OBJECT-ORIENTED SHIP DETECTION FROM VHR SATELLITE IMAGES

Within today's security environment and with increasing worldwide travel and transport of dangerous goods the need of vessel traffic services, ship routing and monitoring of ship movements on sea and along coastlines becomes more time consuming and an important responsibility for coastal authorities. This paper describes the architecture of a ship detection prototype based on an object-oriented methodology to support these monitoring tasks. The system’s architecture comprises a fully-automatic coastline detection tool, a tool for fully or semiautomatic ship detection in off-shore areas and a semi-automatic tool for ship detection within harbour-areas. Its core is based on the client-server environment of the first object-oriented image analysis software on the market named eCognition. The described ship detection system has been developed for panchromatic VHR satellite image data and has proven its capabilities on Ikonos and QuickBird imagery under different weather conditions and for various regions of the world. With the capability of eCognition to combine raster data with imported thematic data it is possible to work with available non-remote sensing based data e.g. detailed harbour GIS information in ESRI shape file format or weather information, which can be attached to the results. Finally the system’s ability of generating customized reports in HTML format and the possibility of exporting results in standard raster or vector format offers new opportunities in the direction of an interoperability of technology where a great number of heterogeneous networks and operators are involved in the surveillance process.