A fast mesh deformation method to build spherical representation models of 3D objects

Abstract

In this paper a new method to build more efficiently spherical representation models of 3D objects is presented. Our approach is applied to a specific representation called LSR/GSR which is a hybrid spherical model where local and global object features are stored. The main contribution of this work is the rearrangement of the iterative mesh deformation process required in any conventional spherical method, in such a way that the high computational cost usually involved in this stage is significantly reduced. A new intrinsic data structure defined over the original mesh is introduced to improve the deformation process: the Modeling Wave Structure (MWS). Thanks to this new concept the deformation process can be seen as a very fast wave propagation phenomenon where displacement of each MWS node only affects to a reduced and fixed set of Neighbor nodes. The method has been applied to synthetic and real 3D closed objects of different shapes. Experimental results for both synthetic range data and data coming from a gray range finder sensor are shown in the paper.