FEATURE SET EVALUATION FOR CLASSIFIERS

KAB LABORATORIES, INC. PROPOSE A PROGRAM TO UTILIZE AND AUGMENT AN EXISTING GENERAL PURPOSE FEATURE SET EVALUATION TOOL, THE ON-LINE PATTERN ANALYSIS AND FEATURE EXTRACTION SYSTEM (OLPARS) DEVELOPED BY PAR GOVERNMENT SYSTEMS CORPORATION, FOR EVALUATION OF FEATURE SETS PROPOSED FOR USE IN AUTOMATIC TARGET RECOGNITION. THE OBJECTIVES OF THIS STUDY ARE TO: (1) IDENTIFY AND PROPOSE A COLLECTION OF FEATURE SET EVALUATION ALGORITHM TOOLS WHICH ADDRESS UNIQUE CHARACTERISTICS OF FEATURE SETS USED IN ATR APPLICATIONS; (2) IMPLEMENT AT LEAST ONE NEW PROMISING FEATURE SET EVALUATION ALGORITHM IN FORTRAN AND INTEGRATE IT INTO THE OLPARS SOFTWARE PACKAGE; (3) DEMONSTRATE THE PERFORMANCE OF THIS ALGORITHM AND COMPARE ITS PERFORMANCE WITH OTHER EXISTING ALGORITHMS ALREADY WITHIN OLPARS USING FEATURE SETS DERIVED FROM BOTH SIMULATED AND REAL E/O IMAGERY; (4) PROVIDE DOD WITH A LICENSED VAX-COMPATIBLE COPY OF THE AUGMENTED OLPARS PACKAGE; (5) DOCUMENT PROPOSED FEATURE SET EVALUATION ALGORITHMS AND TEST RESULTS IN A FINAL REPORT.