摘要
提出一种基于协同表示的新的目标跟踪算法.在贝叶斯框架下,采用基于重构误差的观测似然函数和考虑遮挡的模型更新机制设计一个鲁棒的跟踪器.用l1范数来建模重构误差以更好地容忍奇异值,同时用协同表示对编码系数进行约束.实验结果表明,和其他算法相比,本文算法能够战胜遮挡、尺度变化、光照变化、背景混乱等干扰因素,具有较高的准确度和鲁棒性.
A new algorithm of object-tracking based on collaborative representation is proposed.Within Bayes framework,the robust tracker is presented using the observation likelihood model based on the error reconstruction and the updating scheme accounting for occlusion. The error reconstruction is modelled by the l1 norm to tolerate the outlier,and the code coefficients are regularized by collaborative representation. Experimental results show that the proposed algorithm solves the problems of occlusion,scale changing,varying illumination,and cluttered background with high accuracy and robustness.
出处
《中国科学院大学学报（中英文）》
CSCD
北大核心
2016年第1期135-143,共9页
Journal of University of Chinese Academy of Sciences
基金
山东省自然科学基金高校
科研单位联合专项计划项目(ZR2015FL009
ZR2014FL020)
滨州市科技发展计划项目(2013ZC0103)
滨州学院科研基金(BZXYG1524)资助
关键词
目标跟踪
协同表示
L1范数
观测似然函数
object tracking
collaborative representation
l1 norm
observation likelihood function
