SE COG
n Bremerhaven. The
1999 (DSM, 2003b;
80 analogue images
ie points the three-
re detected. In 2003
alculation of a point
ed with white tape
data was calculated
ments in convergent
paper deals with the
e comparison of the
> wurde im Oktober
nutzte das IPI die
1. Mit zusätzlichen
sionale Punktwolke
de festgestellt, dass
’hotos mit digitalen
'e 1982 gemessenen
1982 wurde das
Punktwolke durch
5.0 Pro. der Firma
die Darstellung der
the cog have been
sure a flexible and
reconstruction and
a free convergent
rammetric image
have been used.
plied cameras the
pto ! : 114.
lrawings of the ten
and exported in
2. MATERIAL AND METHODS
2.1 Object description
The Bremen Hanse cog has been fixed under the roof of the
exhibition hall with a complex construction with several
diagonals made of steel holding the wooden cog together and
fixing it with its lower part on the ground floor. The main
dimensions are 24 x 8 x 7 m. The location of the cog in the
museum allows the positioning of some camera stations on the
first floor, all the other stations are on the ground floor and
restrictd to a maximum distance of 7 m from the cog. Due to
walls and other restrictions most of the camera stations have
been used in the range from 2.5 to 6 m. Additional lighting was
necessary for most of the images.
2.2 Historical results
The historic line drawings contain information about parts of
the wooden surface of the cog as well as the location of the
measured profile points. To prepare the imaging process of the
photogrammetric survey in the year 2003 about 105 tie points
have been fixed on the body of the ship. To compare the
deformation of the measured profiles, the location of the profile
points have been taken from the old line drawings and marked
with white tape on the surface of the cog.
2.3 Digital Camera Application
For the photogrammetric survey two digital cameras have been
used. The 6-Megapixel camera Nikon D100 and the 5-
Megapixel metric camera Rollei D7 metric? had been calibrated
before with a three dimensional and a two dimensional
calibration field (Wiggenhagen, 2002) For the point
signalisation and imaging process about 10 hours were needed.
Most of the time was required for the point and profile
signalisation on the surface of the cog and the image transfer
and image conversion from the compact flash memory cards to
the PC.
Due to the different aperture angles of the cameras and local
conditions in the museum hall over 100 images were taken.
With the Rollei D7 metric? camera 60 images and with the
Nikon D100 camera 39 images were used for the
photogrammetric measurement.
2.4 Coordinate Measurement and triangulation
Prior to the bundle adjustment the achievable point accuracy
was calculated (see. Tab. 1).
Camera Object / Base | Image Accuracy
Distance Scale in x,z,y
ROLLEI mm m /m mm
Sensor width | 8.76 6.9 / 1.8 17929 158, 2.8.107
Focal length | 7.43 2.5/1.3 1377 (el 1 1, 18
NIKON mm m/m mm
Sensor width | 23.7 6.9 / 1.8 1:280. 1.2.2,22, 8.6
Focal length | 24.6 2.8/ 1.8 1: 114 |: 0.90.9, 1.4
Table 1. Reachable point location accuracy
The calculated point location accuracy between | mm and 11
mm fulfilled the required accuracy for the project.
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B5. Istanbul 2004
The software PhotoModeler 5.0 Pro from Eos Systems Inc.
(Eos, 2003) was used for the image coordinate measurement,
the bundle adjustment and the preparation of the results. Most
of the tie points have been measured in 3 and more images (see
Tab. 2). A very stable bundle adjustment resulted from this high
redundancy.
Point Type Number of
points rays/point
Tie point 4 <3
12 3
200 >4
Total 216
Profile point 27 «3
213 3
107 24
Total 347
Table 2. Number of rays per measured point
After all images have been orientated the points of the
signalised profiles could be successfully measured. The point
cloud was visualized in the three dimensional view to obtain an
overview of the measured points and profiles and to detect
points of poor geometry. Most of the profile points could be
measured in three and more images
3. RESULTS
To be able to compare the current results with the survey of the
year 1982 line drawings of the ten profiles had to be processed
at an output scale 1 : 20.
3.1 Coordinates
From the photogrammetric survey 216 tie point coordinates
with an accuracy of +/- 0.5 mm could be calculated. The high
accuracy depends on the fact that the circular targets have been
measured with an automatic point measurement operator
calculating the centre of the imaged ellipse with an accuracy of
0.01 pix in image space.
The profile points had to be measured manually. As the image
coordinate measurement was carried out in monoscopic mode it
was quite difficult to detect the identical points in several
convergent images. The epipolar line visualisation ability of
PhotoModeler 5.0 was very helpful in this cases.
The point accuracy in object space for the 347 points was
calculated with +/- 3 mm in all three coordinate directions.
3.2 Lines
In addition to the photogrammetric survey ten control distances
had been measured. The locations of these distances were given
in the digital images by their start and end points. The
connection between these points have been defined in the
photogrammetric bundle to be considered as independent
reference lines.
To prepare the result of the whole bundle adjustment for later
documentation the different line types were defined in different
layers and with different colors.
For the visualisation of the ten profiles the triangulated points
have been connected with lines.
À comparison between the two measurements in 1982 and 2003
for profile 9 is shown in figure 1.
