Search

• The number of packets is dynamically adjusted based on the moving speed of nodes. • Time on Air (ToA) is used as a selection criterion for modulation parameters. • Distance, moving speed and SNR are used to determine the network margin. Wireless sensor network (WSN) is an important research area in smart agriculture. As an emerging technology, LoRa is used for communication between wireless sensor nodes. When LoRa is used as a mobile terminal node (mobile robot, drone, etc.), the Quality of Service (QoS) of LoRa network will be greatly reduced, which will have an impact on the quantitative decision-making and intelligent control in the future. Therefore, a novel Adaptive Data Rate (ADR) algorithm for agricultural mobile sensor nodes is proposed in this paper. The channel condition changes are effectively sensed based on three LoRa signal characteristics of SNR, RSSI, and frequency offset. It could respond quickly to the complex and changeable radio channel environment, and configured the modulation parameters dynamically and accurately. The Mobile LoRaSim tool and Arduino + SX1278 hardware platform were adopted for evaluation and analysis. The results showed that compared with the standard ADR algorithm and the Blind ADR algorithm, the data extraction rate of the Mobile ADR algorithm was increased by 15 %, the data collision rate was reduced by 45 %, the network energy consumption was reduced by 8 %, and the time on air was reduced by 50 %. The proposed algorithm improves the adaptability and robustness in terms of QoS of LoRaWAN networks. [ABSTRACT FROM AUTHOR]

• Precision agriculture method applicable to remote on-tree pomegranate pixel-based fruit size monitoring. • Comparison of Deep Learning and zero-shot vision Transformer instance segmentation models. • Image-Based size estimation algorithm for occluded fruits with protuberances. In situ automatic fruit monitoring is of great interest for more accurate and cost-efficient decision making in agriculture. For this purpose, the development of computer vision-based tools is essential. Deep Learning techniques have shown good performance in fruit detection and segmentation. Recently, new models based on Transformer architecture have emerged with promising potential and zero-shot inference capability. In this paper, a Deep Learning model, Mask R-CNN, was trained for on-tree pomegranate fruit segmentation and compared with foundational models based on Vision Transformer, Grounding DINO and Segment Anything Model. Results with Mask R-CNN proved a better performance, according to F1 score and AP metrics, and a lower computational cost, according to prediction time. One of the most interesting derived applications from fruit segmentation is fruit size estimation. However, segmented fruit masks are frequently incomplete due to occlusions. Therefore, image fruit size estimation is not a straightforward process. In this work, we also propose a novel algorithm to estimate and monitor the fruit size in pixel units from the automated masks. A median relative error of 1.39% was obtained, demonstrating the potential and feasibility of future fully-automatic fruit size estimators. [ABSTRACT FROM AUTHOR]

• Three kinds of image of cotton spider mites and healthy cotton leaf were collected with single background and natural background respectively to construct the image dataset of cotton spider mites. • This paper adopts the low layer freezing, high training migration strategy utilizing the PlantVillage data set pre-training model. • The focal loss function and exponential linear unit(ELU)are introduced in C-ResNet50, and the attention mechanism module is embedded in different network layers for contrast trials, the optimization model of Dropout regularization method is added. Consider the problems of low accuracy, time consumption and high cost in traditional artificial diagnosis of cotton spider mites. This paper proposed an automatic cotton spider mites' damage grading algorithm for improved deep residual network based on transfer learning (C-ResNet50). In this study, three kinds of image of cotton spider mites and healthy cotton leaf were collected with single background and natural background respectively to construct the image dataset of cotton spider mites. C-ResNet50 is improved on the basis of ResNet50, by first utilizing the PlantVillage data set pre-training model, then the focal loss function and exponential linear unit(ELU)are introduced in C-ResNet50, and the attention mechanism module is embedded in different network layers for contrast trials, at the same time, the optimization model of Dropout regularization method is added, finally compare the recognition effect of the different models. The expermental results showed that when the momentum factor value was 0.90 and the learning rate value was 0.01,compared with the existing models ResNet50, VGG16, MobileNet, Alexnet and SENet, the C-ResNet50 has the best classification effect, the average accuracy rate of identification of cotton spider mites damage level reached 98.10%. [ABSTRACT FROM AUTHOR]

• A navigation line extraction algorithm was proposed, which was suitable for different growth periods and complex farmland environment. • The method of locating corn plant by corn plant core was proposed. • A new spatial pyramid structure ASPPF was proposed. • An improved YOLOv8s model ASPPF-YOLOv8s was proposed for more accurate detection of corn plant cores. • The proposed navigation line extraction algorithm can meet the requirements of visual navigation of corn spraying robot. Aiming at the shortcomings of the existing navigation line extraction algorithm for corn spraying robot in complex farmland environment, such as poor extraction effect and poor adaptability, this paper proposed a navigation line extraction algorithm for corn spraying robot based on improved YOLOv8s network. This algorithm proposed to use corn plant core as the identification target to locate corn plants. Firstly, a new spatial pyramid structure - atrous spatial pyramid pooling faster (ASPPF) was proposed in this paper, and an improved YOLOv8s model ASPPF-YOLOv8s was proposed for more accurate detection of corn plant cores. Secondly, the center coordinates of the network detection box were used to locate the corn crop row feature points. Finally, the least squares method was used to fit the corn crop row lines. Experimental results showed that ASPPF-YOLOv8s network had a good effect on corn plant cores extraction under different growth periods and environmental pressures. Mean average precision (MAP) and F1 of ASPPF-YOLOv8s network increased from 86.4% and 86% of YOLOv7 network, 88.8% and 87% of YOLOv8s network, 88.6% and 89% of ASPP-YOLOv8s network to 90.2% and 91%. The average fitting time and average angle error of the centerline of crop row were reduced from 82.6 ms and 0.97° for SUSAN corner detection method combined with least square method, 74.8 ms and 0.75° for means method combined with least square method, 67 ms and 2.03° for FAST corner detection method combined with least square method to 45 ms and 0.63° by using the center coordinates of corn plant core detection box combined with least square method. The accuracy rate increased from 91.47% for SUSAN corner detection method combined with least square method, 93.6% for means method combined with least square method and 87.35% for FAST corner detection method combined with least square method to 94.35%. It shows that the navigation line extraction algorithm proposed in this paper can meet the requirements of real-time and accuracy of visual navigation of corn spraying robot, and can be used to extract navigation line of corn spraying robot in complex farmland environment. [ABSTRACT FROM AUTHOR]

• A tomato recognition algorithm based on improved YOLOV5s is proposed. • The recognition performance of different algorithms is analyzed and compared. • The tomatoes are located based on the RealSense camera. • The recognition accuracy and location accuracy are improved. In order to meet the requirements of target detection and location for intelligent tomato picking, a recognition and location algorithm based on improved YOLOv5s is proposed in this paper. The CARAFE module structure was used to optimize the upsampling algorithm of YOLOv5s, which increased the receptive field of the network while maintaining the lightweight. EIoU and Quality Focal Loss were introduced to optimize the loss function of the network, which solved the problem of reduced accuracy caused by uneven samples, and at the same time accelerated the training convergence speed and improved the regression accuracy. The improved model is denoted as YOLOv5s-CQE. Compared with YOLOv5s, the mAP_0.5 and mAP_0.5:0.95 of YOLOv5s-CQE was increased by 1.67 and 3.43 percentages points, respectively. The recognition accuracy of YOLOv5s-CQE model in the test set was 99.77%, which had increased by 2.40 percentages points compared with before improvement, and was 4.10, 6.47 and 3.03 percentages points higher than that of the lightweight networks YOLOv4-tiny, YOLOv5-Lite-e and YOLOv5-Lite-s, respectively. Compared with YOLOv7 and Faster RCNN, the recognition accuracy rate was increased by 1.69, 3.97 percentages points respectively. In order to improve the accuracy of positioning, distortion removal and ROI clipping were carried out on the obtained images. The accuracy of location was tested by laboratory positioning test and field picking test. The results showed the total average errors decreased by 6.65 mm compared with those before distortion removal, and the field positioning accuracy was improved by 6.67 percentages points. The experimental results showed that the algorithm in this paper had the advantages of high precision, fast detection speed and strong robustness, which provided a theoretical basis for intelligent tomato picking. [ABSTRACT FROM AUTHOR]

