Improving Vehicle Assistance Systems: Evaluation of Augmented Capabilities through Infrared Thermal Camera Integration

Beg, Mohammad Sojon; Ismail, M Yusri; Badrulhisam, N H; Siswanto, Ibnu; Gunadi, Gunadi · 2025 · Crossref

DOI: 10.15282/ijame.22.1.2025.20.0937

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 challenges associated with nighttime driving, where low visibility and poor obstacle detection significantly increase the risk of severe accidents. While various sensor technologies like LiDAR and visible-light cameras exist, they often struggle with low-light conditions, adverse weather, or high costs. The authors propose integrating infrared (IR) thermal cameras with deep learning algorithms to enhance vehicle assistance systems, specifically aiming to improve object detection accuracy in dark environments where traditional visual cameras fail. The research was conducted on public roads in Pekan, Pahang, Malaysia, covering a total distance of approximately 13.8 kilometers across different road geometries, including T-junctions, standard junctions, and straight roads. The experimental setup involved mounting both a conventional 12-megapixel visual camera and an infrared thermal camera on a vehicle. The system utilized the YOLOv8n deep learning model, trained on a custom dataset of annotated images containing vehicles, motorcycles, and traffic lights. Data processing was performed using a Google Colab platform with a Tesla T4 GPU. The methodology included grey-scale analysis of IR images to understand thermal distribution and temperature profiling to distinguish objects based on their heat signatures. Performance was evaluated using confusion matrices to calculate precision, recall, and accuracy, comparing the IR sensor’s output against the standard visual camera. The results demonstrated that infrared thermal sensors significantly outperformed conventional cameras in nighttime conditions. The IR camera achieved a detection accuracy of 0.98 for traffic lamps and 0.87 for motorcycles and vehicles. Furthermore, the IR system operated at a faster frame rate of 64.94 frames per second (fps) compared to 55.25 fps for the regular camera. Temperature analysis revealed distinct thermal signatures for different objects; for instance, motorcycles and vehicles registered temperatures between 25.8°C and 30.6°C, while traffic lamps ranged from 27.5°C to 28.5°C, allowing for reliable differentiation from the cooler background environment. Grey-scale analysis indicated that IR images provided clearer contrast in low-light scenarios, with temperature variations serving as a robust feature for object identification. The study concludes that integrating infrared thermal cameras with YOLOv8 algorithms offers a superior solution for nighttime object detection compared to traditional visual sensors. By leveraging thermal signatures, the system maintains high detection accuracy and processing speed regardless of ambient lighting conditions. These findings suggest that IR-based assistance systems can significantly enhance road safety by reducing collision risks in low-visibility environments, offering a viable alternative or complement to existing sensor technologies that are limited by darkness or weather conditions.

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.

• visibility analysis litigation