融合权重信息的三维CAD模型聚类研究  

Research on 3D CAD Model Clustering Based on Weight Information

作　　者：汪大涵 王裴岩[1] 张桂平[1] 马伟芳 WANG Da-han;WANG Pei-yan;ZHANG Gui-ping;MA Wei-fang(Human-computer Intelligence Research Center,Shenyang Aerospace University,Shenyang 110136,China)

机构地区：[1]沈阳航空航天大学人机智能研究中心,沈阳110136

出　　处：《小型微型计算机系统》2020年第6期1296-1301,共6页Journal of Chinese Computer Systems

基　　金：辽宁省自然科学基金重点项目(20170540705)资助.

摘　　要：局部区域特征的获取与表达对于研究三维CAD模型聚类至关重要.面向局部区域特征表达问题提出了在现有的六元组方法基础之上将其扩展为七元组,加入了模型中面与面相交形成的边属性信息,从而更好的获得了由局部区域特征构建的词汇本;在聚类阶段,提出了一种模型局部区域加权方法,该方法降低常见局部区域在聚类相似度计算时的最重要程度,从而相对提高了更有区分度的局部区域.实验结果表明,采用本文提出的表达方法能有效支持CAD模型聚类任务,对比基线方法在四种典型聚类算法上得到的NMI值、V-measure值、Purity值均有提升.The acquisition and expression of local region features is very important for the study of 3 D CAD model clustering.To solve the problem of local region feature representation,this paper extend the existing six-tuple method to seven-tuple based on the existing six-tuple method,and add the edge attribute information formed by the intersection of the surface and the surface in the model,so as to obtain the vocabulary book constructed by the local region feature better.In the clustering stage,this paper propose a model local region weighting method,which reduces the common local regions in the clustering phase.Similarity calculation is the most important degree,which relatively improves the more differentiated local areas.The experimental results show that the proposed method can effectively support the clustering task of CAD model.Compared with the baseline method,the NMI,V-measure and Purity values of four typical clustering algorithms are improved.

关 键 词：三维CAD模型 特征扩展 局部区域加权 聚类 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]