| inexpensive
le camera and
is achieved hy
/2" (6.4mm x
nsor elements.
frame grabber
zitized images
ilts in a pixel
8.3um in the
ectural Photo-
ions which are
nmon user (eg.
stabilizer) are
sabled and the
| length. This
store a huge
and allows on-
ead of metric
ution must be
ect space with
;: many images
| arrangements
cted to special
On the other
es rapidly with
advantages of
le to establish
outines.
video images.
nly 4 images
used.
hotographs is
ice is approxt-
mately 20 m). All the object points are imaged in at least
two images and at the average there are 3.3 rays cor-
xv JUI. kpul.srt
responding to each object point.
Xv dU. kp cost
Figure 3.
to 25im (derived photo).
4. MEASURING DEVICE
The DIPAD system (Digital Photogrammetry and
Architectural Design) devoped in ETH (Streilein, 1994)
has been used in this test. DIPAD consists of a Digital
Photogrammetric Station (DIPS, (Gruen and Beyer,
1990)) and a Computer Aided Architectural Design
(CAAD) module) and allows for automatic and semi-
automatic measurement of image coordinates (template
matching (Gruen, 1985)), or manual measurements.
Xv 3.81: kpai.srf xv 341: kn02.srf
Figure 4. The four photos used in the test
Measuring devices of 8.5im and 25im have been simulated by "blowing up" the pixel size from 8.5im(original photo)
Since all the control points used in this campaign were
not signalized ones, only manual measurements on the
zoomed images have been used in the course of this test.
This gives the ability to sub-pixel pointing precision, the
pixel size of the original image being approximately
8.5im.
This measurement resolution is of the order of the
analytical systems used by other participants. Other
institutes have either used measurement devices of lower
accuracies or used digitizers (50-1001 m pointing
accuracies) with enlarged photographs. In order to
simulate such devices, the original photos were also
blown up so that the derived image had a pixel size of
25i m. Such a resolution judged to be rather extreme since
many details started to be too small to be recognized and
measured any more.
5. DATA PROCESSING
5.1 System calibration
It is typical for low-cost non-metric imaging sensors to
suffer from systematic errors. This is especially so for the
camcorders (like the one used in this test) employing off-
the-self CCTV-type lenses with large distortion. Previous
tests (Beyer, et. al., 1992) have shown that only the radial
symmetric distortion of the lens and the adapter used with
the JVC camcorder can reach 50 pixels at the image
corners. Aside of that, the location of the principal point
459
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B5. Vienna 1996
