INTEGRATION OF CONTEXTUAL INFORMATION FOR THE TRANSFER OF BELIEFS IN AN INFORMATION SOURCES FUSION SYSTEM - APPLICATION TO DETECTION AND CLASSIFICATION OF TREES CROWNS

In this paper, we present an approach based on the transferable belief model for the detection and the classification of trees crowns on very high resolution satellite images of forest scenes. The masses resulting from the high resolution image source don't always allow deciding between the classes of occupation satisfactorily. Forest context and information about the structure of the forest species are two key elements in the forest scenes classification process. We expose in this paper the retained modelling of the context concept and the approach of revising masses through a transfer of belief. Then, we give some experimentation that illustrates the given approach.