3D model classification based on multiple features integration

In this paper, we propose and evaluate a novel approach for 3D model classification by integrating multiple efficient shape descriptors. In this approach, first, multiple shape descriptors are passed to different fuzzy SVM classifiers separately, and the fuzzy membership degrees are obtained from each classifier; then, these membership degrees are input into a BP Neural Network, the integrated membership degree and the final classification decision are produced. Experiments show that the proposed classification approach has the better performance than the traditional 3D model classification methods with single feature or single classifier, which proves the validity and potential of the presented approach for 3D model classification.