Integrating Color and Contour Analysis with Deep Learning for Robust Fire and Smoke Detection.

Area of Science:

• Computer Vision
• Artificial Intelligence
• Machine Learning

Background:

• Effective fire and smoke detection is critical for public safety in urban, industrial, and outdoor environments.
• Existing detection methods often struggle with dynamic conditions and varying illumination, leading to potential inaccuracies.
• There is a need for robust and adaptable detection systems capable of identifying both fire and smoke.

Purpose of the Study:

• To develop and evaluate a unique concatenated convolutional neural network (CNN) model for reliable fire and smoke detection.
• To enhance detection accuracy and reduce false positives/negatives through hybrid preprocessing techniques.
• To assess the model's performance against conventional and state-of-the-art methods using a challenging benchmark dataset.

Main Methods:

• A concatenated convolutional neural network (CNN) architecture was developed, integrating deep learning with hybrid preprocessing.
• Preprocessing methods included contour-based algorithms and color characteristic analysis to enhance Regions of Interest (ROIs).
• The model was trained and validated on the D-Fire dataset, which features diverse environmental conditions and illumination levels.

Main Results:

• The proposed CNN model achieved high accuracy (0.989) and recall (0.983) in detecting fire and smoke.
• Experimental results demonstrated superior performance compared to traditional methods and advanced YOLO-based approaches.
• The hybrid architecture effectively reduced false positives and false negatives, improving overall detection reliability.

Conclusions:

• The developed concatenated CNN model offers a highly accurate and resilient solution for fire and smoke detection.
• Its dual capability for detecting both smoke and fire enhances adaptability for various real-world safety applications.
• This study establishes a new benchmark for fire and smoke detection systems, paving the way for future advancements.