Reseach And Implementation Of Localization And Grasping For Robotic Arm Based On Vision Guidance

With the strong support of the manufacturing industry and agriculture by our country in recent years,more and more emerging technologies have penetrated into the field of industry and agriculture.Robots,which have been available for more than 60 years,have always occupied a place in industrial and agricultural production,liberating more and more people from tired and boring work.And the combination of the rapid development of sensor technology and robot makes the robot more and more intelligent,which can make the robot sense the surrounding working environment and make the corresponding changes.Aiming at the limitation that the traditional manipulator can not grasp autonomously,this thesis uses the Xiaomi binocular depth camera and the six-aixs manipulator as reserch objects and applies the vision guidance technology in grabbing task,so that the sensing ability of the manipulator grabbing system is improved and the autonomous grabbing function of the manipulator is realized.Firstly,according to the task requirements of the system,this thesis designs overall scheme of the system,and analyses and selects software and hardware modules in the system.Then the related technology of binocular camera calibration is studied and the relation between image pixel coordinate system and binocular camera coordinate system is established.In order to obtain the relation between the binocular camera coordinate system and the end-manipulator coordinate system,the calibration principle of the handeye system is studied.In view of the noise in the original point cloud obtained by the depth camera,the point cloud is preprocessed.In order to better segment the target point cloud,the point cloud segmentation algorithm combining normal clustering algorithm and the Euclidean clustering algorithm is adopted.In order to recognize the target object,a local template library is established,and the global feature descriptor VFH(Viewpoint Feature Histogram)is used to recognize the target object.In order to obtain the pose information of the object,SAC-IA(Sample Consensus Initial Aligment)rough registration and ICP(Iterative Closest Point)fine registration schemes are used to complete the pose estimation of the target object based on point cloud registration.In order to improve the grasping accuracy,the object grasping pose detection algorithm GPD(Grasp Pose Detection)is studied.The basic principle of GPD algorithm is introduced briefly and the feasibility of this algorithm is verified in two aspects of obtaining point cloud information from point cloud files and depth camera.Finally,a vision-guided robotic arm positioning and grasping system is designed and implemented,and a simple host computer software is designed.The calibration experiment and hand-eye calibration experiment are performed on the Xiaomi binocular depth camera,and the grasping experiment is completed.On this basis,the grasping experiment is completed and the experimental results are simply analyzed.