Abstract
Recently, the advancement of the laser range scanner has made it possible to provide dense and accurate range data of 3D objects. From these point clouds, one can reconstruct a CAD model of the existing old mechanical parts, which is called reverse engineering. However, if we fit the point cloud without any constraints on geometric characteristics, these features will be easily lost on the reconstructed surface. In this paper we introduce novel methods to impose geometric constraints on the B-spline approximants during the iterative geometric fitting process. The geometric constraints include reflectional symmetry of the surfaces of industrial products and feature curves on the aesthetic surfaces provided by designers.