• Yellow peach continuous picking robot optimizes working methods. • Solutions for continuous picking robot. • An improved YOLOv5 model for yellow peach detection. • Collision-free continuous picking path. In recent years, picking robots can be optimized through deep algorithms and path planning. However, the traditional picking robot grabbing work method has a long grabbing cycle. This paper proposes a continuous picking scheme for yellow peaches. First, we proposed a comparison between the calculation method of the continuous picking method and the traditional picking method. It provided a yellow peach-picking robot and a continuous picking end effector. We designed and improved the YOLOv5s model, adding model clipping, ODconv, Global Context Networks and ELAN methods. The best model can be obtained through multiple ablation experiments. We proposed a collision-free continuous picking path (CPP) by Volume variables and a three-dimensional Gaussian kernel. It can be applied to the yellow peach continuous-picking robot. After the improved YOLOv5s network, the MAP and model size reached mAP, and the model size reached 96 % and 3.54 MB. Compared with the original and improved networks, the improved network parameters were reduced by 75.5 %, and the mAP was increased by 4.6 %, reducing the deployment cost. In the simulation experiment, the non-collision continuous picking algorithm was compared with the planning method under the traditional picking method. The continuous picking path is 29.56 % of the original path, the theoretical non-collision rate reaches 96.6 %, the actual non-collision rate is 91.6 %, and the picking success rate is 80 %. The scheme can realize continuous picking of yellow peaches. This scheme provides an essential reference for the continuous picking of fruits. [ABSTRACT FROM AUTHOR]

• A defencing algorithm based on deep learning was proposed. • The algorithm based on U-Net and Pix2pixHD. • Chicken contours occlude by the wire mesh can be recovered. • The accuracy of caged chicken detection is improved after defencing. Cage farming is the mainstream farming mode in China. Accurate individual identification and behavioral detection of caged chickens can provide managers with a better understanding of chicken status. However, for image detection of caged chickens, the cage may affect the accuracy of the detection algorithm. For this reason, CCD (caged chicken defencing), a defencing algorithm based on U-Net and pix2pixHD, was proposed to improve caged chickens' detection accuracy. The proposed defencing algorithm can accurately identify the cage wire mesh and recover the chicken contours completely. In the test set, the detection accuracy of the cage wire mesh was 94.71%, while a structural similarity (SSIM) of 90.04% and a peak signal-to-noise ratio (PSNR) of 25.24 dB were obtained in the image recovery. To verify the practicality of the method proposed in this paper, we analyzed the performance of the object detection algorithm before and after defencing from the perspective of the most basic individual detection in the caged chicken detection task. We validated the defencing algorithm with different YOLOv5 detection algorithms, including YOLOv5s, YOLOv5m, YOLOv5l, and YOLOv5x. The experimental results showed that the defencing algorithm improved the detection precision of caged chickens by 16.1%, 12.1%, 7.3%, and 5.4%, respectively, compared with before defencing. The recall improvement was 29.1%, 16.4%, 8.5%, and 6.8%. To our knowledge, this is the first time that a deep learning-based defencing algorithm has been applied to caged chickens, and the detection accuracy can be significantly improved. The method proposed in this paper can remove cage wire mesh greatly and provide a technical reference for subsequent poultry researchers. [ABSTRACT FROM AUTHOR]

• The method overcomes over-segment and under-segment problems inherent in watershed. • The adaptive-radius circular method had a better accuracy in corner point detection. • Under-segmentation regions were classified by BP neural network. • The method was tested for three rice varieties under complex touching scenarios. • The proposed method achieved high accuracy (94.63%). The ability to segment and count of touching hybrid rice grains can enable the automatic evaluation of seeding performance. In this paper, an algorithm that separates and counts touching rice grain, which consists of the watershed algorithm, an improved corner point detection algorithm, and neural network classification algorithm, is presented. To reduce the over-segmentation regions caused by the watershed algorithm, wavelet transform and Gaussian filter are first applied to enhance the contrast intensity of grayscale image and to reduce noise, followed by an improved corner point detection algorithm based on adaptive-radius circular template. The over-segmentation regions are identified and merged by detecting whether the end points of the splitting lines coincide with the corner points. Considering that regions of different grain quantity vary in appearance and corner point characteristic, a Back Propagation (BP) neural network classifier is employed to classify the under-segmentation regions into five categories: one grain, two grains, three grains, four grains, and more than four grains. The proposed algorithm was tested on three hybrid rice varieties under different realistic touching scenarios formed in the sowing process. The tests results showed that the corner point detection algorithm using an adaptive-radius circular template achieved better corner point accuracy than that using a fixed-radius template, and the over-segmentation regions were more accurately merged. For grain regions of different grain quantity, BP neural classifier achieved an average classification accuracy of 92.4%, which was suitable for counting rice grains in under-segmentation regions. The overall segmentation and counting method proposed in this study could achieve an average accuracy of 94.63%, which was verified by manual counting results. [ABSTRACT FROM AUTHOR]

Highlights • The effect of the local boundary curvature of a non-spherical apple is considered. • Optical coefficients are retrieved using a spatially-resolved imaging technique. • Optical diffusion and firmness measurements are observed for Royal Gala cultivars. • Monte Carlo simulation results and experimental measurements are compared. • Correlation between firmness and estimated reduced scattering coefficient is plotted. Abstract The aim of this paper was to retrieve the absorption (µ a) and reduced scattering (µ' s) coefficients of whole apples which exhibit a complex shape. The effect of the local boundary curvature on the retrieved optical properties was investigated by means of numerical simulations and measurements carried out at the wavelength of 633 nm. A first attempt was made by performing Monte Carlo simulations on an apple-like spheroid model covered with a thin skin layer of thickness 80 µm. Monte Carlo data were then analyzed to depict the changes of photon densities, diffusively reflected images and optical properties as a function of the light source location over the surface of such target. Second, spatially-resolved backscattered images were acquired from 207 'Royal Gala', and the values of µ a and µ' s were retrieved using an inverse algorithm to fit the scattering profiles with a diffusion theory model, in a selected fitting range of 2.8–10 mm. The results confirm the theoretical prediction and show that the absorption coefficient µ a may be overestimated, while the reduced scattering coefficient µ' s is slowly changed when the measurements are performed on these apple species. Finally, experiments carried out on 200 apples still show that µ' s is negatively correlated to the fruit firmness with a correlation coefficient (r) of 0.63. The spatially-resolved technique provides an efficient means for measuring the optical properties of fruits, and may be also useful for assessing the apple firmness. [ABSTRACT FROM AUTHOR]

Highlights • Developed a strawberry harvesting robot with a cable-driven gripper. • The gripper is robust to the localisation errors introduced by the vision module. • The subsystems have been integrated into a complete system. • The system has been evaluated in the farm and the failure cases have been analysed. • The picking cycle time is 7.49 s with a success rate of 53.57% in the farm setting. Abstract This paper presents the development and evaluation of a robot for harvesting strawberries (Fragaria × ananassa) grown on table-tops in polytunnels. The robot is comprised of a newly-designed gripper mounted on an industrial arm which in turn is mounted on a mobile base along with an RGB-D camera. The novel cable-driven gripper can open fingers to "swallow" a target. Since it is designed to target the fruit not the stem, it just requires the fruit location for picking. Moreover, equipped with internal sensors, the gripper can sense and correct for positional errors, and is robust to the localisation errors introduced by the vision module. Another important feature of the gripper is the internal container that is used to collect berries during picking. Since the manipulator does not need to go back and forth between each berry and a separate punnet, picking time is reduced significantly. The vision system uses colour thresholding combined with screening of the object area and the depth range to select ripe and reachable strawberries, which is fast for processing. These components are integrated into a complete system whose performance is analysed starting with the four main failure cases of the vision system: undetected, duplicate detections, inaccurate localisation and segmentation failure. The integration enables the robot to harvest continuously by moving the platform with a joystick. Field experiments show the average cycle time of continuous single strawberry picking is 7.5 s and 10.6 s when including all procedures. Furthermore, the robot is able to pick isolated strawberries with a close-to-perfect success rate (96.8%). However, in farm settings, the average picking success rate is 53.6%, and 59.0% when including "success with damage", testing on the strawberry cultivar of "FAVORI". The failure cases are analysed and most failures are found when picking strawberries in clusters, in which both the detection algorithm and the gripper struggles to separate the berries. [ABSTRACT FROM AUTHOR]

