Forest Fire Detection using Deep Learning Techniques

Forest fires pose a serious threat to both people and the environment. Half of the world's forests are predicted to be destroyed by forest fires by 2030. They are a reason of serious concern since both natural and man-made disasters endanger land, plants, and human as well as animal life which exist in those habitats. To prevent substantial disasters, a quick decision-making mechanism is needed for early detection to save the environment from these fast-spreading forest fires. The latest detection mechanisms make use of artificial intelligence for early forest fire detection. This work proposes a convolutional neural network-based image identification technique for detecting forest fires. The work uses a publicly available satellite image dataset for detecting forest fires, which resolves the classification problem of recognizing images with and without fire. The work proposes the use of deep learning-based CNN models like VGGNet, LeNet5, AlexNet, and Xception and a CNN based model. Several performances criterions are used to evaluate the performance of the suggested methods which are used for identifying fire and no fire and are compared to among the applied CNN models. Where Mobilenet has the lowest accuracy, and the CNN model has the best accuracy. Considering the availability of this new forest fire detection dataset, the results of the suggested approach for classifying forest fires are encouraging.