摘要
针对跟踪过程中运动目标形态不断变化及跟踪不精确导致鲁棒性差的问题,提出了一种运用聚类方法的分层采样粒子滤波算法。通过分层采样把采样空间分成多个部分,使采样点集中于被采样概率密度函数值大的部分,采样误差降低到了原算法的一半;聚类方法利用权重实现合理分配粒子,使粒子的多样性得到保持,因而粒子跟踪的精度得到了提高。实验结果表明,所提算法的跟踪误差不到原算法的一半,每个仿真时间里稳定性都有加强,而且跟踪精度也有所提高。
To solve the poor robustness due to the changing moving target or the inaccurate tracking, a stratified sampling particle filter algorithm based on clustering method was proposed. The sampling space was divided into several parts by group sampling to make sampling points focused on the big probability density value part, thus the sampling error was reduced half of the original; the clustering algorithm could group the particles reasonably by weight, the diversity of particles was kept, thus the tracking precision was improved. The experimental results show that the tracking error of proposed method is less than half of the original one, and the stability has strengthened in each simulation time, as well as the tracking precision.
出处
《计算机应用》
CSCD
北大核心
2013年第1期69-71,共3页
journal of Computer Applications
基金
国家科技支撑计划基金资助项目(2007BAG06B06)
关键词
运动目标
粒子滤波
分层采样
聚类方法
追踪精度
moving target
particle filter
stratified sampling
clustering method
tracking precision