Highlights • A new method is proposed to achieve accurate extraction of tea plant leaf disease saliency map under complex backgrounds. • Simple Linear Iterative Cluster is used as a preprocessing step to separate the salient region and background. • The combination of SLIC and SVM realized the automatic selection of SVM algorithm training set. • Proposed method performs better than the other SLIC based algorithm both in visual effects and quality assessment index. Abstract For the purpose of improving the extraction of tea plant leaf disease saliency map under complex backgrounds, a new algorithm combining SLIC (Simple Linear Iterative Cluster) with SVM (Support Vector Machine) is proposed in this paper. Firstly, super-pixel block is obtained by SLIC algorithm, significant point is detected by Harris algorithm, and fuzzy salient region contour is extracted by employing convex hull method. Secondly, the four-dimensional texture features of super-pixel blocks in salient regions and background areas are extracted, and then the classification map is obtained by classifying the super-pixel blocks with the help of SVM classifier. Lastly, the morphological and algebraic operations are implemented for repairing classified super-pixel blocks. As a result, one accurate saliency map of tea plant leaf disease image is obtained. Through testing based on 261 diseased images, the quality evaluation index, the accuracy, precision, recall and F-value are 98.5%, 96.8%, 98.6% and 97.7%, respectively. It demonstrates that the proposed method performs better than the other three SLIC-based algorithms in visual effects and quality assessment index. Such conclusion can be drawn that the proposed method can effectively extract tea plant leaf disease saliency map from complex background. Consequently, this research is expected to lay a good basis for the study of tea plant leaf disease identification. Last but not the least, the proposed method has good potential that extracts saliency map of crops or plants disease. [ABSTRACT FROM AUTHOR]

Highlights • A model to identify pigs individually based on the Faster R-CNN is proposed and trained. • An algorithm to decide whether a pig is feeding or not is realized. • It provides a solution based on video surveillance for precision pig breeding. Abstract The feeding behavior of each individual pig is an important indicator to determine whether it is healthy or not. Therefore, automatic behavior recognition for individual pig is one of the core problem in precision pig farming. Video surveillance is a common tool for monitoring animal behaviors. To accurately identify each pig from the video sequences is a prerequisite for individual pig behavior recognition. This paper proposed to use Faster R-CNN to locate and identify individual pigs from a group-housed pen. The head of each pig was also located. An algorithm for associating the head of each pig with its body was designed. On this basis, a behavior recognition algorithm based on feeding area occupation rate was implemented to measure the feeding behavior of pigs. Experiment showed that our algorithm can recognize the feeding behavior of pigs with a precision rate of 99.6% and recall rate of 86.93%. [ABSTRACT FROM AUTHOR]

Highlights • A method of extracting key frames from surveillance videos was put forward. • A foreground segmentation method to detect the foreground of edge areas was proposed. • Combined with spatio-temporal continuity, a new region proposal method was proposed. • The rate of accuracy and efficiency of the method was verified by experiments. Abstract As one of the most famous algorithms, Faster R-CNN has been applied to many fields. However, it is unlikely to be used to surveillance videos directly because of its low efficiency and precision. To deal with these problems, this paper puts forward an object detection method which is based on Faster CNN to detect dairy goats in the surveillance video. It includes key frames extraction, foreground segmentation, region proposal and Fast R-CNN. The experimental results show that our method is more than twice as fast as Faster R-CNN and obtains 92.49% average precision. Our results suggest that our key frame extraction and region proposal method are helpful for detecting dairy goats. [ABSTRACT FROM AUTHOR]

This paper proposes a new efficient method for line detection based on known incremental methods of searching for an approximate globally optimal partition of a set of data points A and on the DIRECT algorithm for global optimization. The proposed method was modified for solving the problem of detecting crop rows in agricultural production. This modification can recognize crop rows with a high accuracy, and the corresponding CPU-time is very acceptable. The method has been tested and compared on synthetic data sets with the method based on Hough transformation. The efficiency of this method might be significantly improved in direct application. The proposed method has been used in this paper for the case of two or three crop rows. The generalization to several crop rows is also given in the paper, but was not implemented. Also, the method could be expanded in case when the number of crop rows is not known in advance. [ABSTRACT FROM AUTHOR]

In agriculture, crop monitoring and plant phenotyping are mainly manually measured. However, this practice gathers phenotyping information at a lower rate than genotyping evolves, thus producing bottleneck. This paper presents vitisBerry, a smartphone application for assessing in the vineyard, using computer vision, the berry number in clusters at phenological stages between berry-set and cluster-closure. The implemented image analysis algorithm is an evolution of a previous development, providing 1.63% and 7.57% of Recall and Precision improvement, respectively. The application was evaluated using two devices, taking and analysing 144 images from 12 different grapevine varieties. The Recall and Precision results ranged between 0.8762 and 0.9082 and 0.9392–0.9508, depending on the device. The average computational time required to analyse the 144 images varied from 3.14 to 8.40 s. According to these results, vitisBerry constitutes a tool for viticulturists to acquire phenotyping information from their vineyards in an easy and practical way. [ABSTRACT FROM AUTHOR]

To reduce aquaculture risk and optimize water quality management in prawn culture ponds, this paper uses mechanistic and statistical analytic methods to propose a hybrid water temperature forecasting model based on the water temperature mechanism model (WTMM) with optimal parameters selected by an improved artificial bee colony (IABC) algorithm. Because of existing problems with using an artificial bee colony algorithm in modeling, an improved ABC with a dynamically adjusted inertia weight based on the fitness function value was implemented to improve local and global search abilities. Then, IABC was employed to adaptively search for the optimal combinatorial parameters needed in the WTMM model, which overcomes the blindness of and limits to parameter selection for the traditional WTMM model. We adopted an IABC-WTMM algorithm to construct a non-linear mechanical prediction model. The IABC-WTMM was tested and compared to other algorithms by applying it to the prediction of water temperature in prawn culture ponds. Experimental results show that the proposed IABC-WTMM could increase prediction accuracy and execute generalization performance better than the original water temperature mechanism model (O-WTMM) and back-propagation neural network (BP-NN), but was inferior to the standard LSSVR model. Overall, it is a suitable and effective method for predicting water temperature in intensive aquacultures. [ABSTRACT FROM AUTHOR]

Removal of internal organs such as lungs, liver, and kidneys is a key step required to compute the lean meat percentage from Computed Tomography (CT) scans of live animals. In this paper, we propose two segmentation techniques to remove these organs focusing on pigs. The first method is semi-automatic, and it starts with the first CT slice and a manually defined mask with internal organs. Then, it applies a four-step iterative process that computes the masks of the next CT slices by using the information of the previous one. To find the best boundary it uses a Dynamic Programming-based approach. At each iteration the user can check the correctness of the new computed mask. The second method is fully automatic, and segments each slice individually by using distance maps and morphological operators, such as dilation. It is composed of three main steps which detect the pig’s torso, pre-classify the voxels in different tissues, and segment the internal organs using the information of such classification. Although it has some parameters, user interaction is not required to obtain the results. The proposed approaches have been tested on CT data sets from 9 pigs, and compared with a manual segmentation. To evaluate the results, the precision, recall, and F-score measures have been used. From our test, we can observe that the performance of both methods is very high according to their average F-score. We also analyse how the accuracy of the results in the semi-automatic approach increases when more user interaction is applied. For the automatic approach, we evaluate the dependence of the results on the algorithm’s parameters. If robustness is enough, and high accuracy is not required, the automatic algorithm can be used to segment a whole pig in less than 50 s. However, if the user wants to control the level of accuracy, the semi-automatic algorithm is preferred. Both methods are useful to reduce the time needed to segment the internal organs of a pig from hours (manual segmentation) to minutes or seconds. [ABSTRACT FROM AUTHOR]

