一种多尺度重采样粒子滤波算法

Particle Filer Algorithm Based on Mutil-scale Resampling

粒子滤波过程中通过引入重采样消除粒子匮乏现象,但是重采样过程却削弱了粒子的多样性,导致粒子贫化.为协调粒子多样性和样本贫化之间的冲突,提出一种多尺度重采样粒子滤波算法,粒子空间重采样划分多个尺度,然后重新定义各尺度粒子权重并重采样,用尺度熵值度量重采样粒子的多样性,指导重采样.仿真实验结果表明,多尺度重采样粒子滤波算法有效提高了精度,适用于高精度系统滤波计算,并将应用于精细果业中数据同化.

The introduction of resampling into particle filtering to eliminate particle deficiency weakens the diversity of particles and leads to particle dilution. To reconcile the conflict between particle diversity and sample dilution,a multi-scale-resampling-based particle filtering algorithm is proposed. The particle spatial resampling is divided into multiple scales,and particle weights are redefined for each scale before resampling,using scale entropy as a guide to measure the diversity of resampled particles. The simulation experiment shows that multiscale-resampling- based particle filtering algorithm significantly increases the precision and is practical for highprecision systematic filtering computation,and it can be applied to the data assimilation of precision fruit.