A Lightweight CNN Architecture for Land Classification on Satellite Images

Satellite image classification plays a crucial role in various applications, including urban planning, agriculture monitoring, and environmental studies.Convolutional Neural Networks (CNNs) have shown remarkable performance in image classification tasks; however, their computational complexity and memory requirements often hinder their applicability, especially in resource-constrained environments.In this study, we propose a lightweight CNN architecture tailored for land classification on satellite images.The architecture is designed to strike a balance between computational efficiency and classification accuracy, making it suitable for deployment on platforms with limited computational resources.The proposed model comprises a series of convolutional layers followed by max-pooling layers for feature extraction, augmented by batch normalization and dropout layers for regularization.Furthermore, we employ a novel attention mechanism to enhance the discriminative power of the network, enabling it to focus on informative regions of the input images while suppressing irrelevant details.The attention mechanism adaptively weights the features extracted by the convolutional layers, allowing the network to prioritize salient spatial information crucial for accurate classification.We evaluate the proposed architecture on benchmark satellite image datasets and compare its performance against state-of-the-art CNN architectures.Experimental results demonstrate that our lightweight CNN achieves competitive classification accuracy while significantly reducing computational overhead and memory footprint.Moreover, the attention mechanism enhances the model's interpretability by highlighting regions of interest, facilitating better understanding of the classification decisions.Overall, our proposed lightweight CNN architecture offers a promising solution for land classification on satellite images, particularly in scenarios where computational resources are limited.Its efficiency and effectiveness make it a viable choice for real-world applications requiring rapid and accurate analysis of satellite imagery.