A New Self-adaptive Ball Pivoting Algorithm Driven by an Intrinsic Property Factor of Point Cloud

To solve the shortcoming that there will be some holes and/or redundant triangles when non-uniform point cloud is reconstructed by BPA(ball pivoting algorithm), a new self-adaptive ball pivoting algorithm is proposed. The improved algorithm is driven by an intrinsic property of point cloud, which is initially proposed. And according to the reconstructed surface area, a new method of surface reconstruction quality evaluation is also proposed. Firstly, three isolated points are selected to build a seed triangle, according to the points' vectors, position, spacing, connection and so on. Then, the radius r of the pivoting ball is adaptively calculated based on the intrinsic property of point cloud and front edge length. Finally, a suitable third point is selected by pivoting the ball of radius r around the front edge, to expand the triangulation. Experiments on Dragon and Bunny point cloud show that the proposed algorithm can adaptively reconstruct the surface of both uniform and non-uniform point cloud.Moreover, it is robust, efficient and needs no manual intervention. The reconstructed surface is of high-quality according to the proposed method of surface reconstruction quality evaluation.