International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 12 Issue: 12 | Dec 2025
p-ISSN: 2395-0072
www.irjet.net
Facial Landmark-based Real-Time Driver Drowsiness Detection. Rohit Lodhi¹, Shivansh A. Mishra¹, Saumya Dawande¹, Shivansh Mishra¹, Rajeev Raghuwanshi² ¹Department of Computer Science and Engineering (AI & ML), Oriental Institute of Science and Technology, Bhopal, Madhya Pradesh, India ²Assistant Professor, Department of Computer Science and Engineering (AI & ML), Oriental Institute of Science and Technology, Bhopal, Madhya Pradesh, India ---------------------------------------------------------------------***--------------------------------------------------------------------computer vision techniques. The system employs Eye Abstract - Driver drowsiness is a significant contributor to
Aspect Ratio (EAR) and Mouth Aspect Ratio (MAR) to identify prolonged eye closure and yawning, respectively. To prioritize computational efficiency and real-time deployability on consumer-grade hardware, fixed threshold values are used rather than personalized adaptive models. Although individual adaptation is not addressed, this design choice enables stable performance and reproducibility under controlled conditions.
road traffic accidents worldwide, accounting for an estimated 20–30% of severe crashes. This paper presents a non-intrusive, real-time driver drowsiness detection system based on facial landmark analysis using computer vision techniques. The proposed method computes the Eye Aspect Ratio (EAR) and Mouth Aspect Ratio (MAR) to detect prolonged eye closure and yawning, which are wellestablished visual indicators of fatigue. The system is designed for real-time operation on consumer-grade hardware using a standard RGB camera. Experimental evaluation conducted on a custom dataset demonstrates a drowsiness detection sensitivity of 91% while maintaining a processing speed of 24–28 frames per second. The results indicate that the proposed approach provides a practical and cost-effective solution for real-time driver monitoring under controlled lighting conditions.
The primary objective of this study is to demonstrate that a rule-based, facial landmark-driven approach can achieve reliable drowsiness detection while maintaining real-time performance without the need for specialized sensors. The focus of this work is on practical implementation, performance evaluation, and deployment feasibility rather than the development of new visual features or learningbased models.
Keywords— Computer Vision, Drowsiness Detection, Eye Aspect Ratio, Facial Landmarks, Mouth Aspect Ratio
II. LITERATURE REVIEW Driver drowsiness detection methods can be broadly classified into three categories: physiological monitoring, vision-based analysis, and vehicle behavior assessment.
I. INTRODUCTION Driver drowsiness is a major factor contributing to road traffic accidents, particularly during long-distance driving and nighttime travel. Fatigue impairs reaction time, attention, and decision-making ability, often without the driver being consciously aware of the degradation in performance. As a result, drowsy driving remains difficult to detect and prevent using conventional safety mechanisms.
Physiological approaches, such as electroencephalography (EEG) and electrocardiography (ECG), provide high detection accuracy by directly measuring brain and cardiovascular activity. However, the requirement for invasive sensor placement limits their practicality in real-world driving scenarios [2]. The use of electrodes reduces driver comfort and restricts widespread adoption in everyday driving environments
In recent years, vision-based driver monitoring systems have gained attention due to their non-intrusive nature and relatively low cost compared to physiological sensing approaches. However, many existing systems rely on indirect vehicle-based indicators or rigid thresholding mechanisms that do not generalize well across different drivers and operating conditions. Vehicle-behavior-based methods often detect fatigue only after driving performance has already deteriorated, limiting their effectiveness for early intervention.
Vision-based systems analyze observable facial cues such as eye closure duration, blink frequency, and yawning behaviour. The PERCLOS measure has received extensive confirmation as an effective measure of driver alertness [3]. Soukupova and Cech [1] introduced the Eye Aspect Ratio (EAR) for real-time blink detection using facial landmarks, demonstrating its effectiveness in identifying prolonged eye closure. On the same note, the yawning behaviour is also identifiable in terms of Mouth Aspect Ratio (MAR) and as such, the features can be effectively used in real-time fatigue detection. These metrics provide
This paper presents a real-time driver drowsiness detection system based on facial landmark analysis using
© 2025, IRJET
|
Impact Factor value: 8.315
|
ISO 9001:2008 Certified Journal
|
Page 1028