Enhanced Satellite Image Classification Using Squeeze-and-Excitation Blocks and Deep Learning

Satellite image classification is a vital tool in environmental monitoring, urban planning, and disaster management. This research focuses on addressing key challenges, including data imbalance and variations in image quality, by leveraging machine learning techniques, specifically convolutional neural networks (CNNs). The methodology incorporates a deep learning approach for classifying satellite images into four categories: Cloudy, Desert, Green Area, and Water using the RSI-CB256 dataset. The model is based on a Convolutional Neural Network (CNN) integrated with Squeeze-and-Excitation (SE) blocks, which enhance the feature extraction process by adaptively reweighting channel-wise features. The dataset consists of satellite images that are preprocessed through augmentation techniques, such as rotation, zoom, flipping, and rescaling, to improve model robustness. The model was trained using the Adam optimizer with categorical cross-entropy loss and achieved an impressive accuracy of 92.5% on the test dataset. This performance highlights the effectiveness of the proposed approach in handling complex satellite image classification tasks and demonstrates the potential of using SE blocks for enhancing model accuracy in real-world applications.