AGES
PHS,
ified. The
sed as the
which are
imposed
; per case
at there is
and DEM
. In this
lasses ate
, then by
the field
ntified in
teach the
ypes and
next step,
ver types
d to each
ssification
authors at
ing and
3 at ITC,
y Was to
id by use
4 (Digital
be related
variation in platform altitude, attitude and
(iv)
velocity,
(V) panoramic effects related to the imaging
| geometry, and
(vi) relief displacement due to height difference.
There is possible to remove these distortions and the
relief displacement due to height difference by
introducing ground control points (GCPs) to images
(georeferencing) and use of DEM respectively.
After all geometric corrections, orthoimages with a good
accuracy come out as a result. They are also more
accurate in classification than uncorrected images in
conventional method. By use of aenal photos and the
same DEM we are able to produce the orthophoto of the
same area. This orthophoto can be the result of
mosaicking of several photos covering the whole area,
which are rectified separately. Topographic features
(mainly landuse boundaries) are extracted from this
mosaic. Ultimately, the landuse parcels visible on the
satellite images are integrated in the topographic
network. This network is basically our tools for the
classification improvement.
A part of a full scene of the SPOT XS image with
ground resolution of 20m was used in this case study.
The area of interest was in south of France, and field
data (ground truth) were collected by students of CAR3
of ITC in 1992.
1.1 Location
Our area of interest is located in southern France where
annually fieldwork is conducted by Geoinformatics
students, 1e. CAR3 and PHM3. It lies between 43 49'N
and 43 54N latitudes and 5 16E and S 20E
longitudes that covers the entire area of 10 x 10 km. This
includes the Bonnieux municipality, with various land
covers, partly mountainous and partly flat. Theretore
height difference needs great care.
All steps in the case study were done by digital image
processing techniques. The outputs could be either
thematic maps or tables that are directly usable in a GIS.
1.2 Resources
This set of hardware and software were used in our
study:
Q IBM compatible personal computers
Q Archimedes computer
Q Colour luser printer
Q Colour thermal printer
Scanner
Q ILWIS software
d ALEXANDER software
Q Special purpose software
1.3 Data
These data sets were also used:
M Aerial photographs in 1:15000
Q Part of SPOT XS full scene
Q Topographic maps in 1:25000
O Field data
0 DEM of area
2. PREPARATION AND PLANNING
At the beginning, we had to do the planning, namely
identifying the data sets in the planning and defining the
sources, way of supplying, formats of the data sets. For
example, the area of interest in the region was identified
by matching existing aerial photos, existing topographic
maps, and field data. Then the SPOT XS scene that
covered the whole area was ordered. We chose the
SPOT XS image because its resolution was suitable for
extracting parcel boundaries, roads, and other linear
features.
The other step in planning was design of production
line, defining of events in each step of production line,
timing, identifying the tools and processes. The flow
diagram of production line is depicted in figure 1.
2.1 List of events
2.0 Clipped orthophoto from column 178; center of
photo (OR364) i.c. ORTHO36B
2.1 Mosaicked photo (MOSS from ORTHO37B and
OR364)
3.0 Area of interest (AREA)
3.1 Vector maps (ddbak1/2/3)
3.2 Edited vector map (FINAL)
9.0 GCP file (XSStie2)
9.1 Patched images (pat xs1/2/3)
9.2 Transformation coefficients and control points
(outcoe.ctp and outcoe.coe)
9.3 Geo corrected images without relief displacement
(s1/s2/s3); patched
9.4 Geo corrected images with relief displacement (
s1st/s2st/s3st); patched
9.5 Unpatched 9.3
9.6 Unpatched 9.4
9.7 Stereo image
5.0 Train sample set
5.1 Classified map
5.2 Test sample set
5.3 Rasterized polygons map
5.4 Crossed output map
5.5 Aggregated classified map
5.6 Confusion matrix
4.0 Radiometric corrected images xss1rc, xss2rc and
xss3re
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996
