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

The target material shows up in various regions associated with what appears to
be steeper terrain and the sides of drainages where unconsolidated material often appears;
corresponding to the conditions of the material during collection. This endmember was
given a slightly different margin for MF and infeasibility scores (closer to 5 and 10,
respectively) because unconsolidated material is expected to be mixed with other material
such as vegetation and different mineral mixtures. Therefore, you would expect the pixel
percentages to potentially be lower.
For the repeatability test, the f110512t01p00r08 data set also had positive matches
to targets. In this test, however, there were significantly fewer pixels identified with
feasible mixtures (Figure 33). Since there were so few pixels identified under the
standard ranges, the scatter plot class was increased to encompass MF scores to 52 (the
highest possible range, and as low as 5. Even then most of the pixels were concentrated
in certain areas rather than distributed throughout the imagery as seen in the
f110512t01p00r07 data set. This is a good result however, considering the terrain is very
flat in most of this image. Areas where the endmember corresponds to flat areas appear
to be areas of runoff in both the imagery and looking at Google Earth. The high
concentrations of areas determined to be target material are again associated with regions
of steep terrain where you would expect loosely consolidated material. The majority of
those regions are located, in most instances, around the edges of the landscape features
depicted in Figure 33. These main imagery features appear to be highly lithified rock
structures in the imagery and are confirmed as such to the greatest possible extent using
Google Earth. The areas where the target material has been identified do appear to be
steeper regions with less consolidated surface material as well as deposits at the base of
the structures.
65