Search

In order to improve the accuracy of extracting tree breast diameter from backpack laser scanning（BLS） data, this paper filtered out the noise points whose BLS scan data was more than 10 cm away from the nearest point, and used the irregular triangulation algorithm to extract the ground point, and subtracted the elevation value of the point cloud from the corresponding ground point elevation value to normalize the elevation of the point cloud data. At an elevation of 0.8 to 1.8 m, slice with a step of 0.1 m, and filter out different point clouds to extract DBH. The smallest diameter of the fitting circle of all slices in each tree was selected as the diameter at breast height of the tree, and the results of the test results, measured data and the results of extracting DBH from point cloud only after preprocessed were compared. The results showed that by filtering out low intensity point clouds, the root mean square error（RMSE） of the three plots were reduced from 2.63 cm to 0.99 cm, 5.43 cm to 4.72 cm, and 4.01 m to 2.15 cm, the average error decreased from 1.90 cm to 0.79 cm, 5.24 cm to 4.16 cm, and 3.36 cm to 1.58 cm. In the forest data scanned by BLS, the point cloud with low intensity at breast diameter was a point cloud with large stitching error, which would affect the accuracy of the breast diameter extraction result. Filtering out this part of the point cloud can improve the accuracy. The proportion of large stitching error point clouds was related to the terrain, and the proportion of low quality point clouds on slopes was larger than that on flat ground. [ABSTRACT FROM AUTHOR]

Bamboo chip defect detection is an important process to improve product quality. Using images to automatically detect bamboo chip defects can greatly improve production efficiency. In this paper, an image detection method of bamboo chip defects based on OTSU and CANNY algorithm was proposed, which was suitable for seven bamboo chips such as brushed bamboo chips, cracked bamboo chips, borehole bamboo chips, moldy bamboo chips, bamboo green chips, black knot bamboo chips, and triangular bamboo chips. Firstly, the OTSU algorithm was used to binarize the bamboo pieces, the CANNY algorithm was used for edge detection, and the Hough transform was used for tilt correction to realize the extraction of bamboo pieces. Then, OTSU algorithm was used to detect color defects and CANNY algorithm was used to detect shape defects, and all defective bamboo slices were detected. The research results showed that: this paper detected 350 bamboo defects of seven types, and the average accuracy rate was 95. 14%, which can effectively detect bamboo defects. [ABSTRACT FROM AUTHOR]

Object detection algorithm has an important application in the field of intelligent transportation, but it has two problems: low accuracy of small targets and loss of multi-scale information. To solve these problems, an improved YOLOv3_4d algorithm is proposed in this paper, which can achieve recognition of multi-scale targets in complex traffic scenes. Firstly, in order to solve the problem of information loss of different targets in the common algorithm, a multi-scale detection network is constructed by adding 128128 scale to the detector to mine information with different scales. Secondly, the attention residual unit and feature enhancement module are designed. The attention residual unit extracts semantic feature information of small targets, and the feature enhancement module splices multi-scale feature maps to obtain rich fusion information. Finally, the GIoU and Focal functions were introduced to accelerate the convergence speed and enhance the robustness of the algorithm. The experimental results of the BDD 100K and VOC 2012 show that the mAP is improved by 9. 2% and 4. 0%, and the mAPS of small targets is improved by 2. 3% and 3. 2%, which fully confirms the feasibility of the proposed algorithm. [ABSTRACT FROM AUTHOR]

In order to increase the accuracy of off-line map matching algorithm in transport networks,and avoid matching a wrong link,the idea of iteration at the route node is proposed based on the principle of multi-route. The algorithm includes six steps: data pre-process,subnetwork generation,initial chosen route generation,associative matrix between nodes and links generation,generation of consideration set of routes and final chosen route. The detailed process of the proposed algorithm is referred in this paper,which is simple and easy to implement. The algorithm is used to process GPS data in Minneapolis-St. Paul Metropolitan area. From the application of the algorithm,it can be found that the accuracy is on a high level in total and decrease as the distance getting longer. And compared with the matching algorithm based on the shortest path,it is found that the accuracy of the algorithm in this paper is much better,especially when the distance is long. [ABSTRACT FROM AUTHOR]

To fully explore unmanned aerial vehicle (UAV) imagery and extract forest information efficiently, this study used an orthophoto derived from UA V data to propose a tree counts extraction method combining the Mean Shift algorithm and watershed segmentation algorithm. The method utilized the Mean Shift algorithm to efficiently cluster and smooth the G channel images extracted from RGB images. These images were then fed into the watershed method which combined morphology operation and Euclidean distance for individual treetop detection and tree count extraction. The result showed that the tree counts extraction accuracy of the algorithm in this paper was approximately 92. 74% when the result was compared to ten manually marked and counted plots. The result demonstrated that this method was efficient to detect individual tree and extract tree counts, and had better detection accuracy. [ABSTRACT FROM AUTHOR]

