基于偏差映射聚类的目标关联方法  

Target Association Method Based on Biasmapping Clustering

作　　者：刘甲磊 石志广[1] 张焱[1] LIU Jia-lei;SHI Zhi-guang;ZHANG Yan(National University of Defense Technology,Automatic Target Recognition Laboratory,Hunan Changsha 410073,China)

机构地区：[1]国防科技大学ATR重点实验室

出　　处：《现代防御技术》2019年第3期106-112,共7页Modern Defence Technology

基　　金：国家自然科学基金项目(61302145)

摘　　要：多传感器目标关联是确定不同传感器系统观测的若干信号是否来源于同一目标,它是现代多传感器系统中的一个重要问题。传统的关联方法通过计算不同观测间的关联代价,通过求解代价矩阵最优解来获得关联匹配,但是容易受到环境和传感器性能的影响。提出了一种基于偏差映射聚类(bias mapping cluster,BMC)的目标关联方法,通过对多个传感器间观测目标偏差映射点进行聚类,搜索局部密度最大的映射点集作为传感器间的目标关联结果,走出了利用数学方法求解全局最优解的传统模式。相较其他传感器间目标关联方法,仿真结果表明该方法能有效利用目标观测的空间散布特性,关联正确率更高,并对虚假目标和目标失配等情况具有更强的适应性。Multi-sensor target association is to determine whether some signals observed by different sensor systems originate from the same target or not. It is an important problem in modern multi-sensor systems. The traditional method is usually to calculate the correlation cost matrix,use optimal algorithm to solve matrix and get target correlation pairs between different sensors,which is easily influenced by the environment and sensor performance. We propose a target association method based on bias mapping clustering( BMC). The mesurement bias pair is mapped into a point in the two-dimensional space and these points are clustered. By searching the local point density peak as the target correlation results between sensors,the traditional mode of using mathematical methods to solve the global optimal solution is gotten out. Compared to other methods,the simulation results show that this method can effectively use the spatial distribution of target observations to improve the association accuracy,and has a robust performance at the situation of false target and target mismatch.

关 键 词：多传感器 目标关联 空间偏差特性 峰值聚类 目标失配 

分 类 号：TN98[电子电信—信息与通信工程] TP391.9[自动化与计算机技术—计算机应用技术]