新生目标强度未知的双门限粒子PHD滤波器

传统粒子概率假设密度(PHD)滤波器假定新生目标强度已知,当新生目标在整个观测区域随机出现时不再适用。为解决新生目标强度未知时的多目标跟踪问题,提出了一种基于量测信息的双门限粒子PHD(PHD-DT)滤波器。首先基于似然函数设定门限对存活目标量测进行粗提取,利用上一时刻的目标估计值构建圆形波门进行精细提取,并对门限设定方法进行分析,然后根据提取结果对目标PHD进行分解,得到存活目标和新生目标的PHD预测及更新表达式,最后给出了滤波器的实现方法并同基于量测驱动的PHD(PHD-M)滤波器和Logic+联合概率数据互联(JPDA)方法进行了仿真对比。仿真结果表明,在新生目标强度未知时,PHD-DT可有效避免Logic+JPDA在杂波背景下因航迹起始错误带来的估计误差,并较好地解决了PHD-M的目标数目过估问题,多目标估计性能更优,且杂波越强性能优势越明显。

A dual threshold particle PHD filter with unknown target birth intensity

In situations where the targets can appear anywhere in the surveillance region, the traditional probability hypothesis density (PHD), which assumes that the target birth intensity is known a priori, is inefficient any more. To overcome this problem, a dual threshold particle PHD (PHD-DT) filter with unknown target birth intensity is proposed. First, through an analysis of threshold selection, the threshold based on likelihood function is designed for measurement crude extraction and the circular threshold based on the last state estimation is set for refined extraction. Afterwards, according to the extraction results, the predication and update equations for persistent and newborn targets are derived by the decomposition of PHD. Finally, the implementation of the PHD-DT filter is presented. The simulation results show that PHD-DT filter, which avoids the track initiation error of Logic+JPDA and succeeds in reducing the overestimate problem of PHD-M, possesses higher estimation performance. Moreover, the stronger the clutters, the more significant performance advantage of PHD-DT filter.