Spectral Unmixing Applied to Desert Soils for the - Naval ...

spectrum by atmospheric effects. Reflectance spectrum B shows the
spectrum after removal of the atmospheric effects. .........................................37
Figure 18. This figure shows a series of non-atmospherically corrected spectra
collected with the ASD spectrometer. .............................................................38
Figure 19. This figure illustrates the original sample radiance data (red) and white
reference spectrum (black) collected with the ASD spectrometer (left)
compared to the spectrum calculated using spectral math (right). The blue
reflectance spectrum (right), represents the red spectrum (left) divided by
the black spectrum (left) (solar spectrum removed). .......................................40
Figure 20. This figure from Boardman and Kruse (2011) shows how mixing in a
picture element (pixel) occurs based on 2 (left) and 3 (right) endmember
concepts. The 3 endmember example shows how this occurs both
spatially and spectrally within the pixel...........................................................41
Figure 21. This figure shows the processing methods for spectral mixing analysis
using the N-dimensional approach adapted from Kruse et al., (2003) and
Boardman and Kruse (2011). ...........................................................................42
Figure 22. The plot (A) shows that the eigenvalues calculated for the image drop
toward 1 at approximately eigenvalue band 50, meaning that most of the
data in this band is noise. The bottom figure (B) is a visualization of
eigenvalue band 50, confirming that though there is some signal present,
band 50 is dominated by noise. ........................................................................45
Figure 23. This figure illustrates the number of times a pixel is marked as pure and
shows the leveling off of pixels at around 2000. Analysis of this graph
helps to determine how many pixels to use in visualization and
endmember derivation. ....................................................................................47
Figure 24. This figure is a comparison between the imagery derived endmembers
(top) and the spectra collected from Owens/Death Valley Mazourka
Canyon (bottom). Looking at the two plots, one sees similarities between
the two so it is feasible that using the collected spectral library may be
useful as an endmember input for the AVIRIS imagery..................................48
Figure 25. This figure shows ASD collected soil spectra under various conditions of
impaction/surface disturbance with the continuum removed. Variation in
feature depths and widths at 500 nm, 1100–1125 nm, and 2200 nm are
measureable, supporting the prediction that surface disturbances are
detectable in Hyperspectral imagery. The depths of features show an
overall pattern of decreasing feature depth with increasing disturbance at
500 nm and 2200 nm. At 1125 nm the depth of features seemed to
increase, for the most part, with increasing disturbance. .................................50
Figure 26. This figure shows undisturbed and disturbed BSC spectra collected from
Mazourka Canyon plotted with bare soil. These BSCs have a prominent
absorption feature around 650 nm similar to that observed by Weber et al
(2008) that is useful for discrimination from bare soils. The absorption
feature at 650–700 nm changes in width and depth between BSCs. Also,
as mentioned by Weber et al (2008) the bare soil does not express the
absorption feature. The feature exhibited by the spectra at around 760 nm
xi