Weld Seam Extraction of Intersecting Pipelines Based on Point Cloud Entropy

This paper presents a method for extracting intersecting pipelines weld seam based on point cloud. Firstly, the collected point cloud is filtered by PassThrough filter to extract the region of interest, and the intersecting pipelines model is reconstructed by using the camera calibration results and the pose of robot. Then VoxelGrid filtering and Moving Least Squares filtering are performed on the reconstructed model to smooth the error caused by camera and robot. The Fast Point Feature Histograms of point cloud is used to calculate the entropy map, and the weld seam region is extracted by threshold segmentation. Finally, an iterative algorithm for weld point extraction based on entropy is proposed. Through experimental verification, the algorithm can extract the weld points of the actual intersecting pipelines and obtain the real position of the weld points in the robot base coordinate frame, which has great application value in robot automatic welding area.