Research On Extraction And Correction Algorithm Of Robot 3D Point Cloud Welding Trajectory

Welding robot can replace workers to carry out high-risk welding work,and improve production efficiency and welding quality.It is widely used in automobile,shipbuilding,electronics and other industrial production processes.With the application of 3D sensor and point cloud processing technology in welding industry,welding technology becomes more intelligent,flexible and highly integrated.In the development process of welding technology,as the welding procedures of weldments are varied according to different application scenarios,and the welding is complex and weld types are also diverse,therefore,two welding operations are mainly adopted,one is the real-time guided welding of single line weldment,and the other is the trajectory guided welding for complex weldment.Among them,the extraction of solder joint data and welding trajectory are essential.However,there will exist spatial translation and rotation deviation between the welding trajectory extracted by the algorithm and the theoretical welding trajectory.Thus,it is necessary to carry out realtime correction to avoid serious consequences such as welding failure.In order to realize real-time guided welding of single line weldment,the 3D laser sensor is used to collect the weld data of the weldment in real time,and the data of various weld types are extracted in real time and transmitted to the robot,so as to guide the robot to carry out welding.Meanwhile,an improved contour extraction algorithm based on point cloud plane fitting is proposed for complex weldments.Firstly,the plane fitting operation is conducted for welding trajectory by using the random sampling consistency algorithm.Then,the least square method is used to fit the micro tangent plane to extract the boundary points of the fitting point cloud plane.Furthermore,the outer contour is searched based on the minimum distance to realize the extraction of the effective contour.Finally,to smooth the data,three experiments are carried out for the obtained welding trajectory B-spline fitting curve.In addition,in order to solve the problem of trajectory deviation of welding robot in the actual welding process,a welding trajectory correction method based on local contour feature registration is proposed.As the accuracy of traditional iterative closest point algorithm(ICP)relies heavily on the initial positions of two registration point clouds and is sensitive to noise data,this method extracts the contour feature points of the weldment in the point cloud as registration data.It can effectively simplify the data and eliminate the noise points,so as to improve the registration accuracy.What’s more,this method uses local contour to register the complete contour,it greatly shortens the time of contour extraction and improves the speed of registration.Finally,to test the efficiency of the algorithm proposed in this paper,an experimental platform of robot 3D point cloud welding is built,the software of welding trajectory extraction and correction is developed based on QT.Experimental results show that,compared with the commercial software QS-TRACK,the accuracy difference between the algorithm and the commercial software QS-TRACK is less than 0.1 mm,which can provide accurate solder joint data for real-time guided welding.For the welding of complex weldments,it can meet the requirement of welding after contour extracting with the improved algorithm of welding trajectory extraction and smoothing by three times of B-spline.Compared with the ICP algorithm,the proposed trajectory correction algorithm improves the registration accuracy and speed,and can correct the welding trajectory in real time.