5
curacies by a boosted
demonstrate that the
; balanced (Table 4).
6% and 93.2% for the
the SVM Fusion vary
i results underlines the
d the complementary
ind the multispectral
g maps seem more
for critical classes are
i performance of the
f the highway in the
gnized in the TM data,
sor approach.
iifferent data sets
data sets, containing
imagery has been
training and further
her parametric and
tes including a
We assume the reason for the success of the presented fusion
concept to be the different nature of the data sources. The
different imagery provides diverse information that seem not
equally reliable. Regarding the experimental results it seems
more adequate to define the kernel functions for each data
source individually, instead of using one specific kernel
function on the full data set. Furthermore the pre-classification
by the first SVM can be assumed as class-specific data
transformation. The imagery is transformed into a new feature
space that consists of distance values of the individual SVM
rule images. These values seem better comparable than the
original imagery and each binary rule image is more adequate
to describe the interclass differences.
Moreover it has been shown that SVM outperforms other
classifiers in term of accuracies when classifying single source
data sets. Hence it can be assumed that a modification of the
proposed classification framework could further increase the
classifier performances. In this context the prior image
segmentation and use of multilevel classification concepts
would be worth investigatation. Furthermore the impact of an
increased multispectral time series seems interesting.
Generally, the results clearly demonstrate that a combination of
multitemporal SAR data and multispectral imagery is
worthwhile. Here, the classification accuracy is significantly
increased by such a multisensor data set, irrespective of the
applied classifier algorithms. Thus the development of adequate
approaches for classifying multisensor imagery will be an
ongoing research topic in remote sensing.
REFERENCES
Benediktsson, J. A. and Kanellopoulos I., 1999. Classification
of Multisource and Hyperspectral Data Based on Decision
Fusion. IEEE Trans. Geoscience and Remote Sensing vol. 37,
pp. 1367-1377.
Benediktsson, J. A., Swain, P. H, and Ersoy O. K., 1990. Neural
Network Approaches Versus Statistical Methods In
Classification Of Multisource Remote Sensing Data," IEEE
Trans. Geoscience and Remote Sensing, vol. 28, pp. 540-552.
Blaes, X., Vanhalle, L., and Defoumy, P., 2005. Efficiency of
crop identification based on optical and SAR image time series.
Remote Sensing of Environment, 96, pp. 352-365.
Briem, G.J., Benediktsson, J.A., and Sveinsson, J.R., 2002.
Multiple Classifiers Applied to Multisource Remote Sensing
Data. IEEE Trans. Geoscience and Remote Sensing, vol. 40, pp.
2291-2299
Brown de Colstoun E. C., Story M. H., Thompson, C.,
Commisso, K., Smith, T. G., Irons, J. R., 2003. National Park
vegetation mapping using multitemporal Landsat 7 data and a
decision tree classifier. Remote Sensing of Environment, vol. 85,
pp. 316-327.
Camps-Valls, G., Gomez-Chova, L., Munoz-Mari, J., Vila-
Frances, J., Calpe-Maravilla, J., 2006. Composite kernels for
hyperspectral image classification. IEEE Geoscience and
Remote Sensing Letters, vol. 3, pp. 93-97.
Chen, C.-C. and Lin, C.-J., 2001. LIBSVM: a library for
support vector machines. Software available at:
http://www.csie.ntu.edu.tw/~cjlin/libsvm.
Chust, G., Ducrot, D., & Pretus, J. L., 2004. Land cover
discrimination potential of radar multitemporal series and
optical multispecral images in a Mediteranean cultural
landscape. International Journal of Remote Sensing, 25, pp.
3513-3528.
Fauvel, M., Chanussot, J., Benediktsson, J. A., 2006. A
Combined Support Vector Machines Classification Based on
Decision Fusion. Proceedings IEEE Geoscience and Remote
Sensing Symposium, 2006, Denver, Colorado, USA
Foody, G. M. and Mathur, A., 2004. A Relative Evaluation of
Multiclass Image Classification of Support Vector Machines.
IEEE Trans. Geoscience and Remote Sensing, 42, pp. 1335-
1343.
Freund, Y. and Schapire, R. E., 1996. Experiments with a new
boosting algorithm,” Proceedings 13th Intern. Confe. Machine
Learning, 1996.
Gislason, P.O., Benediktsson, J.A., and Sveinsson, J.R., 2006.
Random Forests for land cover classification,” Pattern
Recognition Letters, vol. 27, pp. 294-300.
Halldorsson, G. H., Benediktsson, J. A., Sveinsson, J. R., 2003.
Support vector machines in multisource classification.
Proceedings IEEE Geoscience and Remote Sensing Symposium,
2003, Toulouse, France.
Huang, H., Legarsky, J., Othman, M., 2007. Land-cover
Classification Using Radarsat and Landsat Imagery for St.
Louis, Missouri. Photogrammetric Engineering & Remote
Sensing, vol. 73, pp. 37-43,
Janz, A., Schiefer, S., Waske B., and Hostert, P., 2007.
imageSVM - A user-oriented tool for advanced classification of
hyperspectral data using support vector machines. Proceed. 5th
workshop of the EARSeL Special Interest Group on Imaging
Spectroscopy 2007, Bruges, Belgium.
Melgani F. and Bruzzone, L., 2004. Classification of
hyperspectral remote sensing images with support vector
machines. IEEE Trans. Geoscience and Remote Sensing, vol.
42, pp. 1778-1790.
Quinlan, J. R., 1993: C4.5: Programs For Machine Learning.
Los Altos: Morgan Kaufmann.
Richards, J. A, 2005. Analysis of Remotely Sensed Data: The
Formative Decades and the Future. IEEE Trans. Geoscience
and Remote Sensing, vol. 43, pp. 422-432.
Scholkopf, B., and Smola, A., 2002. Learning with Kernels.
MIT Press, Cambridge, MA.
Song, X., Fan, F., and Rao, M., 2005. Automatic CRP mapping
using nonparametric machine learning approaches”, IEEE
Trans. Geoscience and Remote Sensing, vol. 43, pp. 888-897.
Vapnik V. N., 1998. Statistical Learning Theory. Wiley, New
York.
Waske, B., Schiefer, S., and Braun, M., 2006. Random feature
selection for decision tree classification of multi-temporal SAR
data. Proceedings IEEE Geoscience and Remote Sensing
Symposium, 2006. Denver, Colorado, USA.
