3D model retrieval based on visual shape and relevance feedback

In this paper, we propose a novel wavelet moment-based surface light source (WM-SLS) descriptor and enhanced Gaussian potential function-based silhouette (GPF-S) descriptor, and on the basis of combination with Adaboost relevance feedback, to build a 3D model retrieval of two-stage strategy. PCA method for model normalization is introduced to obtain optimal bounding box at first. Secondly, the first stage using GPF-S descriptor is to reduce the retrieval scope and the second stage using combined descriptors retrieves precisely. Finally, we apply Adaboost algorithm to improve the retrieval performance by dynamically assigning different weights. Experimental results show that the proposed method is superior to other methods.