Feature Combination for 3D Model Retrieval Based on Evaluation Indicators

Feature combination is an important way to improve the effectiveness of 3D model retrieval. In order to conduct combination more effectively,we propose a novel approach to combine features using quantitative evaluation indicators of the retrieval performance,which adopts depth buffer images,view feature set,normal entropies,and ray-based feature. Firstly,from the training set,quantitative evaluation indicators of the four features are computed which are then used to compute the weights of feature distances. Secondly,for testing set,the weights are used to combine feature distances computed using every single feature. Finally,the similarities between 3D models can be measured. Experimental results show that our method outperforms the traditional DESIRE approach.