Increase of huge amount of data in every application demands an incremental learning technique for data analysis. One of such data analysis task in dynamic environment is to design an incremental classifier for decision making and consequently updating the knowledge base of the overall system. Classifier construction depicts extraction of interesting patterns from the large repository of data and predicts the future trends based on the existing patterns. The time complexity of the classification system increases gradually and the system becomes inefficient while it is learned repeatedly for adding new group of data with the existing one in a certain interval of time. Without learning the same classifier for the whole data, if the knowledge of old data extracted by the classifier is used together with the new group of data to design the updated classifier, called incremental classifier, then time complexity reduces drastically. In the paper, the concepts of Particle Swarm Optimization technique and Association Rule Mining are used to design an incremental rule based classification system. The incremental classifier is suitable to apply on rice disease dataset for disease prediction as the characteristics of rice diseases change in time due to change of climate, biological, and geographical factors. The proposed method has been applied on both simulated rice disease dataset and benchmark datasets and the classification accuracy is measured and compared with various state of the art classification algorithms. The method is also evaluated based on some statistical measures and statistical test is done to establish its significance and effectiveness. [ABSTRACT FROM AUTHOR]

Automatic plant leaf recognition can play an important role in plant classification due to leaf’s availability, stable features and good potential to discriminate different kinds of species. Amongst many leaf features like leaf venation, margin, texture and lamina, leaf shape is the most important one due to its better discriminative power and ease of analysis. One of the most common leaf shape descriptors is Elliptic Fourier Descriptor (EFD). In this paper a new shape descriptor is introduced as “Rotation Invariant Wavelet Descriptor” (RIWD). The performance of RIWD is compared with IEFD using Flavia dataset. MLP neural network is used as the classifier in this work. Results analysis shows better performance of the proposed feature in classification accuracy. Furthermore, an optimum feature vector is constructed using a set of textural and morphological features and the RIWD that reached 97.5% classification accuracy with low computational cost in comparison with many reported results in Flavia dataset. [ABSTRACT FROM AUTHOR]

• A goal point determination method is proposed for the problem of selecting the optimal path point by Pure Pursuit. • An evaluation function is established to select the optimal goal point based on tractor position prediction model. • The performance of the proposed path tracking algorithm is studied by field test. The development of precision agriculture requires the implementation of automatic navigation technology for agricultural machinery. Path tracking control as one of the key steps in automatic navigation technology has great research value. Pure Pursuit algorithm as a popular control algorithm for automatic navigation technology for agricultural machinery, the look-ahead distance is the key to the tracking effect, while the calculation of look-ahead distance has the problem that many influencing factors are difficult to be accurately described by mathematical expressions, which leads to the difficulty to select a suitable goal point tracking path. In order to solve the above problems, this paper proposes a path tracking algorithm for agricultural machinery based on the optimal goal point. The algorithm simulates the look-ahead behavior of the driver and searches for optimal goal point in the look-ahead area according to the evaluation function. The objective of the study is to minimize the lateral error and heading error to achieve the adaptive optimization of the goal point. Finally, the feasibility of the algorithm proposed in this paper was verified in simulations and bumpy field tests under various different conditions, with tracking errors reduced by more than 20% compared to the pure pursuit algorithm. The tracking accuracy is significantly improved. [ABSTRACT FROM AUTHOR]

In this paper, an open algorithm was developed for the detection of cattle’s grass intake and rumination activities. This was done using the widely available inertial measurement unit (IMU) from a smartphone, which contains an accelerometer, a gyroscope, a magnetometer and location sensors signals sampled at 100 Hz. This equipment was mounted on 19 grazing cows of different breeds and daily video sequences were recorded on pasture of different forage allowances. After visually analyzing the cows’ movements on a calibration database, signal combinations were selected and thresholds were determined based on 1-s time windows, since increasing the time window did not increase the accuracy of detection. The final algorithm uses the average value and standard deviation of two signals in a two-step discrimination tree: the gravitational acceleration on x-axis (Gx) expressing the cows’ head movements and the rotation rate on the same x-axis (Rx) expressing jaw movements. Threshold values encompassing 95% of the normalized calibrated data gave the best results. Validation on an independent database resulted in an average detection accuracy of 92% with a better detection for rumination (95%) than for grass intake (91%). The detection algorithm also allows for characterization of the diurnal feeding activities of cattle at pasture. Any user can make further improvements, for data collected at the same way as the iPhone’s IMU has done, since the algorithm codes are open and provided as supplementary data . [ABSTRACT FROM AUTHOR]

Automatic detection of defective oranges by computer vision system is not easy because of the uneven lightness distribution on the surface of oranges. It means that the methods only directly using global segmentation provide unsatisfactory results when orange images present faint defect characters or inhomogeneous surface. The contrast between sound and defective regions can be used to produce more accurate segmentation results, which is more capable of detecting pixels lying around the defect boundary on orange surface based on the local segmentation method. In this paper, we study and propose a sliding comparison window local segmentation algorithm and also presents the detailed image processing procedure including removal of background pixels, image binarization using local segmentation, image subtraction, image morphological modification, removal of stem end pixels for detecting surface defect in an orange gray-level image. This method is an original contribution that allows successful segmentation of various types of surface defects (e.g., insect injury, wind scarring, thrips scarring, scale infestation, canker spot, dehiscent fruit, copper burn, phytotoxicity).The image segmentation algorithm was tested with 1191 samples of oranges. The proposed algorithm was able to correctly detect 97% of the defective orange. Future work will be focused on whole surface and fast on-line inspection. [ABSTRACT FROM AUTHOR]

• The proposed Anti-Gravity Stem-Seeking (AGSS) root image restoration algorithm can quickly and effectively repair root images. • AGSS root image restoration algorithm has good universality and does not need training data and high-performance hardware platforms. • Drought and waterlogging resistances of maizes can be effectively evaluated by the phenotypic parameters of repaired seed roots. Root phenotype detection is key to cultivating seeds with excellent traits, and requires a complete root image. However, soil occlusion, uneven lighting, and other factors cause broken points and segments in the root image. To solve this problem, an anti-gravity stem-seeking (AGSS) root image restoration algorithm is proposed in this paper to repair root images and extract root phenotype information for different resistant maize seeds. First, the obtained root image was processed using uniform illumination, grayscale, binarization, and morphological filtering to separate it from the background. Subsequently, a root skeleton map was generated using a thinning algorithm for pixel-level image processing, and the Taproot junction G was obtained. Subsequently, root pixel coordinates were obtained by traversing the root skeleton map. All root-segment endpoint coordinates were obtained using the endpoint judgment rule and stored in the endpoint list. The lateral and primary root endpoints were separated based on the lateral root judgment rule, and stored in the side root and primary root endpoint lists, respectively. Subsequently, the primary root endpoints were processed and fitted in the order short to long using arbitrary-two-endpoint spacing until all breakpoints of the primary root were found. The coordinates of the top endpoints of each principal root were obtained. Finally, the top endpoint was connected to point G based on the Bezier curve-fitting method to achieve complete root repair. The proposed AGSS root image restoration algorithm was applied to detect the root systems of maize with different resistances and wheat to evaluate its performance against the standard dataset. The results indicated a detection accuracy of greater than 90% for root taproot length and diameter. It was also found that maize drought resistance was positively correlated with root length and diameter, but negatively with the lateral root number. In contrast, the waterlogging and salt resistance traits of maize were positively correlated with the number of lateral roots. In conclusion, the proposed AGSS root image restoration algorithm can quickly and effectively repair root images, is suitable for different resistance evaluations of maize seeds, and is conducive to the detection of root phenotypes. Compared with deep learning methods, this algorithm displays advantages of fast repair, low hardware platform requirements, and less requirement of training images. The algorithm is highly suitable for deploying in small embedded systems, with broad application prospects. [ABSTRACT FROM AUTHOR]

