International Archives of Photogrammetry and Remote Sensing, Vol. 32, Part 7-4-3 W6, Valladolid, Spain, 3-4 June, 1999
138
5.4. Future Work
To improve the extraction of halls, the segmented regions are
planned to be splitted in compact regions and approximated by
right-angled polygons fitted to the contours in the image. This
will also yield a more accurate detection of the trucks. For the
extraction of parking lots, cars, and persons, image processing
operators and their interface to the AIDA system have to be
implemented. Concerning the semantic net, the knowledge base
for the fairground example has to be completed including the
definition of computation and judgement methods. The strategy
for the multitemporal image analysis will be tested in detail and
improved, if necessary.
Additionally, a concept will be developed to allow the
monitoring of landuse changes and detection of new
constructions using again temporal relations to model possible
state transitions. This will be tested for the interpretation and
monitoring of moorland areas near Hannover.
Currently, the uncertainty and vagueness of the data is handled
within the semantic net by a possibilistic judgement approach. It
is planned to develop a second judgement calculus based on a
probabilistic belief network (Bayesian net), which exploits the
nodes and edges of the semantic net. Thereafter, a comparison of
the two judgement approaches will be carried out.
To get more accurate segmentation results, a self-adaptive image
processing module based on agents is currently developed. This
system will select, configure and adapt iteratively an appropriate
image processing operator according to a task description
derived from the expectations and constraints of the semantic net.
Finally, the segmentation results matching best with the given
task description will be returned to the semantic net.
6. CONCLUSIONS
A knowledge based scene interpretation system called AIDA
was presented, which uses semantic nets, rules, and
computational methods to represent the knowledge needed for
the interpretation of remote sensing images. Controlled by an
adaptable interpretation strategy, the knowledge base is
exploited to derive a symbolic description of the observed scene
in form of an instantiated semantic net. If available, the
information of a GIS database is used as partial interpretation,
increasing the reliability of the generated hypotheses. The
system is employed for the automatic recognition of complex
structures from multisensor images.
Currently, extensions are made in order to provide a
multitemporal analysis. The use of knowledge about temporal
changes improves the generation of hypotheses for succeeding
time instances and allows for example the extraction of complex
structures like an industrial fairground. The temporal knowledge
is represented in a state transition graph and integrated in the
semantic net. A new interpretation strategy generates hypotheses
for the successor state of an object in the next image, which are
verified in the sensor data. The first results show that the
knowledge based scene interpretation is a promising approach
for the analysis of multisensor and multitemporal images.
REFERENCES
Clement, V., Giraudon, G., Houzelle, S. , Sadakly, F., 1993.
’’Interpretation of Remotely Sensed Images in a Context of
Multisensor Fusion Using a Multispecialist Architecture”, IEEE
Trans, on Geoscience and Remote Sensing, Vol. 31, No 4, pp.
779-791.
Dubois, D., Prade, H., 1988. ’’Possibility Theory: An Approach
to Computerized Processing of Uncertainty”, Plenum Press, New
York and London, 263 p.
F. Kummert, H. Niemann, R. Prechtel, G. Sagerer, 1993.
’’Control and explanation in a signal understanding
environment”, Signal Processing, Vol. 32, No. 1-2, pp. 111-146.
Liedtke, C.-E., Buckner, J., Grau, O., Growe, S., Tonjes, R.,
1997. ’’AIDA: A System for the Knowledge Based Interpretation
of Remote Sensing Data”, 3rd. Int. Airborne Remote Sensing
Conference and Exhibition, Copenhagen, Denmark, pp.
313-320.
Matsuyama, T., Hwang, V.S.-S., 1990. ’’SIGMA : A
Knowledge-Based Aerial Image Understanding System”,
Plenum Press, New York, 277 p.
McKeown, D., Wilson, A., McDermott, J., 1985. ’’Rule-Based
Interpretation of Aerial Imagery”, IEEE Trans, on Pattern
Analysis and Machine Intelligence, Vol. PAMI-7, No. 5, pp.
570-585.
Mees, W., Pemeel, C., 1998. ’’Advances in computer assisted
image interpretation”, Informatica — International Journal of
Computing and Informatics, Vol. 22(2), pp. 231-243.
Pakzad, K., Buckner, J., Growe, S., 1999. ’’Knowledge Based
Moorland Interpretation using a Hybrid System for Image
Analysis”, In: International Archives of Photogrammetry and
Remote Sensing, Vol. 32, Part 3-2W5, Munich, Germany (to be
published).
Quint, F., 1997. ’’MOSES: A structural approach to aerial image
understanding”, In: A. Gruen, E. Baltsavias, and O.P. Henricsson
(editors): Automatic Extraction of Man-Made Objects from
Aerial and Space Images (II), Birkhauser, Basel, pp. 323-332.
Stilla, U., Michaelsen, E., 1997. ’’Semantic modelling of
man-made objects by production nets”, in A. Gruen, E.
Baltsavias, and O.P. Henricsson (editors): Automatic Extraction
of Man-Made Objects from Aerial and Space Images (II),
Birkhauser, Basel, pp. 43-52.
Tonjes, R., Growe, S., 1998. ’’Knowledge Based Road
Extraction from Multisensor Imagery”, In: International
Archives of Photogrammetry and Remote Sensing, Vol. 32, Part
3/1 , Columbus, Ohio, USA, pp. 387-393.
Tonjes R., Growe, S., Buckner, J., Liedtke, C.-E., 1999.
’’Knowledge Based Interpretation of Remote Sensing Images
Using Semantic Nets”, Special ISPRS Commision III issue of
Photogrammetric Engineering & Remote Sensing, July 1999 (to
be published).
