Antireflective Weld Seam Trajectory Recognition and Localization Method Based on Polarization–Laser Vision

Highly reflective metallic materials such as aluminum alloys significantly degrade the performance of laser vision-based weld seam identification due to strong specular reflections. To achieve fast and robust weld seam trajectory recognition and localization under such conditions, this paper proposes an active laser vision system with a polarization-parameter-based reflective noise suppression method. Exploiting the polarization characteristics of reflection, a masking strategy is proposed to suppress reflective noise while preserving valid laser stripe information. To address residual discrete noise after reflection suppression, a weld seam feature extraction method is designed by integrating probabilistic Hough transform with geometry-based classification. A PCA-based sliding-window filtering strategy is introduced to further enhance robustness. Based on the extracted seam features, a closed-loop weld seam tracking and trajectory reproduction control system is finally established. Experiment results on aluminum alloy fillet and groove welds demonstrate below 0.1 mm three-dimensional localization accuracy, with per-frame processing times of 44 to 46 ms, indicating real-time processing capability.