International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004
accurate in xy. It would be best to load these data in the future
database, after recovering the z-values from the CAD-files.
The first 720 km?, dating from 1988, had far less detailed contents
than the data in the rest of the country. Furthermore, at the rate
that the landscape has changed, there will be not much left
unchanged. We therefore decided that the datacollection in these
areas should be redone completely.
During the following 15 years, from 1989 till now, the
specifications for the data classification, the stereoplotting and the
data structuring have evolved continuously. We retrieved nearly
all instruction-notes with new specifications valid from a certain
date, but not telling whether the map sheets that were being
produced at that time already took in account the new
specifications or not. Did the operators follow a new specification
for the active or the next working area or only for the next map
sheet ? Even though we kept a record of all begin- and end dates
of each major production step in each working area, we cannot
‘answer this question for sure without verifying the data
themselves.
The updating operators at every stage of the work will need to
learn about the initial situation of the data in a certain area so that
whilst performing the update they may pay attention to certain
changes in the specifications. This illustrates the importance of
metadata. The operators will need clear instructions on what they
should adapt to the new specifications and what they should leave
as it 1s.
3.2 The data for the 1:50.000 map
The 50K data cover the whole country and are already being
updated on basis of field survey and orthophotographs before we
start the ‘Seamless GIS of Reference’. By comparing the
production years of the base map data with the first and second
edition of the 1:50.000 map data, and sorting the results, we
learned that because of the 15 years’ difference in age of the data,
it would be most interesting to study the updates in the map 31/3-
4. This study confirmed that it is very useful to glance over the
50K update data whilst updating the 10K database. Those data are
also an important source of information on the number and kind
of modifications that we may expect in the landscape.
3.3 The external data
At the moment, there are but very few external data sets that could
directly be useful for our project. One of these is the high tension
lines file from ELIA, the public utility which manages the
transport of the high tension power in Belgium. The file was
originally made by us, but it was accurately kept up to date by
ELIA. Hence, it would be best if we could load their data into the
future database. A similar co-operation is being organized with
the national railroad company. In the future, we expect to obtain
more and more external update data e.g. building-update-data
from the cadastre and from the regional authorities, which are
collecting large scale data.
4. COMBINING THE BEST PART OF THE
MOST IMPORTANT DATA SETS
Since stereoplotting will remain a very important updating tool,
the data have to be 3D. For most object types it would be best to
load Top10V-GIS in the future database, after giving z-values
to the data. For many object types, the best way to do so would be
744
to project the ToplOV-GIS data on a very good DTM, c.g.
produced by laser scanning. For some other objects, as well as for
nearly all objects in the many areas where we do not have a laser
scanned DTM, the most accurate method is by recovering the z-
values from the parent CAD-files. This way, we combine the
advantages of the two most important data sets: being
topologically structured, identified and 3D.
For recovering the z-values from the parent CAD-files our
colleague Hugues Bruynseels wrote a VBA-script that uses some
predefined functionalities of ArcMap. In a first step, it converts
the elements of the CAD-files into a cloud of 3D points in an
ArcGIS 8.3 Geodatabase. Then the Top10V-GIS-coverages are
imported into the same Geodatabase and a z co-ordinate (with
default = 0) is given to each pair of xy co-ordinates. Since the
original data were subject to manipulations during the structuring;
small displacements, splittings, creation of additional nodes due
to topological cleaning, the script needs to perform a proximity
search: in a third stage, the application goes over every element of
Top10V-GIS and for each encountered vertex it runs a proximity
search and selects the corresponding triplet (x,y,z) from the point
cloud. The distance in this search is a parameter; we have limited
it to 10 meter. At 10 m the search is interrupted. If no
corresponding triplet was found in the point cloud, the z-value is
set to z — -99, in order to make it easy to detect the problem and
correct it manually. The efficiency of this method varies with the
object categories. It works extremely well for buildings, which
solves our main problem. It does not work as well for the linear
elements, though we hope to improve the results by generating
additional points at intersections in the parent data. There will
always remain a number of objects that have to be projected on a
DEM. The best available DEM has to be chosen in function of the
object type and the area. |
5. THE FUTURE DIGITAL
PHOTOGRAMMETRICAL
WORKSTATION CONFIGURATION
For updating the future ‘Seamless GIS of Reference’- database by
stereoplotting, we need our digital stereoplotters to be linked with
a GIS. This is one of the most important requirements for the
updating implementation; it may be a limiting factor in the choice
of the underlaying DBMS. C. Heipke (2004) described how the
ideal GIS should look like in general, from a photogrammetrical
point of view. We would like to point out some more detailed
requirements for the future stereoplotter configuration. It should
answer positively to the following requirements. Many of them
seem obvious, but we point them out explicitly because they are
not as often available as one would expect.
It should be possible to (N=Necessary, D=Desirable):
N1) visualize the DBMS-contents in 3D
N2) edit the DBMS-contents in 3D
N3) automatically check input attribute values against their list
domain or range domain and give warnings when not OK
N4) automatically timestamp each operation
N5) automatically source stamp each operation with a user group
N6) clean overshoots, undershoots and intersections, making a
vertically projected 2D topology without messing up the z-values
nor the db-linkages: in batch
D7) automatically adapt the corresponding stereo-images when
performing a queued locate on the DBMS-contents
D8) shift the stereomodel in x,y,z in order to make it locally
coincide with the vector data
D9) visualize all elements of a 2D-file in 3D at the height of the
Intern
floatii
D10)
D11)
vertic
nor th
eleme
It sho
functi
N12)
witho
‘area’
N13)
- ford
the O
numb
table
- fora
in attr
- for
Active
incren
attribu
and ac
D14) «
object
select:
D15)
*move
'big n
was m
We dc
combi
Pluger
photos
collect
softwa
3D M
topolo
perfori
mainte
solutic
1) Mic
If it w
Stereo
softwa
Micro:
have t
P
Char
Very
Aver
Rare
