摘要
提出了一种改进的自适应新生目标强度的概率假设密度(PHD)滤波算法.首先,对归一化因子进行了分析,在此基础上,提出了一种改进滤波策略,有效解决了归一化失衡问题;其次,在量测点附近通过无迹变换(UT)产生样本点,然后再采用粒子群(PSO)算法寻找最优点,从而得到新生目标概率密度函数的近似估计;最后,在序列蒙特卡罗(SMC)框架下对算法进行了实现.采用一种回溯策略,通过记录新生目标的状态和数目,修正存活目标的估计数目和相关航迹,进而得到每个目标的完整航迹.仿真结果表明:改进算法可以有效跟踪多个机动目标的状态和数目,滤波精度较高,具有较好的工程应用前景.
The improved adaptive target birth intensity for the PHD （probability hypothesis density） filter was proposed in this paper .Firstly ,the filtering strategy was proposed based on the analysis of the normalized factor ,w hich could effectively solve the problem of the normalized unbalance .Second‐ly ,the samples were generated in terms of the received measurements by using unscented transform （UT） .Then the optimum samples were found by using particle swarm optimization （PSO） algorithm . The statistical features of the newborn targets could be obtained .At last ,a sequential Monte Carlo （SMC） implementation was proposed ,and then a recalling procedure for track maintenance was devel‐oped .The estimation number of targets was corrected by recording the number and states of the new‐born targets .So the complete track of each target was obtained by recalling the previous tracks .The simulation results show that the improved algorithm can be used to track the states and number of the multi‐target effectively in clutter ,and it is so easy to be realized because of the simple design ,w hich make it a very popular method with applications prospection .
出处
《华中科技大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2015年第9期66-72,共7页
Journal of Huazhong University of Science and Technology(Natural Science Edition)
基金
国家自然科学基金资助项目(61370180)
关键词
随机集
概率假设密度滤波
量测驱动
粒子滤波
归一化失衡
多目标跟踪
random finite set
probability hypothesis density filter
measurement driven
particles fil-ter
normalized unbalance
multi-target tracking
