In order to exploit information recovered from high spatial resolution satellite images, formal knowledge models or ontologies need to be developed to help complete the “visual” interpretation of the spectral information in the image, with the semantic interpretation of the pixels.
The aim of the whole project is to highlight the benefits in the use of a thematic ontology for automatic regions labeling. The concepts that will be included depend strongly on the existing tools. This ontology for automatic regions labeling will be at the junction of two other ontologies to be developed: an ontology about urban objects (streets, buildings, etc.) and an ontology about the tools that are used for image interpretation and their functionalities
The work we will present here concerns the ontology about image analysis methods and algorithms.
The ontology presented has 97 concepts organized around four main classes (algorithm, image type, land use, sensor) and that form around 30 relationships among them; including spectral and spatial properties of the images with the classification of geographic objects as its aim.
The resulting ontology will be at the base of a knowledge-based system that will perform automated selection of the most adequate algorithm corresponding to the characteristics of an image entered as input.