Extracting invariable features using geometry projection based histogram model for 3D object description

摘要

Extracting invariable features is one key issue for 3D model searching. A novel invariable feature extraction method, namely geometry projection based histogram model, is proposed for 3D model description. Different from the traditional method, one projection plane (or surface) is created for each 3D model, and the points of 3D models are projected to the projection plane (or surface), and then the distribution of distance between the original points of 3D model and the projected points on the projection plane (or surface) is analyzed with histogram model under the different resolution level. Each histogram model has its own projection plane (or surface) for the chosen projection method in advance. Experimental results show that the proposed algorithm obtains the largest similar histogram models for the samples from the same class but the largest discriminative ability for the different class of samples compared with the current popular D1, D2, D3 methods.