Boosting-based visual tracking using structural local sparse descriptors

Yangbiao Liu, Bo Ma*, Hongwei Hu, Yin Han

*此作品的通讯作者

科研成果: 书/报告/会议事项章节会议稿件同行评审

摘要

This paper develops an online algorithm based on sparse representation and boosting for robust object tracking. Local descriptors of a target object are represented by pooling some sparse codes of its local patches, and an Adaboost classifier is learned using the local descriptors to discriminate target from background. Meanwhile, the proposed algorithm assigns a weight value, calculated with the generativemodel, to each candidate object to adjust the classification result. In addition, a template update strategy, based on incremental principal component analysis and occlusion handing scheme, is presented to capture the appearance change of the target and to alleviate the visual drift problem. Comparison with the state-of-the-art trackers on the comprehensive benchmark shows effectiveness of the proposed method.

源语言英语
主期刊名Computer Vision - ACCV 2014 - 12th Asian Conference on Computer Vision, Revised Selected Papers
编辑Daniel Cremers, Hideo Saito, Ian Reid, Ming-Hsuan Yang
出版商Springer Verlag
522-533
页数12
ISBN(电子版)9783319168135
DOI
出版状态已出版 - 2015
活动12th Asian Conference on Computer Vision, ACCV 2014 - Singapore, 新加坡
期限: 1 11月 20145 11月 2014

出版系列

姓名Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
9007
ISSN(印刷版)0302-9743
ISSN(电子版)1611-3349

会议

会议12th Asian Conference on Computer Vision, ACCV 2014
国家/地区新加坡
Singapore
时期1/11/145/11/14

指纹

探究 'Boosting-based visual tracking using structural local sparse descriptors' 的科研主题。它们共同构成独一无二的指纹。

引用此