Toward a Multi-temporal Approach for Satellite Image Interpretation

Multi-temporal remotely sensed images have proved to be of great interest for earth resource assessment and environment monitoring. Automatic or semi-automatic interpretation of these images becomes an important and complex task in computer vision for land use change detection. However, facing the complexity of such images, many classical image interpretation techniques become inefficient. In this paper, the proposed approach, based on multi-agent system and hierarchical blackboard architecture allows an intelligent, concise and flexible control of a multi-temporal scene interpretation. It proposes the combination of semantic network representing the generic description of the scene, and a state transition diagram, modeling the possible state transitions for each one of the classes of interest. This system produces a hierarchic description of the results as well as the structural context of the identified objects including the associated attributes. We illustrate the design and implementation of our system on a set of multi-temporal satellite images SPOT4 representing a center tunisian region for different dates in order to illustrate the potential of the proposed multi-temporal approach.