Real-time Driver Monitoring System with Facial Landmark-based Eye Closure Detection and Head Pose Recognition
DOI: 10.21203/rs.3.rs-3223799/v1
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 presents a real-time Driver Monitoring System (DMS) designed to detect driver drowsiness and inattention using infrared (IR) camera video data. Motivated by increasing regulatory mandates for driver monitoring in semi-autonomous vehicles and the limitations of RGB cameras in varying lighting conditions, the authors developed a system that relies solely on facial landmark estimation. The system addresses two primary safety risks: drowsy driving, identified through eye closure, and inattentive driving, identified through head pose analysis. The methodology employs a custom IR camera developed by CANLAB, capable of operating in low-light conditions with a resolution of 1280x800 pixels. The processing pipeline begins with face detection using the YOLOv7 algorithm, chosen for its balance of speed and accuracy. Following detection, facial landmarks are extracted using Kazemi’s random forest-based algorithm, which is optimized for embedded environments. The system then utilizes these landmarks for two distinct modules. For inattention detection, head pose is estimated using the OpenCV solvePnP algorithm, which calculates rotation and translation vectors to determine gaze direction. For drowsiness detection, the system applies a novel eye-closure recognition filter. This filter binarizes the eye region based on pixel brightness thresholds and scans for specific closed-eye patterns, avoiding the computational overhead and false positives associated with deep learning-based eye state classifiers. The system was evaluated using a custom IR dataset comprising normal and drowsy driving scenarios under both normal and low-light conditions. Face detection achieved 100% precision and recall rates of 98.4% (normal) and 99% (low-light). The drowsiness detection module, which classifies a driver as drowsy if eye closure persists for 50 frames with over 40 detected pixels, achieved 100% accuracy and 99.3% precision. The inattention module, which flags inattention if the head does not face forward for more than 50 frames, also demonstrated over 99% recognition performance. In terms of computational efficiency, the system processed video at 20–25 frames per second (FPS) on a desktop CPU and 10 FPS on an embedded NVIDIA Xavier board using Tiny YOLO, confirming its viability for real-time automotive applications. The significance of this work lies in its demonstration of a robust, computationally efficient DMS that functions effectively in challenging lighting environments without requiring vehicle ECU data. By combining lightweight landmark estimation with specific heuristic filters for eye and head analysis, the system provides a practical solution for enhancing driver safety. The authors conclude that this approach can be integrated with hardware sensor data in future iterations to create a comprehensive monitoring system, offering a viable path for commercial implementation in vehicles requiring continuous driver attention.
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.
| Stage | Outcome | Tool | Model | Prompt | Attempts | Completed |
|---|---|---|---|---|---|---|
| discover | success | Crossref | — | — | 1 | 2026-08-09 |
| archive | success | canonical_url | — | — | 1 | 2026-08-09 |
| extract | success | pdftotext | — | — | 4 | 2026-08-10 |
| clean | success | clean | — | — | 2 | 2026-08-10 |
| chunk | success | chunk | — | — | 2 | 2026-08-10 |
| embed | success | embed | Qwen/Qwen3-Embedding-8B | — | 2 | 2026-08-10 |
| promote | success | — | — | — | 1 | 2026-08-09 |
| summarize | success | llm | qwen3.6-27b-nvidia | summ-v5 | 2 | 2026-08-10 |
| tag | success | vector_similarity | — | — | 17 | 2026-08-11 |
| verify | partial | — | — | — | 2 | 2026-08-10 |
Summary generated by qwen3.6-27b-nvidia on 2026-08-10; verification: verified_with_issues.
Topics
Ranked by relevance to this paper. Hover a topic for its definition.
- drowsiness detection algorithms
- dms validation
- gaze based attention detection
- distraction detection algorithms
- dms regulation
- drowsiness
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