device is
The cam-
) m video
elf video
ment has
ler condi-
round the
the cam-
annot see
the cam-
m above,
‘his leads
oroved.
te images
ansferred
r'IFF-Files
ber. With
rocedures
le adjust-
IP system
for digital
dard PCs
nstant; so
system to
(o the im-
) measure
ally or by
ost of the
| be used,
e of poor
need any
lage rota-
mages are
ch can be
Since the
ackground
difficult to
reliminary
e epipolar
the other
t points in
measured
performed
etermined.
sures, two
oordinates
the object
ferent pen
ith a mini-
| nets (not
sufficient to make a computational object model; cf.
Figure 4). Net 1 is a smaller rectangular cage of 5x5 m?,
net 2 is a rectangular cage of 15x15 m? and net 3 is a
circular cage with a diameter of 15 m. The construction
of net 1 was not as stable as net 2 and 3, also the diver
could not maintain the proposed distance because of
other nearby nets. Therefore the camera positions were
not as good as they should have been.
In all three cases the resulting object coordinates were
compared. The internal accuracies, object sizes, medium
and maximum differences are shown in Table 2.
The differences between the two measurements are in
the range of the internal accuracies which are given as
result of the bundle adjustment. The maximum differ-
ences are still in the range of 3 sigma of the point accu-
racy.
The calculation of projection centres and view vectors for
each of the 2D images used for the photogrammetric
analysis has proven helpful. By comparing the paths
taken by the divers to those planned, it is possible to
improve the efficiency of the data aquisition procedure.
3 CONCLUSIONS
The results obtained clearly demonstrate the feasibility of
and power of the CoSMoLUP/PICTRAN system for the
measurement of large underwater objects. This backs up
earlier synthetic proof of concepts studies.
4 ACKNOWLEDGEMENTS
The authors wish to thank the members of the CoS-
MoLUP consortium namely: Technet GmbH, Germany;
The North Atlantic Fisheries College, Shetland; Wave-
master, Ireland; Johnson Bros, Shetland; Nireus Aq-
uaculture, Greece; and Egersund Traal, Norway.
5 REFERENCES
Krzystek, P., Petran, F., Schewe, H., 1995. Automatic
reconstruction of concept models by using a digital pho-
togrammetric measurement system. In: IAPRS, zurich,
Switzerland, Vol. 30, Part SW1, pp 176-185.
Schultes, M., 1996. Untersuchungen zur Kalibrierung von
CCD-Kameras. Diploma thesis, Stuttgart University,
unpublished.
Technet GmbH, 1995. PICTRAN - Digital Photogramme-
try. Users Manual.
Technet GmbH, 1995b. EASY - Formfinding, static
analysis and cutting patterns for membrane structures.
Users Manual.
6 TABLES AND FIGURES
net |operator | internal point accu- | Difference
racy (med/max)[m] | (medium/max.) [m]
1 1 0.053 / 0.081
2 0.057 / 0.086 0.057 / 0.169
2 1 0.037 / 0.046
2 0.046 / 0.064 0.043 / 0.129
3 1 0.061 / 0.092
2 0.068 / 0.123 0.062 / 0.207
value std. deviation
focal length [mm] 4.7807 0.0024
principal point x [mm] -0.3065 0.0029
principal point y [mm] 0.1295 0.0024
radial dist. (a1) -0.014216 0.000047
radial dist. (a2) 0.000259 0.000002
scale diff. (a3) 0.009329 0.000090
orthogonality (a4) -0.000900 0.000098
Table 1 Calibration results; pixel size is 0.008 mm
Table 2 Internal accuracies of independent meas-
urements and differences
527
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B5. Vienna 1996
