摘要
针对目前目标跟踪算法在复杂条件下跟踪效果差的问题,提出了基于排名的结构稀疏表示目标跟踪算法.该算法首先构造含有目标和背景信息的对象字典,用结构稀疏表示的方法线性表示候选目标,并将相关系数进行组合和排名.然后采用目标残差和背景残差相结合的方法计算各候选目标的残差得分,并进行排名.最后,通过稀疏系数和残差排名得到目标状态.试验证明,文中算法具有跟踪准确度高和鲁棒性强的特点.
As the existing tracking algorithms are of poor performance in complex situations,a target tracking algo-rithm with structured sparse representation is proposed based on ranks.In this algorithm,first,a target dictionary containing targets and background information is constructed.Next,the candidate targets are represented linearly via the structured sparse representation,and the corresponding correlation coefficients are combined and ranked. Then,the residual error scores of the candidate targets are computed by using the residual errors of the target and the background,and the scores are ranked later,from which the target state is finally determined.Experimental re-sults show that the proposed tracking algorithm is of high tracking accuracy and strong robustness.
出处
《华南理工大学学报(自然科学版)》
EI
CAS
CSCD
北大核心
2013年第11期23-29,35,共8页
Journal of South China University of Technology(Natural Science Edition)
基金
粤港关键领域重点突破项目(2011BZ100012)
东莞市高等院校
科研机构和医疗卫生单位科技计划重点项目(2011108101002)
关键词
目标跟踪
结构稀疏表示
残差
稀疏系数
粒子滤波
target tracking
structured sparse representation
residual error
sparse coefficient
particle filtering