image acquisition, and then be continued. However,
during field measurements image acquisition was dated
at short notice, depending on the amount of rainfall
since the last photograph. Ideally, images are captured
immediately after heavy rainfall that could have a mea-
sureable effect on the soil relief. Portable equipment is
needed because field test sites often are not accessible
to cars.
Providing DEMs of the soil test areas (2 or 3 mm grid
spacing, height accuracy better than 0.5 mm) was the
photogrammetric task. Test areas were enclosed by
rectangular metal frames which collected all sheet wash
and such allowed to measure the quantity of water
that caused the erosion. Since the coordinate system
could be chosen freely - as only relative heights were
of importance - the corners of these frames could be
used as reference points, measuring their relative heights
by nivellement. Thus exact surface slope angles could
easily be obtained - a parameter that directly influences
overland flow. Installing a sufficient number of refer-
ence points in the field proved to be nearly impossible,
as test areas lay on farming land with no stable ground.
Considering these circumstances the stereometric came-
ra Zeiss SMK 40 was chosen for this project, a conven-
tional close-range photogrammetric camera that is well
known for its good image quality if calibrated thorough-
ly. An important point was the simple constellation
for image acquisition where no control points were
needed due to the cameras fixed relative orientation
and stable calibration. The projected accuracy, which
was better than 0.5 mm in all three dimensions, in
connection with an object area of 1 m? (field measure-
ment) resp. 0.2 m? (laboratory) could be met by this
camera type using close-range lenses with a focal dist-
ance of about 1.5 m.
After developing the glass plates and reproducing the
photographs on paper positives (or on film transparen-
cies, depending on the scanner), scanning was done
with a resolution of 20 um per pixel, corresponding to
0.4 mm in object space ground coordinates. Digital
images covering the area of interest were about
1,500 x 1,500 pixels wide in the case of the laboratory
measurements and 2,500x 2,500 pixels for field
measurements. The image acquisition process for any
series of about 60 exposed glass plates, including plate
development and scanning, took between one and two
weeks until the material was ready for digital image
evaluation. While two people are needed for the image
acquisition itself thanks to material transport require-
ments — especially for the weighty SMK camera and
tripod - preparations, development, and scanning could
be done by one person only, a method that proved to
be practicable. Meanwhile, products for scanning of
analog photographs have entered the photogramme-
try-oriented production lines of manufacturers - the
Zeiss PS1 photogrammetric scanner recently was
developed right for this task.
There are certain drawbacks of this rather traditional
way. First, bulky equipment has to be carried. Then,
the number of possible images for a project is limited,
since photographic plates have to be inserted into their
cassettes in a dark room. Moreover, image evaluation
may start only days later due to the lengthy photogra-
phic process, so there is no immediate control for image
quality.
Digital Images
But these limitations can be overcome by using direct
digital image acquisition. Immediate image control,
e.g. checking image geometry and radiometry or cor-
relation fitness, is at hand. Plates and film are replaced
by magnetical and optical storage media, thus permit-
ting a virtually unlimited registration capacity. The
equipment for digital image acquisition, if still of
considerable bulk and weight, is considerably easier to
handle than the SMK 40. The cameras themselves are
small and light, no large tripods have to be carried
around. The heaviest parts are the batteries for the
field power supply.
When there are so many advantages of digital cameras,
why did we not use them in this project? The one
simple reason is, that at project start there were none
on the market that met the accuracy demands of this
project. Even with the fast technical evolution in this
field, state-of-the-art digital cameras still do not provide
the image dimensions and the excellent image quality
produced by the SMK 40 - especially they lack its stable
geometry. However, suitable cameras are bound to enter
the market, since CCD chips holding 2k x 2k pixels
(LUHMANN 1991) and cameras with pixel synchronization
for stable image geometry like the VIDEK MegaPlus
are already available, though still not featuring the
needed pixel resolution. Back in 1986/7 when concepts
for the project were made, a rapid development like
this had not been foreseen.
However, possible advantages of digital image acqui-
sition for close range applications were to be researched,
and for this purpose an experimental digital stereo-
image acquisition system (DigiSAS) was designed at
the TU Berlin. This system became operational during
the project work but, due to hardware limitations, was
used only for testing purposes. It is able to capture
images from two CCD cameras with 512 x 512 pixels
(8 bit) each, simultaneously using external line syn-
chronization (Jescuke 1990). Thus only a poor accuracy
of several mm for the project test areas was possible,
which was not appropriate for precise surface measure-
ments. Nevertheless some valuable experiences in hand-
ling of digital cameras for close range applications were
made during the practical tests of DigiSAS. Using non-
photogrammetric cameras brings up the problem of
calibration. Image orientation is also more complicated
- compared to the SMK with its fixed base and camera
directions — and reference points are needed, the more
the better.
For the DigiSAS tests a 3-D metal frame holding some
20 reference points was built. Several test series were
done, checking geometrical stability and the possible
accuracy of the system. During these tests the influence
of often reported (Bever 1987, DAEHLER 1987, LENZ &
FurscH 1988) problems like line-jitter, aliasing, and
warm-up effects on image quality and evaluation results
