Abstract: Spatial data mining is the extraction of implicit knowledge,spatial relations and discovery of interesting characteristics and patterns that are not explicitly represented in the database.It has a promising future in such fields as remote sensing,geographical information systems (GIS),medical imagery and information fusion system and so on.And it is accepted and given emphasis little by little.From the view of the crossover of knowledge discovery,cognizing science and intelligent system,this paper proposes a theoretical model of spatial data mining based on the double bases cooperating mechanism of knowledge base and database,and systematically introduces knowledge types which can be obtained from spatial database and the methods of spatial data mining.Then the paper presents the general framework of intelligent system based on spatial data mining and the fundamental principle of its development.Finally,the future directions of spatial data mining are discussed.