The goal of this research project is to investigate issues involved in the implementation of image databases containing symbols. The issues involve the specification of queries with a preference given to a graphical specification as well as the development of algorithms to process the queries. A query is specified with exemplars of the symbols that are to be found in the database. The symbols in the database are not necessarily classified in advance thereby permitting variation in the quality of the match. Queries to the database involve matching contextual, and spatial constrains. The order in which he different constraints are checked for satisfaction is not fixed, the query optimizer chooses the optimal order. The query images can contain one or more instances of each symbol. The user can specify if all of the symbols must satisfy the spatial constraints or if a maximal subset is also permitted. In the latter case, there is a need for a specification technique for the required subset as well as for the response. The spatial constraints include topological relations such as overlap, encloses, and intersects as well as distance and direction. The symbols are not restricted to point objects thereby also including objects with extent such as lines and areas. The query processing and specification method is integrated with an existing spatial and non-spatial database to enable queries about non-spatial attributes of objects corresponding to the symbols. The results of this research will improve the ability of symbolic image databases and geographic information systems (GIS) to efficiently handle large spatial problems, e.g., adding fiber cables in city planning.
