A Multi-Cue Integrated Adaboost Based Target Tracking Method

Longkui Jiang, Yuru Wang
2017 Innovative Computing Information and Control Express Letters, Part B: Applications  
In order to handle the difficulties of complex tracking conditions in real tracking problems, this paper proposed a novel Adaboost based tracking algorithm. In the proposed method, visual tracking is resolved as a two-class classification problem between fore-ground (the target) and its background. And the classification is resolved by designing an adaptive classifier. In detail, an on-line updated Adaboost classifier is designed to overcome the limitation of single classifier in tracking
more » ... cy and robustness. Specially, the problem of sample's label is considered, and the confidence is assigned to each input sample as its class, so as to realize more accurate classification. In addition, multiple salient features are integrated to represent target's appearance and provided to the classifier, so as to realize the robustness. The experimental results demonstrate the robustness and accuracy of the proposed Adaboost classifier in dealing with tracking problems.
doi:10.24507/icicelb.08.03.609 fatcat:plr2bbrsfzehjnjpzpor5t676q