Searching domain ontologies

Kurzfassung

The interpretation of EO data requires not only data and/or information fusion for better understanding of Earth cover structures, but, at a higher level, needs the aggregation with existing bodies of knowledge specific to the application fields. Thus, from this perspective, the interpretation of EO data is a knowledge driven task. The knowledge consists in the ensemble of existing information, known causalities and other type of associations between information and concepts. During the interpretation process new information and causalities are discovered, thus we have a dual process of knowledge acquisition. To formalize the knowledge both in knowledge driven the interpretation and in the knowledge acquisition we need to define a hierarchy from the point of view of information categories. In the approach we propose now, it is possible to use the knowledge and semantic stored in KIM like systems, to provide to other EO interpretation tools information which is often omitted when humans operate such system because it is clearly an image content element. Thus the idea is to build intelligent interfaces, which explain each other, such to arrive at a compromise for a common understanding. The index system of image databases has a hierarchical structure: image, features, semantics, and domain ontology. The structure can be attached to a tree representation, which is like a union of sub-trees for domain ontology. When a user is training a new semantic label on a image, the system may try to identify the other semantic labels closest to it and give the user the possibility to choose an old one or to define a new one. An entropy measure has been introduced for searching similar subtrees, like Kulllback-Leibler divergence or relative entropy. The method is used for controlling the dynamic index of the archive and for knowledge discovery by searching similar domain ontologies.