International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004
classification concept in this paper, two kinds of practical de Kok, R. and Wever, T. 2002. Automatische
Informationsgewinnung aus einheitlichen Megadatensátzen am
Beispiel des IRS-IC/IRS-1D-Mosaiks von Deutschland. In:
Dech, S. et al (eds.): Tagungsband 19. DFD-Nutzerseminar,
investigations about transferability of rule sets were presented.
The first is a transferability test which objective is to come up
with a number of classification rules for specific feature classes
and test their transferability to other images of each geographic Oktober 2002, pp. 105-112.
region. Following the above mentioned classification rules were
analvsed to set up a base rule set in a specific geographic eCognition User Guide from Definiens Imaging.
region. http://www.definiens-imaging.com/documents/index.htm
T (accessed April 2004).
The extensive transferability test demonstrated the feasibility
and high potential of transferring knowledge-based
classification rules to establish a semi-automatic workflow for
object extraction with a high level of automation. It can be
Esch, T., Roth, A., Strunz, G., and Dech, S. 2003. Object
oriented Classification of Landsat-7 Data for Regional Planning
Purposes. In: The International Archives of the
summarized that when transferring rule bases region specific Photogrammetry, Remote Sensing and Spatial. Information K
adaptations are necessary but they can be done in short time. Sciences, Regensburg, Germany, Vol. XXXIV-7/W9, pp. 50-
This leads to a clear gain of time in comparison to an individual 55.
definition of the rule base for each new image. ‘The pq ; , eri Al
classification rules comprise next to spectral mainly texture Harlich, RM. Shanmugan, R., Dinstein, I. (1973). Textural
properties for area features and shape properties for linear Features DE Image Classification, in; {RES dM System In
features like roads and river. The practical investigations with Man and Cybernetics, Vol. SMC-3, No. 6, November 1973, pp. pr
the base rule set showed also the feasibility to define 610-621. u
knewiedze-based classification rules valid for one specific Hofimann, P. 2001. Detecting buildings and roads from fre
geographic region. IKONOS data using additional elevation information, am
Pause. : FE ui GeoBIT/GIS, 6, 2001, pp. 28-33. ne
The availability of a base rule set brings the following vic
advantages: Kressler, F., Franzen, M., and Steinnocher, K. 2003. fro
Automationsgestützte Erfassung der Landnutzung in Städten ab
- Semi-Automation of the image classification: it is not mittels objektorientierter Auswertung von Lufibildem. In: sin
necessary Lo define an individual rule set for each new Strobl, J., Blaschke, T., and Griesebner, G. (eds.): Angewandte
image Geographische Informationsverarbeitung XV. Beiträge zum
- Gain of time AGIT-Svmposium, Salzburg 2003, Herbert Wichmann Verlag,
- . Cost reduction for data acquisition Heidelberg, pp. 250-255.
- No expert user required In
Mitri, G.H., Gitas, 1.Z. 2002. The Development of an Object- bot
The above advantages come at the expense of the classification Oriented Classification Model for Operational Burned Area des
accuracy which is less than the accuracy of an individual mapping on the Mediterranean Island of Thasos Using Landsat the
definition of the rule set. TM Images. In: Viegas, D.X. (ed.): Forest Fire Research & sci
Wildland Fire Safety, Millpress, Rotterdam, 2002, p. 79 inp
There should be further investigations in an operational data (Abstract), Paper on CD. par
acquisition environment to assess the time necessary for arc:
postprocessing the results from individual and from transferred ACKNOWLEDGMENTS geo
classifications. It is believed that the gain of time resulting from the
the transfer of a rule base will be much higher than potential This work has been made possible because of the Bundeswehr 200
additional work because of the lower accuracy. Other Geoinformation Office (AGeoBw), Euskirchen, Germany, and
investigations will deal with the verification of the base rule set the Federal Office of Defense Technology and Procurement In
in new independent test sites of the same geographic region and (BWB), Koblenz, Germany, financial and continuous support. reg:
the definition of a base rule set for another geographic region, buil
namely desert areas in Algeria. (Me
oft
an (
REFERENCES 200
Woo
Congalton, R. G., Green, K. 1999. Assessing the Accuracy of inve
Remotely Sensed Data: Principles and Practices. Lewis extr
Publishers, Boca Raton, Florida, 1999. and
prof
Corr, D.G, Walker, A, Benz, U,, Lingenfelder, I., and
Rodrigues, A. (2003). Classification of urban SAR imagery In c
using object oriented techniques. In: Proceedings of IGARSS bour
2003 IEEE, July 2003, Toulouse, Volume I, pp. 188-190. adva
upda
Darwish, A., Leukert, K., Reinhardt, W. (2003). Image fusic
Segmentation for the Purpose of Object-Based Classification. reco)
In: Proceedings of IGARSS 2003 IEEE, July 2003, Toulouse, units
Volume Ill, pp. 2039-2041. In (
appr
1064
