MULTIFUSION: A NEW APPROACH TO MOBILE TARGET IDENTIFICATION

THE PROBLEM OF RELIABLY DETECTING AND CORRECTLY CLASSIFYING MOBILE TARGETS IS AN UNSOLVED PROBLEM OF GREAT MILITARY SIGNIFICANCE. IN THIS PROJECT, HNC WILL APPLY ITS NEW MULTIFUSION APPROACH TO MULTIPLE SENSOR FUSION AND A PRIORI KNOWLEDGE EXPLOITATION TO THE PROBLEM OF MOBILE TARGET DETECTION. THE MULTIFUSION APPROACH UTILIZES NEURAL NETWORKS TO CARRY OUT PRE-CLASSIFICATION FEATURE EXTRACTION AND ATTRIBUTE DERIVATION FOR EACH TYPE OF SENSOR PRODUCT EMPLOYED. A PRIORI KNOWLEDGE CONCERNING THE TARGETS AND THE RULES FOR THEIR USE IS PRE-CODED INTO A BELIEF COMBINATION NETWORK. AFTER EXTRACTING THE FEATURES AND ATTRIBUTES FROM EACH OF THE SENSOR OUTPUT IS COMPLETED, THEY ARE ENTERED INTO THE BELIEF COMBINATION NETWORK TO CARRY OUT FUSION. THE OUTPUT OF THE BELIEF COMBINATION NETWORK IS A DETERMINATION OF THE LIKELIHOOD THAT THE OBJECT BEING CONSIDERED IS A MOBILE TARGET. HNC'S MULTIFUSION APPROACH IS UNIQUE IN THAT IT COMBINES CLASSICAL NEURAL NETWORK PATTERN CLASSIFICATION METHODS WITH RULE-BASED METHODS IN A SINGLE UNIFIED STRUCTURE. ANOTHER ADVANTAGE OF THE MULTIFUSION APPROACH IS THAT IT IS COMPUATIONALLY EFFICIENT AND WILL SUPPORT ON-BOARD REAL TIME PROCESSING IN BOTH MISSILES AND AIRCRAFT. WE ANTICIPATE THAT THIS POWERFUL NEW APPROACH WILL PROVIDE SIGNIFICANTLY INCREASED CAPABILITY FOR MOBILE TARGET DETECTION AND CLASSIFICATION.