Line-structured light visual localization for automated welding of reinforcement cages cross joint in prefabricated beams

To improve the accuracy and automation of reinforcement cage welding in prefabricated beams, this study proposes a line-structured light-based visual localization method for robotic welding of cross joints. A structured-light vision sensor is mounted on the robot end-effector to perform 3D scanning of the reinforcement cross joint, enabling precise point cloud acquisition and feature point extraction. The proposed method involves point cloud segmentation, cylinder fitting of longitudinal and transverse reinforcements, and geometric localization of cross joint feature points using the shortest-axis midpoint model. Experimental validation is conducted on an automated welding production line. The results show that the mean localization error is 0.404 mm, with an RMS value of 0.612 mm, and the relative mean similarity error among the reconstructed point clouds was below 0.13% of the model scale. The actual welded specimen exhibited accurate weld localization and good seam formation, confirming the robustness and reliability of the proposed method. The results demonstrate that the line-structured light visual localization approach provides a reliable foundation for precise weld feature recognition in automated reinforcement cage welding.