A 3D-shape retrieval system to improve the text retrieval

This paper describes a 3D model retrieval system based on Google 3D Warehouse, which consists of 3D models in. skp format. Because of the quickly expanding of the 3D models on the web and the insistent demands of 3D models in virtual reality, virtual scene modeling and 3D animation, there is an increasing need for a search engine to help people find them. Traditional text-based search methods are not always effective for 3D data. In this paper, we proposed a 3D model retrieval system combining the text-based search method and the shape-based search method. By means of the Google 3D Warehouse, we start the initial search with inputting some keywords (such as the model´s name). Next, according to the text-based search result, we obtain and download the retrieved models as a database. Then, we convert the SKP models and extract the shape features for the further shape-based retrieval. Finally, we select a model from the text-based retrieved results as a query request and get more reasonable results by shape retrieval. The experiment results show that the proposed system achieves much better performance than Google 3D Warehouse search engine.