factory. Regulating heating modes by tracking and predicting the heating load changes in plant factory is the key to achieve energy saving. Because of the high energy consumption in winter, accurate heating load prediction can improve the energy saving effect of groundwater source heat pump with water energy storage. Changes of heating load in natural light plant factory are dynamic, time-varying, highly turbulent and uncertain. Artificial neural networks is ideal for predicting load changes, especially BP (back propagation) neural network has strong nonlinear mapping ability, which is generally used by many scholars for building heating load prediction, but rarely in plant factory. Given that heating load of both plant factory and building have nonlinear characteristics, we used BP neural network to predict the next day's heating load of plant factory to promote energy-saving control optimization. The BP neural network model has three levels: input layer, hidden layer and output layer. Input parameters include indoor and outdoor air temperature, solar radiation intensity, indoor relative humidity, indoor absolute humidity, indoor wind speed, etc. For plant factory, the next day's weather condition has a significant impact on the heating load. The output variable is determined as the next day's hourly glass greenhouse load value. The number of neurons in the input layer was 9, the number of neurons in the hidden layer was 13, the selected layer number of hidden layers was 1, the learning rate was 0.25 to 0.30, and the initial momentum factor was 0.9. Common evaluation indicators used to determine whether the neural network converges, included standard deviation, coefficient of variation, and expected error percentage. After algorithm steps being determined, the next day's heating load was predicted based on reasonable algorithmic procedures and steps. Experimental data in the paper was obtained from a natural light plant factory powered by groundwater source heat pump with water energy storage system in Chongming National Facility Agricultural Engineering Technology Research Center. Using the neural network toolbox of Matlab to train and simulate the model to process the experimental data from January 19th to 28th, the value of the error function was 0.002 999 94 which was less than the set value of 0.003, so the neural network was convergent. Prediction effect can be drawn by comparison between the actual surveyed value and the predicted value of the heating load. The main heating load was concentrated on 0:00-6:00 and 17:00-24:00 o’clock in the plant factory, and most of these periods were in the cheap electricity price period of Shanghai. Adjusting the operating strategy and operating mode of the energy supply system were based on the predicted heating load，the heat pump operated at full load or high load during the period of cheap electricity prices, and excess heat was stored in the hot storage tank. The hot storage tank provided heat to plant factory during the period of moderate and expensive electricity price. In this case, the energy cost would be reduced. Therefore, it was significantly economical to control start-stop time of the groundwater source heat pump with water energy storage for plant factory heating project. The error was controlled within ±6% basically between the actual value and the predicted value of the heating load, and the standard deviation was 27. 54, deviation factor was 9.4%, expected deviation percentage was 4.35%. Therefore, the results showed that the BP neural network was suitable for the next day's heating load prediction of plant factory. [ABSTRACT FROM AUTHOR]

The double vane pump is a special type of flow vane centrifugal pump. It adopts a design with less blades, which leads to a disadvantage that the performance of the double vane pump is inferior to that of the multi-blade pump at the same specific velocity. Its stability is 3%-8% lower than of a vane centrifugal pump.Therefore, it is necessary to improve the work efficiency by optimizing the hydraulic design. This article took a double-passage sewage pump model 80QW50-15-4 as the research object. The optimization objective was to design the head and efficiency of the flow point. ANSYS CFX(computational fluid dynamics x) was used to perform numerical simulation to obtain performance data. According to the two-dimensional hydraulic drawing of the initial model pump, the three-dimensional modeling software Pro/Engineer5.0 was used to simulate the water body of the impeller and the volute and to perform mesh division and irrelevance verification. The model pump was subjected to numerical simulation and experiment of clear water medium, and the performance curve was obtained and compared. The error analysis showed that the maximum error of head and efficiency was 3.9% and 1.7%, which meant that the performance prediction model established by this method had high accuracy. Partial initial model impeller structure parameters were selected for performance impact analysis. The Plackett-Burman screening test was used to determine the blade wrap angle, blade outlet angle and impeller outlet width were significant factors affecting head and efficiency of design flow. According to Fang Kaitai's unified design table, training samples of RBF(radial basis function) neural network were arranged, so as to establish important structural parameters and performance prediction models, and generated 5 