• It is an efficient, accurate and real-time detection algorithm. • High precision detection technology is not disturbed by field noise. • It provides technical support for research related to variable fertilization. Efficient and accurate detection of fertilizer discharge volume is a prerequisite for high-quality variable fertilization operations, but the detection accuracy of existing fertilizer discharge volume detection systems is greatly disturbed by the noise of field operations. This paper develops an efficient algorithm for converting fertilizer point cloud data into fertilizer volume (P&V Algorithm) by proposing two assumptions that can be used as the basis for the calculation of fertilizer volume: (I. the overall volume of fertilizer is not affected by the field environment. II. The area of change in the geometric characteristics of fertilizer can fully reflect the volume change), and then a new fertilizer volume detection system is designed (P＆V System). The system realizes real-time acquisition of point cloud data of fertilizer geometry by integrating single-line LiDAR and other hardware. The algorithm uses escape value filtering, outlier removal, ordered point cloud set construction, and point cloud smoothing to achieve noise reduction of point cloud data, and uses dynamic region extraction to significantly reduce the computation amount, and transforms point cloud data into volume efficiently by improving Delaunay triangulation division, and then calculates the mass of fertilizer discharge. The test results showed that the measurement error (ME) of the P&V system during field operation was 2.34–4.66% (N), 2.14–4.25% (K) and 1.87–4.59% (P) when the fertilizer drain speed was in the range of 20r/min-70r/min, and there was no significant difference between the detected values of the system and the actual mass of fertilizer discharged (P > 0.01). Under the same field operation conditions, the ME of P&V System was significantly lower than that of the conventional fertilizer apparatus speed detection technology (P < 0.01) and the common fertilizer flow detection technology (P < 0.01). The research results of this paper can provide a new solution for the research of fertilizer discharge detection technology. [ABSTRACT FROM AUTHOR]

A new sensing system to estimate the mass flow of seeds in the grain drill was developed and evaluated in this paper. This was a manipulation of an estimation model developed indoors using fibre sensors; by considering the operational conditions of the machine and environment. Therefore, an original method to install the sensor within the grain drill was tried, and modifications on the estimation model and algorithm were performed. The experiment to test and evaluate the sensing system consisted of running a grain drill-propelling tractor in the farm of Hokkaido University several times to sow rye seeds, where each run was approximately 160 m. Two fibre sensors were installed at two different metering units, that use axial flute rollers, inside the grain drill. Also, the seeds being discharged during the experiment were collected and weighed at the end of each run. The results of 52 trials to estimate the mass flow of seeds indicated that in approximately 98% of the trials the estimation rate exceeded 90%. Moreover, the overall estimation rate was approximately 95%. These results seemed not to be biased by the change of the sensor set nor the sensing location. The modifications on the estimation model were successful in eliminating the overestimation or underestimation bias that resulted by changing the speed in the indoor experiments. In addition, the modifications on the estimation algorithm could eliminate error in the sensor output values that may result from dust, vibration, or variation in the sensor internal resistance. These results showed that the sensing system can be used practically to monitor the seed flow in the grain drill which would have many applications in precision agriculture practices. [ABSTRACT FROM AUTHOR]

The average weight of piglets in lactation can be monitored automatically by piglets’ average weight monitoring systems which are designed based on wireless multimedia sensor networks. Piglets counting in an automatic manner for piglets images is the foundation of these systems. Adhesive piglets may exist in an image due to the social character of piglets, which challenges the image splitting and automatic piglets counting. This paper proposes a segmentation algorithm for adhesive piglets images based on ellipse fitting method. Firstly, ellipse fitting is implemented for a large number of images which have one piglet. Parameters range of ellipses fitted by images with a single piglet on different age is extracted. Secondly, contours of connected components in an adhesive piglets image are extracted. Each contour is segmented based on concave points. Ellipse fitting is implemented for each contour segment. Finally, 5 rules for ellipse merging are put forwarded, which are used to merge anomalous ellipses. After ellipse merging, the number of ellipses equals the number of piglets. The proposed algorithm is applied to adhesive piglets images in Matlab R2012b and the experimental results show that the counting accuracy exceeds 86% when the number of piglets is less than 7. The algorithm provides the foundation for the piglets’ average weight monitoring systems. [ABSTRACT FROM AUTHOR]

Soil available water is an important factor for plant growth. It has been estimated by different soil physical indices, such as the least limiting water range (LLWR), integral water capacity (IWC) and integral energy ( E I ). Moreover, salinity is an important limitation for soil water availability to plants. Despite the advances in the quantification of LLWR, IWC and E I , a comprehensive description of the computational methods, including data management, curve fitting procedures and graphing techniques, is still lacking. The salinity effect on these quantities has still not been implemented in a computer package. In this paper, we present an R package soilphysics and its implementations to determine LLWR, IWC and E I . We described the theory behind each implementation, illustrated the functionalities and validated the outcomes of soilphysics with other software packages for LLWR, IWC and E I calculations (an Excel® algorithm and SAWCal). The salinity effect on soil available water was also employed in the package. The outcomes are basically the same as other software available, with small differences (<4%). The package soilphysics takes advantage of all the power of R for dealing with extensive algorithms and for building high-quality graphics. It is currently available from the CRAN website ( http://cran.r-project.org/web/packages/soilphysics/index.html ). [ABSTRACT FROM AUTHOR]

Accurate mobile robot localisation in orchards relies on precise orchard maps which help the mobile robot to efficiently estimate its position and orientation while moving between tree rows. This paper presents a new method for constructing a local orchard map based on tree trunk detection using camera and laser scanner data fusion. The final orchard map consists of the positions of the trees and non-tree objects (e.g. posts and tree supports) in the tree rows. The map of the individual trees is used as an a priori map to localise the mobile robot in the orchard. A data fusion algorithm based on an Extended Kalman Filter is used for position estimation. Experimental tests were conducted with a small robot platform in a real orchard environment to evaluate the performance of orchard mapping and mobile robot localisation. The mapping method successfully localised all the trees and non-tree objects of the tested tree rows in the orchard. The mapping results indicate that the constructed orchard map can be reliably used for mobile robot localisation and navigation. The localisation algorithm was evaluated against the logged RTK-GPS positions for different paths and headland turns. The average of the root mean square of the Euclidean distance between the ground truth and the estimated position for different paths was 0.103 m, whilst the average of the root mean square of the heading error was 3.32°. [ABSTRACT FROM AUTHOR]

The fieldwork pattern, which is the ordered traversal sequence of in-field tracks by the machine, plays an important role in operational efficiency. A sub-optimal fieldwork pattern is a main reason for lost time in fieldwork operations due to excessive non-working distance and time. Even though algorithms exist that can calculate an optimal route plan of a specific operation that covers the field with minimum non-working distance and time, so far no commercial navigation-aiding system for agricultural vehicles exist that can implement this type of patterns. Yet, the implementation of applicable standard fieldwork pattern in real life operations can provide near-optimal solutions compared to the simple fieldwork patterns generally selected by operators. In this paper, the potential savings from following an alternative fieldwork pattern instead of the one selected by an experienced operator are assessed. We simulate the distance and time corresponding to five selected common fieldwork patterns and compare them with the operators’ selected fieldwork pattern. In order to simulate the fieldwork patterns, four standard turn models, which are formulas for calculating the turning distance of various types of turns, are fitted to actual turns recorded in field experiments, and the turn models together with a model of in-field transport and material reloading are evaluated. As a case study three sequential operations of potato cultivation (bed forming, stone separation and planting) were recorded and analysed in three different fields. The simulation results showed significant potential savings of both distance and time in all three operations and all three case study fields by selecting the best fieldwork pattern instead of the actual selected pattern: For bed forming the savings were up to 18.4% in distance and 32.7% in time; For stone separation the savings of distance and time were up to 35.0% and 60.9%; For the planting the savings were up to 22.6% in distance and 24.8% in time. The increase in time-based field efficiency was up to 2.7%, 7.2% and 7.1% for bed forming, stone separation and planting, respectively. [ABSTRACT FROM AUTHOR]

Trees in orchards are natural landmarks providing suitable cues for mobile robot localisation as they are nominally planted in straight and parallel rows. This paper presents a novel tree trunk detection algorithm using a camera and laser scanner data fusion to enhance the detection capability. The algorithm detects the trees in the orchard and discriminates between trees and non-tree objects (e.g. posts and tree supports). The laser scanner is used to detect the edge points and determine the width of the tree trunks and non-tree objects, while the camera images are used to verify the colour and the parallel edges of the tree trunks and non-tree objects. The algorithm automatically adjusts the colour detection parameters after each test which shown to increase the detection accuracy. Experimental tests were conducted with a small robot platform in a real orchard environment to evaluate the performance of the tree trunk detection algorithm under two broad illumination conditions (sunny and cloudy). The algorithm was able to detect the tree trunks and discriminate between trees and non-tree objects with detection accuracy of 96.64% showing that the fusion of both vision and laser scanner technologies produced robust tree trunk detection. [ABSTRACT FROM AUTHOR]

