Face Tracking Algorithm Based on Online Random Forests Detection

Abstract

The Paper proposed a face tracking algorithm based on online random forests. The algorithm achieved detection-based tracking using online incremental extremely random forests detector, P-N learning is added to correct the detection error, and the dynamic target framework is proposed to maintain the online training set. The proposed algorithm integrated the results of the detector and the P-N learning, and the similarity to the dynamic target framework, thus the tracking position is confirmed. The experimental results show, the proposed algorithm could perform the tracking to any face rapidly and stably in a long-term period and complex background. Thus have practical worth in many application areas.