一种嵌入式环境下的实时人脸跟踪方法  

REAL TIME FACE TRACKING IN EMBEDDED ENVIRONMENT

作　　者：嵇存美[1] 陈伟[1] 

机构地区：[1]浙江大学计算机学院,浙江大学浙江省服务机器人重点实验室,浙江杭州310027

出　　处：《计算机应用与软件》2011年第4期143-145,共3页Computer Applications and Software

基　　金：浙江省科技厅重点实验室建设项目(2008E1004)

摘　　要：针对复杂背景和光照变化条件下的人脸跟踪问题,提出了一种适于嵌入式环境的实时人脸跟踪方法。该方法使用帧差法检测出运动区域,在运动区域内进行AdaBoost人脸检测并进而提取人脸特征点,运用距离约束的金字塔形Lucas-Kanade方法跟踪人脸特征点。当特征点丢失较严重时,重新选取特征点进行跟踪。实验结果表明,该方法有较鲁棒的跟踪性能,在嵌入式环境下能够实时跟踪人脸。For solving the face tracking issue in complicated background and varying illuminating conditions,this paper presents a real-time face tracking method adapted to the embedded environment.Frame difference is used in the method to check movement region.Then AdaBoost face detection is applied in movement region and is followed with the extraction of face feature points.Finally,Pyramidal Lucas-Kanade method with distance constraint is used to track face feature points.When facial feature points are lost seriously,the method re-initialises the features and tracks.Experimental results show that our method has robust tracking performance.Besides,it can track faces in real time in embedded environment.

关 键 词：人脸跟踪 ADABOOST 人脸检测 帧差法 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论]