Feature-Based Precise Identification of Point Clouds in Weld Regions

Abstract

To enhance the accuracy of weld region identification under structured light vision guidance while maintaining method robustness, this paper proposes an adaptive weld identification method fusing local morphological and spatial position features. First, a fused feature space for point cloud data is constructed based on curvature estimation and spatial distance calculation, and the Gaussian mixture model (GMM) is adopted for unsupervised iterative learning of point cloud data to realize the extraction of feature points in the weld region. Second, combining with the hierarchical density-based spatial clustering of applications with noise (HDBSCAN) and voxel connectivity block analysis, the precise segmentation and boundary refinement optimization of the weld region are completed. Finally, experimental results demonstrate that this method achieves a precision of 99.38%, a recall of 88.41%, and an F1 score of 93.57% in identifying V-groove regions in circumferential welds, outperforming comparison methods such as random sample consensus combined with density-based spatial clustering of applications with noise (RANSAC-DBSCAN). Furthermore, the boundary localization error is less than 0.3 mm; parameter tests verify that the proposed method has low parameter sensitivity and strong stability, which meets practical production requirements.

Keywords

structured light / point cloud data / weld region identification / Gaussian mixture model / hierarchical density clustering

References

[1] Liu Z Y, Hu J Q, Wang S, et al. Advance in quality control and in—service performance of laser welding for magnesium alloys in automotive applications[J]. Rare Metal Materials and Engineering, 2025, 54(5): 1377-1396. (in Chinese)

[2] Zhu D H, Wang S Z, Xu Z Y, et al. Research progress in robotic repair machining of complex component defects[J]. Journal of Mechanical Engineering, 2025, 61(11): 1-22. (in Chinese)

[3] He Y Y, Li Y, Xu Y. Segmentation of line laser strip on weld image under strong interferences[J]. Journal of Donghua University (English Edition), 2026, 43(2): 120-127.

[4] Zhu M, Lei R J, Weng J, et al. MIG weld seam tracking system based on image automatic enhancement and attention mechanism deep learning[J]. Transactions of the China Welding Institution, 2024, 45(11): 90-94. (in Chinese)

[5] Jin H R, Sun Z R, Han X Y, et al. Assembly error analysis and control of thin—walled cylindrical parts in flight section of aircraft[J]. Aeronautical Manufacturing Technology, 2025, 68(18): 104—113, 132. (in Chinese)

[6] Wang Z. Review of real—time three—dimensional shape measurement techniques[J]. Measurement, 2020, 156: 107624.

[7] Malhan R K, Gupta S K. Planning algorithms for acquiring high fidelity pointclouds using a robot for accurate and fast 3D reconstruction[J]. Robotics and Computer Integrated Manufacturing, 2022, 78: 102372.

[8] Qian X M, Yang L J, Lou P H. Recognition of weld seam for pipe flange based on improved boundary following algorithm[J]. Transactions of the China Welding Institution, 2016, 37(12): 115-119, 134. (in Chinese)

[9] Ma W F, Wang J L, Zhang J P. Feature extraction from point cloud based on improved normal vector[J]. Science of Surveying and Mapping, 2021, 46(11): 84-90, 146. (in Chinese)

[10] Yang L, Liu Y H, Peng J Z, et al. A novel system for off—line 3D seam extraction and path planning based on point cloud segmentation for arc welding robot[J]. Robotics and Computer—Integrated Manufacturing, 2020, 64: 101929.

[11] Zhang Y K, Jiang Y, Tian X C, et al. A point cloud—based welding trajectory planning method for plane welds[J]. The International Journal of Advanced Manufacturing Technology, 2023, 125(3): 1645-1659.

[12] Gao J, Li F, Zhang C, et al. A method of D—type weld seam extraction based on point clouds[J]. IEEE Access, 2021, 9: 65401-65410.

[13] Zhang Y K, Geng Y S, Tian X C, et al. Feature extraction and robot path planning method in 3D vision—guided welding for multi—blade wheel structures[J]. Optics and Lasers in Engineering, 2024, 176: 108066.

[14] Zheng K, Xia X F, Qian J L, et al. Feature extraction of steel surface cracks based on point cloud multi—feature fusion[J]. Laser Journal, 2025, 46(8): 89-96. (in Chinese)

[15] Ke Y, Chen Y J, Zhang H, et al. Grinding track extraction of athletic shoes based on non—rigid point cloud registration[J]. Journal of Donghua University (Natural Science), 2023, 49(2): 92-98. (in Chinese)

[16] Yang L, Li E, Long T, et al. A novel 3—D path extraction method for arc welding robot based on stereo structured light sensor[J]. IEEE Sensors Journal, 2019, 19(2): 763-773.

[17] Yang B, Wang Z T, Xu Y T, et al. Surface segmentation and weld extraction on noisy point clouds consisting of multiple quadrics[J]. Robotics and Computer—Integrated Manufacturing, 2026, 97: 103100.

[18] Hegedus—Kuti J, Szolosi J, Varga M R T. 3D scanner—based identification of welding defects—clustering the results of point cloud alignment[J]. Sensors, 2023, 23(5): 2503.

[19] Zhou N, Jiang P, Jiang S L, et al. An identification and localization method for 3D workpiece welds based on the DBSCAN point cloud clustering algorithm[J]. Journal of Manufacturing and Materials Processing, 2024, 8(6): 287.

[20] Yang H T, Li X, Cheng Y L, et al. Adaptive segmentation method for weld seam point cloud based on normal features and LMedS[J]. Modular Machine Tool & Automatic Manufacturing Technique, 2025(3): 209—213, 218. (in Chinese)