groups of structural parameters random for neural network testing and error analysis. The head and efficiency performance prediction model trained by radial basis neural network was introduced into the particle swarm optimization algorithm as the fitness evaluation model of particle swarm optimization algorithm. The Pareto optimal solution set of head and efficiency was obtained, and the optimal head and efficiency were selected. In addition, this paper also studied the performance and internal flow field differences of the initial individual, the optimal individual of head and the optimal individual of efficiency when transporting different media. It was known from the performance curve that the performance of individuals was improved when transporting different media. The reason for the performance improvement was revealed by the internal flow field distribution map. In order to verify the practicability of the optimization results, a clear water test was performed on the optimal head and the most efficient individual to obtain a performance curve and compared with the performance curve of the initial individual. Among them, the experimental head of the optimal head at the design flow point increased by 0.96 m than the initial individual, the increase rate reached 5.5%, the efficiency increased by 1.6 percentage point; the efficiency of the best individual increased by 10.11 percentage point, the head decreased slightly but met the design requirements. The test proved that the optimization effect was obvious. This optimization method improves the hydraulic characteristics of impeller and the performance of double vane pump. [ABSTRACT FROM AUTHOR]

In the field of agricultural machine and equipment, gearbox was a key mechanical part that was widely applied in speed regulation and power transmission. The gearbox had a high fault rate in actual operating process due to the severe working condition and its complex configuration. State monitoring and fault diagnosing were of great significance to guarantee the safety and stability of gearbox. A fault diagnosis method based on the improved fireworks algorithm (FWA) and probabilistic neural network (PNN) was proposed to overcome the shortcomings, such as the sensitivity to environmental noise and low fault recognition rate, when conducting the fault diagnosis system of gearbox under complex operating conditions. To enhance the pattern classifying performance of traditional methods based on PNN, a new similarity measure for samples was defined, which made the quality of PNN training data increase in modeling process. An improved FWA-PNN classification method was proposed by combining FWA optimization algorithm with PNN technology. The FWA was applied to optimize the smoothing parameters of PNN to determine the optimal values of network parameters, and thus in some way the pattern classification and identification accuracy of PNN could be improved. The proposed FWA-PNN classification method was applied in fault diagnosis modeling for gearbox under noisy environment, and the complex non-linear mapping relationship between fault characteristic parameters and equipment working conditions was constructed. Experiments were carried out on JZQ250 gearbox in laboratory and the process of the fault diagnosis modeling was summarized as follows: At first, 6 working states of the gearbox that included normal state and 5 typical fault states were simulated during the experiments. And then the vibration signals of the gearbox were gained by using the accelerometers of signal acquisition system under different working conditions. After the pretreatment, time and frequency domain analysis of vibration signals were carried out and 27 time and frequency parameters reflecting the working status of the gearbox were obtained. Kernel principal component analysis (KPCA) method was applied to extract the features of the original high-dimensional data, and 7 characteristic parameters were selected as the fault feature vectors of gearbox at last. As a result, the original fault samples were generated according to the fault feature vectors. Given that the fault vibration signals of gearbox were easy to be interfered by different noises in practice, the fault modeling sample sets were regenerated by adding random noises of different levels in the original fault samples. Two thirds of the samples were randomly selected as the training data and the remaining samples were as the test data to establish fault diagnosis model for gearbox based on FWA-PNN. Next, in order to validate the effectiveness and robustness of this new model, BP (back propagation) neural network (BP-NN), genetic algorithm based BP-NN (GABP-NN) and normal PNN methods were introduced to compare with the improved pattern classification method, and 4 different fault diagnosis models were built. The training results of different neural network models indicated that FWA-PNN had better performance in error convergence speed and precision than GABP-NN and BP-NN, which had an excellent fault tolerance and fault classification capability. Finally, 4 different models were applied in fault diagnosis and classification by using the noise samples as the test data. Comparison results indicated that FWA-PNN model could effectively improve the precision of fault detection due to the smooth parameters of all pattern categories optimized by FWA. Application results showed that by compared with the fault diagnosis models based on BPNN, GABPNN and traditional PNN, the FWA-PNN model had the highest fault recognition rate under different noise levels. In conclusion, a novel fault diagnosis program for nonlinear and complex mechanical systems is provided in this paper. It has good application prospects and popularized value in fault diagnosing for agricultural machinery and equipment. [ABSTRACT FROM AUTHOR]

