基于粒子滤波的检测前跟踪算法的改进

针对基于粒子滤波的检测前跟踪(PF-TBD)技术在弱目标的跟踪定位中,目标检测概率较低的问题提出改进。首先,对重要性密度函数进行重新构造,在只包含弱目标的运动模型预测数据的基础上,将实际观测数据与其一起构成的后验概率密度函数作为改进后的重要性密度函数。其次,在满足该后验概率密度函数的分布中,选取一定数目的粒子,在对目标的下一状态估值中,采用MMSE算法,推导出满足最小均方误差的表达式,而且通过引入概率粒子滤波算法,在计算上避开了积分运算。通过仿真实验表明,改进的PF-TBD算法不仅计算简单,而且提高了弱目标检测概率。

Improvement of tracking-before-detection based on particle detection

In the detection of weak targets ,the tracking-before-detection based on particle detection technique (PF-TBD)was applied.In this algorithm, it faces the low probability of detection.This article proposes improvements in it. Making a new posterior probability density function with the observed data on the basis of the prediction data and motion model of weak targets.In the new posterior probability density function ,a certain number of particles are selected.Using MMSE algorithm to estimate the next state of the target and derive the expression of minimum mean square error. In order to avoid the integral operation in the process of the calculation ,the probability particle filter algorithm is introduced.This algorithm not only avoids the integral operation ,but also improves the detection probability of weak targets.