The object was recorded with 46 convergent images. 23
images covered the object format filling and 23 images
covered 67 % of the object. For image recording a Kodak
DCS 460 with 24 mm Nikon lens was used. The internal
accuracy achieved was 6.1 um in X-direction, 6.3 um in
Y-direction and 10.9 um in Z-direction. Table 4 shows the
RMS difference between DPA-WIN and CMM:
Plate No. RMS X [um] RMS Y [um]
1 6.7 6.5
2 13.1 14.4
3 7.7 7.8
Table 5 shows the scale bar variance comparison:
nominal measured difference
12 - 13 699.909 699.923 0.014
13 - 14 699.859 699.845 -0.014
21-22 700.023 700.029 0.006
22-23 699.660 699.654 -0.006
23 - 24 699.959 699.943 -0.016
32 - 33 699.923 699.941 0.018
33 - 34 699.893 699.867 -0.026
42 - 43 699.841 699.866 0.025
52-53 700.108 700.104 -0.004
Table 4 shows that the accuracy of plate 2 is two times
worth than the accuracy of plate 1 and 3. The error
ellipse in figure 6 show that the accuracies at the edge of
the object are less good than in the centre. This is
caused by the configuration of the image bundles which
are captured in a way that all edge points of the object
are also in the image edges and corners. Due to the fact
that the correction of the lense distortion is not well
defined in the edges and corners these object points are
also not in the same accuracy as in the centre.
GS edo 75 o
© © 660606606065 65 >. ©
9099099889
© 060060 o & Oa
(9 ©9099 €)
o c 000000
9.00000 e. uy 6 ®
00085 000060006
099088999555 09 Q
999999990989 Q (3 (3
906666006066
009008066066 60
d Ooooooooooo (D e 5 Cty
Fig. 6: Error ellipses of reference field
532
3.2 Aircraft industry
In the aircraft production at Daimler Benz Aerospace
Airbus plant Hamburg, Germany large-scale production
devices have to be controlled periodically. This
measurement task is performed with the Rollei ILR
system consisting of the large format camera LFC and
the réseau scanner RS 1 (Dold, Riechmann, 1989). To
enhance the productivity of the measurement team these
tasks should be performed with digital systems in the
future. Therefore a comparison between the actual
measurement systems (ILR and Kern ECDS) has been
performed. The object is captured with 68 images
according to figure 7 from two different heights.
Y
LT. pM
Um
uw
Fig. 7: Arrangement of images
The results were compared with a LFC and ECDS
reference measurement. The reference accuracy is
determined by Daimler Benz Aerospace Airbus to
x=+0.042mm, y=+0.035mm and z=+0.028mm. The
object size is 8800 mm x 6200 mm x 6300 mm. The
coordinate differences between reference values and
values by DPA-WIN in RMS are x=+0.16mm,
y=+0.12mm and z=+0.08mm. These measurements were
performed under realistic environmental conditions on
the shopfloor regarding also a short time window in the
production process for image recording. The results show
a realistic achievable absolute and reproducable
accuracy between 1:50.000 and 1:80.000. Higher
absolute and reproducable accuracies can be achieved if
the whole object can be imaged format filling or if the
environmental conditions are laboratory-like.
4. PRACTICAL APPLICATIONS
Besides the verification tests under laboratory conditions
and under industrial conditions DPA-WIN is in daily use
by a cupple of users for production inspection and quality
control.
4.1 Shape control of submarines
The roundness of submarines is an indicator of the diving
behavior and quality. Therefore every submarine has to
be controlled during the production process as well as at
the final check in the delivery stage. The body of the
submarine is signalized with retroreflecting coded targets
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B5. Vienna 1996
in ev
subn
The
were
adju:
fram
man
appr
The
(e.g.
enhe
imac
brou
phot
syst:
daily
deve
shof
perf
The
and
mea
eval
tran:
We
perf
Gerr
PTB
Dain
Gerr
How
Bey:
Qua
In: (
Gerr
Bós:
ZUr .
Nah
Phol
Bose
tube
Autc
