SATKI. Exercise 2. Supervised classification of multi-
spectral satellite images.
This exercise deals mainly with supervised classification. Two
data-sets, a 4-band Landsat MSS image and a 3-band SPOT XS
image are used. Again CURBOX and SUBSET are used to
generate secondary subsets. Training areas are picked by
DIGSCRN. As in the previous exercise, also these images cover
the area where the University is located, so the students have
a fairly good idea of what is present on the ground. Signatures
are generated with SIGEXT and SIGMAN, and ELLIPSE is
used for an evaluation of the signatures. In MAXCLAS, the
students themselves decide which classification algorithm to
use. Supplementary texts are written by ANNOTAT as shown
in Figure 2. This "Result Screen-Image" contains the original
MSS subset as a False Colour Composite (FCC), the result of
the MSS classification and also the result of the classification
based on the SPOT XS data. The image in the lower right hand
corner of Figure 2 is a small portion of the classified MSS
image, enlarged to the same scale as the XS classification
result. Since the MSS image is geocoded and the XS image has
only been through a system-correction (level 1B), the two
images are slightly differently orientated. It should also be
mentioned that the functions COLORMOD and RECODE are
introduced in this exercise.
Ea
Figure 2. Supervised classification of Landsat MSS and SPOT
XS subscenes.
SATKI. Exercise 3. Satellite mapping / Raster-GIS.
The purpose of this exercise is to demonstrate the close
connection which exists between satellite images in digital
format and other types og geo-data in raster format. The
example in this exercise is taken from a snow mapping project
carried out in collaboration between Fjellanger Widerge A.S. (a
privately owned surveying and mapping company), Norwegian
State Power Board, Norwegian Computing Centre and
Norwegian Hydrotechnical Laboratory. A -.GIS-file showing 6
classes of snowcover is derived from the NIR band in a NOAA
AVHRR image. The image, as well as the -.GIS-file, are
referenced to the UTM coordinate system. The resolution in
terms of pixel size is 1 x 1 km. The State Mapping Authority
has provided a digital terrain model in raster format with
corresponding resolution covering the same area. Originally this
384
DTM is a -.LAN-file. By the aid of CURSES, the studens pick
X, Y and Z coordinates and compare the results with a
topographic map. The terrain-elevation file is also used to
demonstrate various possibilities in the COLORMOD function.
The terrain-elevation file exists also as a -.GIS file with 6
classes. In the last part of the exercise, MATRIX is used to find
those areas which fulfil certain snowcover- as well as terrain
elevation requirements. A third -.GIS-file which defines the
"Nore" catchment area, makes it possible to show only the part
of the actual area which is inside the catchment. Figure 3
shows the "Result Screen-Image".
raster
Figure 3. Digital terrain elevation data and
area in
representation of "Nore" catchment
combination with NOAA AVHRR data.
SATK2. Exercise 1. Satellite image geometry.
In the first part of the exercise, CURSES in combination with
a small scale atlas-map, is used to compare the scale in the
central and in the marginal part of a NOAA AVHRR image
which has not been geometrically corrected. In this way the
effects of a large viewing angle and the curved earth surface
are illustrated. Figure 4a shows the "Result Screen-Image" from
this part of the exercise.
Figure 4a. Geometric properties of satellite imagery illu-
strated with subsets of a NOAA AVHRR scene.