Intelligent fire detection in agriculture using machine learning and embedded systems for risk prevention and...

Area of Science:

• Agricultural Engineering
• Computer Science
• Environmental Science

Background:

• Conventional fire monitoring systems often lack efficacy in remote areas due to limited internet connectivity and infrastructure.
• Increasing fire risks in agricultural and rural settings necessitate advanced, independent monitoring solutions.
• The need for real-time hazard detection and risk prediction is critical for protecting natural resources and farm resilience.

Purpose of the Study:

• To design and implement an autonomous fire detection system adaptable for areas with minimal internet access.
• To enhance fire hazard classification and prediction accuracy using machine learning algorithms.
• To ensure data reliability through anomaly detection and validate system performance.

Main Methods:

• Development of an embedded system using Raspberry Pi 3 B+ with smoke and flame sensors.
• Application of Machine Learning (ML) algorithms, specifically Random Forest and Logistic Regression, for hazard classification and anomaly detection.
• Evaluation of model performance using confusion matrices and five-fold stratified cross-validation.

Main Results:

• The Random Forest model demonstrated superior performance with an average accuracy of 0.9860 ± 0.0172, an F1 score of 0.9740 ± 0.0319, and a recall of 0.9733 ± 0.0327.
• Logistic Regression achieved an average accuracy of 0.9446 ± 0.0600, F1 score of 0.9173 ± 0.0744, and recall of 0.9250 ± 0.0608.
• Anomaly detection techniques successfully identified and corrected potential sensor measurement errors, ensuring data integrity.

Conclusions:

• The developed autonomous system provides a reliable and efficient fire detection solution for internet-deprived rural and agricultural areas.
• The Random Forest model is highly effective for classifying fire risk levels and ensuring timely alerts, significantly reducing agricultural fire risks.
• This research contributes to the transition towards smart agriculture by enhancing risk management, promoting sustainability, and increasing farm resilience to natural disasters.