Multi - View Cloud Registration And Pseudo - Hole Filling For Surface Objects

Multi-view point clouds registration and the detection and filling of pseudo-holes are important parts in the three-dimensional reconstruction. This two sections are studied in-depth in this paper. In terms of the point-cloud registration, the traditional iterative closest point(ICP) algorithm error measurement, Euclidean distance was broken, and a error measurement was presented in this paper. The improved algorithm is based on discrete surface triangular pyramid volume measure as well as the corresponding feature. What’s different from the traditional ICP algorithm is, most traditional ICP algorithms use the corresponding point pairs in the area of the two overlapping point-clouds to obtain the registration, while what the improve ICP algorithm in this paper uses are the effective points and its corresponding triangles of the two overlapping point-clouds clouds. Organize Such discrete features into pairs, wherein, the effective points and the corresponding triangles can form many triangular pyramids, and the volumes of that very triangular pyramids can be used as an error measurement. Experiments showed that the improved ICP algorithm in this paper can not only get a better registration and can obtain faster convergence.Because of the complexity of curved surface, some pseudo-holes can appear in the point cloud; those pseudo-holes should be filled. On pseudo-hole filling, domestic and foreign scholars have made many researches, but their methods are mostly only apply to pseudo-hole in small curvature zones. The algorithm proposed in this paper can be applied to fill pseudo-holes in both small and big in small curvature zones. The method to classify the holes is introduced in this paper, and then the method to fill simple holes is presented. If a pseudo-hole is a complex hole, you can split them into several simple holes firstly, and then fill them and form the result to this complex hole. If the result cannot meet the requirements, then smooth and refine it, and simultaneously update their topology information.