Rural substation in China usually has large scale, but the electrical equipment has low automation and intelligentization. So the substation maintenance in the daily operation requires high amount of man powers. The staffs need to enter the substation to complete the daily patrol, check and maintenance. Video monitor system can be used, however, for remote and real-time monitoring to protect the on duty staff (moving targets) from danger. Moving target detection is one of the critical issues in video monitor system. In order to overcome the disadvantages of the commonly used object detection algorithm, in this paper, we proposed a moving object detection algorithm in rural substation based on multi-domain (time-domain space-domain and frequency-domain fusion TSFF). At first, the consecutive multi-frames were observed in time domain. Then a time-domain signal was constituted by multiple frames pixels in the same location to observe the amplitude of frequency variation by use of short time Fourier transform in frequency domain. Then the horizontal and vertical four directions extension neighborhood of each target pixel was selected in space domain. The amplitude of frequency variation of the pixels in transverse and longitudinal section was also calculated. Eventually the target pixel was detected as the background, noise, or moving target, to complete the moving object segmentation. Comparing with these kinds of algorithm, inter-frame difference (IFD) method can remove noise effectively (mean of accuracy rate was 96.74%), but the object information retention performed poorly (mean of recall rate was 41.01%). The FPCP algorithm performed generally in all aspects (mean of accuracy rate was 65.58% and mean of recall rate was 75.63%). The GRASTA algorithm performed poorly in object information retention when the background noise disturbance was large (the lowest recall rate was 21.83%). The GMM algorithm had the high mean of accuracy rate (85.88%), but for object information reserving had its shortcomings to a certain extent when the back-ground noise disturbance was large (the lowest recall rate was 78.18%). The ViBe algorithm had the low recall rate (mean of recall rate was 65.07%). The TSFF algorithm showed stable performance with different conditions and higher robustness. The TSFF algorithm had the optimal mean recall rate (86.48%) and mean accuracy rate (96.26%). The TSFF algorithm was not only to overcome the RPAC method affected by lights, shadows, and noise, but also to reserve the object information to the maximum compared with the inter-frame difference method and overcome the difficulty in dealing with the high frequency noise compared with the adaptive background modeling algorithm. The experiment results showed that the proposed algorithm reserve the object information well, and maximum removed the background. The TSFF algorithm was suitable for real-time and monitoring substation staff in different circumstances. [ABSTRACT FROM AUTHOR]

The body condition scoring (BCS) is an important tool of assessment method on body condition for sow raising and management. It has been divided into 5 grades that from emaciated to overly fat grade and each grade had a score. The traditional method has negative effect on animals and farmers, and the process is complex with excessive contact. The body size and shape of sows are correlated with their reproductive performance but are difficult to measure manually. There is subjective uncertainty in the process of manual measurement. The Kinect’s 3D reconstruction technology was used to estimate and analyze the buttock shape of sows. A total of 108 images of Large White sows were manually acquired by Kinect camera during the feeding process at Feng Ning Animal Experimental Building in Chengde, Hebei Province of China, from July 25, 2017 to August 15, 2017. The hip height and hip width were measured by using tape and the back fat thickness was measured by using back fat measuring instrument. The hip height, hip width and area of buttock were obtained by analyzing the key points of 48 images. In order to obtain the measurement points of the livestock, several processing steps were taken, and the steps were as follow: 1) The sow stall was removed manually by Geomagic, and the target pig was acquired. In order to improve post-processing speed, the vertex culling algorithm was used to simplify the Three-dimensional model. 2) Since the models acquired were from different angles, the principal component analysis (PCA) was used to acquire new coordinate system. By using the geometrical relationship among the coordinate axes, standard measuring coordinate system was defined in this paper. 