ity. An
led and
atellite
ne key-
: of the
us-data
ility of
omical,
nection
Urban
rridors.
with 25
ı urban
:d on 3
tion of
rds, in
(e.g.
nalysis
spects,
F future
es.
phases.
d GIS
e study
es are
or each
] from
tabases
nilitary
pts the
d level
nalysis
abases,
mpute
pment.
y areas
pment.
carried
n used
pact of
odel is
namics
insport
zoning
ler the
meters.
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004
Here, the aim is both to predict future land use development
under existing spatial plans and policies, and to compare
alternative possible spatial planning and policy scenarios, in
terms of their effects on future land use development.
2 IMAGE-PROCESSING
As geodetic basement we chose topographic maps 1:25k and
existing digital maps 1:5k and 1:1k. Different types of maps
with different accuracy and scaling exist even with different
coordinate-system. Transformation of ellipsoids and projections
took place. Georeferencation and transformation was done in
TopoL-GIS. With partial transformation method we’ve been
able to get most accurate georeferenced maps. With a partial
transformation, the maps are not only georeferenced, the errors
are reduced, as well as a perfect neighbourhoods and a mistake
minimisation of scanning and paper sheets is achieved. A final
Affin-transformation moved the image to the new coordinates.
The first set of satellite imagery of 2000 was IRS/D with 5m
resolution in panchromatic and 25m in multi-spectral mode.
Two IRS-D pan imageries pictures have been selected to cover
the project area as well as 2 MS-Scenes. The images, delivered
in IMG-Format, have been imported into TopoL. Colour-
composition and pan-sharping was done. For the Bosporus area,
IKONOS imageries have been available as a MS set (blue,
green, red and NIR) with 4m resolutions, and as pan with 1m.
Also here a colour-composition and pan-sharping was done.
Georeferencation of the imageries was done by either globally
or by a triangle-network. Always have been picked up ground-
control points on the map or on already georeferenced
imageries. The year 1988 was prepared on base of Russian
satellite photographs from KVR-1000 camera with 2m
resolution and KFA-1000 camera with 5m resolution. These
data belong to the satellite-based spy-campaign of the 80ies.
The pictures have been bought already scanned with high
resolution (8 um). 2 KFA-1000 satellite photographs cover the
entire project area. The colours are different to normal film-
material; it is describes as pseudo colour. The channels are not
separated.
The ground resolution is between 4 and 6m but in detail the
structure of the films corn can be seen. The quality is fine and
can be compared with IRS imageries of the reference year.
However the contrast is smaller and the city-structure in detail
isn't so clear than in IRS. Then they have been automatically
balanced by RGB-colours in the overlapped zone, framed and
mosaiced to a new raster. 4 photos of the high-resolution
satellite camera KVR-1000 cover the central part of the project
area. The ground-resolution is ranging between | and 3m. The
processing of the imagery was very difficult. Only special
transformation algorithm gave good results. In small areas was
the result acceptable. The historical imageries for the years 1968
and 1945 have been created out of aerial photographs. All
photos have 18x18 cm size. Flying altitude was calculated from
focus length and scale. All slides have been scanned with 600
dpi geometric and 8-bit radiometric resolution. The external
orientation was done according to the topographical maps by
identical points on the photo and the map with the heights from
the maps or taken from the DTM. Relative-, height-, position-,
and full-points can be mixed in the bundle-adjustment method
in the used PhoTopoL-Software. The strong land-use-change
made the identification of control-points especially outside the
urban centre difficult. If already 2 points are picked up, the
system pans into the expected new position but to find enough
suitable points is a kind of detectives work. For the process of
orthophoto-calculation, DDE connection between PHOTOPOL
and ATLAS-DMT Software was done to use the TIN-based
DEM for the calculation and so to get most accurate results. The
orthophotos have been resampled to a resolution of 1 m.
3 DIGITALISATION AND INTERPRETATION
All digitisation and interpretation work was done manually on-
screen. Automatic processes for feature extraction in urban
areas fail if such a detailed legendary is recommended.
Figure 1: Example of the digitised transportation network on
m
Tom ER
5 te 2a 13 ho 7020 0 tm te lm hm mim
4
2
LO de s RN NU. FA Mi eoo Ret
the left (highways in red, other road in pink) and the land-use
data set on the right (red = residential, pink = business)
Line-Layer: First the linear elements (lines and polylines), such
as transportation (road and rail) and river/sea-canal network,
have been mapped according the MOLAND legendary on
screen in TopoL. Topological data-structure was recommended.
Limit was given by the visibility in the satellite-imageries. More
wide linear features have been digitised by a middle line and as
far as they have a bigger width than 25 m, they have been
mapped in the area-layer as well. These lines are often
boundaries for area-objects. So far they have been copied to the
file with the area-database of the processed year. The code of
the line has been stored in a related database.
Area-Layer: The lines and polylines of the line-layer have been
used as basement for the area-layer (polygon with label-point).
All other areas with a homogeneous use according to the
legendary have been defined by the boundary-line and the
labelling-point with the attached database in its middle. The
785
