Hits:

Date of Publication:2011-01-01

Journal:光电技术应用

Volume:26

Issue:4

Page Number:56-60,80

ISSN No.:1673-1255

Abstract:A 3D model retrieval algorithm based on multi-view scale invariant feature transform （SIFT） feature is researched. By carrying on the multi-view projection of a 3D model, the omnidirectional 2D projection depth maps are obtained, from which SIFT features are extracted. Individual-k-means and Whole-k-means are used to establish the codebook for shape benchmark respectively. Using the feature quantization codebook, all the features are clustered and generated the histogram as the 3D model feature vector. The similarity matching of 3D model is realized by calculating Kullback-Leibler （KL） divergence according to the feature vector. Experiments show that the algorithm can be adequate to both articulated shapes and rigid shapes, and gain a better retrieval effect.

Note:新增回溯数据