KNOWLEDGE REPRESENTATION FOR WEATHER MAPS AND IMAGERY

TO MOVE TOWARDS MACHINE REASONING ABOUT WEATHER MAPS AND IMAGES, IT IS PROPOSED TO DEVELOP A SCHEME TO REPRESENT KNOWLEDGE ABOUT MAPS AND IMAGES FOR EFFICIENT USE IN AN ARTIFICIAL INTELLIGENCE (LISP PROGRAMMING) ENVIRONMENT. IT IS PROPOSED TO ORGANIZE THE SYSTEM AS A CONTROL ROUTINE TOGETHER WITH THE KNOWLEDGE REPRESENTATION AND THE LOWER LEVEL ALGORITHMS FOR RECOGNIZING AND MANIPULATING MAP COMPONENTS (ISOLINES, FRONTS, TROUGHS, CYCLONES) AND FOR LOCATING IMAGE COMPONENTS (CLOUD BOUNDARIES). THE KNOWLEDGE REPRESENTATION WILL CONSIST OF TWO PARTS: "REFERENCE EXPERTISE" ABOUT THE COMPONENTS AND THEIR STRUCTURAL RELATIONSHIPS, AND "WORKING MEMORY" OF DETAILS ABOUT THE MAPS AND IMAGES CURRENTLY BEING PROCESSED. THE CONTROL PROGRAM WILL INTEGRATE THE MAP AND IMAGE PROCESSING ALGORITHMS WITH THE KNOWLEDGE REPRESENTATION TO COORDINATE AND SEQUENCE THE SYSTEM. THE PHILOSOPHY OF THIS APPROACH IS THAT THE LOWER LEVEL PROCESSING PROVIDES INFORMATION ABOUT THE COMPONENTS, WHICH MUST THEN BE GIVEN A ROBUST, CONSISTENT, SYMBOLIC/STRUCTURAL INTERPRETATION BY THE HIGHER LEVEL KNOWLEDGE-BASED COMPUTATION. CODE WILL BE IMPLEMENTED FOR A SUBSET OF THE KNOWLEDGE REPRESENTATION FOR WEATHER MAPS AND IMAGES. THE PROBABILITY OF SUCCESS IS HIGH BECAUSE THE INVESTIGATORS HAVE CONSIDERABLE EXPERIENCE IN DESIGNING AND IMPLEMENTING SYSTEMS IN WHICH HIGHER LEVEL KNOWLEDGE INTERACTS WITH LOWER LEVEL PROCESSING IN PATTERN AND IMAGE ANALYSIS.