Abstract
Forest and urban fires continue to be a significant issue in numerous regions worldwide. Various solutions exist to combat forest fires. These technologies aim to reduce fire damage through early detection and prevention. This research presents a new approach to fire detection and control using modern technologies. We propose using unmanned aerial vehicles (UAVs) with convolutional neural network-based fire detection technique to patrol fire-prone areas. Through training with datasets, obtain high-accuracy fire image detection that is consistent with fire detection. In this paper, we propose a system for detecting fire using Convolutional Neural Networks (CNN). This allows to identify smoke or fire using still photos or video from drone cameras. This work uses a dataset containing video frames with fire. The data is then preprocessed and used by CNN to build a model to detect fire. A training dataset consists of 1650 photos that have been scaled and reshaped, whereas a testing dataset consists of 1310 images that have been resized and reshaped. To train the model, we employ convolution, activation functions, and the max pooling technique. The model is trained using different batch sizes and epoch settings. As a result, we achieve high precision of 94.78%, recall of 93.84%, and F1-score of 95.76% for fire and smoke detection.