3D Reconstruction Of Weld Seam And Automatic Planning Of Robot Welding Trajectory Based On Stereo Vision

In the context of increasing the level of automation in the global manufacturing industry,robots can automatically identify the position of the weld through sensors to complete the welding,which is of great significance to improve the level of automation and intelligence of welding.This paper proposes a method for 3D reconstruction of welds based on stereo vision and planning of robot welding paths.This method is suitable for V-shaped butt groove welding workpieces of various sizes and angles,and has strong anti-interference ability against noise in the three-dimensional reconstruction process and good robustness.This paper first carried out the hardware equipment construction and point cloud collection work.After analyzing and comparing different stereo vision,this article uses the Kinect depth camera to reconstruct the workpiece,and output a point cloud model.With the stereo camera as the core,a set of hardware system was built.The internal parameters of Kinect's color camera and depth camera were calibrated,and the positional relationship between the two was calibrated,and the corresponding relationship between color image pixels and point cloud was established.Through the measurement of different distances,the accuracy of Kinect is obtained.When the measurement distance is 800 mm,the accuracy is the highest.Then obtain the weld trajectory parameters by means of point cloud processing.A series of point cloud processing is performed on the obtained point cloud,including straight-through filtering,down-sampling,statistical filtering,RANSAC plane fitting to filter the background plane and workpiece surface plane,RANSAC straight line fitting to extract the weld trajectory straight line,and by projecting the point cloud to a line,rotating point cloud to extract the most value to find the two end points of the trajectory line,and combine the normal vector obtained by plane fitting to output the weld trajectory parameters in the camera coordinate system.Then calibrated the hand-eye relationship between the camera and the robot.The pixel coordinates of the calibration points are extracted through image processing operations,including template matching,gray processing,threshold processing,contour finding,ellipse fitting,and the pixel coordinates are converted to three-dimensional coordinates under the camera coordinate.Combining the inputted coordinates of the calibration point in the robot coordinate system,the three-dimensional rigid transformation matrix between the camera and the robot coordinate system is calculated.Use this matrix to convert the weld trajectory parameters in the camera coordinate system to the robot coordinate system.In order to facilitate the verification experiment and subsequent use,a software with3 D reconstruction and trajectory extraction functions was developed.A software based on QT,PCL,Open CV and other libraries was developed.The software realizes the above functions of acquiring color images and point clouds,point cloud processing,hand-eye calibration,and realized data visualization,processing parameter modification,and file reading.And output functions.In the end,a verification experiment was carried out to analyze the specific composition of the error.The error in the length direction of the weld is about 3.230 mm,the error in the width direction of the weld is about 1.765 mm,and the error in the depth direction is 1.665 mm.The errors of the three major processes of point cloud acquisition,point cloud processing,and hand-eye calibration are analyzed in detail.Among them,the error of the point cloud acquisition process accounts for the largest proportion,followed by the point cloud processing.