Research on Point Cloud Segmentation and Weld Seam Extraction based on Adaptive RANSAC Algorithm

Point cloud segmentation and weld seam feature extraction are the core links of the welding automation system, and their accuracy directly affects the quality of the weld. Although the Random Sample Consensus (RANSAC) algorithm exhibits strong robustness, its fixed parameter settings limit its ability to handle noise interference, uneven data density, and boundary-blurring in complex environments. Therefore, an adaptive RANSAC algorithm is proposed in this paper, combined with the weighted least squares method that fuses geometric features, realizes adaptive segmentation of the point cloud, and accurate extraction of weld seams. First of all, the point cloud is segmented by dynamically adjusting the distance threshold and the number of iterations and reassigning neighboring points. Then, a weighted least squares fitting method based on pairwise point matching is used to extract the weld seam. After that, experiments conducted on L-shaped, V-shaped, and I-shaped welding plates validate the effectiveness of the proposed method. The results indicate that the average residual error in planar segmentation is reduced by 10.8% compared with the traditional RANSAC; the average distance error of weld seam extraction is 0.51 mm, and the maximum error is within 1 mm. In addition, the processing speed of the algorithm is 10 times higher than that of manual teaching.