Dynamic Tracking Algorithm For Binocular Vision Of Sorting Robot

The classified transportation of parts in automobile parts factory is an important part of automobile production industry.The traditional parts sorting and transportation operation depends on the assembly line robot,and the automation degree of the sorting operation is not high and depends on the environment.In order to improve the precision and efficiency of auto parts sorting robot,based on binocular stereo vision,combined with image preprocessing,matching and tracking knowledge,a binocular vision tracking algorithm for sorting robot is proposed in this paper.The target recognition and real-time tracking of different parts on the conveyor belt of the production line are realized by the sorting robot.Based on the research of binocular stereo vision model and binocular calibration method,the binocular camera used in the experiment is calibrated in the Matlab calibration box,and the inner and outer parameters of the binocular camera used in the experiment are obtained.Aiming at the problem of obtaining depth information of automobile parts,combining Sift feature point extraction and Harris corner point response function,the feature points of automobile parts image are extracted,and the feature points with high contrast are obtained.The feature point matching and RANSAC mismatching based on Euclidean distance are used to complete the stereo matching of left and right eye images of automobile parts.Combining the calibration results of binocular camera and the binocular stereo vision model of sorting robot,the depth information of matching points on automobile parts is calculated.In order to improve the intelligent level of the sorting robot and realize that the sorting robot can directly identify the target parts in a large number of automobile parts,the template matching of the image of the parts is carried out.In order to improve the speed of template matching,the background image of automobile parts is firstly removed and the target data is simplified.Then,the segmentation method based on edge and the method of maximum inter-class variance are used to segment the image of automobile parts which are removed from the background,respectively,in order to improve the speed of template matching.Get binary image and prepare for subsequent matching.In view of the fact that the boundary of the binary image after image segmentation is not obvious and the holes disappear,morphological corrosion processing is carried out to shrink the image,which makes the lines thinner,the particles smaller,the gaps and holes bigger,and makes the outline of the automobile parts clear.Then the projection method is used to calibrate the image of parts and obtain its two-dimensional position region.Finally,the matching experiments of edge matching and feature point template matching are carried out to realize the recognition and matching of target parts.In this paper,a multi-information fusion target tracking algorithm based on LK optical flow method is proposed to solve the problem of tracking automotive parts on conveyor belt by sorting robot.By means of optical flow tracking and on-line learning detector,the position of the target can be updated after the tracking failure of the tracker when the feature points with high contrast on the target component graph are tracked by the optical flow method and the on-line learning detector is used.The experimental results show that the proposed algorithm can achieve the stability,accuracy,tracking and real-time performance of the moving parts.Based on the above research,a complete set of algorithms is formed to realize the real-time recognition,tracking and positioning of the moving vehicle parts on the conveyor belt by the sorting robot under the complicated background.