A seam tracking system based on a laser vision sensor

It is difficult to ensure robustness and accuracy when the traditional morphological method (TMM) is used to detect weld feature points, especially in an environment with a strong arc light and splash interference. In this study, a novel and robust seam tracking system based on a laser vision sensor is proposed. The feature point is obtained using the traditional morphological method before welding and can thus determine the tracked region. When the welding begins, a spatiotemporal context (STC) tracking algorithm is utilized in order to detect the weld feature points. In the welding process, the STC algorithm is adopted to determine a feature point with a strong arc light and splash interference and the morphological method is used to obtain an accurate weld feature point when the interference decays. As a result, the STC model can be updated in time, so the tracking drift problem can be solved and the robustness can be improved. A model reference adaptive control method is then adopted to improve the robustness of the robot system, which can convert the deviation between the theoretical and actual welding trajectory into a voltage to control the robot's movement. Experimental results show that the tracking error of our seam tracking system is within 0.5 mm even when the distance between the laser stripe and the welding pool is 15 mm, which can completely satisfy the industrial requirements.