Unmixing Space Object's Moderate Resolution Spectra.

摘要

Many non-resolved techniques have been proposed and explored to infer space object's characteristics that may lend insight to object's identification and status. The latter attributes are hard to obtain in the absence of resolved imagery, as for objects not in Low Earth Orbit or for small- sized ones. Spectral unmixing is a non-resolved technique that derives an object's material composition from one or a series of spectra. While spectral unmixing techniques have been tested with space objects and spectrometric visible spectra, with spectral width smaller than 0.4 nanometers and over 100 spectral channels, its success against moderately resolved spectra has not been verified. An example of a moderate resolution sensor is that of a slit-less spectrograph which is desirable because of its simple implementation and arguably better temporal fidelity than systems with a slit. A moderate number of spectral bands are considered a challenge when the number of bands is much smaller than the number of material candidate spectra. We develop the Spectral Unmixing for Space Objects (SUSO) algorithm based on Sparse Recovery optimization techniques to deal with this challenge. Sparse recovery, a specialty area of Compressive Sensing, capitalizes on the knowledge that, while the number of candidate materials is expansive, the external surface of a satellite is effectively composed of a few representative materials. We discuss the technique and show the results of applying it on a number of simulated and measured spectra. Simulated signatures are in the Visible and reflective IR, and measured signatures were made with the Magdalena Ridge Observatory's slit- less spectrograph, which is based on a CCD and grating combination mounted on a 2.4 m telescope. We studied the preservation of temporal information when a time series of spectra is analyzed by SUSO and the prospect of augmenting typical shape recovery with material attribution.