nge
op co tae i UA iue I i unti nd
clearly visible. The advantages between the approaches may seem
only moderate. However, in the major part of the pit, a separation
between the two limestone pit classes seem to be quite simple and
therefore, all approaches yield good results in a large part. Fur-
thermore, it has to be kept in mind that a major source of error
comes from confusion between the limestone pit and natural out-
crops of limestone or open chalky soils. Confusion between these
landcover types narrow the differences between the change de-
tection approaches. It should be noted though, that false positives
based on this source of error are much less for kernel-composition
approaches and more concentrated to single spots. According to
McNemar's test, the advantage in overall accuracy of the kernel-
composition approach over the other approaches are significant.
The reason for the advantage of the kernel-composition approach
and the stacked-features approach over post-classification change
detection is straightforward. While post-classification change de-
tection does not include any information on the change in pixels
intensities between the two points of time, both kernel-composition
and stacked-features do incorporate this information implicitly.
However, the advantage of kernel-composition over the stacked-
features approach is remarkable. Both approaches include in-
formation on the change in pixels intensities. However, kernel-
composition appears to be a better suited technique to exploit
this information. It is assumed that the main advantage lies in
the fact, that the kernel-composition represents this information
in the RKHS, while the stacked-features approach represents it
in the original feature space. Since SVMs operate in the RKHS
when finding their optimal solution, kernel-composition and SVM
seem to be a more suitable combination for representing this im-
plicit information.
6 CONCLUSIONS
Kernel based change detection is a conceptually elegant and use-
ful method for change detection and multi temporal classification.
Standard techniques like image differencing can be executed in
RKHS, thus benefiting from the advantages of kernel based SVM
classification. Changes in landuse for the given dataset from Up-
per Rhine Graben can be visualized and furthermore quantified
with high precision. In future work, the approach will be tested
on more complex change detection problems.
REFERENCES
Almutairi, A. and Warner, T., 2010. Change detection accuracy
and image properties: A study using simulated data. Remote
Sensing 2(6), pp. 1508-1529.
Amari, S. and Wu, S., 1999. Improving support vector ma-
chine classifiers by modifying kernel functions. Neural Networks
12(6), pp. 783—789.
Bandos, T., Zhou, D. and Camps-Valls, G., 2006. Semi-
supervised hyperspectral image classification with graphs. In:
Proceedings of the IEEE Int. Geoscience and Remote Sensing
Symposium, 2006. IGARSS’ 06.
Boser, B., Guyon, I. and Vapnik, V., 1992. A training algorithm
for optimal margin classifiers. In: Proceedings of the fifth annual
workshop on Computational learning theory, ACM, pp. 144—152.
Bovolo, E, Bruzzone, L. and Marconcini, M., 2008. A novel
approach to unsupervised change detection based on a semisu-
pervised svm and a similarity measure. IEEE Transactions on
Geoscience and Remote Sensing 46(7), pp. 2070-2082.
Bovolo, F., Camps-Valls, G. and Bruzzone, L., 2010. A support
vector domain method for change detection in multitemporal im-
ages. Pattern Recognition Letters 31(10), pp. 1148-1154.
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XXXIX-B7, 2012
XXII ISPRS Congress, 25 August — 01 September 2012, Melbourne, Australia
285
Braun, A., Weidner, U., Jutzi, B. and Hinz, S., 2011. Integrating
external knowledge into svm classification - fusing hyperspec-
tral and laserscanning data by kernel composition. Vol. Heipke
C, Jacobsen K, Rottensteiner F, Mller S, SO2rgel U (Eds) High-
resolution earth imaging for geospatial information. International
Archives of Photogrammetry, Remote Sensing and Spatial Infor-
mation Sciences 38 (Part 4 / W19) (on CD).
Bruzzone, L. and Serpico, S., 1997. An iterative technique for
the detection of land-cover transitions in multitemporal remote-
sensing images. IEEE Transactions on Geoscience and Remote
Sensing 35(4), pp. 858-867.
Bruzzone, L., Cossu, R. and Vernazza, G., 2004. Detection of
land-cover transitions by combining multidate classifiers. Pattern
Recognition Letters 25(13), pp. 1491-1500.
Burges, C., 1998. A tutorial on support vector machines for pat-
tern recognition. Data mining and knowledge discovery 2(2),
pp. 121-167.
Camps-Valls, G. and Bruzzone, L., 2009. Kernel methods for
remote sensing data analysis. Wiley Online Library.
Camps-Valls, G., Bandos Marsheva, T. and Zhou, D., 2007.
Semi-supervised graph-based hyperspectral image classification.
IEEE Transactions on Geoscience and Remote Sensing 45(10),
pp. 3044-3054.
Camps- Valls, G., Gomez-Chova, L., Mufioz-Mari, J., Alonso, E.
Calpe-Maravilla, J. and Moreno, J., 2006a. Multitemporal image
classification and change detection with kernels. In: SPIE Inter-
national Symposium Remote Sensing XII, Vol. 6365, p. 63650H.
Camps-Valls, G., Gómez-Chova, L., Mufioz-Marí, J., Rojo-
Álvarez, J. and Martínez-Ramón, M., 2008. Kernel-based frame-
work for multitemporal and multisource remote sensing data clas-
sification and change detection. IEEE Transactions on Geo-
science and Remote Sensing 46(6), pp. 1822-1835.
Camps-Valls, G., Gomez-Chova, L., Mufioz-Mari, J., Vila-
Francés, J. and Calpe-Maravilla, J., 2006b. Composite kernels
for hyperspectral image classification. IEEE Geoscience and Re-
mote Sensing Letters 3(1), pp. 93-97.
Chang, C., Lin, C. et al, 2001. Libsvm: a library for support
vector machines.
Chen, Y., Nasrabadi, N. and Tran, T., 2011. Sparse representation
for target detection in hyperspectral imagery. IEEE Journal of
Selected Topics in Signal Processing 5(3), pp. 629-640.
Coops, N., Gillanders, S., Wulder, M., Gergel, S., Nelson, T. and
Goodwin, N., 2010. Assessing changes in forest fragmentation
following infestation using time series landsat imagery. Forest
Ecology and Management 259(12), pp. 2355-2365.
Coppin, P, Jonckheere, I., Nackaerts, K., Muys, B. and Lam-
bin, E., 2004. Review articledigital change detection methods in
ecosystem monitoring: a review. International journal of remote
sensing 25(9), pp. 1565-1596.
Cortes, C. and Vapnik, V., 1995. Support-vector networks. Ma-
chine learning 20(3), pp. 273-297.
Demir, B. and Erturk, S., 2010. Empirical mode decomposition
of hyperspectral images for support vector machine classification.
IEEE Transactions on Geoscience and Remote Sensing 48(11),
pp. 4071-4084.
Dianat, R. and Kasaei, S., 2010. Change detection in optical re-
mote sensing images using difference-based methods and spatial
information. IEEE Geoscience and Remote Sensing Letters 7(1),
pp. 215-219.