3) According to the geometric feature of the measurement points, the hip height, hip width and area of buttock were obtained. The results showed that the root mean square error of estimated body size was less than 2.1 cm, which meet the requirements of precision. The slice method was used to draw a curve at the highest point of the area of buttock based on the point cloud data. The least square fitting method was used to get the curve of hip contour. The hip radius of curvature was calculated by derivation formula. The results showed that the height-width ratio, area and radius of curvature of the sow’s hip had the correlation with the back fat thickness. The correlation coefficients were 0.567, 0.502 and 0.951 respectively. With radius of curvature as the main parameter, the sow body condition estimation model was built based on the experience of hip morphology. 60 images were selected for validation. By comparing the measured and the estimated values of back fat thickness, the maximum absolute error of back fat detection is 1.3 mm and the average relative error is 3.7%. The accuracy of body condition assessment was 91.7% compared with the traditional methods. All the results mentioned above indicate that this study provides a non-contact body condition assessment method based on 3D reconstruction technology and has certain application potential in the real livestock productive. [ABSTRACT FROM AUTHOR]

Due to the change of vehicle steering attitude caused by road conditions and driver’s intention during driving, location information of detected vehicles relative to the host vehicle is also changed. Aiming at the problem that the method of monocular vision ranging ignores changes in attitude angle in the process of driving, this paper presents a monocular vision ranging model based on inverse perspective mapping (IPM) of variable parameters and road vanishing point detection, which achieves a real-time measurement of longitudinal and horizontal distance during vehicle relative movement by taking advantage of location information of vehicle detection so that it can locate and detect the vehicle on the ground plane as well as provide a good environment perception for advanced driver assistance system (ADAS) and intelligent vehicle system. Firstly, owing to the relationship between changes in attitude angle and the coordinates of road vanishing point, the yaw angle and pitch angle of vehicle motion are calculated in real time through the algorithm for road vanishing point detection, which is based on texture orientation estimation. The algorithm, which possesses a better robustness under different light and road conditions, estimates dominant texture orientation of pixels according to joint activities and confidence measure of Gabor filter with 4 directions, and vanishing point candidates are confirmed by the modified locally adaptive soft voting and particle filter tracking algorithm. On account of the yaw angle which leads to a certain degree of rotation in the top view of IPM and the existence of the pitch angle which leaves the top view of IPM unable to restore the parallel relationship of the top view of actual road, IPM of variable parameters based on the coordinate of road vanishing point is used to compensate for the pitch angle to eliminate the influence of inverse perspective distortion, thereby restoring the parallel relationship of road plane and measuring longitudinal distance between the detected vehicle and the host vehicle using calibrated longitudinal scale factor. Then a modeling analysis of the yaw angle of vehicle motion during the process of IPM is made and the effects of the shape and size of the detected vehicle on ranging model are considered. When the horizontal axis in the lower-right bounding box of detected vehicle is less than half of the number of horizontal pixels in the imaging plane, the detected vehicle would be on the left of the host vehicle and its longitudinal and horizontal distance are calculated in accordance with the coordinate in the lower-right bounding box, while the horizontal axis in the lower-left bounding box of detected vehicle is greater than half of the number of horizontal pixels in the imaging plane, the detected vehicle would be on the right of the host vehicle and its longitudinal and horizontal distance are calculated in accordance with the coordinate in the lower-left bounding box; otherwise, it would be directly in front of the host vehicle with the horizontal distance being zero, and its longitudinal distance is calculated in accordance with the coordinate in the middle base of bounding box. Finally, the vehicle ranging model on the basis of location information of vehicle detection is established to consider compensating for attitude angle. The feasibility and effectiveness of this method are analyzed from 2 groups of contrast experiments on different road environments and ranging methods, and the results show that the proposed ranging model can effectively measure the distance of detected vehicles within about 70 m in the longitudinal direction and 4 m in the horizontal direction, having a measurement error of less than 5%; and the better the road environment, the smaller the error; the ranging error of a good flat road is within 3%, and the average processing speed of this algorithm reaches 40 frames/s. [ABSTRACT FROM AUTHOR]

