3. Istanbul 2004
ents at scale
nents at scale
:cordance with
ance and map
er branches,
are used in
subroutine
ode points.
i
wing segments
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004
/
p
; =
bh
= s
* 7
KL T —
se
n
m
k A
E y
>»
Figure 5. Reduced branches at scale 1: 5000
Figure 6. Smoothed branches at scale 1: 5000
In the first step, after the processing of ‘Counter’ subroutine
calculated branches lengths are sent to ‘BranchReduction’
subroutine. The branches of one segment are deleted. In the
second step, ‘Counter’ subroutine are repeated and it calculates
new lengths of branches, then ‘BranchReduction’ subroutine
deletes branches if they are one or two segment in length. This
process is repeated until any branches that are shorter than the
threshold value do not remain.
After the process, segment coordinates can be saved in a text
file as the points of terrain skeleton lines. On the other hand, all
the following segments groups are not smooth enough for map
representation. They indicate sharp turnings that rarely occurred
in land forms. To solve the problem, line smoothing step is
added to *RidgeValleyAxisPicker' program.
3.4 Line Smoothing
The process can result in targets moving to positions that better
match the line trend. Because the new position is an average of
the neighboring points and itself, the shifting distance is never
more than a grid interval. In this way, the process does not
violate the information provided by the gridded data set (Chang,
et al., 1998). Figure 6 shows the results of line smoothing of
lines in Figure 5.
4. CASESTUDY
In case study, *RidgeValleyAxisPicker' has been used to extract
terrain skeleton lines. All of the DEM points are automatically
collected from a stereomodel in 10 meter grid distance for the
map at the scale 1: 5000 (Figure 7). The points are saved in a
text file with their row and column numbers (ie. grid
coordinates) and their elevations. Grid coordinates are obtained
from land coordinates by means of a transformation.
Figure 7. 10m gridded DEM points sample at the scale 1: 5 000
Figure 8 shows the extracted and the smoothed ridge and valley
line results of the *RidgeValleyAxisPicker' program for the
whole study area.
Although the method for automatic extraction of terrain
skeleton lines can be working with a huge amount of data quite
rapidly, there are some cartographic problems on the extracted
lines. The results of ‘RidgeValleyAxisPicker’ program are
evaluated with the harmony of extracted skeleton lines and
contours. Figure 9 shows the result of digitized valley lines for
the same topographical map. The valley lines are only
demonstrated in these examples, because ridge lines are not
usually digitized for topographical maps.
The numbers of the digitized valley lines and the program
results are different because of the effect of branch reduction
step. Some short valley lines are reduced and not demonstrated
in Figure 8 as a result of branch reduction step. However, there
are some unexpected valley lines on hill sides and no valley line
in valley region.
Moreover, the extracted and smoothed valley lines do not bend
precisely on contours. Figure 10 shows this condition clearly.
Many of the extracted valley lines are not entirely appropriate
to contours.