粗细尺度耦合粒子滤波在多目标跟踪中的应用研究

An Efficient and Accurate Algorithm for Multiple Target Tracking

文章提出了一种进行多目标跟踪的高效率粒子滤波器算法。它利用粗细2种尺度耦合采样信号,粗尺度采样降低算法计算的复杂度;细尺度采样来保证精度,同时不用重采样来消除粒子滤波的退化现象。对信号经过粗细2种尺度处理以后,再用基于MCMC方法的Metroplis-Hasting采样得到多目标系统的最大似然函数参数后验满条件的分布,并对多目标系统进行实时跟踪。仿真表明该算法在处理多目标跟踪问题时具有高效率和精确性。

Aim. We have come to believe that existing algorithms can be improved in efficiency and accuracy. We propose using coarse and fine scale coupling to sample signals so as to obtain a better algorithm. In the full paper, we explain our algorithm in some detail; in this abstract, we just add some pertinent remarks to naming the first three sections of the full paper. Section 1 is： the model of particle filter for multiple target tracking. In section 1, eq. （5） we derive is the most important. Section 2 is： the coupling model for coarse and fine scale sampling. In this section, we use the coarse scale sampling to reduce the computational complexity of the algorithm and the fine scale sampling to guarantee the sampling accuracy. Section 3 is. the particle filtering algorithm using coarse scale and fine scale coupling. In this section, we use the Metroplis-Hasting sampling to obtain the full condition distribution of maximum likelihood posterior function of a multiple target system, thus avoiding the re-sampling and the particle filtering degeneration. Thus we track multiple targets in real time. Also in section 3, we make use of eq. （5） in our mathematical derivation. Finally, we perform the computer simulation of our algorithm. The simulation results, shown in Figs. 1 through 4, indicate preliminarily that our algorithm does enhance the efficiency and accuracy of multiple target tracking.