中文摘要：

提出了一种网格分解二维投影边界点的三维模型检索方法．首先，对三维模型进行位姿对准、投影变换和边界点提取；然后，对提取的投影边界点用网格分解的组合方法参数化，并对该参数化函数进行傅里叶分解，得到图形特征；最后，对特征描述符做相似性匹配，计算特征比较值，得到检索结果．对比实验表明，该方法可以在较短的响应时间获得较优的检索性能．

英文摘要：

Recently, as a result of the development of techniques for modeling, digitizing and visualizing 3D object, the number of available 3D models on Internet and in special databases has been increased, which also affects the development of 3D model retrieval systems. The 3D model retrieval engine is a system which obtains and analyzes the user＇s query object and retrieves several similar 3D objects. Efficient feature preprocessing, feature extracting and dissimilarity measures are needed in 3D model retrieval systems. In this paper, a method for computing shape signature for 3D polygonal models is proposed and analyzed. The key idea is to represent the signature of a model by Fourier descriptor sampled from a shape statistic function. Firstly, position normalization and extracting boundary vertexes of the 2D projection of a 3D model are carried out. Secondly, the boundary vertexes are parameterized as a parameterization function with a method of decomposing using grid, and then the shape signature is obtained by decomposing the parameterization function with Fourier. Thirdly, by computing the Euclidean distance between models, signature descriptors of the query object are matched with all the other models in the database in order to search for similar goals. Finally, the experimental results show that proposed method should achieve better retrieval performance in shorter time.