A PROTOTYPE MODEL-BASED EXPERT-SYSTEM FOR AGRICULTURAL LANDSCAPE ANALYSIS

Agronomic experts frequently use satellite imagery (landuse maps) to m ake diagnoses of the Lorraine region's agriculture. Their diagnoses re ly on landscape analysis and involve various types of knowledge and re asoning methods. Their interest is also in mapping various criteria to validate their field experience at a regional scale. They therefore n eed AI techniques. AI techniques attempt to represent domain knowledge by rules or by ''domain models.'' Knowledge about the relationship be tween agriculture and landscape was represented with a functional mode l of the agricultural landscape; the model components are image region s that have properties and relations whose combination expresses the g lobal functioning of the agricultural system. This model has been impl emented through a multi-agent blackboard-based architecture. The proto type has been applied to images of a small part of the Lorraine region , initially to characterize and classify plots and village areas. Its results are of considerable interest to experts who can then deepen th eir analysis and improve the model.