基于在线词典学习视觉跟踪算法的实现  

Realization of The Visual Tracking Algorithm based on Online Dictionary to Learn

作　　者：刘师师[1] 徐学军[1] 李敏[1] 

机构地区：[1]长沙理工大学物理与电子科学学院,长沙410114

出　　处：《智能计算机与应用》2014年第3期90-91,94,共3页Intelligent Computer and Applications

摘　　要：针对L1跟踪算法仅考虑了稀疏全局表达的不足,基于在线词典学习的视觉跟踪算法将词典池化的优化方法引入到稀疏表示模型中,充分利用了目标模板的局部结构和空间信息,并采用了一种有效的自适应更新方法,本文根据该算法实现了视觉跟踪处理。通过实验对比IYT、MIL和L1这三种目前较先进的跟踪算法,验证了本跟踪处理器的先进性和有效性。For the problem L1 tracking algorithm only considering sparse global express, pooling optimization method in the visual tracking algorithm based on online learning dictionary is introduced into the sparse representation model, which makes full use of local structure and spatial information. After that, using an effective adaptive updating method, this paper realizes the visual tracking processing. By contrast the experiment of 1YT, MIL and L1, these three advanced tracking algo- rithm, the effectiveness of this tracking processor is verified.

关 键 词：视觉跟踪 稀疏表示 词典池化 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]