Research On Workpiece Weld Identification And Location Based On 3D Point Cloud

With the progress of science and technology,more and more industrial robots have been put into modern industrial production,and the appearance of welding robot has solved the problem of fire on the spot and the damage to the welder’s eyesight.In the automatic welding production line,the identification and location of welding joints by welding robot is the first problem to be solved.At present,most of the industrial welding robots are teaching robot or off-line programming robot,the operation is complex and the welding position is fixed,if the operation position changes,it is necessary to redefine the robot’s working track,and the recognition and location of welding joint based on machine vision provides a new solution for this problem.In order to improve the limitation of 2D image,depth camera can be applied to industrial robot to provide more accurate position and pose information for welding robot.In this paper,Kinect V2 sensor is used as the vision sensor of industrial robot,which can obtain the color picture and depth picture of welding scene,the corrected color image and the depth image are mapped into the 3D to obtain the welding scene point cloud data.Taking Visual Studio 2015 as the main integrated development environment,PCL1.8.1 of point cloud library was used to filter scene point clouds by direct pass filter,statistical filter,voxel filter,radius filter and point cloud segmentation,gets the scene point cloud data containing only artifacts.By using Solid Works software to build 3D digital model of welding workpiece,point cloud data of CAD model can be obtained,and feature extraction and description of model point cloud and scene point cloud of workpiece can be done.In this paper,the boundary part of point cloud is extracted by point cloud boundary extraction algorithm,and then the boundary part of point cloud is filtered by point cloud processing software Cloud Compare,the position and pose of the workpiece and its CAD model are estimated by the combination algorithm of SAC-IA coarse registration and P-ICP fine registration,the final average registration error was 4.152 mm and the average registration time was 133.58 seconds.The experimental results show that the proposed method can meet the needs of identification and location of workpiece welds in industrial scenarios.