Improving the 3D model classification based on selecting proper features

Abstract

Classifying 3D models helps to organise databases according to categories. As a result, models are quickly search to recommend models when designing virtual scenes in movies and games. Today, number of 3D models increases sharply and entertainment needs develop quickly. Therefore, classifying 3D models is essential task. Previous studies used many features to improve the accuracy of classifying. It takes a long time for both extracting features and classifying models. In this paper, we select three features and find a suitable classifier to drop the time for computing in classifying 3D models. Proposed features are eigenvalues associated with the principal axes of 3D models. We compare available classifiers to select the best one, Support Vector Machine, to classify models. Experiments are conducted on two benchmark databases including travel means in Princeton Shape Benchmark and animals in Shape Retrieval Contest 2010. Experiment results show our approach is useful for recommendation applications and roughly classifying 3D models.