DOE OSTI · 963240
Large Spectral Library Problem
Abstract
Hyperspectral imaging produces a spectrum or vector at each image pixel. These spectra can be used to identify materials present in the image. In some cases, spectral libraries representing atmospheric chemicals or ground materials are available. The challenge is to determine if any of the library chemicals or materials exist in the hyperspectral image. The number of spectra in these libraries can be very large, far exceeding the number of spectral channels collected in the ¯eld. Suppose an image pixel contains a mixture of p spectra from the library. Is it possible to uniquely identify these p spectra? We address this question in this paper and refer to it as the Large Spectral Library (LSL) problem. We show how to determine if unique identi¯cation is possible for any given library. We also show that if p is small compared to the number of spectral channels, it is very likely that unique identi¯cation is possible. We show that unique identi¯cation becomes less likely as p increases.
Explore related subjects
Keep this discovery
Explore connections, maps & timelines
Chilton, Lawrence K, Walsh, Stephen J. 2008-10-03. Large Spectral Library Problem. https://doi.org/10.2172/963240. https://searchnasa.com/record/osti%3A963240/large-spectral-library-problem
Citations retain the original authors and DOI and include this site’s record URL. If this discovery service contributed to your research, you can also cite Search NASA by Indiciae: BibTeX · RIS.