Hits:

Indexed by:期刊论文

Date of Publication:2012-11-01

Journal:Journal of Convergence Information Technology

Included Journals:EI、Scopus

Volume:7

Issue:20

Page Number:209-217

ISSN No.:19759320

Abstract:In this paper, we propose a visual object tracking algorithm based on projection matrix. Firstly, the variance of pixels in the templates is computed and each template is decomposed into a more discriminative image and a less discriminative image. Then, projection matrix is learned according to the result of the decomposition. Finally, the object tracking results are obtained by using Bayesian MAP (maximum a posteriori probability) estimation. Compared with other popular methods, our proposed method has strong robustness to abnormal changes.