The dimensional synthesis of spherical four-bar linkages mechanism, which contains function generation, path synthesis and rigid-body guidance synthesis, is an important part of linkage design. For integral period dimensional synthesis of spherical four-bar linkages mechanism by using Fourier series theory, firstly the mathematical model of spherical four-bar linkages mechanism is established and the Fourier transform’s function expression is given. The relationship between output function and the basic dimensional type is found out. Then the spherical four-bar linkages mechanism’s harmonic characteristic parameters database is created. Secondly, by using the fuzzy identification theory, the best suitable basic dimensional type of spherical four-bar linkages mechanism for prescribed output function is obtained. Thus, it achieves the purpose of spherical four-bar linkages mechanism function generation. For non-integral period dimensional synthesis of linkage mechanism, the Fourier series theory can not be analyzed and integrated. In many applications, the design requirement is non-integer periodic function. Therefore, the purpose of this study was to propose a new numerical atlas method to apply to the non-integer periodic function generation problem of spherical four-bar mechanism. The wavelet function has a good analysis ability in both time and frequency, which is its unique advantage for non-integral period dimensional synthesis of linkage mechanism. By taking advantage of the multi-resolution characteristic of the wavelet function, this paper presents the Haar wavelet feature parameter method for the non-periodic function generation of spherical four-bar linkages mechanism. Based on the analysis by wavelet theory of non-integral period output function of the spherical four-bar mechanism, the connection between the wavelet feature parameters of output function by translation or scaling and the wavelet feature parameters of primitive function could be discovered. Then a self-adaptive database of basic dimensional type of spherical four-bar linkages mechanism could be created, which contained 148 995 groups of basic dimensional types. Based on the created database and fuzzy recognition theory, a method of matching recognition was proposed by using the relationship between the wavelet feature parameters of basic dimensional type and the wavelet feature parameters of the given non-periodic function. And the given design interval of the spherical four-bar mechanism non-periodic function generation was realized. Based on the recognized result and the immanent connection between output function and wavelet feature parameters, the optimization object function could be created by using genetic algorithm. Then the result was optimized. This paper took the pendulum disc type cutting mechanism of combine as an example to illustrate the effectiveness and practicality of this method. To optimize the spherical four-bar mechanism function output result, the genetic algorithm was adopted, the population size was set as 200, the crossover probability was 0.08, the mutation parameter was 0.06 and the crossover operator was used. The percent error of the first set mechanism was 1.92%, and that of the second set mechanism was 2.31%, which indicated that the pendulum disc type cutting mechanism of combine met the requirement of actual testing, which was 3%. The study for the function generation and the pendulum disc type cutting mechanism of combine provides a very important theoretical basic and reference for the design, analysis and further study on path synthesis and rigid-body guidance synthesis. [ABSTRACT FROM AUTHOR]

In order to improve the adaptability and working efficiency of apple harvesting robot used to promptly pick the ripe apples, the harvesting robot has to work continuously at night. But the night vision image of apple has many dark spaces and shadows besides the low resolution.These negative factors bring difficulties for the harvesting robot to work at night. So this paper proposes an edge preserving Retinex algorithm based on guided filtering to enhance apple night vision image.The illumination component is estimated by using the guided filtering which can be used as an edge preserving smoothing operator, and then it is removed from the original image to obtain the reflection component with its own characteristics.After Gamma correction, the 2 parts of the image are synthesized into a new image. Finally the night vision image of apple is enhanced.The specific implementation process is stated as follows: firstly, an apple night vision image of the RGB (red, green, blue) is converted into HSI(hue, saturation, intensity) color space.Then the intensity of the image is processed by the guided filtering which has a function of edge preserving.This algorithm is able to accurately estimate the illumination of the image in the edge area with high contrast.After that, a single scale Retinex algorithm is used for logarithmic transform to get the reflection image.Then, the Gamma corrections are used for reflection component and illumination component.The 2 parts of the image are synthesized into a new image and the resulting image is converted into RGB color space.Finally, the output of the apple image is the target enhancement image.This paper selects 30 apple night vision images collected under fluorescent lighting to make simulation experiment compared with histogram equalization algorithm, homomorphic filtering algorithm and Retinex algorithm based on bilateral filtering.From the visual effects, the 4 methods have a certain degree of enhancement.By using the histogram equalization algorithm, not only the brightness has improved greatly, but also the reflective part of the apple is magnified.And the apple in the dark area is not fully displayed.After using the homomorphic filtering algorithm, the night vision image is enhanced and the enhancement