Knowledge engineering for image interpretation and classification: a trial run

Image classification based on information on different physical characteristics raises the need for new classification techniques. Rule-based techniques, as used in expert systems, have been introduced. The application of these "knowledge engineering' procedures in image classification is demonstrated in a case study: the development of a "prototype' expert system to assist in the interpretation of soil classes from a SPOT image. Boolean and maximum likelihood classifiers were defined and implemented as decision rules in an in-house developed shell and a commercial one. -Authors