Search

This paper reports the results of a study to differentiate iron ores in terms of their grades, using the hyperspectral (EO-1 Hyperion) image data, covering a mineralized belt in the Noamundi area, eastern India. The study involves hyperspectral data collection, pre-processing (reduction of atmospheric and solar flux effects), generation of spectral curves from the image for the iron ore deposits, extraction of key spectral parameters and linear spectral unmixing for mapping iron ore abundance. Spectral curves for iron ore deposits extracted from the Hyperion image pixels exhibit strong absorption at 850–900 nm and 2150–2250 nm wavelengths, which is typical of iron ores. The strength of the absorption features in the continuum removed spectra varies spatially in the image around the mining areas, indicating differences in composition/grade of the iron ores. Spectral parameters such as the depth, width, area and wavelength position of the absorption features, derived from image spectra in the 850–900 nm and 2150–2250 nm regions, correlate well with the concentration of iron-oxide and alumina (gangue) in the ore samples obtained from the mine face. Well defined correlations are evident between the concentration of iron oxide and (i) the depth of NIR absorption feature ( R 2 = 0.883); (ii) the width of NIR absorption feature ( R 2 = 0.912); and (iii) the area of the NIR absorption feature and ( R 2 = 0.882). Further, the linear spectral unmixing resulted in an iron ore abundance map which, in conjunction with the image- and laboratory-spectra, helped in assessing the grades of iron ores in the study area. Thus, this study demonstrates the feasibility of discriminating grades of iron ores based on spectral information derived from spaceborne hyperspectral imagery. [ABSTRACT FROM AUTHOR]