一种粒子概率假设密度滤波新生粒子采样新方法

A Novel Birth Particle Sampling Method for the Particle PHD Filter

提出一种粒子概率假设密度(Probability Hypothesis Density,PHD)新生粒子采样新方法。以混合高斯分布和均匀分布分别对新生粒子位置和速度分量进行采样,将采样过程置于滤波更新之后,通过最大似然检验多目标状态估计技术提取源于已知目标的量测,避免对这些量测进行新生粒子采样,有效降低粒子数和滤波计算量。结果表明:基于新生粒子采样新机制的粒子PHD滤波,相比于标准方法,在降低计算量的同时提高了多目标状态估计精度。

A novel birth particle sampling method for the particle probability hypothesis density（PHD） is presented.The position and velocity components of the birth particle is sampled from mixture Gaussian and uniform distributions separately,but the kernel of the method is that this sampling process is carried out after the PHD update procedure,as not only improves tracking precision,but also extracts the measurements originating from the existent targets being tracking through maximum likelihood test clustering method,thus the number of birth particles and the computation load of the filter could be reduced.Simulations results show that,compared with the standard particle PHD filter,the tracking accuracy is enhanced significantly with the improved PHD filter based on the proposed birth sampling scheme while a lower computation load is achieved.