一种利用信号幅度信息的光学传感器像平面GM-PHD滤波算法

针对光学传感器像平面多目标跟踪应用,基于有限集统计学理论(Finite Set Statistics,FISST)提出一种利用信号幅度信息的高斯混合概率假设密度(Gaussian Mixture-Probability Hypothesis Density,GM-PHD)滤波算法。该算法通过建立信号幅度似然函数将幅度信息引入到PHD的递推公式中,首先根据光学传感器的特点对输出的信号幅度进行建模,然后分别计算来源于目标和杂波的信号幅度似然函数,在目标信噪比未知的情况下给出一种替代计算方法,最后采用高斯混合方式实现该算法。仿真结果表明,本文提出的算法相比仅利用量测位置信息的PHD滤波器能够显著提高多目标跟踪性能和计算效率,而且在目标信噪比未知时同样具有稳定的性能。

A GM-PHD Filter Using Signal Amplitude Information for Image Plane of Optical Sensor

A probability hypothesis density(PHD) filter using signal amplitude information base on framework of finite set statistics(FISST) is proposed for multi-target tracking in the image plane of optical sensor.Signal amplitude likelihood functions are derived to incorporate amplitude information into this filter.First,the amplitude of signal generated by optical sensor is modeled,and the signal amplitudes of target and clutter are calculated respectively while an alternative method is adopted for the case of unknown target SNR.Gaussian mixture(GM) is used to implement this filter.Simulation results show the significant improvement in multi-target tracking performance and computation cost of our algorithm compared with the generic PHD filter just using location measurements,even when the SNR of target is unknown.