Symbolic and iconic information combination for satellite imagery interpretation

Abstract

Our paper describes a fast algorithm permitting cooperation of different approaches and various data sources (image and out-image data) for classification purposes. We take into account expert knowledge about the context of each searched class (described using out image data) which is used to produce for each point presence potentiality degrees for all classes. This set of potentiality degrees can be considered as one source of information (sort of an “ideal landscape” for the region). On the other hand an usual classification process is applied to satellite image which gives for each point a belonging degree to each possible class. We then consider that we have two sources of information concerning the searched classes and we combine them to reach the final decision.