Real-time Driver Monitoring System Based on Visual Cues

Road accidents have become a concerning event. Every year countless people die and many others suffer from fatal and non-fatal injuries due to road accidents. A lot of these accidents are caused by the driver's distraction of attention as well as micro sleep. There is no doubt that driver's drowsiness is one of the main reasons for road accidents. People's behavior changes when they feel drowsy and show visual symptoms like yawn. So, drowsiness can be detected by analyzing driver's behavior avoiding serious injuries. Therefore, in this paper a real-time driver monitoring system has been developed which analyses the driver's state in real-time and gives alert when drowsy state increases. The system gives three signal which includes safe, intermediate and drowsy state. The system observes eye and mouth state to detect drowsiness. It measures Eye Aspect Ratio, Eye Closure Ratio and yawn rate. It also detects head pose for monitoring driver's attention. The system achieves 97% accuracy on eye blink detection, 96% accuracy on yawn detection and 63.4% accuracy on head pose estimation. In case of testing the combined system, XGBoost outperforms other classifiers. The system runs at 15 fps speed on a computer which is fast enough for real-time surveillance.