International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. Istanbul 2004
2000). In this sense consistency is part of logical consistency as
defined above. Some aspects of completeness cover logical
consistency, too, e.g. attribute completeness which concerns the
question whether all required attributes are stored together with
an object Most aspects of completeness, correctness, and
accuracy must be checked in comparison to reality. It has to be
verified, whether e.g. all objects are registered in the data set
and whether their attributes are set correctly. Accuracy concerns
positional accuracy and temporal accuracy, i.e. currency.
1.2 Quality Management at BKG
A major task of BKG consists in providing the geodata of the
Authoritative Topographic-Cartographic Information System
ATKIS on the territory of the Federal Republic of Germany.
ATKIS' is a trademark of the Working Committee of the
Surveying Authorities of the States of the Federal Republic of
Germany (AdV). The most important components of ATKIS are
object-based digital landscape models (DLM) encompassing
several resolutions and digital topographic maps (DTK). The
ATKIS DLMBasis, ie. the ATKIS data offering highest
resolution, is produced by the 16 surveying authorities of the
federal states of Germany and is delivered to the BKG. Here, at
the Geodata Centre (GDC) of BKG, the ATKIS DLMBasis is
checked with respect to logical consistency, joined to one
homogeneous data set for the territory of the Federal Republic
of Germany, and stored in a database. Since these data are
delivered to customers on the one hand and are used to derive
data of smaller scales within BKG on the other hand, a system
for quality control of the ATKIS data is essential. For testing
logical consistency of the data sets an exhaustive check on the
full coverage of the data is performed at the GDC. This check
includes establishing logical and geometrical consistency of the
different data sets from the surveying authorities at their
borders. Thus, the check of logical consistency is done in an
operational way within the daily production process. Errors
detected during the quality control are reported to the respective
federal state. Since the federal states are producers of the data of
the ATKIS DLMBasis, they are responsible for the appropriate
amendment of the data.
Since nearly all of the changes in the real world are man-made,
information concerning the changes is available very early,
usually already during the phase of planning. Therefore, the
surveying authorities of Germany are forcing topographical
information management to gather information about changes
and make it available in time for the update of geoinformation.
In addition to this well established, broad inspection of logical
consistency of the ATKIS dataset, BKG pushes the increase of
quality control with respect to reality. Therefore, BKG has
initiated a common project with the University of Hannover to
develop a system for automated quality control comparing
orthoimages and the ATKIS DLMBasis (Busch and Willrich,
2002). Since in practice the comparison to the real world still is
far away from being fully automatic, it is implemented as an
interactive procedure based on ArcGIS on the one hand and on
automated image analysis methods on the other hand. The
knowledge-based image interpretation system GeoAIDA
(Buckner et al. 2002) and various methods for feature extraction
form the core of the automated procedures. These programs run
separately in batch mode furnishing results that are imported to
the interactive working environment. While the final decision
about errors is reserved to a human operator, the strategy is to
detect as many coincidences of ATKIS objects and objects
detected in the orthoimages as possible. By filtering these
736
correct situations the human operator can concentrate on the
objects where the automated procedure failed. The comparison
utilises orthoimages of recent date which are an up-to-date
reference of reality and can be used to assess completeness,
correctness, positional and temporal accuracy. Our main interest
concerns objects where most changes arise and that are
important, namely the road network and built-up areas. Other
objects and their attributes can be verified, too, if they are
visible within the images.
The paper focuses on a method for quality control of the ATKIS
DLMBasis that automatically compares the data with reality by
means of images from an independent source. Thus, we look at
quality control as an independent procedure to rate the quality
of geodata by sample and to detect deficiencies within the chain
of production. Depending on the type of image, e.g. airborne or
satellite imagery, different features and attributes can be
verified. Orthoimages of recent date are an up-to-date reference
of reality and can be used to assess temporal accuracy, too.
Orthoimages are a very suitable source to determine the
positional accuracy of features and geometric attributes like the
width of a road. Nevertheless, there exist features and attributes
that are not detectable in the image data. In these cases quality
control has to be based on the topographic information
management.
1.3 Automated Verification and
Databases by Means of Imagery
Update of Spatial
Automatic and semi-automatic feature extraction has been a
focus of international research in photogrammetry and computer
vision for a few decades (e.g. Baltsavias et al. 2001, Heipke et
al. 2004). As a consequence the results are now starting to enter
into the commercial market. Obviously algorithms particularly
give good results if applied to well-defined application areas.
The reason is that all approaches need additional knowledge in
the form of appropriate models, which can more easily be
formulated for restricted situations. Since any automatic feature
extraction algorithm will show a certain error rate, it has to be
integrated to an interactive workflow leaving final decisions to a
human operator. For achieving an efficient workflow, the
algorithms have to be equipped with an appropriate and reliable
self-diagnostics allowing the operator to concentrate on
situations where the automatic procedure failed. Walter (2004),
for instance, developed a system that supports the operator in
quality control of region and line objects in ATKIS by
automatically extracting land cover classes from satellite
imagery by multi-spectral classification, and comparing them to
the corresponding ATKIS objects. He uses prior information
derived from the existing ATKIS dataset for defining training
sets for a supervised classification. ATKIS objects showing a
high probability of differences to the extracted object classes are
indicated as presumed changes. They are visualized for
supporting the human operator’s final interactive analysis. GIS
data in general can provide. a valuable source of prior
information (e.g. Vosselman 1996) and can be used to stabilize
the image interpretation tasks.
The basic concept relies on a knowledge-based system for
image interpretation. Knowledge-based systems have proven to
be a suitable framework for representing knowledge about
objects and exploiting it during the recognition process. Our
system models structural dependencies by semantic networks
(Bückner et al. 2002). It has been successfully applied to land
cover interpretation by means of orthoimages, laser DEMs, and
prior information from a GIS.
Interna
In our
reality,
verifica
charact
e Tt
a |
mi
the
edt 1f
an
e Ot
Thus,
measur
object r
of the
Hence,
In the
algorith
that is c
3. Y Ce
The cot
and th
followit
e Tr.
. Ef
WO
The sys
knowle
from ph
and m:
concept
tasks wi
Current!
by step
research
Hannov
that are
further i
The full
quality «
those ol
Thus, th
automat
consumi
Challen;
the hum
The res
operator
the resu
attribute
