Three-Dimensional Robot Guiding Andtracking Of Weld Seam Based On Laser Stripe Visual Sensing

Localization and tracking of weld workpiece and weld seam is a key issue in intelligent robotic welding.Based on laser stripe sensor,this paper develops a multi-functional vision system which can be used to recognize weld seam,localize initial weld position track weld seam in robotic welding.Matrix transformation approach is established as measuring principle,based on projective homography.As reconstruction procedure,twodimensional image pixels can be transformed to three-dimensional space by laser conversion matrix and hand-eye matrix.Automated calibrating software is developed which only requires a common planar calibration target.Steger method based on Hessian matrix is used to extract sub-pixel level linear feature and its orientation information.Then profile model is established to describe weld joint profile by lines with different orientation,which utilize result of Steger method to further reject welding noise.RANSAC method is used to fitting lines as profile model and center of weld seam is extracted.Based on point cloud data,KD-Tree model is used to substract background points,then RANSAC based plane fitting can segment planes of workpiece.Weld seam and initial weld position can be localized from point cloud segmented workpiece.For typical weld joint like T-joint and V-joint,experiment is designed and implemented to verify the functionality of vision system and error of tracking.Taught trajectory as ideal weld seam position is used to be compared with tracking trajectory to get tracking error curve along welding direction.Weld quality and tracking error show that system can accurately finish guiding and tracking procedure on weld seam as a robotic welding system.