NEURAL NETWORK TARGET TRACKING AND RECOGNITION BASED ON MOTION VISION APPLIED TO ANTISATELLITE WEAPON SYSTEMS ACQUISITION TRACKING

LNK IS APPLYING A NEURAL ARCHITECTURE TARGET TRACKING AND STATE ESTIMATION SYSTEM BASED ON MOTION VISION INFORMATION FROM A PASSIVE 2D SENSOR. LNK IS DEVELOPING A HYBRID 2D SILHOUETTE AND 3D RELATIVE DEPTH MODEL-BASED TARGET RECOGNITION SYSTEM BASED ON THE ABOVE MOTION VISION INFORMATION. THE TRACKING AND STATE ESTIMATION PERFORMANCE HAVE PROVEN VERY ROBUST: DOWN TO SUBPIXEL TARGET SIZES, AS SLOW AS SUBPIXEL INTERFRAME SPEEDS, WITH UP TO 100 SIMULTANEOUSLY OVERLAPPING TARGETS, USING COMPLEX BACKGROUNDS. THIS WORK EFFORT EMPHASIZES CONTROLLED TEST AND EVALUATION TO SUPPORT PREDICTION AND OPTIMIZATION OF THE TRACKING AND RECOGNITION PERFORMANCE. THE RESULTS INCLUDE DEFINITION OF THE OPTIMAL CONCEPT, TEST AND EVALUATION RESULTS KEYED TO RESOLUTION AND S/N, APPLICATION DERIVED ENGINEERING REQUIREMENTS, AND A PLAN FOR FOLLOW ON DEVELOPMENT. THESE RESULTS ARE IN SUPPORT OF THE ANTISATELLITE ATR APPLICATION.