A new robust visual tracking algorithm based on Transfer Adaptive Boosting

作者：Wu, Songtao; Zhu, Yuesheng^*; Zhang, Qing

来源：Mathematical Methods in the Applied Sciences, 2012, 35(17): 2133-2140.

DOI：10.1002/mma.2643

• visual tracking
• historical instances
• transfer AdaBoost

A new robust visual tracking algorithm, based on Transfer Adaptive Boosting, is proposed in this paper. The algorithm is able to produce a higher accuracy hypothesis for more effective and adaptive tracking by using certain useful historical instances of the video. The theoretical upper bound for the classification error of the samples and its corresponding theorem are also given. The theoretic analysis and simulation results show that the new method achieves a better performance compared with other methods in terms of avoiding drifting and miss-tracking, even with complex variations and alterations of the object's original appearance, such as occlusions, illuminations, and shape deformations.

• 出版日期2012-11-30
• 单位北京大学深圳研究生院