Smart real-time evaluation of tunnel fire risk and evacuation safety via computer vision
(2024) In Safety Science 177.- Abstract
The distribution of vehicles during a tunnel fire is a crucial factor that affects fire development and hazards, as well as the following evacuation and rescue operations. This work proposed a novel method using computer vision for assessing the real-time tunnel fire risk and evacuation safety by considering the classification and entry flow of vehicles. The proposed system utilizes YOLOv7 and DeepSORT for vehicle detection, classification, and tracking to enable a real-time digital twin for tunnel fire safety management. Vehicles are divided into 10 categories, in terms of their size, usage, number of passengers, fuel load, and peak fire HRR. After monitoring the vehicle flow at the tunnel portals, the real-time vehicle and fire load... (More)
The distribution of vehicles during a tunnel fire is a crucial factor that affects fire development and hazards, as well as the following evacuation and rescue operations. This work proposed a novel method using computer vision for assessing the real-time tunnel fire risk and evacuation safety by considering the classification and entry flow of vehicles. The proposed system utilizes YOLOv7 and DeepSORT for vehicle detection, classification, and tracking to enable a real-time digital twin for tunnel fire safety management. Vehicles are divided into 10 categories, in terms of their size, usage, number of passengers, fuel load, and peak fire HRR. After monitoring the vehicle flow at the tunnel portals, the real-time vehicle and fire load distribution are predicted. Then, the real-time tunnel fire scenarios and the safety of the evacuation process are evaluated based on the distribution of vehicles. The system is demonstrated in real road tunnels with traffic video cameras and exhibits a robust performance. The proposed vision-based real-time tunnel fire risk evaluation enables intelligent daily fire safety management and supports fire emergency response and decision-making.
(Less)
- author
- Zhang, Xiaoning
LU
; Chen, Xinghao
; Ding, Yifei
; Zhang, Yuxin
; Wang, Zilong
; Shi, Jihao
; Johansson, Nils
LU
and Huang, Xinyan
- organization
- publishing date
- 2024
- type
- Contribution to journal
- publication status
- published
- subject
- keywords
- Digital twin, Fire evacuation, Fire load distribution, Machine learning, Tunnel fire safety, Vehicle fire
- in
- Safety Science
- volume
- 177
- article number
- 106563
- publisher
- Elsevier
- external identifiers
-
- scopus:85195817515
- ISSN
- 0925-7535
- DOI
- 10.1016/j.ssci.2024.106563
- language
- English
- LU publication?
- yes
- id
- ddfd088b-4db5-446d-8dd3-1456f6eaa12b
- date added to LUP
- 2024-07-03 11:11:53
- date last changed
- 2024-07-03 11:12:49
@article{ddfd088b-4db5-446d-8dd3-1456f6eaa12b, abstract = {{<p>The distribution of vehicles during a tunnel fire is a crucial factor that affects fire development and hazards, as well as the following evacuation and rescue operations. This work proposed a novel method using computer vision for assessing the real-time tunnel fire risk and evacuation safety by considering the classification and entry flow of vehicles. The proposed system utilizes YOLOv7 and DeepSORT for vehicle detection, classification, and tracking to enable a real-time digital twin for tunnel fire safety management. Vehicles are divided into 10 categories, in terms of their size, usage, number of passengers, fuel load, and peak fire HRR. After monitoring the vehicle flow at the tunnel portals, the real-time vehicle and fire load distribution are predicted. Then, the real-time tunnel fire scenarios and the safety of the evacuation process are evaluated based on the distribution of vehicles. The system is demonstrated in real road tunnels with traffic video cameras and exhibits a robust performance. The proposed vision-based real-time tunnel fire risk evaluation enables intelligent daily fire safety management and supports fire emergency response and decision-making.</p>}}, author = {{Zhang, Xiaoning and Chen, Xinghao and Ding, Yifei and Zhang, Yuxin and Wang, Zilong and Shi, Jihao and Johansson, Nils and Huang, Xinyan}}, issn = {{0925-7535}}, keywords = {{Digital twin; Fire evacuation; Fire load distribution; Machine learning; Tunnel fire safety; Vehicle fire}}, language = {{eng}}, publisher = {{Elsevier}}, series = {{Safety Science}}, title = {{Smart real-time evaluation of tunnel fire risk and evacuation safety via computer vision}}, url = {{http://dx.doi.org/10.1016/j.ssci.2024.106563}}, doi = {{10.1016/j.ssci.2024.106563}}, volume = {{177}}, year = {{2024}}, }