A 3d Engineering Model Retrieval Algorithm Based On Relevance Feedback And Features Combination

Ting Zhuang,Xutang Zhang,Zhenxiu Hou

DOI: https://doi.org/10.4028/www.scientific.net/KEM.579-580.340

2014-01-01

Abstract:In order to reuse 3D models and design knowledge efficiently, a number of 3D model retrieval algorithms based on content features of models have been proposed in recent years. Although, the features-based methods have achieved some progress, there are two limitations stilly. The first, single content feature can't be suit for all kinds of 3D models; different features have different strengths and weakness. The second, "semantic gap", the semantic of model is independent from low-level characteristics. For those two issues, we present a 3D engineering model retrieval algorithm based on relevance feedback and features combination in this paper. The proposed method takes advantage of multiple features by allying them with weights. In the retrieval process, our method utilizes the Particle Swarm Optimization to update the weights dynamically based on user's relevance feedback information in order to narrowing the gap between high-level semantic knowledge and low-level content features. The Experiments, based on publicly available 3D model database Engineering Shape Benchmark (ESB) developed by Purdue University, suggested that the proposed approach has better retrieval ability than traditional ones.