Point cloud modeling using algebraic template

Point cloud reconstruction is a fundamental and important research topic with many applications in the fields of geomatics and computer graphics. In this paper, a novel approach for reconstructing point clouds by a hierarchical template model is presented. The template model is composed of three types of primitive geometric shape in a hierarchical manner. Compared to previous approaches which are based on an iterative fitting process, the primitive shapes are represented in algebraic form and fit to a point cloud by solving a least-square linear system. This non-iterative process makes the proposed approach feasible and robust for modeling huge amounts of point data. Furthermore, some geometric constraints are integrated into the least-square fitting system to retain the geometric relations between the primitive shapes in the template model, which can improve modeling quality. The experiment results for various point clouds show that the proposed approach is capable of handling point clouds with both noise and sharp features.