摘要： 针对复杂变化背景中的人脸跟踪问题,提出了一种基于Adaboost的人脸跟踪算法。该算法采用背景消除、肤色检测来提取可能包含 活动人脸的区域,用Adaboost算法做人脸验证,用Kalman滤波为运动模型实现对人脸的跟踪。实验表明,该算法具有快速、可靠的特点,速 度超过80fps,完全满足实时系统的需要。

关键词: 人脸检测, 人脸跟踪, 肤色检测

Abstract: In this paper, a face tracking algorithm for complex dynamic background is presented. This algorithm is based on the powerful Adaboost face detection algorithm, it uses background subtraction and skin detection to pick out the probably face areas, and uses Kalman filter as motion model to track the face. Experimental results show that it is fast and reliable, it achieves aboves 80fps,and can fully meet the requirements of real-time system.

中图分类号: 

• TP391.41