Integrating Context Priors into a Decision Tree Classification Scheme

Scene interpretation systems need to match (often ambiguous) low-level input data to concepts from a high-level ontology. In many domains, these decisions are uncertain and benefit greatly from proper context. This paper demonstrates the use of decision trees for estimating class probabilities for regions described by feature vectors, and shows how context can be introduced in order to improve the matching performance.

Authors:

• Kasim Terzic
• Bernd Neumann

Keywords:

• Classification
• Decision Trees
• Interpretation
• Vision

References:

[1] F. Fusier, V. Valentin, F. Bremond, M. Thonnat, M. Borg, D. Thirde, and
J. Ferryman, "Video understanding for complex activity recognition,"
Machine Vision and Applications (MVA), vol. 18, pp. 167-188, August
2007.
[2] L. Hotz, B. Neumann, and K. Terzi'c, "High-level expectations for lowlevel
image processing," in Proceedings of the 31st Annual German
Conference on Artificial Intelligence, Kaiserslautern, September 2008.
[3] S. Wenzel, M. Drauschke, and W. F¨orstner, "Detection of repeated
structures in facade images," in 7th Open German / Russian Workshop
on Pattern Recognition and Image Understanding, E. Michaelsen, Ed.
Ettlingen: FGAN-FOM, August 2007.
[4] J. ˇ Cech and R. ˇS 'ara, "Language of the structural models for constrained
image segmentation," Czech Technical University, Prague, Tech. Rep.
Technical Report TN-eTRIMS-CMP-03-2007, 2007.
[5] M. Mohnhaupt and B. Neumann, "Understanding object motion: recognition,
learning and spatiotemporal reasoning," pp. 65-91, 1993.
[6] B. Hummel, W. Thiemann, and I. Lulcheva, "Scene understanding
of urban road intersections with description logic," in Logic
and Probability for Scene Interpretation, ser. Dagstuhl Seminar
Proceedings, A. G. Cohn, D. C. Hogg, R. M¨oller, and B. Neumann,
Eds., no. 08091. Dagstuhl, Germany: Schloss Dagstuhl - Leibniz-
Zentrum fuer Informatik, Germany, 2008. [Online]. Available:
http://drops.dagstuhl.de/opus/volltexte/2008/1616
[7] L. Hotz and B. Neumann, "Scene interpretation as a configuration task,"
KI, vol. 19, no. 3, pp. 59-, 2005.
[8] K. Terzi'c, L. Hotz, and B. Neumann, "Division of work during behaviour
recognition - the SCENIC approach," in Workshop on Behaviour
Modelling and Interpretation, 30th German Conference on Artificial
Intelligence, Osnabr¨uck, Germany, September 2007.
[9] B. Leibe, E. Seemann, and B. Schiele, "Pedestrian detection in crowded
scenes," in Computer Vision and Pattern Recognition, 2005. CVPR 2005.
IEEE Computer Society Conference on, vol. 1, 2005, pp. 878-885 vol.
1.
[10] D. G. Lowe, "Distinctive image features from scale-invariant keypoints,"
International Journal of Computer Vision, vol. 60, pp. 91-110, 2004.
[11] M. Drauschke and W. F¨orstner, "Comparison of adaboost and adtboost
for feature subset selection," in PRIS 2008, Barcelona, Spain, 2008.
[12] M. Drauschke and W. F¨orstner, "Selecting appropriate features for
detecting buildings and building parts," in 21st Congress of the International
Society for Photogrammetry and Remote Sensing (ISPRS),
Beijing, China, 2008.
[13] U. Steinhoff, D. Omercevic, R. Perko, B. Schiele, and A. Leonardis,
"How computer vision can help in outdoor positioning." in AmI, ser.
Lecture Notes in Computer Science, B. Schiele, A. K. Dey, H. Gellersen,
B. E. R. de Ruyter, M. Tscheligi, R. Wichert, E. H. L. Aarts, and A. P.
Buchmann, Eds., vol. 4794. Springer, 2007, pp. 124-141.
[14] F. Korˇc and W. F¨orstner, "Interpreting terrestrial images of urban scenes
using discriminative random fields," in Proc. of the 21st Congress of the
International Society for Photogrammetry and Remote Sensing (ISPRS),
2008.
[15] J. Hartz and B. Neumann, "Learning a knowledge base of ontological
concepts for high-level scene interpretation," in IEEE Proc. International
Conference on Machine Learning and Applications, Cincinnati (Ohio,
USA), Dec 2007.
[16] D. Heesch and M. Petrou, "Markov random fields with asymmetric
interactions for modelling spatial context in structured scenes," Journal
of Signal Processing Systems, to appear, 2009.
[17] F. Korˇc and W. F¨orstner, "eTRIMS Image Database for interpreting
images of man-made scenes," Tech. Rep. TR-IGG-P-2009-01, April
2009.
[18] V. A. Bochko and M. Petrou, "Recognition of structural parts of
buildings using support vector machines," in Pattern Recognition and
Information Processing, PRIP2007, 2007.
[19] L. Hotz, B. Neumann, K. Terzi'c, and J. ˇSochman, "Feedback between
low-level and high-level image processing," Universit¨at Hamburg, Hamburg,
Tech. Rep. Report FBI-HH-B-278/07, 2007.
[20] B. Neumann, "Bayesian compositional hierarchies - a probabilistic
structure for scene interpretation," Universit¨at Hamburg, Department
Informatik, Arbeitsbereich Kognitive Systeme, Tech. Rep. FBI-HH-B-
282/08, May 2008.
[21] A. R. Webb, Statistical Pattern Recognition, 2nd Edition. John Wiley
& Sons, October 2002.
[22] D. Poole, A. Mackworth, and R. Goebel, Computational intelligence: a
logical approach. Oxford, UK: Oxford University Press, 1997.
[23] L. Breiman, J. Friedman, R. A. Olshen, and C. J. Stone, Classification
and Regression Trees. Monterey, CA: Wadsworth and Brooks, 1984.
[24] T. Joachims, "Making large-scale support vector machine learning
practical," in Advances in kernel methods: support vector learning,
B. Sch¨olkopf, C. J. C. Burges, and A. J. Smola, Eds. Cambridge,
MA, USA: MIT Press, 1999, pp. 169-184.
[25] B. Zadrozny and C. Elkan, "Obtaining calibrated probability estimates
from decision trees and naive bayesian classifiers," in In Proceedings of
the Eighteenth International Conference on Machine Learning. Morgan
Kaufmann, 2001, pp. 609-616.
[26] L. R. Bahl, P. F. Brown, P. V. De, and R. L. Mercer, "A tree-based
statistical language model for natural language speech recognition,"
vol. 37, no. 7, Jul 1989, pp. 1001-1008.
[27] W. Buntine, "Learning classification trees," Statistics and Computing,
vol. 2, pp. 63-73, 1992.
[28] B. Cestnik, "Estimating probabilities: A crucial task in machine learning,"
in ECAI, 1990, pp. 147-149.