Cl PA 2003 XIX th International Symposium, 30 September - 04 October, 2003, Antalya, Turkey 
341 
2.3 Orthophoto production in Larisa’s ancient theatre. 
Larisa’s ancient theatre has been discovered a couple of years 
ago under a street and twelve apartment buildings in the city 
centre. The buildings have been demolished and the theatre now 
is under reconstruction. 
Detailed plots of 1:50 scale were necessary showing current 
state. Mr. Tziafalias, the responsible archaeologist for the 
restoration and conservation of the theatre, have been informed 
about orthophotographs and directly asked for such product. 
The total area of the theatre was 5400 square meters, with a 
height difference of 13.5. In particular, within the scene there 
were height differences of 2.5 meters in the walls, which from 
22.5 m which was the selected flying height, develop a 1:9 ratio 
of object depth : camera-object distance. The expected 
occlusions were very big and therefore excessive number of 
photographs was taken over the scene. 
In order to correctly orthorectify the photographs a detailed 
DTM was collected. It was even more intense and detailed than 
3d plots of features, since all features had to be double plotted 
with one line describing the upper edge and another on the 
lower one. These lines were then divided in 5 cm steps, creating 
points in these intersections. Additional points were measured 
in order to describe the surface exactly. The final DTM has 
been created using only points, without any breakline 
information. Triangles created were not mixing information 
from different levels due to their high density in comparison to 
simple height points. DTM collection was fully manual and 
time consuming 
In addition, almost all photographs over the scene were rectified 
and the best ones were selected for mosaicking. A lot of work 
was necessary in order to connect many orthophotographs so 
that the final result to be acceptable and the walls vanished from 
the final product (fig. 4). The final mosaic was retouched for 
tone balancing and elimination of stretching in some areas with 
strong relief. 
Figure 4. Final mosaic of Larisa’s ancient theatre (original 
mosaic of 1:50 scale). 
2.4 Facades and sections over the scene of the ancient 
theatre of Larisa. 
As part of the theater’s recording and restoration, sections and 
facades of the scene at 1:25 scale were also needed. Since the 
surfaces were rather flat and in order to cut down the cost, 
simple rectification of photographs was selected as the most 
appropriate approach. 
21 independent sections and facades were requested. This 
project was the most intense in terms of photography. A 
specialized group of three people was working for twenty hours 
in order to stick 321 control points in inner and outer surfaces of 
the scene, measure the network and the control points and take 
131 photographs. 
Since the control points should be measured with 1.25 cm 
precision, a traverse with eleven stations was established and 
measured with reference to a previously established network of 
four stations. From each station at least four angles and two 
distances (measured at least twice) were measured to known 
and unknown stations. Observations were used for least squares 
solution. The solution had residuals of 0.6, 0.9 and 1.3 cm in X, 
Y and Z respectively. From these stations the 321 control points 
were measured using angle intersections or distance and angle 
measurements. Relative accuracy the control points is less than 
2 cm. 
In order to perform projective transformation on the 
photographs, simple in-house software was created and used to 
calculate the best fit (using least squares) vertical plane passing 
through the control points of each section or facade. The new 
coordinates were calculated, with the Z being the distance from 
the best-fit surface. 
Although special care has been taken in order to view at least 
four control points in all photographs, where this was 
impossible a foul triangulation was taking place. Although the 
photographs were strongly converged aerial software had no 
problem at all solving the bundle adjustment. The exterior 
orientation was then used to produce orthophotographs at zero 
elevation. 
Figure 5. Three meters high corridor. The dark part was under 
shadow and processed with wallis filtering, prior to mosaiking. 
3. CONCLUSIONS AND DISCUSSION 
Experience is an expensive and invaluable asset in such 
projects. Expensive because it is gained the hard way; by the try 
and failure method. This is mentioned particularly for the first 
project. Absence of the onboard video camera lead to a triple 
visit for what today seems an extremely easy job. 
The communication between the helicopter operator and the 
‘navigator’ who looks the video and triggers the photograph 
needs a lot of work. Guiding helicopter movement by seeing 
only the video is difficult, not to mention that the operator can 
only rarely hover on the point the ‘navigator’ wants. That’s the 
main reason why the layout of the photographs is almost 
random. 
In all cases the aerial bundle software performed well. Large 
attitudes of up to 75 degrees on phi and omega did not cause 
any problems at all. The main problem is the uncomfortable 
viewing of the operator during three-dimensional plotting. A 
self-levelling mechanism of the photographic camera would be 
welcomed, along with a lock mechanism of the flying height. 
Variations in flying height cause scale discrepancies, which 
deteriorate the situation. These are our next considerations for 
improvements.