Abstract:  Weld seam identification and segmentation is an important part of automated welding. The burr produced during spot welding process will bring serious noise in the image, thus affecting the weld segmentation results. Taking the shape of weld seam into account, the paper proposed a weld seam level set segmentation method based on straight line prior information. Since the weld seam is always a straight bar, the proposed method takes the curvature of weld seam as a prior information regularization term, and added it to the active contour models. This is very useful to filter out pseudo-seam, thus it improves the robustness of the algorithm. Compared with the traditional CV model and the improved LBF model, the proposed method is of better segmentation results and higher accuracy.


Key words:  weld seam segmentation, level set, prior knowledge