Reassembly of fractured objects using surface signature

Abstract

As 3D object acquisition technology has improved, research has been actively conducted on the virtual reassembly of broken 3D objects. However, because fractured surfaces are bumpy and complicated, it is difficult to extract salient features from them. In this paper, we propose a new simple descriptor, the surface signature, to delineate a fractured surface and to find the counterpart fractured surface effectively. This descriptor is based on the convex/concave information of a point on the fractured surface. To apply the descriptor to reassembly, feature curves are extracted from the boundaries of the surface signature. The similarity between two fractured surfaces is calculated based on the spin images of feature curve points and the distance and normal deviation between two feature curves to find the matching fractured surface. Several reassembling experiments using the surface signature are performed on real-world objects, and it is shown that the proposed descriptor can be effectively applied to the reassembly process.