Design Of Target Recognition And Positioning System Based On Machine Vision

The use of machine vision technology to detect and identify targets on the production line can greatly improve the intelligent level of the production line and the quality of the product.Target recognition and positioning is an important research direction of machine vision.Based on the products of robot automatic sorting production line,this paper studies the vision-based target recognition and positioning technology in robot vision servo sorting system.main tasks as follows:Firstly,the camera calibration and image processing algorithms in visual servo system are summarized and experimentally studied.The camera calibration was realized by Zhang Zhengyou calibration method.The image processing algorithms such as image filtering,edge detection and morphological processing were compared,which provided guarantee for subsequent target recognition and positioning.Secondly,for the identification and location of static targets,the ORB algorithm is used for target recognition.Aiming at the problem that the ORB algorithm features matching accuracy is not high,an ORB feature matching method based on multiple constraints is proposed.The method can effectively eliminate the mismatched feature points.When the coarse matching correct rate is less than 50%,the matching correct rate can reach 99% after multi-constraint purification.After the feature matching is completed,the homography matrix between the template and the target image is calculated by the random sampling consistency algorithm to achieve the coarse positioning of the target.Then,through the operations of target segmentation,edge detection,and finding the minimum connection rectangle of the target,the coarse positioning result is optimized,and the precise positioning of the target is achieved.Thirdly,for the recognition and location of the dynamic target,after the target detection and feature matching identify the target,the dynamic target is subjected to real-time positioning based on Cam-shift tracking.For the Cam-shift algorithm,only a single color feature is used,and when the background has similar color interference,the tracking effect is poor.The adaptive fusion of edge features and color features is used.A Cam-shift dynamic target real-time localization method combining Kalman prediction and multi-feature fusion is proposed.When there is similar color interference in the background,the method can still accurately track the target and greatly improve the tracking efficiency.Finally,the position-based visual servo simulation is carried out by Matlab robot toolbox and Simulink.The simulation experiments of static target location and dynamic target tracking are completed.The feasibility of target recognition and localization method in robot vision servo system is verified..Then based on OpenCV computer vision library and VS2013 software development environment,a machine vision-based target recognition and positioning system was designed.The system can accurately identify the corresponding target and calculate its attitude information on the plane.