International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004
following: black color layer (map grid, objects, text, map sheet
description, spot heights, etc), blue layer (hydrography), brown
layer (contours and terrain forms), green layer (vegetation) and
some other special layers (masks for hydrography, vegetation
and road classifications). For building DTM only first three
layers are used.
MST ARRETE
J ^ Y oW CRX
qi Su ka tinack, Ix
One of the basic characteristics of 1:25000 topographic maps is
that there is a very long cycle of updating these maps for the
whole territory. Therefore, for some of these maps the last
update was done twenty years ago. Fortunately, this is
something that is not so dangerous when DTM that
corresponds to 1:25000 maps is considered. Of course, there
are certain areas such as open mine areas that would require
special updating of existing map data.
The major data acquisition method for height data mapping for
1:25000 maps was aerial photogrammetry and most of the data
was captured from 1955 to 1965. The aerial photography was
done in 1:33000, 1:30000, and 1:26000 photo scales. Absolute
orientation of photo models for some maps was done without
prior aerial triangulation, so this could be one of the important
sources of errors. Direct contouring in dynamic mode was
mostly done by using analog photogrammetric plotters. Major
contour interval is 10m, but in flat areas there are also contours
on 5m and 2.5m interval.
3. DTM DATA ACQUISITION
Map layers containing height data were scanned together with
all the other map layers within a single campaign. Drum
scanner was used. Scanning resolution for most map layers was
508 DPI. This unusual scanning resolution is chosen to get map
pixel with some logical metrics (0.05mm on map, 1.25m on the
field).
There were two options for digitizing height data. One of them
was to use separate map layers with height data, and the other
was to use color paper map with all the layers merged. First
option was preferable, and it was selected, because it was
much more convenient to use automatic and semi-automatic
procedures for contour vectorization without prior color
separation, which was required for the merged color map.
Disadvantage was that there were only four map corner marks
available for georeferencing height map layers. This was the
case for all the map layers, except for the black layer (contains
map grid in Gauss-Kriiger projection, and meridians and
parallels). This was a problem, because it was not possible to
use advanced methods for determination and removal of map
distortions caused by distortions of the map material and
scanning errors. Some tests were carried out and it has been
estimated that these errors were about 4-5m (0.16-0.20mm on
a map). Since it was not possible to detect and remove these
errors using only four corner marks, these errors would
contribute to the overall error budget of the DTM. For map
layers with available map grid these errors were analyzed and
systematic part of the total error, regardless of the source (map
distortion or scanning errors) was successfully removed from
the map image.
Contour vectorization was done by using semi-automatic, line-
following technique. This was relatively fast, and most suitable
considering the type of available contour maps. Contour
heights were assigned manually to the digitized contours
during vectorization. Problems that were experienced during
this stage were related to the vectorization of maps for certain
terrain types featuring lot of stones, extremely high slopes and
lot of small pits. However, this process was rather
straightforward in general, and it can be said that there were
very few critical issues. Several workstations were dedicated
for this task, working 16 hours per day. Also, all available spot
heights are manually digitized. Spot heights data will be useful
for improving DTM quality, but also for the detection of gross
errors made during initial data capture and during contour
vectorization and height assignment and spot height
digitization. The whole process for contour vectorization and
spot height digitization took about 18 months, including data
verification.
PLUS E MN
LE MTS cs
Na
LI
n -- tay
c o P. «
’ v
PIG)
2
AS = .. : > E = lh
Ve ^o WI C Me . mr pe RENI
M n : me d M t
M P AA A - d 3 - Sa .
E
ES
; ii STAN
Figure 2. Typical example of problematic area for contour
vectorization from contour map layer
^
"f
A
7,
st
After all map layers with contours were vectorized contours
were checked and their connections on map borders are
corrected. This was done by using standard CAD tools.
4. SOFTWARE FOR DTM ANALYSIS
Since it was the necessary to provide as many workstations for
DTM capture, verification and editing as possible, it was
decided to use some relatively low-cost DTM software. The
other requirements were related to the possibility of unlimited
software customization and development of new algorithms for
652
Intern
terrain
has m
softwa
softwa
specifi
survey
are SU]
require
RDBM
projec
develo
Facult
planne
develo
countr
SurfIn;
analys
modul
all the
Visual
is a ste
interfa
DTM
(Triang
conforı
terrain
form
accom]
triangle
triangle
standar
interpo
interpo
data co
5. D
Consid
operato
data di
stage o
probler
procedi
