A Formal Representation of Landscape Spatial Patterns to Analyze Satellite Images
Résumé
Today french farm enterprises fast increase their area and modify their farming systems. Furthermore they modify the land use and eventually the environment. Satellite data are useful to follow and to forecast these evolutions because of their quantity and frequency. Agronomic researchers use the satellite information about land-use categories and their spatial organization both to diagnose agricultural systems and to forecast environmental problems. They search for land-use patterns to speed up their analyze. We defined satellite patterns of landscapes at various scales. These patterns were defined as sets of agricultural spatial objects related with spatial relations. We worked on the modelling and computing of these patterns from a satellite image of the Lorraine region, in the east of France. The work was a cooperation between agronomists and computer scientists. It resulted in formal patterns of satellite landscapes and associated computation methods. The formalism was based on qualitative descriptions of space. The patterns were implemented with an object-oriented language.