Vision-based detection algorithm for monitoring dynamic change of fire progression

• Published in: Journal of big data Vol. 12; no. 1; pp. 134 - 19
• Main Author: Suh, Yongyoon
• Format: Journal Article
• Language: English
• Published: Cham Springer International Publishing 01.12.2025; Springer Nature B.V; SpringerOpen
• Subjects: Accident prevention; Accidents; Accuracy; Algorithms; Artificial neural networks; Big Data; Communications Engineering; Computational Science and Engineering; Computer Science; Computer vision; Construction; Convolution neural network; Data Mining and Knowledge Discovery; Database Management; Deep learning; Emergency response; Fatalities; Fire incidents; Fire prevention; Fire progression; Information Storage and Retrieval; Machine learning; Mathematical Applications in Computer Science; Networks; Neural networks; Patterns of fire events; Sensors; Supervisors; Vision-based pattern; Fire progression; Convolution neural network; Patterns of fire events; Vision-based pattern; Fire incidents
• ISSN: 2196-1115; 2196-1115
• DOI: 10.1186/s40537-025-01211-9

Fire incidents in industrial settings often result in hundreds of worker fatalities, severe injuries, and substantial financial losses. To minimize the impact of industrial fire accidents, it is essential to establish response strategies that adapt to fire progression. This study aims to define vision-based patterns of fire events to identify multiple objects that contribute to different types of fire accidents. To achieve this, a convolutional neural network (CNN) based on deep learning is applied to detect fire events through vision-based patterns. Flames and smoke are trained as multiple objects to recognize fire event patterns, while their size and position are visualized to assess fire severity. The results offer valuable insights for industrial supervisors, academic researchers, and fire accident investigators, enhancing their understanding of fire incidents and their progression within industrial environments. This vision-based approach provides a more effective method for detecting and forecasting fire development, contributing to improved fire safety and emergency response strategies.