IS in close
VIN. Major
er friendly
| scanners
the Rollei
his paper
Jrammetry
trology) in
Is for easy
jrammetry
ich 1993).
e concept
asurement
ind a user
om a wide
of coded
surements,
Js camera
of points,
ed images
essing of
tems and
erpretation
and the
inagement
| with free
especially
ments for
|, quality
DPA-WIN
ocessor, a
software
ement and
Fig. 1: DPA-WIN
The DCS 460 camera with its high resolution CCD
sensor (3048 x 2024 pixels of 9 um x 9 um size) provides
an internal slot for PCMCIA drives as image store with a
capacity up to 54 digital images. The sensor/storage unit
is mounted on a slightly modified Nikon camera body. All
the standard Nikon lenses can be applied with the
restriction of a slightly reduced viewing angle due to the
1.8 times smaller photosensitive area of the CCD chip
(18.4 mm x 27.6 mm) compared to a 24 mm x 36 mm
film. The digital image of 6 MByte size can be transfered
into a PC via a PCMCIA drive or an internal SCSI
interface. Finally, special drivers supplied by Kodak for
use with the Adobe Photoshop or Aldus PhotoStyler
image processing software allow data acquisition and
visualization.
2.2 Targetting
The demand for a high degree of automation in industrial
applications has led to the development of coded
retroreflecting targets. The coded targets allow an image
measurement without any interactive point detection and
identification. These targets, first presented in Schneider
(1991), consist of a circular target in the centre and a
code ring outside. The code represents a range of point
numbers from 150 up to 3000 pointnumbers in black and
white images and up to 9000 pointnumbers in colour
images. The coded targets can be detected, identified
and measured in the images fully automatical. In addition
the code ring contains sophisticated error checks to
ensure that even partly covered points will get the correct
pointnumber. Figure 2 shows a coded target in a digital
image.
Fig. 2: Coded target in digital image
531
As well as coded targets also non coded targets can be
measured completely automatically. After measurement
of the coded targets the image bundles will be pre-
oriented by bundle adjustment. Then the non-coded
targets which have already been detected during the
automated image measurement can then be identified
using identification algorithms.
2.3 Processing tools
After the automated image measurement the orientation
of the image bundles is performed with the software
packages NAWE and PROMPT by Rollei Fototechnic,
Germany (Fellbaum, Godding, 1995) using robust
balanced L1-norm estimation. The derived
threedimensional object coordinates can be displayed
and printed for interpretation with several add-on
software packages integrated in the DPA-WIN as there
are roundness control, plain control, variance
comparison packages.
3. VERIFICATION TESTS
Due to the wide range of achieved accuracies presented
in various publications (from 1:50.000 up to 1:250.000)
using the same camera hardware, the achievable
accuracy of the DPA-WIN is verified in comparisons with
higher accuracy absolute reference data.
3.1 Verification test at the PTB (German federal
institut for calibration and metrology)
The PTB has set up a special test object for the
verification of digital photogrammetric systems. This
object consists of three plane plates with 36 reference
points on an area of 800 mm x 800 mm and 5 scale bars
with an extension from 2130 mm to 3350 mm. Figure 3
shows the arrangement and extension of the test object.
The coordinates of the reference points were measured
by an optical coordinate measurement machine (CMM)
with an accuracy better than 10 pmicrons.
Y^
0 : [|
. e000 00 lo
Plate? 1222.7
2m | | 20.000
N ©0000 0 00 00 00 e
Peel as .
xd edi oo 0 00 0 0 0 0 00 U
Platte 2 Platte 3
| y
| [pv
3m
X
Fig. 3: Arrangement of reference objects
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B5. Vienna 1996