effect of the highlight reflective area is poorer.After using the bilateral filtering algorithm, the edge is also maintained as this proposed algorithm, but there is a circle of white halo at the area of the apple gradient mutation.However, after using the proposed algorithm to enhance the images, the apple fruit is more prominent.Its details are clearly visible in the dark areas and there is no phenomenon of over enhancement.There are more obvious visual effect and clearer outline of the target fruit, and the halo part has been well suppressed.According to the objective performance indices of experiment results, it shows that the mean grey value of the 30 images after processed by the proposed method, compared with the original image, histogram equalization algorithms, homomorphic filtering algorithm and Retinex algorithm based on bilateral filtering, increases averagely by 230.34%, 251.16%, 14.56% and 7.75%, respectively, the standard deviation on average increases by 36.90%, -23.95%, 53.37% and 28.00%, respectively, the information entropy increases averagely by 65.88%, 99.68%, 66.85% and 17.53%, respectively, and the gradient on average increases by 161.70%, 64.71%, 139.89% and 17.70%, respectively. The enhancement effect of proposed algorithm is superior to other 3 algorithms.In addition, compared with the Retinex algorithm based on bilateral filtering, the proposed algorithm has an average reduction of 74.56% in processing time, which reflects the timeliness and efficiency. In conclusion, this algorithm has an unique advantage for night vision image enhancement. So it can satisfy the actual demands and realize the continuous operation of apple harvesting robot at night. [ABSTRACT FROM AUTHOR]

The rapid development of modern agriculture in China has put forward higher requirements for agricultural machinery operation. In terms of area measurement, GPS (global positioning system) has become an important measuring tool, completely changed the traditional mode of operation, liberated the labor force, and improved the operation efficiency. Field operation is still basically in the stage of manual operation, so it is inevitable that there is much repetitive operation and missing operation. How to accurately measure the area of operation, this is a necessary issue. In this paper, the adaptive Kalman filter was used to improve GPS positioning accuracy for accurately measuring the tractor operation area. The adaptive Kalman filtering algorithm was mainly to solve the problem of the degradation of the system’s filtering accuracy and the divergence of the system in the case of noise statistics being unknown or not accurate. In order to achieve the system noise estimation of adaptive filtering, we used the covariance matching technology and the Kalman filter residual error to realize the algorithm. In this research, the LABVIEW software was used to get latitude and longitude data of GPS receiver. And then the Gauss projection algorithm was used to change the longitude and latitude data into plane coordinates to calculate the area. To test and verify the influence of different ways of operation on the operation efficiency, back tillage, spindle tillage and alternative plough method were chosen. Firstly, this research used MATLAB to identify the operation trajectory, then used different color to show the area of operation, and used the image processing method to calculate the effective operation area, the actual operation area, and the missed and repeated tillage rate, which were used to evaluate the operation efficiency of the tractor. In order to verify the feasibility of the algorithm, the accuracy of single point positioning and the accuracy of area measurement were tested. The single point positioning experiment showed that the Kalman filter improved the accuracy of GPS single point positioning. The numerical changes at the x and y direction before filtering had relatively big fluctuation, and became flat after filtering. The mean value of coordinates changed after filtering, and the mean square error became smaller. The x coordinate reduced from 0.06317 to 0.05807 m after filtering, and the y coordinate reduced from 0.07901 to 0.04097 m after filtering. In the test of GPS area measuring precision, which was the preparation for its work in the measurement of field area, this research used the GPS to measure some regular and irregular figures. The result showed that the relative error of area measurement was 2.09%. Finally, the field experiment was conducted. The result showed that Block 1 missed tillage rate was 14.29%, and repeated tillage rate was 6.19%; Block 2 balk rate was 10.72%, and backset rate was 5.54%; Block 3 balk rate was 1.81%, and backset rate was 6.82%. With the measurement area increasing, the measurement accuracy was higher. The most efficient farming method was alternative tillage, and the second was spindle tillage. Back tillage’s balk rate was the highest, and its operating efficiency was the lowest. Image processing method was used to calculate the backset and balk acreage in this paper. Different colors were used to display normal area, repeated tillage area and missed tillage area, through which it could visually display missed and repeated tillage locations, and then calculate the working efficiency. We can use this method to guide the actual agricultural production operation, and select the operation mode with high efficiency. [ABSTRACT FROM AUTHOR]