Abstract: In engineering applications, retrieval of existing triangular mesh models and reuse of their design information can reduce time and cost. A similarity assessment approach of triangular mesh models based on region segmentation is described. A triangular mesh model is segmented into some regions based on spherical image. For each region, a 10-dimensional vector is used to describe its geometric properties and topological attributes. A group of 10-dimensional vectors act as the shape descriptor of the triangular mesh model. Each 10-dimensional vector is regarded as a node with attributes. Two groups of nodes that correspond to the two shape descriptors can be seen as a complete bigraph. By seeking a matching of maximum total weight from the weights of the bigraph the similarity between two shape descriptors is assessed. Experiments show that this algorithm can efficiently give the similar degree of triangular mesh models.

Key words: triangular mesh model, similarity comparison, region segmentation, spherical image

摘要： 为了在工程应用中检索已有的三角网格模型，以便重用相应零件的设计信息，节省设计和加工成本，提出一种基于区域分割技术的三角网格模型相似性比较算法。依据三角网格模型的球面图像将模型分割为若干个区域；对每个分割得到的区域用一个10维向量表达其形状的几何特征和拓扑特征，一个三角网格模型的特征即可通过各分割区域所对应的10维向量组成的向量组表达；将该向量组作为三角网格模型的形状描述子，两个三角网格模型的相似性可通过相对应的形状描述子间的相似性表达。将描述子中的每个向量看成是一个带有属性的节点，通过两组节点组成的完全二分图的最优匹配进行两个形状描述子之间的相似性比较，实现两个三角网格模型之间的相似性比较。实验结果表明，该算法有效可行。

关键词: 三角网格模型, 相似性比较, 区域分割, 球面图像