Rule-based Identification of Revision Objects in Satellite Images

A rule based method for detection and identification of map “revision objects” in satellite data is presented. High resolution panchromatic data from the SPOT satellite are used to identify new objects in comparison to an existing map database. Medium resolution data from the wide field sensor (WiFS) of the IRS-1C satellite are tested for change detection. Fuzzy logic is used in the rules for detection and identification of the potential revision objects. Existing map data help in differentiating between new objects and objects which have been previously mapped. The development is focused on the forested parts of Sweden, and revision objects under consideration are new roads and forest “clear-cuts”. Rules with different degrees of complexity are compared and the results are evaluated against visual interpretation of the satellite data. It is concluded that the rule-based method is successful for identification of clear-cuts and that the complexity of rules does not significantly affect the result. Roads are also identified, but not as successfully as clear-cuts.