Computer vision based fire detection and automated suppression system

This project presents a real-time, computer vision-based fire detection and suppression system designed to enhance safety across diverse environments. The system employs a robust multilayered detection algorithm that begins with color-based flame clue extraction, followed by three refined filtration stages: Centroid Analysis, Histogram Analysis, and Variance Analysis. Validated on benchmark fire video datasets, the algorithm achieved an impressive accuracy of 95.26%, with a true positive rate of 91.61% and a true negative rate of 98.91%, outperforming existing methods by an average of 7.95% in accuracy and 9.43% in precision.

The system also features a unique fire localization mechanism that pinpoints the fire’s position within the video frame. This information is used to trigger an Arduino-based suppression module, enabling fully autonomous fire detection and extinguishing. Real-world laboratory tests confirmed the system’s reliability with a precision of 99.51% and a recall of 95.93%, making it suitable for personal, industrial, and environmental fire safety applications.

| Paper |

• Mondal, Md Safwan, et al. "Automating fire detection and suppression with computer vision: a multi-layered filtering approach to enhanced fire safety and rapid response." Fire Technology 59.4 (2023): 1555-1583.