Research On Workpiece Disordered Sorting Technology Based On Point Cloud Model

In modern intelligent manufacturing,the sorting of workpiece by robot is an essential link.Stereovision has the advantages of comprehensive information,strong flexibility and high precision,which can greatly improve the flexibility of robot sorting.Stereoscopic vision model has two typical methods: image reconstruction and scanning point cloud construction.In this paper,the disordered sorting technology of workpiece based on point cloud model is studied to realize the recognition,pose estimation and sorting operation of scattered workpiece.The main work of this paper includes:1 Calibration of point cloud data of scanning camera was carried out.The camera is calibrated by MATLAB to obtain the internal reference,external reference and light plane parameters of the camera.Camera calibration the camera and robot vision system are calibrated according to the "eye hand" method to obtain the conversion relationship between the two coordinate systems.2 Obtain point cloud data and optimize it.For the original point cloud data,RANSAC algorithm was used to fit the object platform plane in the point cloud,and the point cloud data of the object platform was filtered by straight-through filtering.Statistical filter was used to filter outliers from point cloud and voxel grid was used to simplify point cloud.The KD-tree algorithm is used to construct the topological relation between points.Through the pre-processing operation of point cloud data mentioned above,noise points,outliers and useless points in point cloud are filtered out,the size of point cloud is reduced,the quality of high point cloud is improved,and the posture of target object is prepared for subsequent identification.3 Design pose estimation algorithm based on disordered workpiece point cloud model.Identify and obtain the pose information of the target object.3D Harris algorithm was used to extract key points,and SHOT local feature descriptors were established at key points.The 3D Hough voting algorithm is used to identify the target workpiece,and the recognition results are used as the input of ICP algorithm for precise matching,and finally the pose information of the target object is obtained.4 Design and build a disorderly sorting platform for robot workpiece.The camera collects the point cloud data of the target object,identifies the target object to obtain its pose,and the robot picks up the target object according to the pose.The robot absorbs the target object according to its position and pose.Therefore,it is necessary to select the hardware such as camera and robot according to the environment and target object,and design the end-effector of robot to achieve the absorption of target object.Finally,the disorderly sorting test of workpiece with different posture is carried out,and the test results show that the sorting success rate of the system is more than 90%.