基于机器视觉的焊缝跟踪系统设计与实现OA

With the rapid advancement of industrial automation,welding robots are increasingly widely used.However,weld seam deviations caused by workpiece assembly errors and thermal deformation seriously compromise welding quality.A machine vision-based weld seam tracking system is developed to address this is-sue.The EFORT ER6-1400 welding robot is adopted as the execution platform,and the laser structured light vision sensing scheme is applied.For image processing,a complete preprocessing procedure consisting of median filtering for denoising,image enhancement based on the improved dark channel prior,binarization via maximum-median filtering and Canny edge detection is designed.Automatic recognition of four types of weld seams,namely butt,corner,lap and groove seams,is realized by means of the ResNet-50 network.The pixel-to-robot base coordinate mapping is established via camera calibration and hand-eye calibration,and a PID trajec-tory correction algorithm with feedforward compensation is designed.Experiments show that the system achieves a mean tracking deviation below 0.3 mm and a response time of approximately 35 ms,with notably improved weld formation quality over conventional teach-playback methods.