Unifying measured point sequences of deforming objects

Abstract

Recent progress in digitizing technologies is making it possible to capture the 3D shapes of moving objects. To efficiently utilize time series records of spatial data, the information must be unified to yield coherent deforming models. We present a general method that unifies unregistered 3D point sequences to generate deforming mesh models. The method does not assume any specific kinematic structure, and is applicable to any digitizer. The method first polygonizes the initial points and then deforms meshes to best fit the subsequent data points while minimizing the deformation energy. Experiments are conducted on real measured data and CG data, and successful results are obtained. As an application of the method, we examine data compression and achieve a 380 fold reduction rate for a measured data sequence.