VIDEO SURVEILLANCE USING FACIAL FEATURES-BASED TRACKING

In this work, we present a real time system for facial features detection and tracking in image sequences. We are interested in developing a human tracking system that can improve human-computer interaction and benefit video surveillance problems by making it invariant to rotation, illumination and subject’s movement. However, since human body and face movements are very complicated to detect and track, all available cues that can narrow the search space should be considered. This paper describes a novel strategy for both face tracking and facial feature detection. Face detection is important because it reduces the search space and consequently saves time for further face processing, e.g., recognition or transmission. Moreover, facial feature detection enables face normalization which leads to size invariant face recognition. The proposed face tracker resembles human perception in that, initially, it utilizes motion as the major cue and thereafter, searches for the eyes in the areas likely to contain human faces. The presence of a face is determined using an eye tracker. Eyes are important facial features due to their relatively constant interocular distance. In this work, efficiency improvement focuses on two points: reducing template matching area and speeding up the matching process. Our method initially detects two rough eye candidate regions using a feature based method. All other processes are thereafter performed inside the candidate regions. In addition, we can evaluate the size of eye template according to the size of the regions. Altogether, the proposed method combines the accuracy of template based methods and the efficiency of feature based methods in the visual spectrum. To prove the effectiveness of this approach, we performed comparative experiments using real video images. We achieved a real time detection accuracy of about 96%.