correlation procedure, the third one explains the determined terrain
models, in the fourth one the attained accuracy is presented and the
last one gives some final remarks.
2. THE DATA
2.1 Metric camera pictures
From a strip covering parts of Bavaria, the photos N° 885 and N° 887
were selected. The exposures were taken on the 5. of december 1983 at
9.00 GMT on Kodak 2045 double-x aerographic film. Unfortunately in this
time of the year the maximum sun inclination in Germany is limited to
about 30°, what results in illumination conditions with long shadows.
This led together with the turbid atmosphere and the snow cover of the
surface to a reduction of contrast and information content in the pio-
tures.
Furthermore some failures of the camera equipment had to be noted during
the flight. For some photos the film had to be transported manually and
didn't match the plane of exposure, what results in geometrical distor-
tions and reduced sharpness. Image N° 885 was affected by these influen-
ces.
The selected images form a stereo-model of about 60% overlap in flight
direction containing southern parts of Bavaria and north-eastern parts
of Austria. The model area ends at the eastern side in the Bavarian Wood
with heights of about 1000 m, at the southern boundary in the alpes and
covers in the remaining parts a hilly country with height differences of
about 300m in maximum. Most parts of the surface are formed from forests
and snow covered fields interrupted by some valleys and small towns.
2.2 Digital data
First preparational step for the image correlation is the digitization
of the analog photographies. This procedure has to be accomplished with
care to attain an acceptable accuracy. Due to the bad experiences with
conventional drum scanners /Boochs, 1984b/ this is done here with help
of the Perkin Elmer PDS 2020 microdensitometer. This flat bed scanner is
constructed for the digitization of astronomic pictures and meets the
corresponding high accuracy and reliability requirements. The impressing
weight of about 4 tons gives an idea about the construction principle
which assures good results.
As the investigations of /Tucholke, 1983/ have shown some minor errors
are to expect. However, the lack of orthogonality doesn't exceed 10 um
for a distance of about 50 cm, the parallelity of the movements is
within 1-2 um for a distance of about 40 cm, the positional accuracy
during longer time periods is better than 1 pm and thermal effects after
warm up phase led to deviations smaller than 2 pm within the whole image
plane. Therefore the reproducibility of the digitization process should
be better than 2 um within the whole image.
The disadvantage of a such precise instrument is the fairly long digiti-
zation time for images with dimensions of aerial photographies. In the
present case the pictures were digitized with a spot size of 50 * 50 um?
what consumed 2.5 hours measuring time.
- 9p -
