! 2004
rule of
on the
is size
n like
get the
1
LL]
such a
dividual
e range
nes and
a much
yn roads
2 pixels
is also
colour
complex
ip scale.
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004
Figure 5: IKONOS pan
size of element on road: 2 x 3
pixels
Figure 6: IKONOS pan: size
of road separation lines
0.2 pixels
In figure 4, upper right, in the multispectral IKONOS image
with 4m ground pixel size, buildings with a red roof can be
identified even if they do have only a size of 2 x 2 pixels. The
neighbourhood of the buildings do allow also a save
interpretation. Without the support by the colour the
identification of the buildings is quite more difficult. In figure
4, lower right, the multispectral IKONOS image has been
changed to grey values. In this image the detection and
interpretation of single buildings is quite limited and needs a
size of 5 pixels.
IRS-1C
Figure 7: streets in urban areas
examples in figure 8 do show very clear the difference between
detection and interpretation — if we do have the information
about the location of a road from the SPOT image, we can see it
also in the IRS-1C image. The visible fractions of the road can
be connected if we do have some information about the
location.
IRS-1C
Figure 8: roads in rural areas
SPOT 5
Topographic maps with smaller scale do not show individual
buildings, only building blocks or even only the build up area.
The identification of the build up area is not a problem with all
used space images. The identification of building blocks even
can be made with ASTER images having 15m ground pixel
size.
The identification of the road network is very important for
topographic maps in the scale range of approximately 1:50000.
As obvious in figure 3, in the ASTER and the TK350 images
the major roads can be identified but not the minor roads. This
is quite different in the images starting with IRS-1C and smaller
pixels. In the build up areas in IRS-1C images not in any case
the streets can be seen, but the structure of buildings includes
the information of a street between two lines of buildings
(figure 7 left). The slightly smaller pixel size and better image
quality of SPOT 5 (figure 7 right) shows quite better the details
of the street network.
In rural areas not 10096 of the roads could be identified in the
IRS-1C image (figure 8, left). Here we do not have major
problems with SPOT 5 (figure 8, right). On the other hand, the
Of course with the better resolution of IKONOS and KVR1000
there are no problems with the identification of the minor road
network. The IKONOS images are always in the range of a
competition with aerial images. Standard aerial images do have
a photographic resolution of approximately 40 line pairs/mm.
Based on experiences this can be compared with 80 pixels/mm
or 12pm pixel size in the image. Corresponding to this a ground
pixel size of 1m is available in aerial images with a scale | : 80
000, or QuickBird images do correspond to a scale of the aerial
photos of 1 : 50 000.
Figure 9: water courses in
near infrared band of
ASTER
In the test area Zonguldak not so many smaller water courses
are available. For the mapping of water courses the spectral
range is very important. In the near infrared band, there is
nearly no reflection of the energy from the water bodies, that
means, the water courses are black and do have a very good
contrast to the neighbourhood (figure 9).
In Wegmann et al 1999 the information contents of an IRS-1C
image has been compared with aerial photos 1 : 12 000. The sun
elevation of the used space image was very low, so the quality
was not so good like in the area of Zonguldak. In the IRS-1C
image 56% of the road length was recognised and correctly
classified. 9% has been classified as path and not as road, so
finally 35% could not be seen. A higher percentage of the not
visible roads by error were just identified as field separation. A
smaller percentage was covered by trees in a forest. By this
reason also in the large scale aerial images 6% of the road
length could not be seen. 5% was classified as path. Compared