Research On Weld Seam Recognition And Tracking Based On Laser Vision

With the rapid development of intelligent manufacturing and automation,the demand for small batch,diversified and complex workpieces is increasing day by day.Traditional welding relies on manual online teaching or offline programming is too inefficient and inaccurate to meet the actual production requirements for product quality,so vision guidance technology is widely and rapidly used in welding robots,which is significant for the robot to achieve intelligent and autonomous welding.This paper focuses on weld seam identification and weld seam tracking in the field of robotic autonomous welding technology.This paper proposes a typical weld seam feature point extraction algorithm and a weld seam tracking algorithm based on a high-precision 2D line laser sensor as a visual device to obtain 2D point cloud data of the welded workpiece surface.The main contents include.1.Based on the KUKA KR5 ARC robot,we construct a standard D-H coordinate system and a D-H parameter table,study the forward and inverse kinematics of the robot and solve the inverse kinematics analytical solution,and analyze various expressions and conversion methods of the robot's posture during the motion.2.By analyzing the transmission format of the laser sensor data stream and using the Visual Studio 2015 platform to complete the parsing of the data,we obtained the2-dimensional point cloud data of the laser sensor and used the 5-point calibration method to calibrate the robot hand-eye system.3.A method based on PCA principal component analysis and least squares method for weld seam feature identification is proposed.For V-type welds,the point cloud data is managed in the form of a K-D tree structure and a nearest neighbor search is established,the approximate normal vector of the point cloud data is calculated by PCA,the angular difference between the normal vector of corner points and the normal vector of other points is used to cluster the point cloud,and finally the least squares fitted line is used to obtain the feature point location information,and the recognition rate is verified to be around 96.6%.For I-type welds,the slope is used to find the upper and lower surface boundaries and use its midpoint as the weld feature point,and the recognition rate is verified to be around 96.6%.4.The weld data scanning and weld tracking methods are proposed based on the KUKA robot control system.The significance of each coordinate system and the transformation relationship during the robot motion are analyzed.The laser scanning data strategy and weld seam tracking strategy are based on the criterion that the laser is perpendicular to the weld seam and the laser center point of the weld seam coincides with the laser center point.