Real-Time Eye Locating and Tracking for Driver Fatigue Detection

Assistant driving systems have attracted more and more attention during recent years. Among them fatigue detection plays a key role because of its close relationship with accidents. In this paper, we propose a novel method which uses eye locating and tracking technique to detect driver fatigue. The present method consists of four steps. First, we employ Adaboost and Haar-like features to construct a robust classifier which can detect eye corner points. Second, we use extended parabolic Hough transformation to construct the parabola curves of upper and lower eyelid. Then, particle filter algorithm is used to track eye corner points in video sequences. Finally, the driver fatigue state is estimated through computing the frequency of eye opening and closing intervals. Experimental results from real environment datasets are given in our discussion as well.