Face Detection And Tracking Fusing Texture Information And Kalman Filtering

Biometric recognition technology,which can be widely applied to many fields such as defense security,intelligent surveillance,and electronic commerce etc.,plays an important role in human computer interaction(HCI)technology.Among biometric recognition technologies,face recognition is paid attention to because of its advantages such as stability,friendliness,and convenient collection etc.Images containing faces are an indispensable part of face processing research such as HCI based on intelligence vision,face recognition,face tracking,pose estimation,and expression recognition etc.For constructing full-automatic face image analysis system,robust and effective face detection and tracking algorithms are urgently required.In this thesis aiming at the shortcomings of traditional continuously adaptive mean shift(Cam Shift)tracking method some strategies are used,and a full-automatic face detection and tracking method is proposed.Some detailed work is as follows:(1)The development and present situation of face detection and tracking technologies are detailedly introduced.Face detection technologies are summarized four classes including knowledge-based method,feature invariant-based method,template matching-based method and appearance-based.Face tracking technologies are classified into detection-based method and moving objects tracking-based method.Difficult problems about face detection and tracking technologies are indicated.(2)Aiming at adverse effect of feature model of traditional Cam Shift method such as similar color background,occlusion,and light change,fusing texture information of local binary pattern(LBP)and varied local edge pattern(VLEP)to construct feature model based on original color feature is proposed,which is beneficial to realizing robust and steady tracking.(3)Aiming at adverse effect of sharp pose change of object in traditional Cam Shift method,kalman filter is used to provide feedback and prediction information.Fusion feature model is combined at the same time.A face detection and tracking method which can overcome the adverse effect of sharp pose change is proposed.Experimental results verified the effectiveness and feasibility.