论文部分内容阅读
重采样算法是粒子滤波器的重要组成部分,不同的重采样算法对滤波性能的影响也不同。本文对三种常用的重采样算法(多项式重采样、分层重采样、系统重采样)进行了理论分析,并结合免聚类粒子概率假设密度(Free Clustering Particle Probability Hypothesis Density,FCP-PHD)滤波器在设定的多目标跟踪场景下进行比较,重点对比了三种算法的滤波精度和计算时间。仿真结果表明,在不同的采样粒子规模下,多项式重采样的计算量最大,而分层重采样与系统重采样的滤波精度均好于多项式重采样,多项式重采样的计算效率也相对最低。另外分层重采样与系统重采样的计算量相近,随着粒子数的增大系统重采样的优势也逐渐明显,当粒子数较多时其计算效率也相对最高。
Resampling algorithm is an important part of the particle filter, different resampling algorithm on the filtering performance is also different. In this paper, three commonly used resampling algorithms (polynomial resampling, layer resampling, system resampling) are theoretically analyzed, and combined with Free Clustering Particle Probability Hypothesis Density (FCP-PHD) filtering In the set multi-target tracking scenarios to compare, focusing on the three algorithms compared to the filtering accuracy and calculation time. The simulation results show that the polynomial resampling is the most computationally intensive under different sampling particle sizes, while the filtering accuracy of both hierarchical resampling and system resampling is better than that of polynomial resampling, and polynomial resampling is also the least efficient. In addition, the stratified resampling is similar to the system resampling. With the increase of the number of particles, the advantage of the system resampling becomes obvious. When the number of particles is large, the computational efficiency is also relatively high.