Satellite imagery curation and preprocessing approaches for urban semantic segmentation supporting smart cities Applications in Ethiopia

Abstract Smart city development needs a smart strategies and techniques to successfully implementing it. Extracting detailed urban insights supports for urban planners for their effective development. Since satellite imagery is very rich in content and can cover large surface of the urban we are benefited to apply image analysis techniques such as Semantic segmentation of satellite imagery as a critical tool for extracting detailed urban insights to support smart city development. In Ethiopia, rapid urbanization and unique urban challenges, such as densely packed informal settlements, variable environmental conditions, and limited data availability, demand specialized approaches to curate and pre-process satellite imagery for accurate segmentation of urban features like buildings, roads, green spaces, and other urban features of the cities. This article presents a comprehensive methodology for collecting diverse satellite imagery with varying resolutions, environmental conditions, and object types, coupled with a state-of-the-art pre-processing techniques designed for urban semantic segmentation. The results demonstrate significant improvements in segmentation performance, supporting applications in urban planning, mobility, environmental management, and disaster resilience. This work provides a scalable framework for geospatial analysis in Ethiopia, contributing to sustainable smart city development in resource-constrained settings.