Driver Drowsiness Detection System Using Computer Vision

Drowsiness detection is an important issue in road safety and other areas that require sustained attention. Some individuals fall asleep behind the wheel because of inadequate sleep or rest, or however fatigued they may feel-which is dangerous for self and other road users. Studies suggest that drowsiness is a significant contributory factor to road accidents. This research paper is based on a computer vision approach in which we use CNN for enhancing quality of frame for better facial detection and facial landmark analysis to detect Drowsiness. The method extracts important facial features, with a special emphasis on the lips and eyes, using a pre-trained 68-point facial landmark detection model. The method determines eye closure and yawning behaviors, two important markers of exhaustion, by examining variations in the Eye Aspect Ratio (EAR) and Mouth Aspect Ratio (MAR). Live video input is captured by a webcam, and OpenCV uses the Viola-Jones algorithm to detect faces in each frame. To detect drowsiness, we use these extracted features to analyze the conditions. An alert is sounded to wake the driver if the EAR stays below a predetermined threshold for an extended period of time or if the MAR above its threshold. This approach helps to reach the accuracy of 90.2-92% and detect drowsiness at real time without need of any wearable sensors.