Iot-Enabled Real-Time Driver Monitoring System for Drowsiness and Alcohol Detection Using Arduino and OpenCV

Rani, B Usha; Mohammad, Imran; Nishanth S.; Vikas, Naik; Ashok, P · 2026 · DOAJ

DOI: 10.1051/epjconf/202636302002

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 safety issue of driver impairment, specifically targeting drowsiness and alcohol intoxication, which contribute to approximately 30–35% of serious road crashes globally. The authors identify a gap in existing solutions, which typically address these risks in isolation or rely on intrusive methods like EEG or driver-initiated breath tests. To bridge this gap, the study proposes a real-time, non-intrusive, and low-cost system that concurrently monitors both impairment modalities using computer vision and embedded sensing. The system architecture integrates a Python/OpenCV software pipeline with an Arduino Uno hardware platform. Drowsiness detection utilizes the Eye Aspect Ratio (EAR) derived from dlib’s 68-point facial landmark regression, processing video at 8–15 frames per second. Alcohol detection employs an MQ-3 semiconductor gas sensor calibrated to the 0.05% Blood Alcohol Concentration (BAC) legal threshold. The system implements a three-stage hierarchical response model: Stage 1 triggers an auditory alert; Stage 2 adds controlled deceleration; and Stage 3 initiates autonomous left-lane parking with GSM emergency notification. The total hardware cost is estimated between USD 62 and 78. Experimental validation involved ten participants across varying lighting conditions, yielding 1,443 annotated test frames. The drowsiness detection algorithm achieved 92.4% accuracy, an F1-score of 89.7%, and an ROC AUC of 0.962 under normal lighting. Performance remained robust under nighttime simulation (87.5% accuracy) and with prescription glasses (89.3%), though it dropped to 74.2% when face masks occluded facial landmarks, triggering a fallback head-nod detection mode. The alcohol sensor demonstrated 100% detection reliability at legally intoxicated levels (BAC ≥ 0.08%) and 91.3% at borderline levels, with a 3.1% false-positive rate mitigated by baseline correction. The system exhibited a mean response latency of 3.2 seconds. The significance of this work lies in its simultaneous coverage of drowsiness and alcohol detection with autonomous corrective actuation at a fraction of the cost of deep learning-based alternatives. While the system offers practical deployability for broad vehicular safety applications, the authors note limitations including performance degradation under direct sunlight, sensor warm-up delays, and the need for real-vehicle validation of the autonomous parking sequence. Future work aims to integrate deep learning models for improved low-light robustness and V2X communication for coordinated traffic management.

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-prismaquant on 2026-06-26; verification: verified.

Topics

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

• alcohol detection systems
• drowsiness 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).

• Methodological Resource: tool software