中文摘要：

提出了一种基于骨架树进行机械零件三维模型检索的方法。检索分为两个阶段。第一阶段首先提取机械零件三维模型骨架,然后将骨架转换成骨架树并用邻接矩阵来描述骨架树的拓扑结构特征,通过比较邻接矩阵特征值之和迅速完成零件拓扑结构匹配,实现零件的初步筛选。将大量与待匹配模型拓扑结构差异较大的模型过滤掉,极大地减少了第二阶段的匹配计算量。第二阶段首先寻找匹配的骨架子树,其次在匹配子树的基础上搜索骨架枝匹配对,进而采用空间离散曲线的曲率和弗朗内特标架进行空间曲线相似性计算,得到整个骨架形状相似度。通过实例验证与试验分析,该方法快速有效,具有较高的准确性和良好的鲁棒性。

英文摘要：

A method for 3D model retrieval of mechanical parts based on skeleton tree is put forward. Retrieval is divided into two stages. The first stage is to extract the 3D model skeleton of mechanical parts, and then convert the skeleton into a skeleton tree. Using the adjacency matrix to describe the topological structure of the skeleton tree, the initial selection of mechanical parts is realized by comparing the sum of eigenvalues of the adjacent matrix. A large number of models which have a big difference with the matching model in topological structure are filtered out, which greatly reduces the matching computation of the second stage. The second stage, find the matched skeleton subtree firstly, search matched skeleton branches based on the matched subtree secondly, and then using the curvature and Frenet frame of space discrete curves to calculate the skeleton branches similarity, and then get the whole skeleton tree shape similarity. Through example analysis with the experiment, this method is effective and has high accuracy and good robustness.