Eye detection for a real-time vehicle driver fatigue monitoring system

With the vast amount of vehicles on roads worldwide on any given time of day, the severity of fatigue related accidents have become a major concern. The obvious solution to prevent or at least decrease fatigue related accidents, is to ensure that the driver rests frequently. However, the simple fact of the matter is that frequent resting periods cannot be effectively enforced, and as a result there is a need for a system to monitor the level of driver fatigue in real-time. The ultimate goal of this research is to develop a camera-based driver fatigue monitoring system, centered around the tracking of driver's eyes, since the eyes provide the most information with regards to fatigue. The most critical aspect of eye tracking is to first accurately detect the eyes, and although a number of eye trackers have already been illustrated in the literature, the process of eye detection has seldom been described in much detail. Given a number of possible eye candidate sub-images, eye detection is in essence the classification of these sub-images as either eyes or non-eyes. To the knowledge of the authors, different classification techniques have not been directly compared for the purpose of eye detection, and therefore the aim of this paper is to evaluate different classification techniques to determine which technique will be the most suitable for a driver fatigue monitoring system. The classification techniques that have been considered are artificial neural networks (ANN), support vector machines (SVM) and adaptive boosting (AdaBoost). Results have shown that AdaBoost will be the most suitable eye classification technique for a real-world driver fatigue monitoring system.