Camshift based real-time multiple faces match tracking

Abstract

A novel CAMShift based multiple faces tracking algorithm with weighted greedy tracking and match pursuing is proposed in this paper. The greedy tracking overcomes the disadvantage of CAMShift, which tends to lose target or reiterate tracking for overlapped targets. The match pursuing identifies the tracked targets by using additional information. In addition, a whole frame PDI is introduced to monitor the change of object number in the scene and re-initialize the tracked objects if the change happened. Experimental results indicate the proposed method can track multiple faces robustly in real time.