Detection of 3-D feature points by using multiresolution surface model

Geometric hashing method is the efficient technique to realize the matching based on feature points. Since the performance of the method, however, depends on the accuracy of the detection of feature points, a method for detecting feature points of a 3-D object is proposed in this paper. The feature points that are the local maxima of Gaussian curvatures can be obtained by using a multiresolution surface model that can reconstruct NURBS surface model at an arbitrary resolution level. The experiments indicate that the level of the surface model has to be determined appropriately to obtain the desirable number of feature points. Therefore the proposed method is demonstrated to be able to determine a desirable level by using an absolute resolution to perform the efficient matching in the geometric hashing method.