An Adaptive Seam-Tracking System with Posture Estimation for Welding

In this paper, a robotic seam tracking system with welding posture estimation is proposed that can adapt to different conditions encountered in the field of welding, such as uncertain working positions and complex workpiece shapes. In the proposed system, the target coordinate system of the welding torch is established in real-time at each welding position, and the rotation angles are obtained to change the welding posture of the robot. Gaussian kernel correlation filter is used to track the weld feature in real-time that improves the accuracy and robustness of welding seam tracking. Compared with morphological methods, this method can quickly and accurately find the position of the weld from the noisy image. Finally, experimental results show that the method can be used to calculate the welding posture, which meet the welding requirements of a complex environment.