一种三维弹道目标跟踪的粒子滤波方法∗

摘要

针对三维弹道目标，给出了一种有效的基于粒子滤波的跟踪算法。这种算法以标准的粒子滤波算法为基础，根据贝叶斯原理利用局部线性化技术获得最佳近似的重要性密度函数以避免粒子退化现象，并且利用 Metropolis-Hastings(MH)采样构造的马尔科夫链得到更加符合目标分布的样本，从而最小化重采样后的粒子枯竭问题。此外，这里采用 Kullback-Leibler 距离(KLD)指标对不同粒子滤波算法的性能进行评估。仿真结果表明，该三维弹道目标跟踪算法粒子群与参考粒子群(近似真实目标概率分布的粒子群)之间的KLD比标准粒子滤波与参考粒子群之间的KLD更小，因此，能获得比标准粒子滤波算法更好的跟踪效果。%A particle filter algorithm for three dimensional(3D)ballistic target tracking is given.Based on the standard particle filter(sampling/importance resampling,SIR),this algorithm uses an optimized im-portance function which is based on Bayes principle and local linear technique to combat particle degeneracy. Meanwhile,it incorporates a Metropolis-Hastings(MH)move step which is possible to gain the more suit-able samples to reduce particle impoverishment.Furthermore,the Kullback-Leibler divergence(KLD)inde-xes are used to evaluate the performance of different particle filters.Simulation results demonstrate that the KLD’s estimator between the particle clouds of the particle filter algorithm for 3D ballistic target tracking and the reference particle clouds(optimized truth)is smaller than the SIR.Therefore,the algorithm can be better than the standard particle(SIR)in tracking.