A new approach for satellite image analysis by means of a semantic network

This paper investigates synergy effects between satellite images and a digital topographic database. The use of such topographic databases, which are actually built up in many countries (e.g. ATKISDLM200 data in Germany), can support the satellite image analysis. This digital database offers a geometric as well as a semantic prediction for every object in the satellite image. Therefore, an automated, knowledge based feature extraction can be performed. Because of the relatively poor geometric resolution of common satellite sensors the project will concentrate on main areabased landuse classes like ’settlement’, ’forest’, ’water’ or ’agriculture’. The result of this feature extraction is a symbolic description of the image content. Both, the symbolic description and the digital database, are transfered in a semantic network, a compact formalism for structuring the entire knowledge. Beside this, the semantic network is used for verification and classification of the above mentioned objects. Therefore, suitable valuation and analysis functions have to be created. In the last step, a change detection between satellite image and database information will lead to an update process of the topographic database.