di
|
i
qi
4
il
Î
1
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV. Part B2. Istanbul 2004
3.3.2 Second step - reading camera calibration and
measuring fiducials — to determine IO. Camera calibration
parameters are stored in the database and by connecting the
image to its corresponding camera calibration, the user must
measure pixel coordinates for all fiducials found in the image
(fig.1, red crosses). To facilitate the manual measurement,
Mapcheck pans and zooms in, close to the image location of
the fiducials — one by one. Then the operator only has to
point out its exact image pixel location.
3.3.3 Third step — verification of image rectification and
archiving metadata. It is possible to verify the precision of
the VT parameters (IO and rotation) calculated for an image,
by draping the database vector data as was used for image
rectification. This makes it possible to judge, if points are
measured correctly. When ground control points and fiducials
are accepted, metadata is put in the database.
3.4 VT Control procedure examples
To illustrate different aspects of the implemented VT
methodology, examples are presented and discussed next:
3.4.1 Image-overview and -loading. With model vector
data loaded as a map in Mapcheck, it is possible to get a
visualization of all images that are ready for VT module
prepared, and cover all or part of the model (fig.2).
From this, one can choose the best image offering best
coverage of the current model. When "right clicking? with
the mouse, it is possible to load that raw image, having its
photo center closest to the mouse cursor position (fig.2). It is
possible to toggle between all raw images that are ready for
VT presentation in this way. Loading a raw image with
vector data on top, only takes a few seconds.
3l eis aee MAS A. Ga HO RR RA VE HIA ROTE jus 0s
Figure 2. Shown is a block of update-data to be controlled -
together with the images, ready for photogrammetric
transformation of vector data (X, Y, Z) in that area. Image-
coverage are shown with red lines. These are shown
photogrammetrically, from the photocenter of the active photo
3.4.2 Single object error. Having vectordata active and
loading a raw image in VT mode, the vector data are draped
on the image as defined in the photogrammetrical metadata
stored in the database (calculated IO and D, as in 3.3, from all
ready verified TOP10DK objects).
756
As shown in fig.3, the VT module gives a very high
correlation between vector data and the raw image, and most
objects are matched exactly to their appearances. Vector data
is at their positions even when they are not straight lines
because the photogrammetric displacement is adapted. It can
be seen that when a object is digitized with a wrong Z
coordinate, then the vector-object will be shown displaced
away from its correct image-position (fig.3, a small building
at top). The height of that displaced building object is
measured as 0 meters, where as the correct building height
should be around 70 meters asl.
3.4.3 Systematic object errors. When displacements are
general in a model, the reasons could be failures in
parameters for the photogrammetric transformation of vector
data or it could be systematic error concerning the image-
orientations (AT setup) used by contractors. VT-rectification
parameter errors can be checked by using neighboring images
covering the same area or alternatively using independent
ground control points for the calculation.
Systematic errors due to errors in the original image-
orientation (AT) could be hard to recognize in this workflow,
because data from the database, might include errors, is used
for calculating the rectification parameters. If there is any
doubt about the database data-quality, then control has to be
done by use of independent ground control points.
Figure 3. Example of Vector Transformation on raw image. The
vector building shown close to silo has wrong Z value (0 m should
be 72 m), and is displaced from its correct position by the road.
sion
UM
DRM ;
Figure 4. Same area as shown in fig. 3, from the polynomial
rectified image (same fiducial and GCP°s). Big displacements.
Internat
3.4.4
rectifiec
shown i
differen
far fron
The san
rectified
It is ob
better fi
345 1
nature €
exact vi
angels e
Another
differen
measure
measure
objects :
the sar
methode
because
the ima;
forest bx
To over
where \
values f
using th
“straigh
their ori
Figure ©
height sy
top and «
Bottom:
data fro:
Top + Be
