Abstract
Well-designed and effective in-mine robots can expedite miner self-rescue during emergencies and reduce fatalities. These in-mine robots for miner self-rescue can carry out diverse tasks such as scouting (including object detection and autonomous navigation), and payload delivery. However, robots that can effectively detect humans in a dark underground mine do not yet exist. This paper investigates challenges in the design of object detection algorithms for in-mine robots using thermal images, especially to detect people in real-time, in low-light conditions. The research team collected 500 thermal images in the Missouri University of Science & Technology Experimental Mine with the help of student volunteers using the FLIR TG 297 infrared camera, which they pre-processed and split into training and validation datasets with 450 and 50 images, respectively, using tenfold cross-validation. The research retrained two state-of-the-art, real-time object detection models, namely YOLOv5 (You Only Look Once version 5), and YOLOv8 (You Only Look Once version 8), using transfer learning techniques on the training dataset for 50 epochs. On the validation dataset, the re-trained YOLOv8 outperforms the re-trained YOLOv5. These trained models as well as the original models were then applied to a simulated mine fire emergency to assess their performance in emergency situations. The results show that the mAP of the YOLOv8 variants improved drastically after transfer learning. For instance, YOLOv8n improved from 13.90 to 74.70%. And that of the YOLOv5 variants improved significantly. For instance, YOLOv5n improved from 42.10 to 68.30%.
Recommended Citation
C. Addy et al., "YOLO-Based Miner Detection Using Thermal Images In Underground Mines," Mining, Metallurgy and Exploration, Springer, Jan 2025.
The definitive version is available at https://doi.org/10.1007/s42461-025-01249-6
Department(s)
Computer Science
Second Department
Mining Engineering
Keywords and Phrases
Object detection, Miner detection, Thermal image, YOLO, and Underground mine
International Standard Serial Number (ISSN)
2524-3470; 2524-3462
Document Type
Article - Journal
Document Version
Citation
File Type
text
Language(s)
English
Rights
© 2025 Springer, All rights reserved.
Publication Date
01 Jan 2025
Comments
Centers for Disease Control and Prevention, Grant #U60OH012350 - 01–00