Registration of pointcloud and CAD model based on extended Gaussian sphere

Abstract

There are some problems in the registration of point cloud and CAD model, such as definition initial corresponding position of point cloud and CAD model, registration method's efficiency. In order to solve them, a new fast template matching method based on extend Gaussian sphere is proposed. Trans-forming the extended Gaussian space and Cartesian space solves the initial corresponding position. In order to promote the efficiency of registration, the CAD direct object point cloud is used in the registration of point cloud and CAD model. When two point clouds had been matched, the point cloud and CAD model had also been in the same coordinate system. Because using the searching method to get the closest point in point cloud replaces the optimization method between the point cloud and NURBS, the efficiency of the method has promoted. Some cases are done to prove the method's correction.