基于稀疏表示的目标跟踪算法  

Object tracking algorithm based on sparse representation

作　　者：储向锋[1] 朱春[1] 胡晓飞[2] 

机构地区：[1]南京邮电大学通信与信息工程学院,江苏南京210003 [2]南京邮电大学地理与生物信息学院,江苏南京210023

出　　处：《南京邮电大学学报（自然科学版）》2016年第4期101-106,112,共7页Journal of Nanjing University of Posts and Telecommunications：Natural Science Edition

基　　金：国家自然科学基金(61271082)资助项目

摘　　要：文中针对目标跟踪领域中有严重遮挡、光照变化、背景复杂等挑战性问题,提出一种基于稀疏表示联合外观模型的目标跟踪算法。在生成模型中,将得到的块结构的稀疏编码系数进行对齐汇集操作,并将汇集后的稀疏系数进行联合加权处理,这样保留目标的空间结构和局部信息来提高目标定位的精度与跟踪的鲁棒性,获得一个更加具有鲁棒性的外观模型。采用大量的测试视频和不同的主流算法进行对比,实验证明,文中提出的算法具有更高的准确性和鲁棒性。A new target tracking algorithm based on the sparse representation combined with the appearance model is proposed for the problems of serious occlusion,illumination variation and complex background. In the generative model,the sparse coding coefficients of the alignment operation together in the block structure and the sparse coefficients are combined with weighted processing,thus preserving the spatial structure of the target and local information to improve the robustness of target positioning precision and tracking,and a more robust appearance model is obtained. Compared with test video and different mainstream algorithms,the experiment proves that the algorithm has higher accuracy and robustness.

关 键 词：稀疏表示 目标追踪 外观模型 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]