中文摘要：

在视觉目标跟踪（video tracking）过程中,当跟踪图像存在背景杂波、图像噪声（如图像遮挡、图像快速移动）时,算法往往不能取得很好的图像追踪效果.为解决该问题,在经典L1-tracker追踪算法的基础上,针对目标遮挡、目标消失等严重影响跟踪效果的情况进行研究,提出加入拓展模板（固定模板和近况模板）的策略来提高跟踪精度和抗遮挡能力.固定模板保持追踪目标最初的图像特征,防止错误的追踪结果在模板更新时引入错误的特征,进而导致识别目标偏移.近况模板记录目标的最新跟踪结果,避免由于点模板的大量使用而造成遮挡的误识别.通过对多个标准数据集的实验测试,证明加入新策略的L1-tracker算法,在不破坏原有L1-tracker优势的基础上,显著地提升了L1-tracker算法应对遮挡问题的能力.

英文摘要：

In video tracking applications,most algorithms often fail in such conditions as object occlusion,disappearance. To address this issue,an improved L1-tracker algorithm is proposed. Expanded templates（ including the fixed template and the evolved template） are applied to improve the tracking accuracy and robustness.The fixed template keeps the original features of the target,and prevents the error introduced by false tracking result with the template update. The evolved template recording the latest tracking result is used to avoid massive use of trivial templates which will lead to the false recognition of occlusion. The experimental results on a number of standard data sets prove that the application of expended templates improves the ability of L1-tracker algorithm to deal with the occlusion problem.