this method, the most important thing is
definition of extreme value in the enclosed area.
An undulation curve was used for extreme value
estimation in this study. An undulation curve is
expressed a relationship between radius from top
of a mountain and the undulation. The undulation
means difference between minimum elevation and
maximum elevation. It is represented one of the
topographical feature. If geological structure is
homogeneity, the undulation curve becomes almost
same with neighbor mountains. So, when acreage
of enclosed area could be calculated, maximum
elevation can be estimated.
Finally, a complete DEM can be generated
from combination of previous two interpolated
DEMs. This method doesn't need searching profile
or definition of calculation window size. So, this
will be more reliable method. Figure 6 shows
shaded image of generated DEM by this developed
method using previous contour map. In this figure,
there are no noisy area.
4. RESULTS AND EVALUATIONS
Evaluations of buffering method were
carried out by comparison with existing methods
which are profile method and window method.
Items for evaluation are not only elevation but also
topographical feature such as slope gradient and
slope aspect, stream pattern and slope stability.
Verification data were generated from 5m
grid DEM by profile method using 1:25000
topographical maps. The map had enough number
of contour line to apply any interpolation methods.
The generated 5m grid DEM were resampled to 50m
grid size by taking the average in order to make
more suitable DEM for verification. 100m, 200m,
300m and 400m interval contour line maps were
created from the verification DEM. Those contour
line maps were used for DEM generation by using
each interpolation method from each contour map.
4.1 Elevation
An index of elevation accuracy is used
percentages of correct pixels on the whole pixels. In
case of elevation evaluation, correct pixel means
difference with verification DEM indicates inside
of 20m. Figure 7 shows relationship between
contour line interval and correct percentage in each
method. In this figure, accuracy has tendency to
drop with increasing contour interval. And
buffering method is always located the highest
accuracy in all contour intervals. Though window
method shows almost same accuracy with
Correct percentage (96)
buffering method in 100m contour interval, the
accuracy becomes about 10% less in other contour
intervals. Profile method shows the worst results,
because radiated noises influence to accuracy.
90.00 m
— -$- - - Profile
80.00 | --_y-- Window
70.00 L —3J»— Buffering
60.00 FE
50.00 %
40.00 EL
i.
30.00 } “eg,
7
20.00 A L 1 L L i A i
50 100 150 200 250 300 350 400 450
Contour line interval (m)
Figure 7 Relationship between contour line interval
and correct percentage of elevation
42 Slope Gradient
A slope gradient can be expressed from DEM,
which is one of the most important items for
topographical analysis. In this study, the slope
gradient means maximum inclination at one target
pixel. An index of slope gradient accuracy is also
used percentages of correct pixels. In case of slope
gradient accuracy, correct pixel means difference
with verification slope gradient data indicates
inside of 20 degree. Figure 8 shows relationship
between contour line interval and correct
percentage in each method. Buffering method is
always located the highest accuracy in all contour
intervals. Though profile method shows almost
same accuracy with buffering method in 100m
contour interval, the accuracy becomes 10% less in
other contour intervals. The window method shows
the worst results, because steep slope along contour
line influenced to accuracy.
80.00 = ;
—-$- - - Profile
70.00 | --Jü-- Window
60.00 L —f— Buffering
40.00 } Wo .
= VIDE S25
- i]
30.00 } —e
20.00 L i 1 L A L ri —
50 100 150 200 250 300 350 400 450
Contour line interval (m)
Figure 8 Relationship between contour line interval
and correct percentage of slope gradient
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996
