DEVELOPMENT OF ALGORITHMS USING SPECTRAL AND TEXTURAL ANALYSIS FOR AUTOMATIC DETECTION OF SURFACE MINES

THIS PHASE I PROPOSAL SUGGESTS THE APPLICATION OF TWO-DIMENSIONAL TRANSFORMS TO THE SPACE DOMAIN DATA HOLDING SURFACE MINE IMAGERY TO PRODUCE A TWO-DIMENSIONAL SPATIAL SPECTRUM OF THE MINE TEXTURE AND SHAPES AS WELL AS OF THE BACKGROUND. SPATIAL, FREQUENCY MATCHED FILTERS CAN BE DEVISED THAT ARE INVARIANT TO OBJECT LOCATION, ROTATION AND SIZE (MAGNIFICATION), BUT OTHERWISE DISCRIMINATE AGAINST THE BACKGROUND ON THE BASIS OF OBJECT SHAPE AND TEXTURE. BY USING ALGORITHMS BASED UPON SUCH TRANSFORMS, ONLY THE DESIRED SPATIAL FREQUENCIES ARE ACCEPTED. INVERSE TRANSFORMS APPLIED TO THESE FILTERED OUTPUTS CAN THEN RE-CREATE THE MINE IMAGES IN THEIR PROPER LOCATIONS IN A BACKGROUND THAT HAS BEEN CLEANSED OF INTERFERING FEATURES. THE APPLICABILITY OF SEVERAL TRANSFORMS WILL BE INVESTIGATED, AND POSSIBLE PERFORMANCE LEVELS WILL BE ESTIMATED.