Image 1: 3-D view of part of test area
made with VirtuoZo's Drape function after a seamless ortho
map and DEM was derived (Kubik, 1993). To the North-West
some steeper hills were covered with light forest. The test area
was covered by wide angle aerial photography at five different
flying heights (image scales 1:1600, 1:5000, 10 000, 25 000
and 50 000; flying heights 245m, 760m, 1520m, 3800m and
7600m resp.). High resolution analogue photography was used,
which can be digitised at different pixel resolution in order to
simulate different digital camera types. Both black&white and
colour photography was available. Ground control and a
selected number of check points were measured by Differential
GPS techniques. Accuracy evaluations within the tests were,
however, usually made relative to the most accurate
photogrammetric restitution (1:1600) in order to eliminate
outside error sources. À joint block adjustment was performed
in order to establish homogeneous orientation parameters for
each photogrammetric model. The individual tests within the
continuing project include Traditional vs. Softcopy
Photogrammetry, Accuracy vs. Photoscale, Accuracy vs. Scan
Resolution, and Accuracy vs. Compression ratio (Cock,
1995). This work was done under contract to QDPI.
2. TRADITIONAL PHOTOGRAMMETRY VERSUS
DIGITAL PHOTOGRAMMETRY
These tests have been done on the largest scale (1:1600)
photography with a flying height of 245m above the terrain, in
order to assess the potential accuracy performance of soft copy
photogrammetry. A traditional Digital Elevation Model (DTM)
with regular 10 m grid was collected for one stereo model using
our ZEISS PLANICOMP C100 Analytical Stereo Plotter. Grid
points falling on obscured ground under trees were omitted,
leaving over 500 points for use. No other editing apart from an
automatic blunder elimination process within the DEM
comparison programme was performed. The measurements
204
were repeated on the ZEISS C100 over the same area in order
to assess the repeatability of measurements. A RMS value of
0.034m (21 micron) was computed for the differences between
the two height models.
The film dia-positives were now scanned in a high quality
Helava DSW 100 at a resolution of 25 micron pixel size. Using
the same controls for absolute orientation , a 10 m DEM was
created over the same grid of points by soft copy
photogrammetry on our Helava DSW 750. The differences
between the original C100 DEM and the Softcopy DEM have
an RMS value of 0.21m. This is very large indeed, but closer
inspection learned that a systematic bias was present of 0.18m,
reducing the random component to 0.11m . This 18cm bias in
height can - in our opinion - be attributed to the tufted grass and
crop stubble which covered most of the area. The experienced
operator pushed the floating mark into the terrain to define the
true ground whereas the automatic correlation was done on the
top of the tufts. This experiment taught us that it is difficult to
make meaningful comparisons between analytical and digital
photogrammetry as some of the basic operational assumptions
are different and thus hard to compare.
3. ACCURACY VERSUS PHOTO SCALE AND PIXEL
SIZE
For the purpose of this test the stereo image pairs were
digitised, which cover a terrain area of 1*4 km from four
different flying heights (1:5000 to 1:50 000). The images were
digitised at pixel sizes of 12.5, 25, 50 and 100 micron. The
Helava DWS 100 was used for the high resolution scans and a
HP 3C scanner for the lower resolutions. No compression was
applied.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B2. Vienna 1996
Digi
by V
nine
using
furth
were
secti
parai
orde:
Proc
Virt
2 list
mod:
We |
work
mach
note
numl
resol
wind
mate
resol
autor
main
impo
operz
Table
the d
The |
ofm
the Ç
note,