Mastitis is an infectious disease associated with the increased number of somatic cells in cow’s milk, and it is one of the most relevant cause of economic losses in dairy farming industries. In this paper, we propose a method capable of determine, with 99.7% accuracy, the number of these cells in microscope slide images. This level of accuracy is achieved by changing the image original RGB format to Lab color space and applying k -means clustering algorithm to remove debris and other background features. A new gray level thresholding is proposed, and the remaining bound cells are separated in the final segmentation step applying Watershed transform. Many microscope slide images with debris, contrast, and hue variation were used to validate the experimental results. Comparison between the proposed method and manual counting indicates that this new approach is a robust and promising solution to be incorporated in a future automated somatic cell counting system using optical microscopy. [ABSTRACT FROM AUTHOR]

• An algorithm for identifying and locating sugarcane stem nodes is proposed. • Use the horizontal projection function of binary image to intercept the region of interest. • Use vertical projection function and local pixels to locate sugarcane stem nodes. • The algorithm has good performance in recognition rate, accuracy and time-consuming. Aiming at the difficulty of sugarcane nodes identification and location during automatic cutting of sugarcane seeds, based on machine vision system, this paper proposed a sugarcane nodes identification algorithm based on sum of local pixel of minimum points of vertical projection function. Firstly, according to the color and texture characteristics of yellow sugarcane, the captured RGB image of sugarcane was converted into HSV color image. Then, the S-component image in the HSV image was extracted, and the S-component image was binarized by the Otsu algorithm to obtain the binary image, and then the binary image was processed by morphological closed operation, which can eliminate the noise and fill holes of binary image. Subsequently, the sugarcane area was segmented as the region of interest through the horizontal projection of the binary image, which lowered the amount of calculation while reducing interference. Finally, the vertical projection function of the binary image of the region of interest was established, and the function was continuously derived to obtain the minimum points, and then the position of sugarcane nodes was preliminarily determined by the obtained minimum points. Then, the final position of the sugarcane nodes were determined according to the number of nodes to be identified and the sum of pixels of each 5 columns on both sides of the minimum points. The pixel column positions corresponding to the minima of the sum are the accurate nodes positions determined by the proposed algorithm. The experimental results show that the algorithm proposed in this paper has a single node identification rate of 100%, with an average time consumption of 0.15 s, and a position deviation of less than 0.34 mm; a double nodes identification rate of 98.5%, with an average time consumption of 0.21 s, and a position deviation of less than 0.42 mm. Compared with other nodes identification algorithms mentioned in this paper, it has higher identification rate and accuracy. [ABSTRACT FROM AUTHOR]

The tilt angle of a machine is important for its stability and technological operations. Most sensors are gravity or inertial based, and their outputs can be affected by vibration induced errors, random errors or drift. A sensor that works on a different physical principle in collaboration with these sensors could decrease the errors and thus increase the trustworthiness of the measurement. This paper deals with the optical determination of the machine’s tilt angle, using camera images of the trees. This method assumes the surrounding forest, in the machine’s location, as a base for the analysis and it is focused on the feasibility of the idea. The Sobel algorithm for edge detection is used to process the inputs from the camera. A simple and fast algorithm inspired by the fringe pattern has been designed. The results show that the initial assumption of the upright position of the trees in the direction of gravity is sufficient and can be used to determine the machine’s position on a slope, with an average error in the range of 0.63–1.63°. This method can also be utilized to determine a vehicle’s or robot’s inclinations in an urban environment, which is possible without the modification of the method. [ABSTRACT FROM AUTHOR]

Plant leaf area index (LAI) is an important structural parameter for describing the ability of the canopy to intercept solar radiation, and it plays an important role in the estimation of forest biomass. To overcome the difficulties of current ground measurement instruments in automatically collecting plant LAI, we have designed a continuous LAI measurement system: LAINet. The main part of LAINet includes the distributed Below Node (BN), which is deployed below the canopy and receives transmitted solar radiation, the Above Node (AN), which is used to receive downward radiation above the canopy, and the Central Node (CN), which is used as a data reception and control node. Communication among the nodes is implemented through the Zigbee protocol, while data exchange between the CN and the remote data center server is completed through the General Packet Radio Service (GPRS) network. The design of the LAINet framework and the algorithm for estimating the LAI from a multi-point directional transmittance is presented in this paper. The sensors of LAINet are calibrated on two clear days, and the performance of the proposed system is compared in time series field experiments. Comparison of the nodes’ sensor results indicates that under the condition of the same solar radiation, the response between the different sensors is of high consistence, and each node has a stable calibration coefficient across different dates. In the field coniferous forest experiment, upon comparing with TRAC (Tracing Radiance and Architecture of Canopies) instruments and the results obtained using different data processing algorithms, it is found that the LAINet system can produce the best results with a minimum root mean square error (RMSE = 0.34) and a higher determination coefficient ( R 2 = 0.81). We adopt a multi-day window for aggregating the daily observed LAI to remove the noise that occurs from daily weather variation, and it indicates that the aggregated time series LAI can capture the process of the coniferous needles’ growth and defoliation properly. Thus, it is confirmed that LAINet can be used as a feasible method to conduct automatic ground observations, especially in the scenario where continuous observation is required. [ABSTRACT FROM AUTHOR]

• The state-of-the-art image-based researches in plant phenotyping were reviewed. • Effective image-based solutions and pipelines for plant phenotyping were introduced. • Various modern imaging techniques in plant phenotyping were discussed. • The challenges and future perspectives of image-based plant phenotyping were stated. Plant phenotype plays an important role in genetics, botany, and agronomy, while the currently popular methods for phenotypic trait measurement have some limitations in aspects of cost, performance, and space-time coverage. With the rapid development of imaging technology, computing power, and algorithms, computer vision has thoroughly revolutionized the plant phenotyping and is now a major tool for phenotypic analysis. Based on the above reasons, researchers are devoted to developing image-based plant phenotyping methods as a complementary or even alternative to the manual measurement. However, the use of computer vision technology to analyze plant phenotypic traits can be affected by many factors such as research environment, imaging system, research object, feature extraction, model selection, and so on. Currently, there is no review paper to compare and analyze these methods thoroughly. Therefore, this review introduces the typical plant phenotyping methods based on computer vision in detail, with their principle, applicable range, results, and comparison. This paper extensively reviews 200+ papers of plant phenotyping in the light of its technical evolution, spanning over twenty years (from 2000 to 2020). A number of topics have been covered in this paper, including imaging technologies, plant datasets, and state-of-the-art phenotyping methods. In this review, we categorize the plant phenotyping into two main groups: plant organ phenotyping and whole-plant phenotyping. Furthermore, for each group, we analyze each research of these groups and discuss the limitations of the current approaches and future research directions. [ABSTRACT FROM AUTHOR]

• Continuous Point Slope Blur is convenient for data processing. • Characteristics of seedling band obtained by slope virtual clustering. • Iterative clustering algorithm for obtaining seedling band trajectory. In this paper, a method for identifying rice seedling band tracks by using LIDAR data is proposed, and the method provides navigation information for straight-line driving to minimize seedling injuries when paddy field management machinery enters a rice field. In this paper, the reconstruction of seedling band tracks is realized by constructing a three-dimensional (3D) point cloud image by collecting LIDAR data. Based on a two-dimensional servo platform, a mean-shift clustering algorithm is designed with slope virtualization, iterative zone clustering and linearization processes. The point cloud data of rice field on the 35th day of transplanting were simulated by Matlab. In the environment 2–3 m in the forward-looking direction and a standard plant spacing of 21 cm, the reconstructed seedling trajectory was compared with the real seedling trajectory. The maximum parallelism of the seedling belt trajectory was 45 mm, and the maximum median deviation was 7 mm. These values meet the needs of paddy field management machinery for achieving nondestructive seedling driving trajectories and verify the feasibility of the slope-blurred mean-shift clustering algorithm for seedling identification. [ABSTRACT FROM AUTHOR]

