A Novel Approach for Robotic Welding Trajectory Recognition Based on Pseudo-Binocular Stereo Vision
Hao Wang,Chao Chen,Yu Liu,Boqiao Ren,Yaowen Zhang,Xiaohui Zhao,Yupeng Chi
DOI: https://doi.org/10.1016/j.optlastec.2024.110669
IF: 4.939
2024-01-01
Optics & Laser Technology
Abstract:For plane welds, positioning seam, positioning arc-starting point, positioning arc-extinguishing point, and programming trajectory are often achieved through inefficient teaching-playback mode. For solving this series of problems, a novel approach of 3D welding trajectory recognition based on pseudo-binocular stereo vision is proposed to endow robots with autonomy in the welding non -standard plane welds or batch plane welds. Firstly, a 3D coordinate recognition model of pseudo-binocular stereo vision based on a single CCD camera is innovatively constructed to accurately identify 3D coordinates of spatial points in the robot base coordinate system, and the average identification accuracy of the model is 0.2251 mm, which is sufficient to satisfy welding requirements; Secondly, positioning seam, positioning arc-starting point, positioning arc-extinguishing point, identifying type of weld, and identifying ROI are transformed into a unified YOLOv5 ' s inference task, which greatly improves the robot's ability to perceive spatial objects; On this basis, algorithms such as Sobel operator, modified template matching algorithm, point cloud filtering, B-spline curve fitting realize the recognition and 3D reconstruction of welding trajectory from irresistible, dynamic and irregular interference information. Notably, compared with the traditional image matching algorithm, the modified template matching algorithm improves the correct matching rate of pixels from 86.667 % to 100 %. To evaluate the feasibility of the novel approach, the accuracy and robustness are verified using experiments on the S-shaped weld and the folded linear weld. The results demonstrate the proposed pseudo-binocular vision welding strategy give the robot the autonomy of welding plane welds in the actual industrial environment.