Advancing Satellite Image Classification Using Deep Learning

This research presents a novel deep learning-based approach to satellite image classification using the RSI-CB256 dataset, which consists of four distinct classes: cloudy, desert, green area, and water. Unlike previous methods, our approach integrates adaptive data augmentation techniques and hybrid feature extraction using multiple deep learning architectures, including ResNet50V2, InceptionResNetV2, EfficientNetB3, VGG16, and Xception. Our extensive experiments show that our method achieves a state-of-the-art classification accuracy of 99%, outperforming existing approaches. The outcome of this research has significant implications for environmental monitoring, disaster assessment, and automated land cover mapping.