Decision Based Data Fusion Techniques for the Analysis of Settlement Areas from Multisensor Satellite Data

Despite the population stagnation in many industrialized countries, we experience a steady sprawl of urban/suburban space. This fact is based on the migration out of the inner cities, increasing demands on living space, and set-aside of land for commercial developments. In developing countries, this urban growth occurs at dramatic speeds and costs to the environment. Often, planning offices have no or only outdated information about new settlements and their composition in their jurisdiction. The University of Osnabrueck’s Institute for Geoinformatics and Remote Sensing has developed a hierarchical method for decision based fusion to automate the delineation of new residential and commercial built-up area. The procedure is based on fusion techniques at the decision level using differently weighted parameters for texture, shape, and spectral characteristics. It was applied to multisensor satellite images from sensors such as SPOT-5, Kompsat 1, Landsat ETM, and Aster. Using this method, accuracies exceeding 90% could be achieved.