改进的高斯粒子概率假设密度滤波算法

高斯粒子概率假设密度滤波在预测和更新时需要进行粒子近似和重新采样，这在一定程度上降低了算法的精度和实时性．针对这一问题，提出一种改进的高斯粒子概率假设密度滤波算法．算法通过粒子的方式表示并传递目标的概率假设密度（PHD）预测值，然后直接利用这些表征PHD预测值的粒子进行更新，最后利用具有最大似然性的粒子将更新后的PHD表示为混合高斯形式．仿真实验表明，和高斯粒子概率假设密度滤波相比，改进算法的多目标误差距离减少了约30％，运行时间减少了约50％．

Modified Gaussian particle probability hypothesis density filtering algorithm

The Gaussian particle probability hypothesis density filter needs particle approximation and resampling in the prediction step and the update step; this lowers the accuracy and deteriorates the real-time performance of the algorithm to some extent. To solve this problem, a modified Gaussian particle probability hypothesis density filtering algorithm is proposed. This algorithm expresses and transfers the predicted probability hypothesis density (PHD) of targets in the form of particles, and then directly updates these particles representing the predicted PHD. Finally, the algorithm approximates the updated PHD into a Gaussian mixture function by using the particles with greatest likelihood. The simulation experiments show that the modified algorithm reduces the multi-target error distance by nearly 30% and cuts the running time by nearly 50% in comparison with Gaussian particle probability hypothesis density filter.