Research On Object Tracking And Grasping System Of Robot Arm Based On Vision

In order to improve the target grasping ability of the robot in complex scenes,this paper established a target tracking and grasping system based on vision by series robot grasping the moving stacked objects on the conveyor belt,and carried out research on the key technologies such as target detection,tracking and grasping in the system.Firstly,this paper summarizes the research status of robot location grasping system and key technologies,and introduces the research content and research framework.After analyzing the tasks and difficulties of the robot grasping system,a grasping scheme based on stereo vision was designed,and the principle of stereo vision in the scheme was analyzed and verified by experiments.Then the target detection technology with deep learning is studied.In this paper,a target detection method based on deep learning is used to filter the target region of the image,and the traditional target detection process based on image segmentation is optimized by reducing the detection range of the target in the image.Based on feature points of image edge detection method for process optimization,first of all,on the basis of the outline of the object matching pixel image reconstruction and form a new target detection area,then the corner detection of target detection area and get the details of the feature point coordinates,the object edge feature extraction problem is transformed into angular point detection,By matching feature points in binocular image,coordinate reconstruction of target feature points space can be realized quickly.Then,the key technology of tracking and grasping the moving object is studied.In order to deal with the diversified attitude of the object,the coordinate systems of different attitudes of the model are created and the corresponding grasping attitude is designed through the three-dimensional geometry operation of the target feature points.Aiming at the tracking and grasping tasks of dynamic targets on the production line,a research scheme based on the visual servo system based on position was adopted,and the tracking and grasping path of the end-effector was optimized to improve the grasping efficiency of the robot.In order to improve the location accuracy of the end-effector in the tracking process,the position estimation algorithm was applied in the tracking capture method,and the Kalman filter mathematical model based on the Lucas-Kanade parameter optimization was established.Finally,the hardware platform and software operating interface of the experiment are built.The coordinate conversion between the stereo vision system and the robot grasping system was realized through the hand-eye calibration experiment,the grasping pose was solved,and the 3D pose estimation of the target object was completed.The effectiveness of the key technology algorithms in this topic is demonstrated through the experiments of object search,pose extraction and tracking grab for moving objects in different stacked states.