Deep Learning Based Land Cover Classification Using Satellite Imaginary for Sustainable Urban Planning

Abstract

Land cover classification plays a critical role in understanding urban development, environmental changes, and spatial patterns required for sustainable urban planning. The rapid availability of high-resolution satellite imagery has created opportunities for automated and large-scale analysis of urban and natural land-cover conditions. However, manual interpretation and conventional image-processing techniques are time-consuming and may have difficulty distinguishing complex land-cover patterns. This literature survey examines deep learning-based approaches for automated land cover classification from satellite imagery, with particular emphasis on Convolutional Neural Networks (CNNs), spatial feature extraction, image preprocessing, and multi-class classification. Existing machine-learning and deep-learning methods are analyzed according to their classification capabilities, advantages, limitations, computational requirements, and applicability to urban environments. The survey further examines the use of satellite-derived information for identifying buildings, roads, vegetation, water bodies, and other land-cover categories. Based on the identified limitations, an integrated artificialintelligence framework is formulated that combines satellite image acquisition, preprocessing, feature extraction, deep learning-based classification, accuracy assessment, and urban planning decision support. The framework is intended to support automated land-cover mapping, urban expansion monitoring, environmental-change analysis, and detection of urban sprawl. The study also identifies challenges related to dataset diversity, spatial resolution, class imbalance, computational complexity, seasonal variations, model generalization, and explainability. Overall, deep learning-based satellite image classification provides a promising foundation for intelligent, efficient, and sustainable urban land management.