|
True
color image
This image shows an approximate true color image of the scene under study.
This is an (~10kmx10km) area in an agricultural center near Winters, CA.
The town of winters is visible in the lower center of the region.
The creek running vertically thru the image is Putah creek, and the interstate
running horizontally is I-5. North is to the left. This imagery is
from an AVIRIS overflight from 92/08/20. AVIRIS imagery covers the
wavelength range of .4-2.5 microns at 10 nm increments. There are
614x512 pixels with a pixel size of about 20m. |
Variations
in surface roughness
This image shows the scale variations in surface roughness. The map
shows the distance of any of the pixels in the image from a large change
in surface type. The surface changes are in relation to vegetation
amounts. In the image, the colors correspond to different proximities
of the pixels to such a boundary:
|
Color
|
Approx. Distance (m)
|
|
yellow
|
20
|
|
green
|
100
|
|
cyan
|
140
|
|
blue
|
400
|
|
magenta
|
800
|
|
red
|
1000
|
It can be seen from the image, that the majority of the scene is in close
proximity to a boundary. This implies very little large (100-500m)
scale homogenutiy in the image. The fact that this is an agricultural
image contributes highly t this heterogenuity. |
Smaller scale variations
This is an image intended to
show some of the smaller scale variations in the image. This image
shows the log of the standard deviation of a vegetation parameter in a
localiazed area about each pixel. The area is a 5 x 5 pixel area
(100m x 100m). The log of the standard deviation was chosen because
at crop edges, the standard deviation is very large, and much less so in
the center of the fields. The log scale brings out more of the finer
scale variations. The range of vegation values is from 0-255.
The range of the std. dev in the 5 x 5 sub areas is in the range of 0-110.
The mean value is ~ 20, with a long tail into the 110 range. The
colors in the image correspond to the following std. dev.
|
Color
|
Approx. Std Dev
|
|
white
|
0
|
|
yellow
|
2
|
|
green
|
6
|
|
cyan
|
10
|
|
blue
|
25
|
|
magenta
|
60
|
|
red
|
100
|
Green crops like alfalfa and sugar beets, have low std dev, drier crops
and litter has a slightly higher std. dev, while orchards tend to have
higher still std devs. Areas around the edges of the crops have the
greatest values. |
|
Classification
by ML of SMA
This image shows the different kinds of crops grown in this area.
The map was done by applying a maximum likelihood classifier to the
image of the spectral mixture analysis. The ML classifier uses 12
training sites and 5 textures, eg. green vegetation, soil, NPV, shadow-std.
dev. and green vegetation-std. dev.
In the resultant image seven different classes are shown.
| Color |
Crop Type |
| Bright Green |
Low Crops |
| Dark Green |
High Density trees |
| Olive Green |
Low Density Trees |
| Dark Brown |
Soil type 1 |
| Light Brown |
Soil type 2 |
| Yellow |
Low Litter |
| Peach |
High Litter |
These areas show relatively homogenous areas of certain surface types and
textures. Areas with very high variance were masked out. These
are areas that correspond to large changes in surface texture, and are
not heterogenous. |
Correlations
between endmembers in SMA
This shows the ML correlation results of Spectral Mixture Analysis
of the three endmembers over four different groups
1:Greenstone grassland
2:Chaparral
3:Evergreen woodland
4:Lake |
