摘要
用积分直方图技术提出了一种低复杂度的粒子滤波跟踪方法和基于积分颜色直方图的观测似然模型.采用积分方向直方图建立一个检测响应图,用其上的观测信息构建建议分布函数,在状态空间中似然概率较大的子空间中进行粒子采样.仿真结果表明,在采用大量粒子跟踪大目标时,提出的跟踪方法的计算复杂度明显低于直方图直接提取的粒子滤波方法,而且在光照变化的条件下,比采用传统颜色Mean-Shift的跟踪方法准确.
A particle filtering tracking method is presented with low computational complexity using integral histogram technique.Integral orientation histogram is adopted for the construction of a response map.Based on this response map,a proposal is constructed such that particles are drawn from regions of the state space with high likelihood.Also,an integral color histogram-based likelihood model is proposed.Simulation results show that the proposed tracking method has lower computational complexity than that using straightforward histogram extraction method when using large number of particles for tracking large objects.The results are much accurate compared with that using traditional color Mean-Shift method under illumination change conditions.
出处
《光子学报》
EI
CAS
CSCD
北大核心
2011年第11期1761-1766,共6页
Acta Photonica Sinica
基金
国家重点研究发展计划(No.2009CB724005)资助
关键词
视觉跟踪
粒子滤波
建议分布函数
积分直方图
颜色直方图
方向直方图
Visual tracking
Particle filtering
Proposal distribution
Integral histogram
Color histogram
Orientation histogram
