International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B5. Istanbul 2004
Figure 6: These 2 mm DEM grids are from the same area that
can be seen in figure 4.
3.4 Image mosaic
The image mosaic was created with Geomatica OrthoEngine
software using 25 color images with about 5% overlap. The
gridlines, which were visible in the edges of the images, were
bounded off so that they would not be visible in the mosaic.
The images were orthorectified using the digital elevation
model. The cut lines between the images were selected to be the
edge features and the color balance to be adjusted in the
overlapping areas of adjacent images. The orthorectified
geocoded images were then automatically joined to the image
mosaic. The images seemed to fit well together and the mosaic
looked seamless.
The resolution of the mosaic we produced first was 0.1 mm
although the images were scanned with pixel size of 20 um.
The full resolution mosaic would have been too heavy to be
handled in our computers (over 2 Gb). If the best resolution is
used, an image pyramid for multi resolution representation
should be created.
3.5 Triangular network
A triangular network with point spacing of 4 mm was created.
Coordinates of the digital elevation model were exported to
ImageStation Stereo Display software and triangulated. The
image mosaic was draped over the network. As a result we got a
textured 3D surface model, which can be rotated and examined
from different angles.
Figure 7. The image mosaic draped over the triangular network.
3.6 Stereo mosaic
The image mosaic was converted to stereo pairs to be viewed
both in a digital stereo workstation as well as an anaglyph
image. The stereo pairs were created using the digital elevation
model. First the DEM was converted to left and right side
DEMs shifting the pixels to left or right in proportion to the
elevation. In other words the x-coordinates were changed in
proportion to the z-coordinates. We used proportions + 3/10 and
+ 6/10. The bigger the shift of the x-coordinates is the more the
stereoscopic effect is emphasized. The new coordinate values
were interpolated to equidistant points to form new left and
right side DEMs. These DEMs were used to create new left and
right side ortho images from the original images. The new ortho
images were merged to new left and right side image mosaics
from which the anaglyph version was composed
4. DISCUSSION AND CONCLUSIONS
The photogrammetric reproduction of the Map of Mexico was
created using similar techniques as in terrain mapping. In
general these techniques worked. well and the developed
procedure is usable for creating accurate reproductions of maps.
However in our project the interior orientation has need for
improvement. Because the original edges of the images were
cut off the principal points could not be defined. We didn't
either know the other camera parameters than the focal length.
In order to produce accurate reproduction, camera calibration
and use of original image format is crucial.
The poorly defined interior orientation as well as possible errors
in the control information affects the block geometry. This can
be seen as edges in the digital elevation model. The edges are
still slightly visible in the seam areas of adjacent DEM pieces,
though the average elevation values were used in the
overlapping areas and the edges were filtered with a smoothing
algorithm.
The full resolution digital reproduction of the Map of Mexico
(pixel size 20 pm) is planned to be shown in exhibitions as a 3D
surface model or as anaglyph images. The user interface will be
interactive so that users can get a close view into the details and
have a feeling of touching the map. The digital reproduction
will give an experience of being in front of a high-resolution
image of the map.
Interne
Inter
Brown
Techni
Librar
denote
althou
skin, W
the ma
skin.
Larsso
' Aziec:
León-|
Tecno
1986,
Refer
Haggr
http://
graph