bul 2004
Cory high
ind most
ctor data
zht lines
d. It can
vrong Z
lisplaced
building
bject is
g height
ents are
lures in
f vector
| image-
ification
' images
pendent
image-
kflow,
.ds used
ds any
as to be
SE miss
ri2ix
ge. The
ı should
road.
o
em
'nomial
ents.
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. Istanbul 2004
3.44 Vector Transmission compared to polynomial
rectified image. An example of Vector Transformation is
shown in fig.3 for a rural area in Denmark with height-level
differences up to 100 meters in that image. The area shown is
far from the image center and close to the corner fiducials.
The same area is shown in fig.4 — but here from a polynomial
rectified image (same GCP's as used for the VT calculationl).
It is obvious that Vector Transmission is showing a much
better fit to objects in the image.
3.4.5 Vector Transformation misalignments. Due to the
nature of the transformation, it is not possible to measure
exact values of the displayed vector data (lengths, areas,
angels etc.) because vector points are individually displaced.
Another problem to be treated is when object types have
different height specifications. Some object types are
measured at object bottom (roads, rivers etc.) and some are
measured at the top of the object (trees, buildings etc.). If two
objects of different height specification (bottom/top) are from
the same geographical coordinates (X, Y), then VT
methodology will show the two at a distance to each other,
because their height is used for calculating their location in
the image. An example is shown in fig 5 (top) where road and
forest border are crossing each other.
To overcome this problem, a function has been implemented
where Vector Transformation is calculated by using height
values from a DTM (already present in Mapheck) instead of
using the object heights. This means that object vectors are
"straightened" out, but also that objects are displaced from
their original position in the image (fig.5 bottom).
Figure 5. Top: Displaced objects caused by different object
height specifications — by Vector Transformation (white trees in
top and orange roads at bottom).
Bottom: VT with objects relatively straightened out, using height
data from DTM
Top + Bottom: Notice cross on “top” of the pipe (in blue circle)
757
3.4.6 Vector transformation, automation. To meet the
needs for expanding the control of geometry to 3D and still
verify large update-areas with little cost and manpower, an
effort in automating the system has been made. Some semi-
automatic procedures are to be mentioned.
Suggestions and automatic zoom are given for the:
- Placement of fiducials, to be used for IO
- Placement of visually good building corners (no shadow
or trees) with good distribution in the image (GCP's), to
be used for calculation of the rotation matrix
Other facilities in the VT rectification workflow are
automatic extraction of TOPIODK data from the database,
covering raw images that are loaded to be rectified (area is
calculated from image GPS-center and image-size). In the
rectification workflow, there is also automatic adjustment of
vector-data when ground control points are measured.
4 SUMMARY AND CONCLUSION
Experience has shown that control of object geometry can be
done with good results in 3D using mono environments, by a
software solution with photogrammetric functions as shown.
The Vector Transformation module shows potential for
further development to manage object measurement and
database updates from the 2D environment. It is planned to
develop a system to use measurements of object pixel
coordinates directly in a raw image in VT mode.
Transformation of the object-pixel-coordinates to UTM
should then be by the VT methodology, using the DTM-
height in addition with a “surface-element-height” given by
the operator.
It is argued that data will gain in quality by control procedure
evaluation in a scheduled plan — also taking into account the
current and end-goal quality of the database content.
REFERENCES
Frederiksen, P., Grum, J., Joergensen , L.T., 2004, Strategies
for updating a national 3-D topographic database and related
geoinformation. /n Proceedings of ISPRS XX.Congress
Istanbul, Turkey.
KMS, 2004a, TOP10DK - Specifikation for fotoflyvning.
KMS, 2004b, TOPIODK - Specification for Geometrisk regi-
stering.
Olsen, B., 2004, Automated change detection for validation
of digital map databases [In Proceedings of ISPRS
XX. Congress Istanbul, Turkey.
AKNOWLEDMENTS
A number of people has been working on implementation of
the Vector Transmission module in Mapcheck at KMS. These
are Jon C. Olsen, Peter Hejholt Sorensen and Svend Wied.
