Thermal imaging for the detection of driver impairment: evidence from a high-fidelity driving simulator study

Göbel, Jan-Philipp; Mertens, Jan Cedric; Riener, Andreas · 2026 · Crossref

DOI: 10.1080/15389588.2026.2624006

archive: archived pipeline: cataloged verified

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

Summary

This study addresses the critical safety issue of driving under the influence of alcohol (DUI), which remains a leading cause of fatal traffic crashes globally. Motivated by increasing regulatory pressure from organizations like Euro NCAP for in-vehicle impairment detection, the research evaluates the feasibility of using thermal imaging as a noninvasive method to detect alcohol impairment. While previous studies relied on small samples in controlled laboratory settings, this work aims to provide the first large-scale assessment of thermal video-based DUI detection in a realistic, high-fidelity driving environment. The researchers conducted an experimental study with 120 participants (105 complete cases) in a high-dynamic driving simulator featuring a real vehicle and panoramic visual projection. Participants completed baseline and alcohol-impaired driving sessions targeting a blood alcohol concentration (BAC) of 0.6–1.0‰. Facial thermal data were captured using an Infra Tech PIR uc 605 camera mounted in the cabin. A novel processing pipeline was developed using the RetinaFace algorithm to extract temperature data from five specific facial regions: cheek, temple, ear, forehead, and nasal tip. To mitigate noise and environmental confounds, ambient cabin temperature was recorded and used to normalize facial readings via linear correction. Several machine learning classifiers, including Logistic Regression, Random Forest, Support Vector Machine (SVM), and Gradient-Boosting Models, were trained on these features using five-fold subject-wise cross-validation to prevent data leakage. The results demonstrated significant temperature changes in specific facial regions, particularly the cheek, ear, temple, and nasal tip, under the influence of alcohol. Among the evaluated models, Logistic Regression achieved the highest average classification accuracy of 62%, while SVM showed the most stable performance across validation folds. The model exhibited a conservative bias toward predicting the non-impaired class, which reduces the risk of false positives. The study confirmed that environmental conditions were stable and that the normalization process effectively isolated physiological changes from ambient temperature fluctuations. The significance of this work lies in demonstrating the feasibility of thermal imaging for in-vehicle DUI detection under ecologically valid conditions. By creating a unique dataset from 120 participants and developing a robust facial temperature processing pipeline, the study provides a comprehensive comparison of classification models. The findings suggest that thermal imaging is a promising complementary modality for future driver monitoring systems, offering a noninvasive alternative to traditional methods like eye movement analysis or driving behavior monitoring. This approach supports the development of reliable, real-time safety systems capable of detecting impairment in realistic automotive environments.

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 2026-06-09; verification: verified.

Topics

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

• drowsiness detection algorithms
• alcohol detection systems

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: validation psychometrics, tool software