基于KL距离的异类传感器动态数据关联算法  

Dynamic Data Association Algorithm Based on KL Distance for Heterogeneous Sensors

作　　者：吕丽平[1] 白鑫[1] 张玉宏[2] LYU Li-ping BAI Xin ZHANG Yu-hong(Department of Information Engineering, Shengda Economics Trade ＆ Management College of Zhengzhou, Zhengzhou 451191, China College of Information Science and Engineering, Henan University of Technology, Zhengzhou 451000, China)

机构地区：[1]郑州升达经贸管理学院信息工程系,郑州451191 [2]河南工业大学信息科学与工程学院,郑州451000

出　　处：《电光与控制》2017年第11期43-48,共6页Electronics Optics & Control

基　　金：河南省科技厅自然科学项目(152102210261)

摘　　要：针对异类传感器多目标跟踪系统中的动态数据关联问题,提出了一种基于KL距离的异类传感器动态数据关联算法。该方法将静态多维分配问题推广到动态跟踪中,通过对量测集合与航迹集合的合并,将各传感器量测估计位置的概率密度函数与目标航迹一步预测值的概率密度函数之间的KL距离作为关联代价函数,继而代入多维分配模型求解,从而实现量测到航迹的动态关联。仿真实验表明,所提算法能更精准地反映量测与航迹之间关联的可能性,且能够对多目标进行快速稳定的跟踪。To solve the problem of dynamic data association in the multi-target tracking system of heterogeneous sensors,a dynamic data association algorithm for heterogeneous sensors is proposed based on KL distance.This method extends the static multi-dimensional assignment to dynamic tracking.The KL distance between the probability density function of the estimated position and the probability density function of the predicted position of the target track is taken as the association cost function,by merging the measurement set with the track set.Then,it is substituted into the multi-dimensional assignment model to realize the dynamic correlation between measurements and tracks.Simulation results show that the proposed cost function can reflect the association probability between measurements and tracks more accurately and can track multiple targets quickly and steadily.

关 键 词：数据关联 多目标跟踪 粒子滤波 KL距离 

分 类 号：TN953[电子电信—信号与信息处理]