基于特征相似度聚类的空中目标分群方法  

Aerial Target Grouping Method Based on Feature Similarity Clustering

作　　者：柴慧敏[1,2] 张勇 方敏[1] CHAI Hui-min;ZHANG Yong;FANG Min(School of Computer Science and Technology,Xidian University,Xi'an 710071,China;Science and Technology on Electro-Optical Information Security Control Laboratory,Tianjin 300308,China)

机构地区：[1]西安电子科技大学计算机科学与技术学院,西安710071 [2]光电信息控制和安全技术重点实验室,天津300308

出　　处：《计算机科学》2022年第9期70-75,共6页Computer Science

基　　金：装备预研重点实验室基金(6142107190106)。

摘　　要：针对采用聚类算法进行目标分群时需要给出聚类个数和对初始中心选择敏感的问题,提出了一种基于目标特征相似度聚类的分群方法。该方法首先计算目标间的相似度值,构建相似度矩阵;然后计算相似度矩阵的连通分支,获取群中心结构和孤立目标点,识别的群中心结构个数为聚类个数;最后将不属于群中心结构和孤立点的目标归类到与其最相近的群中心结构中,使得聚类过程不再过多地依赖于聚类初始中心的选择。实验结果表明,所提方法能够正确识别出多种形态的群中心结构,并能检测出孤立点,且目标聚类正确率均高于其他4种聚类算法。In order to solve the problems that the number of clusters needs to be given and the sensitivity to the initial positions of the cluster centers while clustering algorithm is utilized for target grouping, a novel aerial target grouping method based on feature similarity clustering is proposed.Firstly, the similarity between targets is calculated and the similarity matrix is constructed.Then, the connected branches of the similarity matrix are calculated to obtained the group center structure and the isolated target points are detected.The number of group center structures is the number of clusters.Finally, the targets which are not belonging to the group center structure and the isolated points are clustered into the closest group center structure.It makes the clustering process no longer depend too much on the initialization of the cluster centers.Experimental results show that the proposed methodcan correctly identify the group center structure and detect the isolated points.Furthermore, its the clustering accuracy is higherthan that of other four clustering algorithms.

关 键 词：空中目标分群 聚类算法 目标相似度 群中心结构 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]