基于随机游走与最优匹配的装配体相似性分析

Similarity Assessment of Assemblies Based on Random Walks and Optimal Matching

为了提高利用拓扑信息进行装配体检索的能力,提出一种基于随机游走和最优匹配的相似性分析方法.首先用图节点表示装配体中的零件,将装配体转化为属性连接图并进行简化;然后分析零件连接装配关系的紧密程度并构建转移概率矩阵,利用随机游走模型分析各零件的拓扑特征;最后依据拓扑和属性特征进行零件的两两相似性评价,建立对比装配体的二分图并用最优匹配获得装配体的整体相似性.实例验证了文中方法是正确和有效的.

In order to improve the utilization of topology information in assembly model retrieval,an efficient similarity analysis method based on the random walks and optimal matching is presented in this paper.Firstly,the parts of assembly are represented as graph nodes,and the assembly is represented by the attribute adjacent graph which is simplified by removing fastener.Secondly,the probability matrix is built based on the connection relationships analysis,and the topological signature of part is computed by using the random walks model.Thirdly,the bipartite graph for assembly is built by the multiple comparisons of topological and attribute signature,and the similarity of assemblies is assessed by solving bipartite graph optimal matching problem.Finally,the validity and rationality of the method is demonstrated with the study cases.