A hybrid machine learning model for forest wildfire detection using sounds

Abstract

Forest wildfires pose a significant threat to ecosystems, human settlements, and the global environment. Early detection is important for effective mitigation and response. This paper introduces a novel approach to forest wildfire detection by harnessing the unique sound signatures associated with wildfires. Our proposed model combines the strengths of deep learning techniques with heuristic optimization algorithms. The deep learning component focuses on recognizing the intricate patterns in the sound data, while the heuristic optimization, based on a Particle Sworm Optimization (PSO) algorithm, ensured the model’s adaptability and efficiency in diverse forest environments. Preliminary results indicate that our hybrid model outperforms traditional methods and existing machine learning models in terms of accuracy, sensitivity, and specificity, demonstrating robustness against ambient forest noise, ensuring fewer false alarms.