Classification of urban SAR imagery using object oriented techniques

Abstract

This paper describes the development of techniques for the production of urban mapping data from interferometric polarimetric synthetic apertures radar (SAR) data. The information contained in the radar originates from four types of data properties: radiometric, i.e. the channel intensities; polarimetric, e.g. decomposition properties entropy and alpha; interferometric, e.g. coherence and interferometric height; and geometric, e.g. shape and area. A multi-scale analysis, using the infrastructure provided by eCognition image analysis software, enables these different sources of information to be brought together. Ambiguities that result from the use of radiometric and polarimetric information alone are eliminated. A map product with broad classification information is produced. Validated results for urban scenes are presented, that were produced using data from the Deutsche Luft und Raumfahrt system E-SAR.