Robust weld seam tracking method based on detection and tracking of laser stripe

In the automatic welding system based on structured light vision, the precision of detection of the welding feature point in the weld image plays a critical role. Due to various interferences captured in a complex welding environment, it is essential to extract the feature point accurately. A practical and robust welding seam tracking algorithm by considering the welding feature point detection problem as the detection and tracking of laser stripes is proposed in this paper. In the initial image, the laser stripe candidates are detected by searching the peak of the gray distribution and the actual laser stripe is extracted by the similarity between real image and simulated image. The laser stripe is tracked based on the improved mean shift tracking method with the patch-based representation in the sequent images. The real-time performance and accuracy of the proposed algorithm are demonstrated by comparison with other methods through experiments.