Abstract: Small Rock Outcrops in Vegetated Terrain Identified by Digital Processing of LANDSAT Data
Anthony Morse, John b. Mckeon
The spatial resolution (1 pixel) of standard LANDSAT multispectral data is 58 × 79 m (0.46 ha). In heavily vegetated terrain, such as that of the humid continental United States, outcrops typically are too small to use as training sets for computer processing and categorization of standard LANDSAT data. A special technique, Bendix restoration, was developed to increase the spatial resolution of LANDSAT data by means of discrete deconvolution, to produce cells that are smaller than the original pixel, in this case 28.8 × 28.8 m (0.08 ha). This technique was applied to the raw data in a preprocessing step to improve the homogeneity of training sets by exclusion of "boundary" cells at the margins of the outcrops. Spectral signatures obtained by this technique, for oth dark metasediments and light granites, were subjected to multivariate categorical processing on a Bendix MDAS to produce maps at a scale of 1:24,000 for an area of 15 × 15 minutes. Maps were generated both with and without restoration to permit a comparative evaluation. The results are displayed with low-altitude photography to provide further evaluation of the technique. Using this technique to produce a preliminary map of small rock outcrops, the exploration geologists can plan field activities more efficiently and thus spend more time at the outcrops.
AAPG Search and Discovery Article #90961©1978 AAPG Annual Convention and Exhibition, Oklahoma City, Oklahoma