基于MCMC粒子直接跟踪算法

Direct tracking by MCMC-based particle algorithm

针对运动目标直接跟踪粒子滤波器存在粒子贫化问题,提出一种基于马尔科夫链蒙特卡洛(Markov chain Monte Carlo,MCMC)粒子直接跟踪算法。首先建立一个基于时延和多普勒频移的直接跟踪观测模型,然后在贝叶斯滤波迭代中引入典型的MCMC方法—Metropolis Hasting(M-H)抽样算法,选择符合拒绝接收率的采样样本作为新的粒子群,避免了传统粒子滤波器采样枯竭。仿真结果表明,在复杂度相同的条件下,所提算法比直接跟踪粒子滤波算法具有更好的跟踪精度。

A direct tracking algorithm based on Markov chain Monte Carlo(MCMC)is proposed to solve the particle depletion problems of the direct tracking of moving target by particle filter.Firstly,the algorithm builds a direct tracking model based on time delay and Doppler frequency shift.Then,the standard MCMC method,Metropolis Hasting(M-H)sampling,is incorporated into the Bayesian filtering iterations to select the sampling samples that meet the rejection-reception rate as the new particle group,which avoids the problem of sampling exhaustion of the standard particle filter.The simulation experiment showed that under the same complexity condition,the proposed algorithm has much better performance in tracking accuracy than the direct tracking particle filter algorithm.