Object-oriented detection of urban areas from TerraSAR-X data.

Abstract

Urban environments represent one of the most dynamic regions on earth. Even in developed countries the yearly conversion of natural or agricultural space into residential, industrial or transport areas frequently exceeds 100 ha. Due to these rapid changes in landuse short-term data collection is demanded. Thus, remote sensing satellites and particularly the new German radar system TerraSAR-X with its independency of weather and daytime and a revisit of up to 3 days can serve as a valuable instrument. The goal of this study was to show the potential and the limitations of TerraSAR-X in view of the automated object-oriented detection of human settlement. In view of the limited significance of single-polarized radar intensity data a second layer is integrated which provides textural information derived from an analysis of speckle statistics. Two urban environments were analysed in this study: the Upper Rhine Valley and the region of Munich in Germany. The presented approach towards the detection of urban areas by means of TerraSAR-X data achieves overall accuracies around of 95% and 89% - an indicator for the high potential of TerraSAR-X data in terms of identifying built-up areas.