Multi-approach system based on multi-sensors fusion for land cover classification
Résumé
Images satellites interpretation is in full evolution permitting the extraction of useful information. But the process of the decision making in interpretation systems is marked by several kind of imperfections such as imprecision or/and uncertainty. So, for a better images comprehension, we must take in account this problem.
In this paper and in order to resolve such problems, we propose an approach using three different approaches: probability, possibility and evidence and based on Decision Support System (DSS) architecture. The DSS will help us to recognize the type of soil occupation and the most adapted method of fusion for the image en entry. This task is made by refering to a learning process, using appropriate geometric, radiometric and topologic rules and consulting knowledge and cases bases.
We validated our approach on optical images of the satellite SPOT4 and radar images of the satellite ERS2 representing a central tunisian region. The results obtained consist in the production of five classes of land use.