基于AdaBoost和Camshift的人脸检测与跟踪

Face Detection,Tracking Based on AdaBoost and Camshift

为提高视频序列中人脸跟踪的准确性,提出了一种使用Camshift算法,结合人脸检测实现实时、自动的人脸跟踪方法。利用图像的Haar特征,结合Ada Boost算法训练得到人脸分类器,进行人脸的检测来初始化人脸跟踪窗口;利用Camshift算法对人脸进行跟踪,并在跟踪过程中引入距离约束条件,使跟踪的结果更加稳定。实验结果表明,该方法能够对视频帧中出现的人脸实现自动跟踪,有效地改善了传统Camshift跟踪方法中窗口发散、跟踪丢失的问题。

This paper proposes a method based on Camshiftand face detection, to achieve a real - time, automatic tracking system for human face. The algorithm uses the Haar feature and AdaBoost to train a classification for face detection which used to initialize the tracking window. Face tracking will be done based on Camshift algorithm, and distance constrains is also used to make sure the tracking result is stable. Experiments show that the method in the paper is more effective than traditional facetracking ways based on Camshift.