A Ridge-Valley Line Extraction Method on Hole Filling of Groove Point Cloud

Among several teaching methods of welding robot, the one based on 3D point cloud path extraction is a much more intuitive approach. However, due to reflectivity, corrosion and insufficient illumination, etc., the point cloud on the workpiece is squeezed into a small hole that hinders the path planning. Thus, this paper puts forward a hole segmentation and filling algorithm on the basis of ridge-valley line extraction. Firstly, the discrete curvature of the point cloud surface is estimated by Weingarten mapping properties, and the ridge and valley feature points are identified. Then, an improved connection algorithm is used to connect the ridge and valley points, and the feature lines obtained by fitting divide the holes into simple sub-holes. Finally, after the rough filling of each sub-hole with triangulation and subdivision algorithm, the positions of the newly added vertices are adjusted to the RBF (Radial Basis Function) implicit surfaces to complete the hole filling with sharp and curved features preserved. The experimental results show that this method can restore the sharp and curved features well on different groove point clouds.