Knowledge and context based SAR classification for map-updating using fuzzy logic and GIS-methods

The reliable classification of monofrequent, monopolarized SAR data often requires additional context information. In this paper a new context based scheme for the updating of existing maps is presented. The implemented procedure contains a fuzzy expert system which enables a knowledge based, automated evaluation of SAR data with a high degree of accuracy. The procedure is based on an adaption mechanism, which adjust the expert system to the actual condition of each data set. This procedure is required because of an incomplete modelling of system interdependences, caused by the lack of information and unsuitable context data. The procedure was applied for the purpose of flood monitoring and compared to standard methods of evaluation. This work is part of the project KORAL, which is sponsored by DLR (former DARA).