基于MDk-DPC的空中目标自动分群方法  

Aerial target automatic grouping method based on MDk-DPC

作　　者：马钰棠 孙鹏[1] 张杰勇[1] 闫云飞 赵亮 MA Yutang;SUN Peng;ZHANG Jieyong;YAN Yunfei;ZHAO Liang(Information and Navigation College,Air Force Engineering University,Xi’an 710077,China)

机构地区：[1]空军工程大学信息与导航学院,西安710077

出　　处：《北京航空航天大学学报》2024年第10期3219-3229,共11页Journal of Beijing University of Aeronautics and Astronautics

摘　　要：空中目标分群本质上是一个类数未知的聚类问题,也是战场态势估计领域中的研究热点。针对未知的空战场环境,从聚类角度提出一种基于流形距离和k近邻采样密度的MDk-DPC算法。引入流形距离代替欧氏距离,以增加同一流形中目标的相似性;利用k近邻计算目标的局部密度,使其能更真实地反映目标周围分布;通过自适应选取聚类中心方法确定聚类中心,并运用密度峰值算法指定剩余点类别完成分群。仿真实验表明,所提方法在人工合成数据集和UCI真实数据集上均有更好的聚类性能,同时通过对空战场仿真数据进行分群验证了所提方法的可行性和有效性。Air target grouping is a popular topic for research in the area of combat scenario assessment and can be thought of as essentially an uncountable class clustering issue.Aiming at the unknown air battlefield environment,a MDk-DPC algorithm based on manifold distance and k-nearest neighbor sampling density is proposed from the perspective of clustering.First,manifold distance is introduced to replace Euclidean distance to increase the similarity of objects in the same manifold.Secondly,the target's local density is determined using the k-nearest neighbors method,allowing the local density to more accurately represent the distribution surrounding the targets.Finally,an adaptive cluster center selection method is proposed to automatically determine cluster centers,and the DPC algorithm is used to specify the remaining point categories to complete the clustering.Simulation experiments show that the proposed method has better clustering performance on both artificial synthetic datasets and UCI real datasets.At the same time,the feasibility and effectiveness of the method are verified by clustering the simulated air battlefield data.

关 键 词：态势估计 目标分群 流形距离 K近邻 密度峰值聚类 

分 类 号：V247.5[航空宇航科学与技术—飞行器设计]