Multi-temporal analysis of urban areas using textural information in space-borne imagery

High level "objects" are proposed as instruments for the temporal morphological analysis of urban zones using spaceborne imaging sensors. These objects, that correspond to the various urban categories as defined in the CORINE land cover nomenclature, are defined as texture appearances of low-level image classes derived from imagery with a low spectral content and a relatively high resolution. In order to describe the texture appearances of the low-level classes, co-occurrence matrices are utilised. For the feature selection and successive model definition a typical pattern recognition methodology is followed, aiming at finding a robust set of features that allows successful identification of the categories of interest. Preliminary experimental results indicate the usefulness of the direction taken.