Implementation of 3D model retrieval feature extraction

This paper mainly realizes the key-steps of 3D model retrieval, which are model preprocessing, feature extraction and similarity measure. Firstly, area weighted and random point projection to deal with translation and scale normalization of 3D model is employed. Secondly, feature extraction of 3D models is analyzed by studying the algorithm based on the content. The two algorithms based on shape histogram, Osada algorithm and the Ankerst algorithm are realized in this paper. The result of two algorithms are compared based on 100 models. We found that Osada algorithm is better than Ankerst algorithm. For Ankerst algorithm, using quadratic distance to calculate similarity is better than the way using Euclidean distance.