inbul 2004 
had to be 
raphs was 
of the map 
ed camera 
^ Sinacron 
2,5 cm, so 
25 cm. 
with grid 
reads and 
ges. Each 
S" color 
2.5) and 9 
e between 
6 cm. The 
50 that the 
'aphs. The 
ly and for 
sional, an 
'e mosaic. 
:graphy of 
ziewing of 
corded for 
th was 22 
yrresponds 
eters. One 
w oblique 
1g. In total 
'ere taken. 
h a grid 
or 20 um. 
size of 0.1 
wich 5 x 
he original 
this was a 
would be 
1e images. 
s for the 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B5. Istanbul 2004 
  
  
  
Figure 3. The 20 cm x 25 em grid cell is visible in the image. 
3.2 Orientation of images 
Totally 24 images and their 28 stereo pairs with overlap of 55% 
were oriented to one block with Pictran software. These images 
were colored except two black and white images, which had to 
be used because the corresponding color images were missing. 
The resolution of the images were scaled half of the original 
size, to pixel size of 40 jum, so that the orientation process 
would not become too heavy. 
3.2.1 Interior Orientation: Because the original edges of 
the images were cut off, the exact coordinates of principal 
points could not be defined. In this case even the estimate that 
the principal points would be in the center of the images was 
not very accurate because some edges were cut more than the 
others and all the images were also of different size. That is 
why we defined the initial coordinates of the principal points to 
be (0,0) and let the value be a free variable in the block 
orientation. We didn’t either know the other internal camera 
parameters than the focal length and they were thus set to zero. 
3.22 Block Orientation: All the 52 images were oriented 
to one block. The gridline intersections, which were visible in 
all the images, were used as control information (Figure 4). 
Totally 5 xyz control points and 4 z control points were 
measured to the edges and to the center of the image block. 
Also 527 unknown tie points were measured manually from the 
images. The standard deviation of image coordinates was 0.68 
pixels. 
    
Aix > ah 
Figure 4. The gridline intersections were used as control 
information. Because the grid was lifted a few 
centimetres up from the map surface, the 
intersections appear to be a little bit different places 
on different images. The size of these pieces is about 
2,0 cm x 3,5 cm and the same area can be seen in 
the upper right corner of the figure 3. 
3.3 Digital elevation model 
Geomatica OrthoEngine software was used for generating a 
digital elevation model (DEM) of the map. First the original 
images (pixel size 20 um) were resampled to epipolar images. 
Then the elevation models were extracted automatically from 
the overlapping epipolar pairs. The pieces of the extracted 
elevation models were geocoded and combined automatically 
together to form a mosaic DEM. In the areas where adjacent 
DEM pieces had overlap, the average elevation value was used. 
After the combination there were some visible edges left in the 
seam areas of the adjacent DEM pieces. They were bounded 
and filtered with a smoothing algorithm. The pixel size of the 
final DEM is 2 mm and it includes 219 000 points. The 
elevation varies within 2 cm. 
  
Figure 5: The digital elevation model of the Map of Mexico. 
Light areas are higher than dark areas.