Vision-Based Real-Time Drowsiness Monitoring System Using OpenCV

Driver's sleepiness is a leading cause of crashes, and sometimes even leads to fatalities. In this paper, we propose an effective and real-time drowsiness detection framework to track and analyze the important facial indicators with the help of state-of-the-art computer vision and machine learning algorithms. It uses the OpenCV and the Dlib libraries for fast reliable facial landmark detections and tracking, allowing to monitor three relevant physical drowsiness markers, namely, eye closure, yawning, and head position. Eye Aspect Ratio (EAR) is calculated to recognize the blink period of eye, whereas Mouth Aspect Ratio (MAR) is employed to recognize yawn by detecting elongated mouth openings. Head pose estimation is also used to track attentiveness and detect nodding or head tilting as signs of sleepiness. The system runs in real-time and can accommodate changes in illumination, face poses, and driver appearance. Experimental studies show that its accuracy and reliability are quite high under different conditions, which indicates that it is a feasible and cost-effective non-invasive method for an ITS. The proposed design can shed light on the design of Advanced driver assistance system (ADAS) to bring down fatigue related crashes and thus to improve road safety.