Driver drowsiness detection system

Dalvi, Spandan; Bankar, Sarthak; Pimpalkar, Aarti · 2026 · Crossref

DOI: 10.30574/ijsra.2026.19.2.0900

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the critical road safety issue of driver drowsiness, a leading cause of accidents that significantly impairs reaction time and decision-making. Motivated by the limitations of traditional vehicle safety systems, which monitor mechanical factors rather than the driver’s physical state, the authors propose a real-time, non-intrusive Driver Drowsiness Detection System. The study aims to provide a cost-effective solution using computer vision to continuously monitor driver alertness and prevent fatigue-related incidents, particularly during long-distance or night driving. The methodology employs a computer vision pipeline using OpenCV and Haar Cascade classifiers. The system captures live video input, converts frames to grayscale to reduce computational load, and applies histogram equalization to enhance contrast. Face detection identifies the driver’s region, selecting the largest detected face to isolate the subject. Eye detection is then performed on the upper half of the face (region of interest). The core detection mechanism relies on temporal analysis: the system tracks the duration of eye closure across consecutive frames to distinguish between natural blinking and prolonged closure indicative of fatigue. A novel fatigue percentage model quantifies drowsiness levels (0–30% alert, 30–70% moderate, 70–100% high) based on the ratio of eye closure time to a maximum threshold. When the closure duration exceeds a predefined threshold (set at 3.5 seconds in testing), the system triggers audio and visual alerts. Experimental results demonstrate the system’s effectiveness across various scenarios. Under normal conditions, the system achieved a precision, recall, and accuracy of 0.95. Performance remained robust but slightly decreased under challenging conditions: low light (accuracy 0.90), head movement (accuracy 0.89), and wearing glasses (accuracy 0.86). The fatigue percentage model successfully provided a continuous measure of alertness, allowing for proactive warnings before critical drowsiness levels were reached. The system effectively detected prolonged eye closure and triggered alerts as intended, though detection accuracy was noted to be lower in low-light environments or when reflective glasses caused occlusion. The study concludes that the proposed system is a reliable, cost-effective, and practical solution for real-time driver monitoring. By focusing on eye closure behavior and utilizing a graduated fatigue metric, the system offers a significant improvement over binary detection methods. The authors highlight its suitability for deployment in intelligent transportation systems, noting that while advanced deep learning methods exist, this approach balances accuracy with low computational requirements, making it accessible for widespread implementation. Future work may involve integrating advanced machine learning techniques to further enhance performance in adverse lighting and occlusion scenarios.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-nvidia on 2026-08-10; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• drowsiness detection algorithms
• dms validation
• gaze based attention detection
• drowsiness
• microsleep
• distraction detection algorithms

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).

• Empirical Findings: physiological data
• Methodological Resource: tool software, measurement protocol