Research On 3D Model Retrieval Based On Semantic Tree

With the continuous improvement of the computer in the multimedia technology and virtual reality technology, 3D model is becoming increasingly popular in various areas, such as computer-aided design, mechanical engineering and entertainment. With the improvement of computer graphics hardware and the application of software of geometric modeling. The number of 3D models created is growing at an alarming rate. This greatly enriched the 3D model resources, and a large number of 3D model libraries are occurred. 3D model retrieval is beginning to attract more and more researcher's attention.3D model retrieval can be divided into three categories, based on keywords, based on content and based on semantic. The 3D model retrieval which is based on keywords is the simplest and it can be widely used. The 3D model retrieval which is based on content is the most important point of the research. Feature extraction and similarity calculation of 3D model are both important and difficult points. However, 3D model retrieval which is based on semantic is the inevitable development trend of 3D model retrieval.The main contents of this paper include the following aspects:(1) Establish semantic tree. Since the semantic web cannot show the hierarchical relationships between the various concepts, but the tree has this advantage. So the semantic tree is inclined in this research. From the deep understanding of Word Net structural features,take advantage of this characteristic structure. Using the Princeton Shape Benchmark from the Princeton University's 3D Warehouse, and then establish semantic tree.(2) The expansion of keywords and the semantic retrieval method. Provide keyword synonyms and other words, ant this can reduce cases semantic tree without keywords and in the case of semantic tree under free keywords. According to the keyword semantics, it can compute the similarity of model semantics. It returns the model which is highly relevant semantic node, reducing the possibility that the search results are empty.(3) Using relevance feedback mechanism to improve the search results. Using the user's feedback to the second query to improve retrieval efficiency is commonly used method.However, this paper on the basis of improved. Mainly concentrated on the search results using the model of the geometry of the secondary features to retrieve, return the models more in line with user's retrieval intention.This paper combines the semantic tree with the research on 3D model retrieval, and corresponding work are carried on. This method can improve the retrieval efficiency in some way, and the search results are also optimized.