A Rule Based Fuzzy Approach to the Classification of Man Made Objects in Satellite Image Data

The architecture and implementation of a rule-based fuzzy approach to the detection and classification of man made objects in satellite image data are presented. In a first processing step fuzzy clustering is used to obtain an initial coarse presegmentation of the data. Subsequently, a rule-based fuzzy system performs the classification. Its rules are generated both manually (based on common-sense or higher level expert knowledge) and automatically by a neural network evaluating image features which are difficult to extract or to evaluate directly. Some experimental results, a discussion of the potential of the approach and directions of possible future research conclude the paper.