Remote Sensing Scene Classification

Remote sensing scene classification aims to automatically identify the content of images captured from satellites or aircraft, facilitating applications in environmental monitoring, urban planning, and resource management. Current research emphasizes improving classification accuracy and robustness, particularly in scenarios with limited labeled data, using techniques like self-supervised learning, few-shot learning, and continual learning. Convolutional neural networks (CNNs) and vision transformers (ViTs), often enhanced with feature fusion methods or attention mechanisms, are prominent model architectures. These advancements are crucial for handling the complexities of high-resolution imagery and the ever-increasing volume of remote sensing data.