Comparison Analysis of Satellite Images for Wildfire Detection using Convolutional Neural Network (CNN)

Abstract

This research aims to compare the effectiveness of two datasets, the Kaggle Wildfire Prediction Dataset, and the SAT-4 dataset, in detecting wildfires using Convolutional Neural Networks (CNNs). The study analyzes satellite images relevant to wildfire incidents from both datasets. The Kaggle dataset, which uses satellite images to detect historical trends, has a test accuracy of 93.76%. In contrast, the SAT-4 dataset, used by Google Research, has a higher accuracy of 95.97%. Both datasets demonstrate high levels of accuracy, but the SAT-4 dataset is faster and more accurate in its implementation. Combining the historical data from Kaggle with the real-time efficiency of SAT-4 can further enhance the system's ability to detect and prevent wildfires more effectively. Integrating CNN-based image processing with remote sensing data improves the accuracy and effectiveness of wildfire identification and monitoring. The system leverages Telegram to notify users once a wildfire is detected, ensuring timely alerts and informed responses. The findings from this research have the potential to significantly improve existing approaches to wildfire management, bringing them to a new level of efficiency and accuracy.