Low-resolution urban area outlining in satellite SAR images

Urban areas represent a vital and highly dynamic environment, and monitoring their growth provides important input to decision and policy makers at large scale. Detection and outlining of urban areas from satellite sensors, though less precise, is faster than any on-site data collection. Various techniques have been proposed to classify urban areas, based on their typical features like textural features or backscatter intensity. In this paper a fuzzy-connectedness technique proposed for coastline detection, based on interferometric coherence and backscatter intensity, has been tested and adapted to urban area boundary detection. The modified method appear to be suitable for urban area extraction at ERS-like resolutions, for which long historical records are available allowing to reconstruct urban area growth in the past.