Object extraction and attribution from hyperspectral images

Imaging spectrometers acquire imagery in many, narrow and contiguous spectral bands with the aim of collecting “image radiance or reflectance spectra” that can be compared with field or laboratory spectra of known materials. Imaging spectrometry has been widely used in geological mapping, specifically in so-called hydrothermal alteration systems. These are areas where the composition of host rocks is altered through the circulation of hot fluids giving rise to the formation of new mineral assemblages in a predefined order in (3-D) space. Surface mineralogical information can be derived from imaging spectrometer data by comparing imaged reflectance spectra of unknown composition to data from spectral libraries. This comparison is mostly done on a pixel-by-pixel basis. In general, a matching is done to express the similarity between the unknown pixel spectrum and known spectra from spectral libraries. As a result, in geology, information on surface mineralogy can be derived from imaging spectrometry data, which in turn can be incorporated into geologic models. Field and laboratory spectra have been used to relate absorption features to chemical composition of samples in both the areas of soil science and mineralogy, as well as in the area of vegetation science. For the analysis of hyperspectral image data, there are several techniques available to derive surface composition (e.g. surface mineralogy) from a combination of absorption-band position and depth. However, no such technique provides spatial information on the variation of absorption-band depth, position and shape despite the fact that these parameters are of vital use in quantitative surface compositional mapping.