Towards an Ontological Approach for Classifying Remote Sensing Images
Samuel Andrés, Damien Arvor, Christelle Pierkot · 2012
Interpretation of satellite images is a complex issue. Remote sensing experts and the maticians interpret and use information contained in satellite images depending on their knowledge and expertise in a given application domain. This knowledge is usually ambiguous and consequently cannot be used in an automatic process. Formalizing expert knowledge thus appears as a prerequisite toward an automatic semantic interpretation of remote sensing images. In computer sciences, ontologies have proven to be efficient for formally expressing remote sensing expert knowledge. This paper aims to demonstrate how expert knowledge explanation via ontologies can improve automation of satellite image exploitation. We argue that ontologies can be used to link this knowledge with the content of remote sensing images by conceptually describing them. For this purpose, we first built an image ontology for describing image segments based on spectral, pseudo-spectral and textural features. Then we used those concepts to build a remote sensing knowledge ontology describing the way experts identify land cover classes in satellite images. Third, image ontology is also used to describe image facts which populate image ontology. We finally tested a concrete application of our approach using an automatic reasoner for classifying remote sensing images.