Auto Iterative Repairing Of Surface Point Cloud Sequences With Missing

There often appears a phenomenon in the 3D sampling process of actual object that the sampling data are incomplete, causing that the actual object represented by discrete surface generated from the sampling data is partially missed. Accurately detecting and legitimately repairing the missing parts is a pivot research subject in both computational geometry and digital graph processing. Currently, there are two fundamental ways to repair the discrete surface partially:one is based on the template, needing artificial implementation and the priori-information additionally; another is by filling the high dimensional hole which is intensive and complex computationally. Both methods have used few of the close correlation and motion consistency among the frames of the sequences, which provide more information for repairing the partial surfaces. Thus, new approach is needed to extract and utilize this information.For discrete point cloud sequences generated by time-space sampling, in this paper, we propose a complete novel algorithm for repairing a surface point cloud sequence. We first pre-registration the adjacent frames in the sequence, then detecting the relative lack of data. Second, determine the neighborhood of missing portion, and then do local registration for local repair. When forward iterative repairing is completed, backwards iterative repairing should be done immediately. By doing this, we completed the restoration of the entire point cloud sequence. This method only requires input the sequence of point cloud data, without knowing any additional information of the sequence, such as the topology of the models, templates, standard frames, etc. The algorithm requires no manual intervention and can be automatically performed by system itself.In the first chapter, the background and significance of the work is simply elaborated; The second chapter is devoted to introduce the principles of registration of discrete surfaces and K-means clustering algorithm and its improvement. In the third chapter, we introduce a novel approach to repair the surface with the point cloud sequence and describe the effectiveness with numerical examples.