In order to monitor the real time situation of forest fire, this paper studies the image segmentation algorithm applied to the recognition of forest fire image. The forest fire image with complicated background and interference targets, using the single threshold Otsu method, the segmentation accuracy of fire image is poor. Therefore, the whale optimization algorithm to optimal threshold of multi-threshold image segmentation is studied in this paper. Whale Optimization Algorithm (WOA) is searching for the optimal solution of the high efficiency of the algorithm, which has the features of fast convergence and high accuracy. Aiming at the problem of accurately identifying and segmenting the flame area in forest fire image, WOA is used to optimize fitness degree function of Otsu, and simulation experiments are carried out in the fire image (Red Green Blue, RGB) model. The experimental results show that the proposed algorithm outperforms the traditional single threshold Otsu algorithm, can obtain more accurate segmentation threshold and higher segmentation efficiency, even has strong engineering practicability. [ABSTRACT FROM AUTHOR]

Aiming at the problem of simultaneous tracking of other vehicles in the process of intelligent vehicles driving on the road, a multi-object tracking method based on binary particle swarm optimization is proposed. In this paper, the state equation and measurement equation of tracking target are established based on radar data, the measurement data of multiple scanning periods of radar are correlated to form a variety of assumptions with possible orbit, and the orbit is scored by likelihood ratio to form a mixed integer linear programming problem, which is solved by discrete binary particle swarm optimization algorithm. The problem of correlation between the measured value and the target at each time is solved, and the track discrimination of the tracking target is realized. Finally, through the simulation in MATLAB, the data results verify the effectiveness of the algorithm. [ABSTRACT FROM AUTHOR]

In order to explore the stitching accuracy and efficiency of regional scale unmanned aerial vehicle (UAV)forest images, SIFT algorithm, SURF algorithm, ORB algorithm and KAZE algorithm were used to study of the UAV images stitching for different canopy density forest stand. Firstly, four algorithms were used to extract feature points from UAV images, match feature points and fuse the images with gray weighted fusion, so that multiple UAV images could be assembled into one image. Then, image feature extraction, feature matching, matching accuracy and image stitching consumption under different canopy density were compared and analyzed. The results showed that ： for the two images stitching, the number of feature points extracted from low canopy density stand images was lower than that in high canopy density stand images. The feature matching rate of ORB algorithm was significantly lower than other algorithms, but the time consumed was the shortest, so if the accuracy is not high, the ORB algorithm can be used to quickly stitching images ； for medium find high canopy density stand images stitching, the correct matching ratio of SIFT algorithm and SURF algorithm was about 45% , and KAZE algorithm was about 55%. Therefore, if the image stitching requires higher precision, the KAZE algorithm is suitable； from the average time of stitching 64 images and the correct matching rate of the average feature points, since the stitching images was mostly medium and high canopy density stand, the correct matching rate of the average feature points of SIFT algorithm, SURF algorithm and KAZE algorithm was better, but the time spent by SURF algorithm image mosaic was 577. 8 s, which was significantly lower than 876. 3 s of SIFT algorithm and 806.8s of KAZE algorithm； from the aspect of visual effect, SIFT algorithm, SURF algorithm and KAZE algorithm have better visual effects, while the ORB algorithm has obvious dislocation and deformation phenomena. The results of this paper can pro­vide some reference for the selection of UAV image stitching algorithms with different stands and different canopy density. [ABSTRACT FROM AUTHOR]

Different blade species have different blade characteristics information, whether it is based on leaf information to identify tree species or accurately detect leaf diseases, image segmentation is particularly important. This paper proposes a blade image segmentation method based on the new addition algorithm, which is used to add the edge contour of the binary image extracted by the Canny detector to the original gray image, and the inverted binary image is quadratic with the first added image. Before and after adding, the image format and size of the blade image should be kept consistent, and the image addition condition can be satisfied to realize the segmentation of the blade image. The experimental results show that the method retains the complete feature information of the blade and the blade is completely separated from the background. It provides solid theoretical support for subsequent image feature extraction and tree species identification, the addition method can also be applied to other fields of image segmentation. [ABSTRACT FROM AUTHOR]

In order to improve the safety of the engine, this paper uses BP neural network for engine fault diagnosis. BP neural network hidden layer nodes are chosen according to the empirical formula, and the effects of training methods are considered. By comparing the error of each scheme, the best BP neural network is determined. Because there are many problems to be optimized in BP neural network, particle swarm optimization is used to optimize BP neural network. First, the corresponding parameters in PSO as weights and thresholds of the network are extracted. Then the parameters of network evolution are set. Finally, the numerical value extracted from the particle swarm is assigned to the neural network for network training. The fault diagnosis test of two algorithms is carried out by the engine fault diagnosis platform. Comparison results showed that the learning speed and correct rate of optimized BP network using particle swarm optimization algorithm are better than no optimization BP neural network. [ABSTRACT FROM AUTHOR]