A MODEL-BASED SYSTEM FOR CROP CLASSIFICATION FROM RADAR IMAGERY

A technique is described, which uses models of agricultural crop growth and of microwave interaction with the Earth's surface, together with clustering and machine-learning methods, to provide a classification of a remotely-sensed agricultural scene into crop-type. It is demonstrated that overall accuracies of up to 75% can be obtained for classification of five different crops, using data obtained from an airbone radar scatterometer during the 1988 AgriSCATT campaign. The technique is easily extendable to the use of non-microwave data and could also be used for other applications, such as land-use or sea-ice monitoring.