Research on a hole filling algorithm of a point cloud based on structure from motion.

ABSTRACT

A point cloud can be obtained from three-dimensional measurement reconstruction based on fringe projection. However, there are holes in a point cloud due to objects with complicated shapes and the defect of the method. The holes have a profound impact on the subsequent data processing. A fitting approach to fill the holes based on structure from motion (SFM) is proposed in this paper. First, fringe projection with a two-dimensional phase is used to extract the hole boundary. Second, the registration of the SFM point cloud and the fringe projection point cloud is carried out. Then supplementary points are extracted. Third, the holes are filled based on a radial basis function on the point cloud added with the supplementary points. This method has been proven to be robust by experiments, and information of complex surface holes can be restored sufficiently.