基于主成分分析的粒子滤波器目标跟踪方法  

Particle Filter Algorithm Based on Principal Component Analysis

作　　者：王欣[1,2] 赵连义[1] 薛龙[1] 

机构地区：[1]吉林大学计算机科学与技术学院,长春130012 [2]吉林大学符号计算与知识工程教育部重点实验室,长春130012

出　　处：《吉林大学学报（理学版）》2012年第6期1156-1162,共7页Journal of Jilin University:Science Edition

基　　金：国家自然科学基金(批准号:60905022);吉林省科技发展计划项目(批准号:201105016)

摘　　要：提出一种基于主成分分析的粒子滤波器目标跟踪新方法.该方法将主成分分析法和传统的粒子滤波方法相结合,避免了传统粒子滤波器的过度重采样,提高了目标跟踪精度.实验结果表明,该方法对单个目标跟踪精度高,且对多障碍物下的目标跟踪精度也较高,适用于复杂背景下的人脸跟踪.与传统粒子滤波方法相比,该方法提高了目标跟踪的精度和鲁棒性,避免了粒子退化和粒子贫化.A new principal component analysis particle filter algorithm proposed in this paper combines principal component analysis method with the traditional particle filter method. The proposed method avoids undue resampling and improves the accuracy of object tracking. Experimental results demonstrate that this algorithm can acquire better precision in single object tracking and multiple- obstacle object tracking, and this algorithm could track human face accurately in complex backgrounds. Compared with present particle filter algorithm, this algorithm performs better in robustness and accuracy and avoids particle degeneration and particle impoverishment.

关 键 词：目标跟踪 粒子滤波算法 主成分分析法 重采样 

分 类 号：TP317.4[自动化与计算机技术—计算机软件与理论]