Y. Zhyr, D. Radchuk. Improvement of the operation algorithm for automatic fire protection systems. P. 367-378.

Improvement of the operation algorithm for automatic fire protection systems

 

Zhyr Yehor

Dnipro University of Technology

https://orcid.org/0009-0001-3701-1710

 

Radchuk Dmytro

Dnipro University of Technology

https://orcid.org/0000-0001-8034-541X

 

DOI: https://doi.org/10.52363/2524-0226-2026-43-23

 

Keywords: fire safety, artificial intelligence, machine learning, fire detection, computer vision

 

Аnnotation

 

 

A systematic review and comparative analysis of computer vision architectures and deep ma-chine learning models yielded systematized data on the limitations of existing solutions based on convolutional neural networks. Their significant dependence on the direct line of sight of the fire source and high sensitivity to complex lighting changes was proven. Based on these shortcomings, the theoretical concept of an optimized two-stage algorithm for a computer vision system was devel-oped for early fire detection in enclosed industrial premises under conditions of hidden flames. It was established that fire localization efficiency can be significantly increased by combining direct flame recognition methods and analysis of indirect visual fire signs. The approach was justified by theoreti-cal modeling of the decomposition of short video fragments into frames with a fixed frequency for subsequent mathematical calculation of temporal fluctuations of overall brightness, specific glares, and dynamic light flickering. The expediency of using a sharp decrease in lighting intensity due to sudden smoke was recorded as a defining indirect sign for localization. The prospect of forming spe-cial time series of derived parameters instead of traditional processing of entire image pixels was justi-fied, followed by their transmission to the input of neural network models to recognize unique chaot-ic-periodic patterns of optical changes. Only the theoretical foundations and technological parameters of the proposed architecture were formed, while software development, neural network model train-ing, and their practical testing based on the datasets were defined as tasks for subsequent articles. The results confirm the possibility of future modernization of existing video surveillance systems without significant additional hardware costs.

 

References

  1. Vasconcelos, R. N., Franca Rocha, W. J. S., Costa, D. P., Duverger, S. G., Santana, M. M. M. d., Cambui, E. C. B., Ferreira-Ferreira, J., Oliveira, M., Barbosa, L. d. S., & Cordeiro, C. L. (2024). Fire detection with deep learning: A comprehen-sive review. Land, 13(10), 1696. doi: 10.3390/land13101696
  2. Ministry of Regional Development of Ukraine. (2014). Fire protection sys-tems: DBN V.2.5-56:2014. Kyiv.
  3. Kurasinski, L., Tan, J., Malekian, R. (2023). Using neural networks to detect fire from overhead images. Wireless Personal Communications, 130, 1085–1105. doi: 10.1007/s11277-023-10321-7
  4. Labed, S., Touati, H., Herida, A., Kerbab, S. (2023). An AI-based image recognition system for early detection of forest and field fires. European Journal of Forest Engineering, 9(2), 48–56. doi: 10.33904/ejfe.1322396
  5. Mozaffari, M., Li, Y., & Ko, Y. (2025). Automatic flame detection: Evalua-tion of deep learning algorithms using a custom thermal image dataset. Fire Technol-ogy, 61, 5789–5813. doi: 10.10694/025-01810-1
  6. Renuga, S., Pailochan, N., Rajkumar, Riyaz, M. (2023). Deep learning based fire detection and fire extinguisher using sound wave. International Journal for Re-search in Applied Science & Engineering Technology (IJRASET), 11(VIII), 605–609. doi: 10.22214/ijraset.2023.55033
  7. Refaee, E. A., Sheneamer, A., & Assiri, B. (2024). A deep-learning-based approach to the classification of fire types. Applied Sciences, 14(17), 7862. doi: 10.3390/app14177862
  8. Zell, O., Pålsson, J., Hernandez-Diaz, K., Alonso-Fernandez, F., Nilsson, F. (2023). Image-based fire detection in industrial environments with YOLOv4. Pro-ceedings of the 12th International Conference on Pattern Recognition Applications and Methods – ICPRAM, 379–386. doi: 10.5220/0011689400003411
  9. Xie, X., Chen, K., Guo, Y., Tan, B., Chen, L., Huang, M. (2023). A flame-detection algorithm using the improved YOLOv5. Fire, 6(8), 313. doi: 10.3390/fire6080313
  10. Zhang, Z., Tan, L., Robert, T. L. K. (2024). An improved fire and smoke detection method based on YOLOv8n for smart factories. Sensors, 24(15), 4786. doi: 10.3390/s24154786
  11. Maraveas, C., Loukatos, D., Bartzanas, T., Arvanitis, K. G. (2021). Appli-cations of artificial intelligence in fire safety of agricultural structures. Applied Sci-ences, 11(16), 7716. doi: 10.3390/app11167716
  12. Danyk, Y. O., Kirichenko, D. O. (2025). Application of artificial intelli-gence systems to solve fire safety problems. Mechanics and Mathematical Methods, 7(1), 152–172. doi: 10.31650/2618-0650-2025-7-1-152-172
  13. Wang, M., Jiang, L., Yue, P., Yu, D., Tuo, T. (2022). FASDD: An open-access 100,000-level flame and smoke detection dataset for deep learning in fire de-tection. Earth System Science Data Discussions. doi: 10.5194/essd-2022-394
  14. Singh, H., Ang, L. M., Srivastava, S. (2025). Active wildfire detection via satellite imagery and machine learning: an empirical investigation of Australian wild-fires. Natural Hazards, 121, 9777–9800. doi: 10.1007/s11069-025-07163-w
  15. Grammalidis, N., Dimitropoulos, K., Cetin, E. (2017). FIRESENSE data-base of videos for flame and smoke detection (Version v1) [Data set]. Zenodo. doi: 10.5281/zenodo.836748

 

Received by the editorial board: 10.03.2026

Accepted for publication: 14.04.2026

Date of publication (release): 31.05.2026