Abstract: This paper uses product of maximum distance of feature points and curvature feature points to define the function of the corresponding ball of 3D model.Global features are used to define the function of the corresponding ball of 3D model for weighted value of local features.Then，spherical harmonic transform is used on the sphere.The energy of the coefficients of spherical harmonics forms the feature descriptions which are used for search.Experimental results show that using this method has a good search result.

Key words: 3D model retrieval, harmonic function, curvature, feature descriptor

摘要： 利用特征点最远距离与特征点曲率的乘积对3D模型定义了对应的球函数，使用全局特征对局部特征的加权值定义了3D模型对应的球函数，然后对此球函数进行调和分析，得到一组平移、旋转、缩放不变的特征描述符，使用此特征描述符进行检索。实验结果表明使用此方法具有较好的检索效果。

关键词: 3D模型检索, 调和函数, 曲率, 特征描述符