For intelligent surveillance a system should locate, track and identify individuals. In this paper, we propose a surveillance system with face recognition that performs human detection and face tracking and recognition with static and Pan-Tilt-Zoom(PT...
For intelligent surveillance a system should locate, track and identify individuals. In this paper, we propose a surveillance system with face recognition that performs human detection and face tracking and recognition with static and Pan-Tilt-Zoom(PTZ) camera. Human are located by background subtraction method in the static camera. The proposed background modelling is insensitive to shadow and extracts the accurate human silhouette using MeanShift segmentation. Using homography between the two cameras face regions of high resolution are acquired in the PTZ camera using frontal face detection based on Viola and John''s method and tracked by CAMShift and adaptive skin color estimation. For robust facial feature extraction we combine Generic and Person Specific Active Appearance Model(AAM) and the feature is used to recognize a face. The experiments shows that the proposed face tracking and recognition algorithms are robust to head pose and motion comparing to face normalization method using 2D eye position.