A review of the approaches for enhancing driver safety through camera-based monitoring

Road accidents caused by human error continue to result in a rising number of serious injuries and fatalities each year. Automated Driving Assistance Systems (ADAS) offer a promising approach by continuous monitoring of drivers to promptly identify and address unsafe behaviors. A further technical challenge is solely relying on camera sensors to determine crucial human deficiencies such as drowsiness and distraction. In our research, we provide a review of relevant papers from the field of computer vision for driver monitoring. Most of these approaches rely on Machine Learning (ML) algorithms and Convolutional Neural Networks (CNNs) for both image processing and driver state classification. Our findings indicate that Hidden Markov Models (HMMs) and CNNs consistently outperform other techniques in terms of accuracy. In conclusion, we present our perspective on the essential features that a comprehensive driver monitoring system (DMS) should include.