AUTOMATED FEATURE EXTRACTION AND PATTERN RECOGNITION ALGORITHMS

THIS PHASE I SBIR PROGRAM RESULTS IN AUTOMATIC FEATURE EXTRACTION AND MENSURATION OF CARTOGRAPHIC FEATURES FROM MULTISENSOR AND CARTOGRAPHIC (DFAD, DTED) TYPES OF DATA. TWO MAJOR INNOVATIONS ARE EXPLOITED FOR THIS PROGRAM -A NONLINEAR DYNAMICAL (NLD) PROCESSING APPROACH CAPABLE OF INTEGRATING AND PROCESSING "N-DIMENSIONAL" DATA SETS, AND A SOPHISTICATED MEDIAL AXIS TREE-GENERATION ROUTINE WHICH PRODUCES AN INVERTIBLE, ORTHONORMAL, HIGH FIDELITY REPRESENTATION OF THE FEATURES OF INTEREST. THESE TWO METHODOLOGIES ALLOW EXPLOITATION OF THE TREE REPRESENTATION TO RECOGNIZE CERTAIN CHARACTERISTICS OF THE NLD-GENERATED "N-DIMENSIONAL" DATA ASSOCIATED WITH A GIVEN FEATURE.