• This paper implements minimal non-working distance in intra- and inter-fields. • The moist fields constrained harvesters′ type in mathematical model. • A hybrid algorithm composed of ALNS and tabu search are employed. • The results indicate the methods are effective to reduce non-working distance. Reducing the non-working distance for agricultural vehicles is beneficial for reducing the total operational costs for farmers. In this study, we optimize the non-working distance in intra- and inter-fields simultaneously for harvesters while harvesting rice. A mixed integer programming is presented for optimizing the rice harvesting problem, where a similar working time for all harvesters are considered based on a real case. Besides, in this model, two kinds of fields, namely moist and non-moist fields are considered, where the first can only be harvested by crawler harvesters while the second can be harvested by any type of harvester. This model aims to find an optimal schedule for the harvesters. A hybrid algorithm using both adaptive large neighborhood search and tabu search techniques is implemented to optimize the problem. A case study is considered to verify the performance of the hybrid algorithm. The results indicate that the hybrid algorithm is an effective method for solving the rice harvesting problem because the total non-working distance could be improved significantly by 15.3%. The proportion of moist fields is analyzed because it influences the schedule significantly. When compared to the traditional method for optimizing the non-working distance for harvesters, the total non-working distance in this method could be reduced by 6.2%. This paper provides farmers with a general framework for reducing the non-working distance in intra- and inter-field logistics, which results in a more efficient performance of the harvesters in the fields. [ABSTRACT FROM AUTHOR]

• An algorithm for detecting ripe fruit under natural illumination and occlusion conditions is proposed. • A graying factor K is proposed to remove the complex background of colorful ripe fruits. • Improving the Hough Transform by selecting equidistant points and increasing judgement threshold. • A real contour Hough Transform was proposed to detect ripe fruits or occluded circles. For an accurate detection of ripe fruits under uneven illumination and ubiquitous occlusion conditions, this paper proposes a method to detect and locate ripe fruits based on machine vision. There are four key steps in this method including image graying and background removal, binary image optimization, true contour fragments extraction, and fruit fitting. For testing the proposed method, field experiments were conducted with tomato and citrus, and the ripe fruits in complex environments were successfully detected and located. From the detection experiments, it showed that the recognition rate for ripe fruits in the near zone of the proposed method was higher than 97.44%, the average time consumption was 0.2966 s, and the positioning error was less than 4.41%. In addition, it can be concluded from the comparative experiment that the proposed method is superior to conventional Hough transform, random Hough transform, and other methods based on deep learning in terms of detection rate, time performance and positioning accuracy. Therefore, it can be applied to picking robots for real-time detecting and locating ripe fruits. [ABSTRACT FROM AUTHOR]

• Implemented Smart Farming increases farmers' profit by improving product quality. • A BPMN-standardized coffee value chain model provides farm usability and management. • Data Analytics Layer favors the implementation of machine learning algorithms. • Edge Layer detects outliers avoiding farm stakeholders operate with unreliable data. • Perception Layer compares IoT technologies by performance, usability, and price. Currently, the adoption of smart technologies for sustainable farming systems creates a distinct competitive edge for farmers, extension services, agri-business, and policy-makers. However, selecting the most appropriate technologies from a wide range of options is never an easy job. In this context, several authors consider Smart Farming as the best solution. However, they fall short in providing more information to recommend the most appropriate IoT technology, the options to manage the IoT infrastructure, and the services to crop management plans and crop production estimation. This paper implements a Smart Farming System based on a three-layered architecture (Agriculture Perception, Edge Computing, and Data Analytics). In the Agriculture Perception Layer, we evaluated Omicron, Libelium, and Intel technologies under criteria such as the price, the number of inputs for sensor connection, communication protocols, portability, battery life, and harvesting energy system photovoltaic panel. We evaluated edge-based management mechanisms in the Edge Layer to provide data reliability, focusing on outlier detection and treatment using Machine Learning and Interpolation algorithms. We recommend the Isolation Forest algorithm for classifying outliers in the monthly temperature dataset (99% of precision) and the Cubic Spline technique for effectively replacing the data classified as outliers (RMSE lower than 0.085). In the Data Analytics Layer, we evaluated different machine learning algorithms to estimate coffee production. The results show that the measured error values of the XGBOOST algorithm keep the values lower than the other models (RMSE 0.008, MAE 0.032, and RSE 0.585). The www.iot-agro.com platform offers farmer services such as weather variables monitoring, coffee production estimating, and IoT infrastructure setting. Finally, stakeholders, researchers, and engineers validated our Smart Farming Solution through a Colombian coffee farm case study. The test evaluated the usability, the straightforward interpretation of data, and the look feel of the web application. [ABSTRACT FROM AUTHOR]

Alternaria blotch, Brown spot, Mosaic, Grey spot, and Rust are 5 common apple leaf diseases that severely impact apple production and quality. At present, although many CNN methods have been proposed for apple leaf diseases, there are still lack of apple leaf disease detection models that can be applied on mobile devices, which limits their application in practical production. This paper proposes a light-weight CNN model that can be deployed on mobile devices to detect apple leaf diseases in real time. First, a dataset of apple leaf diseases composed of simple background images and complex background images, which is called AppleDisease5, is constructed via data augmentation technology and data annotation technology. Then a basic module called MEAN block(Mobile End AppleNet block) is proposed to increase the detection speed and reduce model's size by reconstructing the common 3 × 3 convolution. Meanwhile, the Apple-Inception module is built by introducing GoogLeNet's Inception module and replacing all 3 × 3 convolution kernels with MEAN block in Inception. Finally, a new apple leaf disease detection model, MEAN-SSD(Mobile End AppleNet based SSD algorithm), is built by using the MEAN block and Apple-Inception module. The experiment results show that MEAN-SSD can achieve the detection performance of 83.12% mAP and a speed of 12.53 FPS, which illustrates that the novel MEAN-SSD model can efficiently and accurately detect 5 common apple leaf diseases on mobile devices. [ABSTRACT FROM AUTHOR]

Abstract: Segmentation is the first step in image analysis to subdivide an image into meaningful regions. It directly affects the subsequent image analysis outcomes. This paper reports on the development of an automatic adjustable algorithm for segmentation of color images, using linear support vector machine (SVM) and Otsu’s thresholding method, for apple sorting and grading. The method automatically adjusts the classification hyperplane calculated by using linear SVM and requires minimum training and time. It also avoids the problems caused by variations in the lighting condition and/or the color of the fruit. To evaluate the robustness and accuracy of the proposed segmentation method, tests were conducted for 300 ‘Delicious’ apples using three training samples with different color characteristics (i.e., orange, stripe, and dark red) and their combination. The segmentation error varied from 3% to 25% for the fixed SVM, while the adjustable SVM achieved consistent and accurate results for each training set, with the segmentation error of less than 2%. The proposed method provides an effective and robust segmentation means for sorting and grading apples in a multi-channel color space, and it can be easily adapted for other imaging-based agricultural applications. [Copyright &y& Elsevier]

Abstract: Spatial–temporal modelling of environmental systems such as agriculture, forestry, and water resources requires high resolution input data. Assembling and summarizing this data in the appropriate format for model input often requires a series of spatial analyses which can be extremely time-consuming, especially when many large data sets are involved. In this paper we investigated the ability of high-performance computing techniques to improve the efficiency of spatial analysis for model data assembly. We implemented an array-based algorithm to calculate summary statistics for long time-series daily grid climate data sets for 11,575 climate–soil zones across the Australian wheat-growing regions for input into a crop simulation model. We developed a zonal statistics algorithm using Python’s Numpy module then parallelized it and processed it using a shared memory, multi-processor system. We assessed algorithm performance with a varying number of CPU cores, and assessed the influence of load balancing on the efficiency of parallel processing. Compared with traditional desktop GIS software, the serial and parallel (32 cores) implementation achieved about 180 and 1440 times speed-up, respectively. We also found that the most efficient computation occurred when not all of the available CPU cores were used, and the chunk size of jobs also had an important influence on computing efficiency. The algorithm and the parallel processing scheme provides a useful approach to address computing challenges posed by spatial analysis of numerous large data sets for large scale environmental modelling. [Copyright &y& Elsevier]

