04
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. Istanbul 2004
300
ATKIS: Apa = 3m, Awa = 3m, just extracted road objects.
100 ATKIS segments assessed.
ATKIS: Apa = 3m, Awa = 3m, extracted road objects and
rows of trees. 364 ATKIS segments assessed.
Figure 5: Results for Incorrect ATKIS Data
relations between any ATKIS segment and the assigned
objects are assessed. The given evidence is collected and
combined using Hint-Theory which is an approach to the
Dempster-Shafer Theory of evidence. By this means any
ATKIS segment obtains a certain portion of support and
plausibility expressing its compliance to the model.
First results show that the quality of ATKIS road vector
data is reflected by means of this approach. In further
work it will be investigated how a final verification de-
cision (whether an ATKIS object will be accepted or re-
jected) can be derived from the support and plausibility
measures for the segments. It was shown with the exam-
ples that for an acceptance both values must exceed a cer-
tain threshold, the definition of this threshold is a matter of
further research.
Moreover it has to be investigated whether a substitution of
the simple equipartitioned statistical model for the analysis
of topologic relations by a more individual density function
does improve the analysis.
Another potential improvement concerns the global net-
work aspect: the road network is designed to connect im-
portant places by an optimal path and every road object
gives a certain contribution to this network. In the pre-
sented evidential framework this contribution can be judged
and considered for the assessment.
R09
Concerning the update of the road network, i.e. the de-
tection of roads not currently contained in the database, it
is to investigate to what extent the given approach can be
used. For example, it will be possible to formulate hy-
potheses for new roads based on the accepted network and
additional information from other sources or from road ex-
traction algorithms. Such hypotheses may then be judged
in a similar way the given road network is assessed.
Last but not least the evaluation of object extraction algo-
rithms is very important. In this paper a formal frame-
work for the assessment of road vector data is given, but
the overall result still depends on the quality of input infor-
mation.
ACKNOWLEDGMENT
This work funded by the German Federal Agency for Car-
tography and Geodesy (BKG).
REFERENCES
Dempster, A., 1967. Upper and lower probabilities in-
duced by a multivalued mapping. Anneals of Mathematical
Statistics 38, pp. 325-339.
Gerke, M., Butenuth, M., Heipke, C. and Willrich, F.,
2004. Graph supported verification of road databases.
[SPRS Journal of Photogrammetry and Remote Sensing
58(3-4), pp. 152-165.
Gordon, J. and Shortliffe, E. H., 1990. The dempster-
shafer theory of evidence. In: (Shafer and Pearl, 1990),
pp. 529 —539.
Hinz, S, 2003. Automatische Extraktion urbaner
StraDennetze aus Luftbildern. PhD thesis, Fakultát für
Bauingenieur- und Vermessungswesen der TU München.
Hinz, S. and Baumgartner, A., 2000. Road extraction in
urban areas supported by context objects. In: IAPRS,
Vol. 33, pp. 405-413. Part B3.
Kohlas, J. and Monney, P., 1995. A Mathematical The-
ory of Hints. An Approach to Dempster-Shafer Theory of
Evidence. Lecture Notes in Economics and Mathematical
Systems, Vol. 425, Springer-Verlag.
Mayer, H., 1998. Automatische Objektextraktion aus
digitalen Luftbildern. professoral dissertation, Deutsche
Geodátische Kommission. Series C, Vol. 494.
Shafer, G., 1976. A Mathematical Theory of Evidence.
Princeton University Press, Princeton.
Shafer, G. and Pearl, J. (eds), 1990. Readings in Uncertain
Reasoning. Morgan Kaufmann Publishers.
Winter, S., 1996. Unsichere topologische Beziehun-
gen zwischen ungenauen Fláchen. PhD thesis, Deutsche
Geodátische Kommission. Series C, Vol. 465.
Winter, S., 1998. Uncertain topological relations between
imprecise regions. Technical report, Fachbereich Geoin-
formation, TU Wien.
