基于GM-PHD平滑器的检测前跟踪技术

Track-Before-Detect Algorithm Using GM-PHD Smoothing Filter

雷达观测下弱小目标的检测前跟踪(TBD)问题中,针对基于高斯混合概率假设密度滤波器(GM-PHD)的TBD算法在信噪比降低时,存在目标的数目估计不准确、状态估计精度下降的问题,提出了基于GM-PHD平滑滤波器(SGM-PHD)的检测前跟踪算法(SGM-PHD-TBD)。该算法在TBD标准多目标观测模型框架下,采用平滑递归方法,利用多个量测数据对滤波值进行平滑,在牺牲一定运算效率的基础上提升了算法的估计精度。仿真结果表明,该算法在信噪比较低的情况下对目标数目估计的准确度和目标状态估计的精度均优于基于GM-PHD的TBD算法。

For the weak target track-before-detect （TBD） problem in radar sensor, when the signal-tonoise ratio is reduced, the TBD algorithm based on Gaussian mixture probability hypothesis density （GM- PHD） cannot estimate the number and state of the target precisely. In order to solve this problem, a new TBD algorithm based on the Gaussian mixture PHD smoothing （SGM-PHI）-TBD） algorithm is proposed. Within the framework of the TBD standard observation model, the algorithm adopts the smooth recursion and use a quantity of measurement data to smooth the filtering value. The simulation results show that under the condition of low SNR, the proposed algorithm can improve the estimation of target numbers and the target state to a certain extent and is better than the TBD algorithm based on GM-PHD filter.