A HYBRID NEURAL NETWORK/EXPERT SYSTEM APPROACH TO MULTIPLE TARGET RECOGNITION

MULTIPLE TARGET RECOGNITION SYSTEMS ARE REAL-TIME INFORMATION MANAGEMENT SYSTEMS WHICH PROCESS ;IND ASSESS MULTI-SENSOR DATA, AND PRESENT THE BEST OPTIONS TO A DECISION MAKER. THE EFFECTIVENESS OF CURRENT MULTIPLE TARGET RECOGNITION SYSTEMS IS LIMITED BY THE INCREASING SPEED OF MODERN WEAPON SYSTEMS, AND THE INCREASING NUMBER OF DIFFERENT TYPES OF SENSORS. THE MOST IMPORTANT CHARACTERISTIC OF SENSOR DATA IN MULTIPLE TARGET TRACKING IS THE UNCERTAINTY OF ITS ORIGIN. THIS UNCERTAINTY CAN ARISE NOT ONLY FROM CLUTTER, INTERFERENCE AND MULTI)3ATH EFFECTS, BUT ALSO FROM MULTIPLE TARGETS IN THE SAME NEIGHBORHOOD. A HYBRID ARTIFICIAL NEURAL NETWORK AND EXPERT SYSTEM APPROACH CAN SIGNIFICANTLY ENHANCE THE PERFORMANCE OF CURRENT MULTIPLE TARGET TRACKING SYSTEMS. HERE, WE PROPOSE A HYBRID NEURAL NETWORK/EXPERT SYSTEM APPROACH TO MULTIPLE TARGET RECOGNITION BASED ON OUR IN-HOUSE HYBRID NEURAL NETWORK/EXPERT SYSTEM DEVELOPMENT TOOL NUEX. IN PARTICULAR WE PROPOSETO DEVELOP A HYBRID NEURAL NETWORK KNOWLEDGE BASED MULTIPLE TARGET RECOGNITION SYSTEM, AND DEMONSTRATE FEASIBILITY BY IMPLEMENTING A PROTOTYPE DEMONSTRATION. OUR PROPOSED SYSTEM WOULD CORRELATE MULTI-SENSORDATA TO PROVIDE TARGET CLASSIFICATION AND THREAT ASSESSMENT. SUCH A TACTICAL DECISION WOULD GREATLY EASE THE WORKLOAD ON T.E SHIP COMMANDER.