Detection and reconstruction of absorption bands on visible/NIR reflectance spectra of minerals using the wavelet transform and the Modified Gaussian Model

The reflectance spectra of asteroids and meteorites in the visible/NIR range are characterized by the presence of charge transfer and crystal field transition bands that can be used to infer the composition of the materials. Sunshine et al. [1]  have shown that these bands can be modeled using the Modified Gaussian Model, and thus that the reflectance spectra these materials can be fitted by a number of gaussian bands superimposed onto a continuum. In order to apply this model to the analysis of reflectance spectra of asteroids and planetary surfaces in general it is necessary first to make initial guesses about the number and position of the bands to be fitted. This initial step is often complicated by the presence of noise, which can mask the presence of some features. Also, even relatively small amounts of noise  can lead to non-unique solutions to the fitting procedure [2]. Both problems can, to some extent, be solved by the use of spectral wavelet analysis [3] to the reflectance spectra of such materials, since it can autonomously detect absorption features in the presence of noise and the use of the wavelet coefficient of the features across multiple scales can provide more robust constraints to the fitting procedure. This work presents tests of the robustness of the algorithm using synthetic spectra and its application to meteorite reflectance spectra from the RELAB database. 

 

References:

[1] Sunshine  et al. 1990.J ournal of Geophys. Res. 95, 6955-6966.

[2] Mothé-Diniz et al.  2008. Icarus 195,277-294.

[3] Starck et al. 1997. Astrophys. J. 483, 1101-1020.