A surface descriptor of surface matching base on normal angle of surface data

Three dimensional surface matching is an important problem in computer vision. The process of surface matching is to match similar regions or correspondences across multiple surfaces with unknown relative poses. Many surface descriptors for finding correspondences have been proposed. However, most of these approaches have been used in triangle mesh surfaces. In this paper, a local surface descriptor constituting moving least square (MLS) surface representation, differential angle calculation, and Zernike moments is proposed to find correspondences in unconstraint point set surfaces. The MLS projection is applied to generate a differential angle map described the geometric information. Then, Zernike moments are used to provide surface descriptors with rotation invariant. In addition, a congruent consistency method is proposed to find the reliable correspondences. To conclude, the experimental results show that the approach outperforms the original spin image method [2], and has slightly better performance than a modified spin image method presented in this paper. Moreover, the proposed approach has good matching performance both in terms of matching surface by using variety of descriptor size or matching surface among different resolutions. Keywordssurface descriptor; MLS surface; surface matching; congruent consistency analysis; Zernike moment