<![CDATA[Driver fatigue and drowsiness are among the major factors causing traffic accidents, leading to decreased alertness and slower reaction time. This study aims to develop a computer vision-based driver drowsiness detection system by integrating the You Only Look Once (YOLO), Eye Aspect Ratio (EAR), and Mouth Aspect Ratio (MAR) methods, as well as to evaluate driving performance quantitatively. The system is designed to operate in real time using a camera as the visual data source, which is processed through face detection, facial landmark extraction, and the calculation of EAR and MAR parameters as indicators of eye condition and yawning activity. The experimental results show that the system is capable of classifying driver conditions into three categories: normal (alert), drowsy, and microsleep, with an average F1-score of 0.92. In addition, this study proposes the Alertness Index (AI) as a composite indicator calculated based on drowsiness frequency, eye closure duration, and yawning intensity to represent the driver’s alertness level more comprehensively. The system is also able to operate in real time with stable performance, supporting implementation in real-world conditions. The results demonstrate that the integration of YOLO, EAR, and MAR not only improves detection accuracy but also enables objective and data-driven driving performance evaluation. This system can be applied as a driver monitoring solution to improve road safety.]]>
Copyrights © 2026
