International Archives of the Photogrammetry,
Roadway
Cont. Stripe
gv1
ext1
p1
Tw
part-of t ; part-of N part-of
left boundary part-of central part-of right boundary
[ Roadway | [Lane Marking] [Double | [Lane Marking] _ | Roadway |
|MargnLeft| S | tef | 5 [Whiteline T | Right | 5 Margin Right
Cont. Stripe — _ | Periodic Line! < _| Cont.Line — Cont.Line € Cont. Stripe |
bee o o «Gi
gv6 a gv2 a gv2 Ed gv5 = gv6
ext = ext2 Pe 2*ext2 © ext5 o ext6
p1 2 .! [D nh | 7 p1 | = up?
Figure 8. Instance Net for Dual Carriageway for 0.9 m/pel
In the next stage of the net adaptation process, the net consists
merely of the roadway itself - the only feature that is still left
detectable at that scale. The feature extraction operator
connected to the roadway node will now be called to extract a
continuous stripe, as the roadway has become the bottom node.
At last, the object type changes from continuous stripe to
continuous line before the line vanishes and there is no roadway
or part thereof extractable in the example scene.
Roadway |
Cont.Stripe |
gvl
ext1
p1
partof 77 part-of Es partit
left boundary central right boundary
“Roadway
Margin Right |
Cont Line
1 gv6
"Roadway | f Double |
| Margin Left | | White Line
Cont.Line Cont.Line
= > a
gv6 gv2
ext6 2"ext2 ext6
Df LL pt
Figure 9. Instance Net for Dual Carriageway for 1.7 m/pel |
left-of [2*d1]
|
right-of [2*d1]
|
|
6. CONCLUSION
In this paper an approach to derive object models for low
resolution images from models created manually for high
resolution images was presented. After an overview about the
general strategy of the procedure we focussed on the
composition of the semantic nets and suggested some
constraints, in order to be able to handle the semantic nets
automatically regarding scale adaptation. The prediction of the
scale behaviour of object types requires investigations on the
scale behaviour of feature extraction operators, which we
presented for three operators. At last, we described an example
for scale change events observed in a scene and their impact on
the semantic net. This example demonstrates the suitability of
the proposed kind of semantic net to follow the scale space
events in digital images, and thus, its applicability in an
automatic approach.
Future work will deal with the specification of the exact steps
of an automatic scale adaptation of semantic nets. Furthermore,
we intend to work on extensions of the described semantic nets
to new object types, and the impacts on the semantic net
creation rules. In addition we want to work on the
implementation of the nets into the knowledge based system
GeoAIDA (Bückner, 2002, Pahl, 2003, successor of AIDA).
Regarding the investigation of the feature extraction operators
(section 4) an exact simulation of sensor data in different
resolutions would require the incorporation of more complex
models than we used. We assume that the used procedure is
sufficient for our task. Yet, this assumption still has to be
Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004
verified by using real sensor images. Eventually, the
comparison of our results with predictions from scale space
theory is surely interesting.
7. ACKNOWLEDGEMENTS
This work has been funded within a project by the Deutsche
Forschungsgemeinschaft under grant HE 1822/13.
8. REFERENCES
Baumgartner, A., 2003. Automatische Extraktion von Strafen
aus digitalen Luftbildern. DGK, Reihe C, Dissertationen, No.
564, München, 91 p.
Bückner, J, 2003. Ein wissensbasiertes System zur
automatischen Extraktion von semantischen Informationen aus
digitalen Fernerkundungsdaten. Schriftenreihe des TNT der
Universität Hannover, Band 4, ibidem-Verlag, Stuttgart, 162 p.
Canny, J., 1986. A Computational Approach to Edge Detection.
IEEE Transactions - on Pattern Analysis and Machine
Intelligence, 8(6), pp. 679-698.
Deriche, R., 1987. Using Canny's Criteria to derive a
Recursively Implemented Optimal Edge Detector. International
Journal of Computer Vision, 1(2), pp. 167-187.
Liedtke, C.-E., Biickner, J., Grau, O., Growe, S., and Tónjes,
R., 1997. AIDA: A system for the knowledge based
interpretation of remote sensing data. 3rd. Int. Airborne Remote
Sensing Conference and Exhibition, Vol. Il: pp. 313-320:
Lindeberg, T., 1994. Scale-Space Theory in Computer Vision.
Kluwer Academic Publishers, The Netherlands, 423 p.
Lindeberg, T., 1998. Edge Detection and Ridge Detection with
Automatic Scale Selection. International Journal of Computer
Vision, 30(2), pp. 117-154.
Niemann, H., Sagerer, G., Schröder, S. and Kummert, F., 1990.
ERNEST: a semantic network system for pattern understanding.
IEEE Transactions on Pattern Analysis and Machine
Intelligence, 12(9), pp. 883-905.
Pahl, M., 2003. Architektur eines wissensbasierten System zur
Interpretation multisensorieller Fernerkundungsdaten.
Schriftenreihe des TNT der Universität Hannover, Band 3
ibidem-Verlag, Stuttgart, 145 p.
Pakzad, K., 2001. Wissensbasierte Interpretation von
Vegetationsflächen aus multitemporalen Fernerkundungsdaten.
DGK, Reihe C, Dissertationen, No. 543, München, 104 p.
Schiewe, J., 2003. Auswertung hoch auflósender und multi-
sensoraler Fernerkundungsdaten. Habilitationsschrift,
Fachgebiet Umweltwissenschaften, Schwerpunkt
Geoinformatik, Hochschule Vechta, 159 p.
Steger, C., 1998. An Unbiased Detector of Curvilinear
Structures. IEEE Transactions on Pattern Analysis and
Machine Intelligence, 20(2), pp.113-125.
Tonjes, R., 1999. Wissensbasierte Interpretation und 3D-
Rekonstruktion von Landschaftsszenen aus Luftbildern.
Dissertation, Universitit Hannover, Fortschritt-Berichte VDI,
Reihe 10, No. 575, VDI-Verlag, Düsseldorf, 117 p.
Witkin, A., 1986. Scale space filtering. In Pentland, A. (Ed.):
From Pixels to Predicates. Ablex Publishing Corporation, New
Jersey, pp. 5-19.
KEY \
ABSTI
In this
Auckla
unsupe
image
isolate:
If no r
found,
a certa
disjoin
width |
threshc
Roads
within
update
meet t
remote
resolut
platfor
imager
and of
extract
renewe
road ng
Extract
accom
is subj
recogn
incom]
study :
consun
inform
though
Autom
challer
the sp
extract
object:
Vehicl
difficu
So far
roads
catego!
progra
