Skip to Main Content
An intelligent system is being developed which integrates remote sensing data from aircraft and satellites with raster and vector geographic information systems (GIS). This System of Experts for Intelligent Data Management (SEIDAM) responds to queries or to product requests to select the appropriate mix of sensors, data processing methods and GIS to provide the answers. Recently, natural language processing was introduced into SEIDAM as one of the modes by which queries can be asked. The other two modes are selection of a stored query from a library or selection of subjects and attributes to create a query. Since SEIDAM contains more than one terrabyte of data, queries must also deal with meta data about the platforms, their sensors, and their properties. The long-term knowledge for SEIDAM is stored in frames which are object-oriented structures with multiple inheritance. Examples of queries are: "How much forest is there in this test site?" or "What are the wavelengths of the MODIS airborne simulator?" or "What remote sensing data are available for Clayoquot Sound before 1985?". A query is parsed in order to extract a set of goals that are passed to SEIDAM's reasoning system. Case-based reasoning and goal-regression are applied together to form a plan that, when executed, will satisfy the goals. A plan may involve the use of several expert systems that understand the use of GIS and the analysis of remotely sensed imagery. For the query "How much forest...", for example, SEIDAM creates a plan that would check the forest inventory GIS and evaluate the timeliness of the corresponding GIS files (maps). Older GIS files would be updated using satellite data such as Thematic Mapper, prior to responding to the query. The presentation format of the answers must be matched with the characteristics of the questioner. SEIDAM also allows the user to specify one or more products to be created. These products likely meet the need of a collection of queries. Products supported include: updated GIS files, temporal change products, maps (forest cover, timber volume, canopy chemistry, infected trees, etc.), enhanced imagery, values and tabular summaries, text, processing and dissemination histories. This paper describes the flow of a query in SEIDAM, the structure for - processing this query, and the application of case-based planning.