抑制模板漂移的目标跟踪算法  

A Template-Drift-Resistant Object Tracking Algorithm

作　　者：潘吉彦[1] 胡波[1] 张建秋[1] 

机构地区：[1]复旦大学电子工程系,上海200433

出　　处：《电子学报》2009年第3期622-627,共6页Acta Electronica Sinica

基　　金：国家重点基础研究发展规划(973计划)项目(No.2006CB705700)

摘　　要：本文提出了一种解决目标跟踪中模板漂移现象的新算法.模板漂移现象指的是,由于不断进行模板更新,目标会逐渐移出模板,而背景物体则逐渐占据整个模板.本文为了解决这个问题,提出了一组能够精确描述模板漂移对于模板更新滤波器的影响的表达式,以使得考虑了模板漂移影响后的模板更新滤波器能够在空间及时间上自适应地实现及时更新目标外观与避免模板漂移之间的近似最优平衡.大量的包含不同目标外观变化类型的实景视频流的实验结果证实了本文算法的有效性.We propose an algorithm addressing the problem of template drift in object tracking. Template drift is a common phenomenon in which the target gradually shifts away from the template and the template is gradually occupied by background objects due to template utxtate.In order to solve this problem,this paper derives expressions that accurately evaluate the influence of template drift on the template-updating filter, so that the filter achieves a quasi-optimal balance between reducing template drift and keeping track of target appearance variations in a spafiotemporally adaptive manner. Experimental results based on a wide range of real-world video sequences containing diverse degrees of target appearance variations confirm the effectiveness of our algorithm.

关 键 词：目标跟踪 模板漂移 模板匹配 自适应Kalman滤波 

分 类 号：TN911.73[电子电信—通信与信息系统]