Drowsy driving causes road accidents with severe consequences, and drowsiness causes sensory and cognitive impairments. Current systems, however, often base detection on vehicle data and ignore the direct behavior of the driver. Through this project, a non-invasive camera system is introduced that detects driver drowsiness using computer vision to monitor facial features, especially eye and mouth closure. The system detects the driver\'s faces, extracts facial landmarks, computes the Eye Aspect Ratio (EAR), and computes the Mouth Opening Ratio (MOR) using static and dynamic thresholding techniques. Abnormal EAR or MOR or a hand gesture such as head nodding or mouth covering will immediately trigger an alert for warning the driver. The system thus provides an inexpensive and easy-to-implement solution for increased road safety.
Introduction
Drowsy driving is a dangerous condition caused by sleepiness and impaired alertness, often due to insufficient sleep, sleep disorders, medication, alcohol, stress, fatigue, or long hours of driving. It reduces vigilance and reaction time, increasing the risk of accidents.
To address this, a deep neural network model called Faster Region Convolutional Neural Network (FRCNN) is proposed for real-time detection of driver behavior. The system uses multi-stream video inputs (side and front views, optical flows) and processes these to detect distracted or drowsy driving.
The system includes several components:
A driver face enrollment and recognition module capturing and processing face images.
A distracted driver monitoring system that detects facial cues such as eye closure and yawning using CNN and feature extraction.
A warning system that alerts the driver and notifies vehicle owners via alarms, voice alerts, steering wheel signals, and emails.
The architecture supports robust detection under various lighting conditions and real-time video processing. Multiple test cases confirm the system's effectiveness in detecting distracted or drowsy behavior, adapting to different environments, handling multiple drivers, and integrating alert notifications reliably.
Conclusion
The facial landmark-based system for detecting driver fatigue offers an effective and non-intrusive solution to enhance road safety. By continuously monitoring key facial features such as eye closure, yawning, and head position, the system can accurately identify signs of drowsiness or inattention. Leveraging advanced image processing techniques and machine learning models, it ensures real-time detection and timely alerts through visual, audio, and haptic signals. This approach not only minimizes the risk of accidents caused by fatigue but also ensures proactive intervention before the driver\'s condition worsens. The integration of such a system into vehicles contributes significantly to intelligent transportation systems and supports the broader vision of safer roads and reduced human-error-related incidents
References
[1] J. D. Slater, \"A definition of drowsiness: One purpose for sleep?\", Med. Hypotheses, vol. 71, pp.641-644,Nov.2008,[online]
[2] Available: http://linkinghub.elsevier.com/retrieve/pii/S0306987708002946.
[3] \"Drowsy driving and automobile crashes\", vol. 808, pp. 707, 1998.
[4] L. S. Arnold and B. C. Tefft, \"Prevalence of self-reported drowsy driving United States: 2015\", 2015, [online]
[5] Available: https://www.aaafoundation.org/sites/default/files/PrevalenceOfSelf-ReportedDrowsyDrivingReport.pdf.
[6] Violence and Injury Prevention and World Health Organization Global Status Report on Road Safety 2015, 2015.