Point cloud scanning-based identification and reconstruction of welds

One of the key challenges for the weld grinding wall-climbing robot is to accurately obtain the specific location and the detailed profile size and height of the welds before grinding the hull surface. Given the substantial height of ships, it is difficult for operators to precisely determine the location of welds with the naked eye alone. Therefore, the weld grinding wall-climbing robot designed in this paper is equipped with a depth camera, the D435, which scans and identifies the point cloud of the weld. This paper introduces the principle of the camera and conducts camera calibration work. After scanning and identifying the weld point cloud in three dimensions, the paper uses statistical and bilateral filtering to denoise the ordered and scattered point clouds during the weld point cloud pre-processing, resulting in a smooth weld point cloud map. Finally, the contour of the weld point cloud is reconstructed using the triangulation projection method.