基于在线特征选择的实时压缩跟踪.pdfVIP

第 22 卷第3 期 光学精密工程 Vo l. 22 No.3 2014 年 3 月 Optics and Precision Engineering Mar.2014 文章编号 1004-924X( 20 14)03-0730-07 基于在线特征选择的实时压缩跟踪 毛 征\袁建建，吴珍荣，曲劲松，李红岩 (北京工业大学电子信息与控制工程学院，北京 100022) 摘要:针对原始压缩跟踪算法无法自适应选择特征易导致目标丢失或者漂移这一现象，提出一种基于在线选择特征的 实时压缩跟踪算法。首先，在初始化阶段生成两个互补的投影矩阵，利用该投影矩阵提取特征构造候选特征池。然后， 通过特征置信度度量策略实时从候选特征池中选择置信度较高的特征，并融合所选特征对应的置信度构造分类器。最 后.利用该分类器对候选样本做二元分类处理，获得最高分类器响应值的样本作为最后跟踪结果;而上一帧跟踪结果用 来在线更新特征池和分类器为后续跟踪做准备。本算法在多个公共测试视频序列与原始压缩跟踪算法进行了对比，结 果表明其跟踪精度和鲁棒性方面得到了改善，对于 320 pixel X 240 pixel 大小的视频平均处理帧速为 25 frame/s ，满足实 时跟踪要求。 关 键 词:在线特征选择;压缩感知;子区域特征;实时压缩跟踪 中图分类号:TP391 文献标识码:A dOÎ:I0. 3788/0PE. 0730 Real-time compressive tracking based on online feature selection MAO Zheng* , YUAN Jian-jian , WU Zhen-rong , QU Jin-song , LI Hong-yan (College of Electron lnformation and Cοntrol Engineering , Beijing University of Technology , Beij ing 100022 , China) Corresρonding author ， ιmail: maozheng@bjut.edu.cn Abstract: As original compressive tracking algorithms can not select adaptively the object futures , it will result in drifting or tracking lost when the object is occluded or its appearance changes. To ad dress this problem , this paper proposes a real-time compressive tracking algorithm based on online feature selection. First , two complementary projection matrixes were generated in an initial phase , and the projection matrixes were used to extract the feature to construct a feature poo l. Then , fea- tures with high confidence scores were selected from the feature pool by a confidence evaluation strate gy and these discriminating features and their corresponding confidence sco