Abstract
<title>Abstract</title> <p>To automatically extract welding trajectories for planar welds on reflective materials, this paper proposes a three-dimensional active vision-based welding trajectory extraction methodology that is based on the fusion of features. First, to accurately segment the laser stripes at the weld seam, a saliency feature fusion-based laser stripe segmentation model (SFF_LSM) has been developed. The SFF_LSM model effectively suppresses background interference features, accurately segmenting the general features of the laser stripe distribution along the weld seam. Next, the gray-gravity algorithm with secondary smoothing is employed to extract the center of the laser stripe. Finally, the point cloud data at the weld seam is filtered, and the least squares fitting algorithm is employed to extract the welding trajectory automatically. The results of the experiment demonstrate that the proposed method is an effective means of overcoming the interference caused by reflections, with the maximum fitting errors of the weld seam welding trajectory being 0.51 mm in the X direction, 0.40 mm in the Y direction, and 0.29 mm in the Z direction.</p>