摘要
针对由单传感器概率假定密度滤波到多传感器情形推导困难的问题,提出了一种有序粒子概率假定密度跟踪算法。首先,推导出集中式多传感器粒子概率假定密度滤波模型,再根据集中式融合系统的特点,选取与多传感器相关的重要性密度函数,通过多传感器多步更新重采样粒子,从而实现多传感器多目标有序粒子概率假定密度跟踪。仿真结果表明,该算法的跟踪误差距离差要小于单传感器粒子概率假定密度跟踪算法,且具有更优越的跟踪性能。
For the problem of the difficulty in extending single sensor Probability Hypothesis Density（PHD） filtering to the multi-sensor case,a new sequential particle-PHD tracking algorithm is proposed.First,the general theoretical model of centralized multi-sensor particle-PHD filtering is deduced.Then,the importance density function with regard to multiple sensors is chosen according to the characteristics of centralized fusion system.The resampling particles are updated via multiple sensors.So multi- target multi-sensor sequential particle-PHD tracking is implemented.Experimental results show that the tracking miss distance of the proposed algorithm is less than single sensor particle-PHD tracking algorithm and it has better tracking behavior.
出处
《计算机工程与应用》
CSCD
北大核心
2009年第25期58-61,64,共5页
Computer Engineering and Applications
关键词
多传感器
随机集
融合系统
多目标跟踪
multi-sensor
random set
fusion system
multi-target tracking