Abstract: Remote diagnose of fish diseases for farmers is unrealized in China, but use of mobile phones and remote analysis based on image processing can be feasible due to the widespread use of mobile phones with camera features in rural areas. This paper presents a novel method of classifying species of fish based on color and texture features and using a multi-class support vector machine (MSVM). Fish images were acquired and sent by smartphone, and the method utilized was comprised of the following stages. Color and texture subimages of fish skin were obtained from original images. Color features, statistical texture features and wavelet-based texture features of the color and texture subimages were extracted, and six groups of feature vectors were composed. LIBSVM software was tested using leave-one-out cross validation to find the best group for classification in feature selection procedure. Two multi-class support vector machines based on a one-against-one algorithm were constructed for classification. The feature selection results showed that the Bior4.4 wavelet filter in HSV color space achieved greater accuracy than the other feature groups. The classification results indicate that only the DAGMSVM meets the requirement of time efficiency for the system. The results of this study suggest that the best classification model for fish species recognition is composed of a wavelet domain feature extractor with Bior4.4 wavelet filter in HSV color space and a one-against-one algorithm based DAGMSVM classifier. [Copyright &y& Elsevier]

Abstract: This paper introduces a real-time labor monitoring system (LMS) with the ability to track and record individual picker efficiency during manual harvest of specialty crops. This system utilizes existing commercial harvest equipment and integrates a digital weighing scale, RFID reader, computational unit, and a portable datalogger carried by pickers. The RFID reader, digital scale and computational unit are assembled on a common portable chassis. As pickers transfer fruit into a standard collection bin, the system reads the picker’s ID (RFID tag) and the weight of fruit. Weight data can then be transmitted wirelessly to the picker’s datalogger which records and displays the incremental and total weight of harvested fruit. An algorithm was developed in Matlab® to record, process and store the data, as well as to transmit wirelessly the weight value to the wearable datalogger. System prototypes were assembled and field-tested for accuracy and reliability during commercial harvest of sweet cherries (Prunus avium L.) in the Pacific Northwest. The LMS reliably calculated the harvest rate, picking cost, weight of harvested fruit, number of harvested buckets, range in fruit weight per bucket, and mean fruit weight per bucket, in real time. The mean harvest rate (±standard error) in a ‘Chelan’/Mazzard sweet cherry orchard trained to a steep leader architecture was 0.53±0.13kg/person/min. Harvest rate was similar for the same genotype trained to a steep leader (3 leaders) at 0.50±0.10kg/person/min and for harvest of ‘Tieton’/‘Gisela®6’ trees trained to a central leader architecture (0.53±0.15kg/person/min). When surveyed after using the LMS, every picker indicated a preference for knowing the exact weight of fruit they harvested compared with the current system of reimbursement per bucket or bin. Using the LMS, reliable data on worker efficiency can be collected with minimal interference with standard commercial practices. [Copyright &y& Elsevier]

Abstract: This paper presents an algorithm for analyzing barley kernel images to evaluate cereal grain quality and perform grain classification. The input data comprised digital images of kernels obtained from an optical scanner. The algorithm identified individual kernels’ smooth and wrinkled regions, described their orientation relative to the axis of symmetry, crease visibility and germ location. We were also able to determine the size of the wrinkled and smooth areas on a grain’s surface, which allowed automatic grain classification and kernel quality assessment. The proposed algorithm was tested using barley grain images, and validated by comparison with the evaluation results of a professional assessor. The validation of the algorithm confirmed that it is efficient and robust allowing accurate description of over 93% of kernel samples in comparison with the expert. [Copyright &y& Elsevier]

Abstract: Bruise damage is a major cause of fruit quality loss. Bruises occur under dynamic and static loading when stress induced in the fruit exceeds the failure stress of the fruit tissue. In this article the potential of an artificial neural network (ANN) technique has evaluated as an alternative method for the prediction of apple bruise volume. Neural bruise estimation models were constructed to calculate Golden Delicious apple bruise volume with respect to fruit properties. The neural models were built based upon impact force and impact energy as the main input parameters including fruit curvature radius, temperature and acoustical stiffness. Optimal parameters for the network were selected via a trial and error procedure on the available data. In this paper, the performance of Basic Backpropagation (BB) training algorithm was also compared with Backpropagation with Declining Learning Rate Factor algorithm (BDLRF). It was found that BDLRF has a better performance for the prediction of apple bruise volume. It is concluded that ANN represents a promising tool for predicting apple bruise volume in comparison to regression model. [Copyright &y& Elsevier]

Abstract: The changes occurring in the dynamics of sugar concentration in grape berries are fairly significant during maturation, whereby they are commonly used as a marker of their development. In view of the importance this parameter has for wine producers, this paper designs several models for predicting the must’s probable alcohol level using both meteorological variables and those specific to the vineyard. Presentation is made of a comparative analysis of learning and meta-learning algorithms for the selection of variables and the design of useful predictive models for estimating this level. The models are designed according to data gathered at different locations within the Rioja Qualified Designation of Origin (DOC Rioja, Spain) under different climate conditions, as well as involving different grape varieties. The models designed in this study provide very good results, and following their validation by experts, they have been proven to make a major contribution to decision-making in vine growing. Finally, considering the indices of analysis studied, it has been observed that the ensemble-type model based on the Bagging algorithm with REPTree decision trees records the best results, with a root mean squared error (RMSE) of 8.1% and a correlation of 84.9%. [Copyright &y& Elsevier]

Abstract: Hyperspectral scattering image is an advanced technology widely used in non-destructive measurement of fruit quality. To develop a better prediction model for apple firmness, the present study investigates a model fusion method coupled with wavelength selection algorithms. The current paper first discusses two wavelength selection algorithms, namely, uninformative variable elimination (UVE) and supervised affinity propagation (SAP). The selected effective wavelengths are then set as input to the partial least square (PLS) model. Six hundred “Golden Delicious” apples were analyzed. The first 450 apples were used as sample for the calibration model, whereas the remaining 150 were used for the prediction model. Compared with full wavelengths, the number of effective wavelengths based on the UVE and SAP algorithms decreased to 34% and 35%, but the correlation coefficient of prediction (Rp) increased from 0.791 to 0.805 and 0.814, whereas the root mean-square error of prediction (RMSEP) decreased from 6.00 to 5.73 and 5.71N, respectively. A fusion model was then developed using UVE-PLS and SAP-PLS models coupled with backpropagation neural network. A better prediction accuracy was achieved from the fusion model (Rp=0.828 and RMSEP=5.53N). The model fusion provides an effective modeling method for apple firmness prediction using hyperspectral scattering image technique. [Copyright &y& Elsevier]

Abstract: This paper evaluates the feasibility of applying visible-near infrared spectroscopy for in-field detection of Huanglongbing (HLB) in citrus orchards. Spectral reflectance data from the wavelength range of 350–2500nm with 989 spectral features were collected from 100 healthy and 93 HLB-infected citrus trees using a visible-near infrared spectroradiometer. During data preprocessing, the spectral data were normalized and averaged every 25nm to reduce the spectral features from 989 to 86. Three datasets were generated from the preprocessed raw data: first derivatives, second derivatives, and a combined dataset (generated by integrating preprocessed raw data, first derivatives and second derivatives). The preprocessed datasets were analyzed using principal component analysis (PCA) to further reduce the number of features used as inputs in the classification algorithm. The dataset consisting of principal components were randomized and separated into training and testing datasets such that 75% of the dataset was used for training; while 25% of the dataset was used for testing the classification algorithms. The number of samples in the training and testing datasets was 145 and 48, respectively. The classification algorithms tested were: linear discriminant analysis, quadratic discriminant analysis (QDA), k-nearest neighbor, and soft independent modeling of classification analogies (SIMCA). The reported classification accuracies of the algorithms are an average of three runs. When the second derivatives dataset were analyzed, the QDA-based classification algorithm yielded the highest overall average classification accuracies of about 95%, with HLB-class classification accuracies of about 98%. In the combined dataset, SIMCA-based algorithms resulted in high overall classification accuracies of about 92% with low false negatives (less than 3%). [Copyright &y& Elsevier]

Abstract: The objective of this paper is to develop an algorithm that could be used in order to reduce the spread of chicken hatching in industrial incubators for chicken eggs. The approach that is used is based on frequency analysis of sounds recorded inside the industrial incubator and aims at identifying the time at which all the eggs inside the incubator have reached the internal pipping stage. The developed algorithm is able to be calibrated automatically in order to adjust for sounds around the incubator and the acoustics of every incubator. The algorithm has been implemented in a Digital Signal Processor and applied in real-time in an industrial environment. It is shown that the algorithm can correctly identify the time at which 93–98% of the eggs have had been in the internal pipping stage. This level of accuracy is considered adequate for a practical application focusing on reduction of the hatching window. [Copyright &y& Elsevier]