融合形状特征的MRG骨架树三维检索方法

3D model retrieval based on both shape feature and MRG skeleton

提出一种基于MRG骨架树的三维模型检索方法。根据多分辨率Reeb图(MRG)的原理,提取反映模型拓扑特征的Reeb图骨架并且映射成树形结构,分析了节点的拓扑属性。针对拓扑属性在形状特征上的表达能力不足,在节点相应区域提取离散曲率和面积比例描绘局部的形状特征。有效地结合了模型的拓扑特征和形状特征计算模型的相似度。该方法突出了模型的整体拓扑特征和形状特征,实验结果表明了该方法的高效性和鲁棒性。

A 3D retrieval method based on MRG skeleton tree structure is proposed. Based on multi-resolution Reeb graph （MRG） theory, the skeleton of Reeb graph which indicates the topological structure of 3D model is extracted and generated. The method transforms Reeb graph to a tree structure. The skeleton of Reeb graph accounting for only topology features are unable to compare model at a local level of shape detail. The method acquires the geometrical features by computing the discrete curvatures and areas of corresponding segments for each critical point in skeleton. Combining the topology structure and local geometric details, an optimized similarity and dif- ference functions are proposed. The experimental results show the robustness and efficiency of this method in 3D models similarity matching.