CIPA 2003 XIX th International Symposium, 30 September - 04 October, 2003, Antalya, Turkey 
Figure 4. Map sheet, scale 1/500 
Up to now, around 17.000 ground points were captured to 
produce the 1/500 scale map sheets. 
4. DIGITAL ELEVATION MODELS (DEM) 
Once a digital elevation model of a landscape is available, such 
a model can serve as a base for many applications, like contour 
line generation, volume calculation, visibility analysis, 
calculation of cross sections and all other applications which 
need a digital representation of the terrain as an input, like 
virtual flights and walk throughs, for instance. 
//Transformation mit 5 Passpunkten ( 1000,4000, 5000, 11U0U, 9000} 
Arc: transform contours fivepp 
Transforming coordinates for coverage contours 
Scale (X.Y) = (0.997,0.999) Skew (degrees) « (0.081) 
Rotation (degrees) « (-0.056) Translation = (-2710.751.-1121.360) 
RMS Error (input,output) = (0.888.0.885) 
Affine X - Ax + By + C 
Y « Dx ► Ey * F 
A = 0.997 B = 0.002 C = -2710.751 
D = -0.0C1 E * 0.999 F - -1121.360 
tic id input x input y 
output x output y x error y error 
1000 
5000 
4000 
11000 
9000 
Arc : 
12727.471 
10000.00C 
11485.780 
8760.518 
11516.838 
8790.385 
13955.460 
11221.840 
13001.104 
10272.820 
11147.184 
10000.000 
10911.637 
9766.446 
10067.302 
8922.818 
11076.767 
9928•892 
10980.450 
9833.549 
-0.662 
0.864 
-0.073 
1.068 
-1.196 
0.308 
-0.191 
0.057 
-0.125 
-0.049 
Table 2. Results of an affine data set transformation 
from the paper map and ending in the data fusion with data sets 
obtained from independent field surveys. 
4.2 Handling of Concurrent Information 
The 1/5000 scale paper maps cover a much larger area than the 
1/500 scale maps do. The smaller scale maps enclose the actual 
research area. Following the progress of the archaeological 
research digital elevation information for more and more parts 
of the research area will be captured in future by tacheometric 
high quality measurements. This procedure of continuously 
extending the ground survey unavoidably will result in 
concurrent and possibly in contradictory elevation information 
as compared with the already available digital elevation 
information obtained from the paper maps. In our case we 
assume the accuracy of the ground survey results to be 
substantially higher than the paper map results. Thus, we 
decided to supersede paper map results by ground survey results 
whenever a ground survey was available. 
4.3 Quality Checks 
In our area there are mainly two sources of digital heights 
available, namely the results of the tacheometric field surveys 
and a set of 1/5000 scale paper maps (Siebold, 1998). Both 
sources were used to create a Triangular Irregular Network 
(TIN). Three-dimensional point co-ordinates were taken from 
the tacheometrically measured ground points and from digitised 
contour lines of the paper maps, respectively. When performing 
the fusion between these two data sets several items have to be 
addressed. 
4.1 Common spatial reference frame 
The local reference frame (see section 2) was to be used as the 
one common base system for all data sets. That is why, in a first 
step, all data sets available in other reference systems had to be 
transformed into that target system. For the results of the 
tacheometric surveys this task could be easily performed by 
using a set of common reference points which were used in all 
field campaigns in the same way. The TIN’S which had been 
generated from the paper maps were transformed by using a set 
of points on the maps which could be identified in the field. 
Table 2 shows the residuals of a redundant affine 
transformation with 5 control points. As can be seen the root 
mean square error is in the range of 1 metre, which equals 0,2 
millimetres in 1/5000 map scale, a satisfying measure which 
proofs the quality of the whole data conversion process starting 
The concurrent elevation information available in all areas 
covered by ground survey results so far was used for an in 
depth quality check by analysing the differences of heights 
calculated from the concurrent TIN’s: Heights for the complete 
set of around 17.000 tacheometrically measured 3D points were 
calculated from the paper map TIN. If one compares these 
calculated heights with the tacheometrically measured point 
heights one gets a sound insight into the properties of both data 
sets. Figure 5 shows the spatial pattern resulting from a 
classified plot of height differences. As one can see a striking 
fact is that the hillocks (Buyukkale, KUctikkale, Zegreg Tepe 
and others) in general show positive height differences against 
the pseudo true values of the ground survey TIN. This means 
that the heights calculated from the paper map TIN are lower 
than the heights obtained from ground surveys in that areas. In 
other areas we have to do with the opposite facts: the heights 
obtained from the paper map TIN are larger than the ground 
surveyed heights. Both facts can be explained: in the paper 
maps sometimes there were no spot heights available at the top 
of a hillock, which means, that the contour lines with the 
highest locally available height numbers define the top. The 
remaining height difference up to the real top of the local 
hillock is missing, whilst it is present in the ground survey data. 
The fact that there are areas where the surface of the paper map 
TIN is located systematically above the real ground surface 
may be caused by the map production process: data automation