A Real-Time Driver Drowsiness Detection Using Computer Vision and Machine Learning

Driver drowsiness is one of the main factors that lead to road accidents, thereby causing serious injuries and deaths. Driver monitoring systems that are traditionally adopted include physical or physiological measures like EEG, ECG, or wearables, which are cumbersome, expensive, and not suitable for practical use. In order to overcome such disadvantages, a new non-intrusive approach for driver drowsiness detection based on their vision is presented. In the proposed approach, Haar cascade classifiers are utilized, resulting in effective facial feature detection like face, eye, and mouth, based on live webcam feed. All detected regions are further processed using a YOLO-based classification model for classifying and deciding the driver’s status like eye-open, eye-closed, yawn, and combined conditions of eye closed with a yawn. On the basis of these classifications, the level of alertness of the driver is effectively determined. In case of drowsiness, a three-level alert system is implemented in the system, viz. buzzer sound, voice, and video recording. If drowsiness persists, the recorded video is saved and sent via email.