A Video Based Fire Smoke Detection Using Robust AdaBoost.

Area of Science:

• Computer Vision
• Artificial Intelligence
• Fire Safety Engineering

Background:

• Early fire detection is crucial for minimizing damage and ensuring safety.
• Traditional smoke detectors have limitations in certain environments.
• Camera-based detection offers a supplementary approach for fire monitoring.

Purpose of the Study:

• To develop and evaluate a novel camera-based system for fire smoke detection.
• To enhance the accuracy and robustness of smoke detection algorithms.
• To investigate the effectiveness of combining static and dynamic visual features.

Main Methods:

• Feature extraction including texture, wavelet, color, edge orientation histogram, irregularity (static), and motion direction, change of motion direction, motion speed (dynamic).
• Development and application of a robust AdaBoost (RAB) classifier for improved training and classification.
• Extensive experimentation and validation on challenging datasets.

Main Results:

• The proposed system demonstrates satisfactory performance in fire smoke detection.
• The combination of static and dynamic features improves detection accuracy.
• The robust AdaBoost classifier enhances both training and classification capabilities.

Conclusions:

• Camera-based fire smoke detection using advanced feature extraction and classification is feasible and effective.
• The proposed robust AdaBoost classifier offers a promising solution for reliable smoke detection.
• Further research can explore real-world deployment and integration with existing safety systems.