Driving Fitness Detection : A Holistic Approach For Prevention of Drowsy and Drunk Driving using Computer Vision Techniques

Abstract

This paper presents a holistic, non intrusive approach for driving fitness detection by checking drowsiness of the driver and loss of vehicle control due to potential influence of alcohol, using computer vision techniques of facial landmark detection and motion detection. A smart phone camera is used to capture a real-time video feed of the driver in the vehicle. The proposed system continuously analyzes this feed to detect drowsiness by checking for persistent eye blinks using a single scalar quantity, Eye Aspect Ratio (EAR) which characterizes eye opening. The system simultaneously checks for the orientation of the head and body, with respect to the steering wheel, to flag potential drowsiness or drunkenness or both, leading to loss of control of the vehicle. This is done by detecting motion using the Differential Images technique, which operate in real-time. Both of these analyses cumulatively yield a severity score which indicates whether the person is fit to drive or not. In case of positive results, an alarm is sounded and for suitable severity, the geographical location of the smart phone is sent to the respective kin and concerned authorities.