Early Detection of Forest Fire Using Mixed Learning Techniques and UAV.

Area of Science:

• Environmental Science
• Computer Science
• Artificial Intelligence

Background:

• Increasing frequency of forest fires due to deforestation and global warming poses significant threats to ecosystems.
• Effective forest fire detection is crucial for timely management and mitigation efforts.
• Technological solutions are needed to address the growing challenge of forest fire management.

Purpose of the Study:

• To propose a cost-effective deep learning model for predicting forest fires.
• To develop an integrated system utilizing unmanned aerial vehicles (UAVs) for forest fire detection.
• To enhance forest fire management through advanced prediction techniques.

Main Methods:

• A mixed learning technique combining YOLOv4 tiny and LiDAR was developed.
• The model was deployed on an onboard unmanned aerial vehicle (UAV) for forest surveillance.
• The system was trained on data from both dense and rainforest environments.

Main Results:

• The proposed model achieved a classification time of 1.24 seconds.
• High accuracy of 91% and an F1 score of 0.91 were recorded.
• The system demonstrated real-time data transmission and 3D modeling capabilities.

Conclusions:

• The developed deep learning model offers a cost-effective solution for forest fire prediction.
• The UAV-based system provides efficient and accurate real-time forest fire detection.
• This approach outperforms traditional methods in forest fire detection and prediction.