融合遮挡感知的在线Boosting跟踪算法

摘要

提出融合遮挡感知的在线Boosting跟踪算法，该算法对跟踪结果实时进行遮挡检测，根据检测结果自适应调整分类器更新策略。该方式能够有效维护分类器特征池的纯净，提高算法在遮挡环境下的顽健性。实验结果表明，与传统的在线Boosting跟踪算法相比，改进的算法能有效解决目标遮挡问题。%Online Boosting tracking algorithm combined with occlusion sensing was presented. In this method, occlusion sensor was introduced to check the tracking results, and classifier updating strategy was adjusted depending on the occlu-sion checking results. By this way, the feature pool of the classifier can be kept pure, which will improve the tracking ro-bustness under occlusion. Experimental results show that compared with traditional Boosting tracking algorithm, im-proved algorithm can solve the problem of occlusion very well.