The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B5. Beijing 2008
1102
requires too much manual effort. Indeed, it has been proven by
the photogrammetric practice that reconstruction of such
complex objects using automatic DSM extraction algorithms
(area/feature based matching) is in most cases impracticable,
due to occlusions and big scale differences among images.
Furthermore, special treatment is needed regarding the planning
of the imagery, since the strong relief of the ground along with
the height variations of the monuments (Parthenon, Erechtheion,
Propylaia and the temple of Athena Nike) may lead to occluded
areas in the images. On the other hand, the rough anaglyph
requires the exploitation of specialized “true-orthophoto”
methods for the orthophotomosaic production, since
conventional orthorectification programs suffer from double
projections and displacement artefacts.
2. DATA ACQUISITION
2.1 Imagery
Organization skills were required in order to handle the huge
amount of data and bypass the difficult outdoor conditions
which are the main difficulties in such applications. Several
difficulties had to be handled like heat and especially windy
weather on the hill, many obstacles (scaffoldings and cranes),
the rough terrain with rocks and scattered marble pieces, the
lack of a closed or wind-protected place for parking of the
balloon during night, very irregular surface with occlusions, and
especially on walls holes and vegetation, trees obscuring sight
especially for terrestrial laser scanning, and many and moving
tourists.
A group of five people constituted the field team, which was in
charge of controlling the digital camera equipment. It is stressed
here, that in order to follow the general rule which prohibits the
use of any motor vehicle (helicopter or UAV) above the
Acropolis monument, a balloon system had to be employed,
consisting of a gyroscopic GPS-supported base (Figure 1)
where the camera was mounted on, and a helium balloon with
three meters diameter carrying the equipment in the air, fact
that made the image acquisition cumbersome and time-
consuming. The 22 MP (5336 x 4008 pixels) Mamiya ZD
camera was used, having a medium format of 48 x 36mm, 9pm
pixel size, 12 bit radiometric resolution and a wide angle lens of
45mm. The base was hanging much lower than the balloon to
minimize effects of balloon sudden movements due to wind on
the base and the camera. The flight plan was designed and
controlled by Aerotopol software (www.aerotopol.de) allowing
the real time monitoring of the system through a bluetooth
connection between the GPS and the laptop. The correct
positioning of the balloon according to the flight plan was
controlled with the wireless transmission of GPS measurements
to the laptop. In some cases, when GPS signal was lost due to
disturbances in Acropolis, a WEB camera on the balloon and
wireless transmission of images on a monitor were used to
guarantee that the camera image to be taken was centered at the
correct position. The image of the WEB camera was compared
with a pre-computed orthophoto on the laptop or a person
standing at the pre-planned nadir, to ensure the correct nadir
position of the image to be acquired. A correct kappa of the
acquired images was performed by manually turning the
balloon via strings. The camera could be tilted in the vertical
direction manually to be able to image other objects like the
walls (s. Figure 1 right). The average flying height of the
camera for the top view was chosen at 22m above ground
(image scale ~ 1:500) giving a ground pixel size of 5mm,
adequate for the resolution of the final orthophotomosaics. The
images had 75% forward and sidelap, ensuring the multi-image
coverage of all areas in the site and, therefore, an accurate DSM
production and no occluded areas in the orthophotomosaic. A
special flight planning was adopted in areas with strong
anaglyph and/or large monuments with scaffolds and cranes
(Parthenon, Erechtheion, Propylaia).
Figure 1. The balloon system (left) and the gyroscopic device
mounted with the camera (right).
The architectural and historical aspect of the site is not
concentrated only on the visible erections of the hill, but also on
the surrounding walls, valuable for the quantity of material
from various Acropolis structures built into them (Moulou and
Mavromati, 2007). Thus, the walls (north, south, east and west)
were recorded at an average distance of 5m (image scale ~
1:100), with a pixel size smaller than 1mm on the ground. The
overlap between adjacent images was preserved at 65% in x
direction, while a 35% (conventional image configuration) was
selected for the y direction, since multi-image coverage was not
a demand (the wall surface is modelled by laser scanning). The
images were to be used for orthophotomosaics, texturing of the
laser 3D model and possibly fill-in of gaps in the laser 3D
model by photogrammetric stereo measurements. Furthermore,
the balloon was used to take images of the surrounding rock
from the top, while the Mamiya was also used to take images of
the interior side of the walls.
Additional digital cameras were used in the project. Two Canon
5D, one for taking high dynamic range (HDR) images by using
multiple exposures for texturing the Erechtheion (El-Hakim et
al., 2008), and one for image-based 3D modeling of the
Erechtheion (Remondino et al., 2008). An additional Canon
400D was used with the laser scanner (see Section 2.2),
especially for texturing the vertical parts of the Acropolis rock.
For the Mamiya, in total, 1300 images were collected for the
top view (Figure 2), while 1700 images were captured for the
outer and inner part of the walls. Lastly, about 500 images were
collected for the Erechtheion and the surrounding rocky area
around the wall of Acropolis. To ensure a good quality of the
image measurements, radiometric corrections and enhancement
took place through histogram matching.
A very dense network of signalised ground control and check
points has been established and geodetically measured for the
top view images. About every other second image nadir point