Fast Unsupervised Multi-Scale Characterization of Urban Landscapes Based on Earth Observation Data

Most remote sensing studies of urban areas focus on a single scale, using supervised methodologies and very few analyses the “neighborhood” scale. The lack multi-scale analysis, together with scarcity training validation datasets in many countries lead us to propose fast unsupervised method for characterization areas. With FOTOTEX algorithm, this paper introduces texture-based characterize at three nested scales: macro-scale (urban footprint), meso-scale (“neighbourhoods”) micro-scale (objects). combines Fast Fourier Transform Principal Component Analysis convert texture into frequency signal. Several parameters were tested over Sentinel-2 Pleiades imagery Bouake Brasilia. Results showed that image better assesses footprint than global products. images allowed discriminating neighbourhoods objects texture, which is correlated metrics such as building density, built-up vegetation proportions. best configurations each scale analysis determined recommendations provided users. open algorithm demonstrated strong potential scales areas, especially when data are scarce, computing resources limited.