WIDE AREA MONITORING USING SATELLITE IMAGERY RETRIEVAL AND MINING

Seeing the expected technical improvements as to the spatial and spectral resolution, satellite imagery could more and more provide a basis for complex information systems for recognizing and monitoring even small-scale and short-term structural features of interests within nuclear facilities, for instance construction of buildings, plant expansion, changes of the operational status, underground activities etc. The analysis of large volumes of multispectral satellite data from different sensors will then definitely require a high degree of automation for (pre-) processing, analysis and interpretation in order to extract the features of interest. In order to process the image analysis steps on the numerous satellite imagery datasets stored in a database, the relevant areas of interests are usually retrieved by a query on the metadata (coordinates, acquiring date, etc.) first and analyzed afterwards. As a more advanced approach the analysis is performed within the retrieval process as an image content-based query. Thus, queries like “find areas with huge construction projects”, “find areas with significant changes of industrial sites between 2002 and 2004”, “find all areas with industrial sites located nearby water”, and others can be realized for the database image data. However, in this paper we developed an image retrieval and mining approach for nuclear-safeguards related, wide area monitoring using satellite imagery. The given monitoring system will enable the users to apply their specific analysis task directly to the satellite imagery in the database.