图结构上下文及在图聚类中的应用  

A New Graph Clustering based on Graph Structure Context

作　　者：施汉琴[1] 罗斌[2] 汤进[2] 江波[2] 

机构地区：[1]淮北师范大学计算机科学与技术学院,安徽淮北235000 [2]安徽大学计算机科学与技术学院,安徽合肥230601

出　　处：《淮北师范大学学报（自然科学版）》2015年第2期21-25,共5页Journal of Huaibei Normal University：Natural Sciences

基　　金：淮北师范大学科研项目(700692)

摘　　要：图结构特征的提取以及距离度量是计算机视觉和模式识别中的重要内容.针对传统的图上下文描述子中只考虑每个柱形区域内的特征点数目而忽略特征点之间的结构信息的问题,提出一种图的结构上下文描述子及距离度量方法.首先对图的所有顶点建立图结构上下文描述子;其次,利用二次卡方(QC)距离方法实现图的距离度量;最后,采用最小生成树聚类算法实现图聚类.实验表明,该方法对于一些非刚性变换的图聚类有较好的效果.Feature extraction and similarity measurement for graphs are important issues in computer vision and pattern recognition.However,traditional graph context method only considered the number of feature points and ignored the structure information of feature points in each quadrant area,so a new graph structure context（GSC）descriptor and its similarity measurement method were proposed.Firstly,graph structurecontext descriptor was presented based on the feature point set.Secondly,graph distance was measured by using Quadratic Chi-square（QC）distance. Finally,minimum spanning tree clustering algorithm was adopted to realize graph clustering.The experimental results demonstrate that the proposed method performs bette rfor the graphs under some non-rigid transformation.

关 键 词：相似度量 图结构上下文 二次卡方距离 最小生成树 图聚类 

分 类 号：X703.1[环境科学与工程—环境工程]