A 3D model retrieval method based on multi-feature fusion

Abstract

3D model retrieval is a hot topic in information retrieval, and it is of great importance to fuse multi-feature of 3D models to achieve high quality retrieval. Therefore, in this paper, we propose a novel 3D model retrieval method based on the multi-feature fusion technology. Motivation for the proposed 3D model retrieval method lies in that we convert the 3D model retrieval problem to a discriminative feature space mapping problem. The framework of the multi-feature fusion based 3D model retrieval system contains two main modules: 1) model normalisation; 2) multi-feature fusion. The proposed 3D model retrieval method is designed based on multiple feature fusion and online projection learning. In order to effectively fuse multiple features, we train a model to learn a low dimensional and discriminative feature space from the multiple views of 3D models. Particularly, to effectively retrieve the newly added samples, we propose an online projection learning algorithm, which learns a projection matrix by handling the least square regression model. Experimental results show that the proposed method can achieve higher precision for a given recall than others methods, that is, the proposed method can obtain higher quality 3D model retrieval results than state-of-the-art methods.