Olaf Hellwich
For the second example, 23 ERS-1/2 intensity data sets were acquired during the growing seasons of 1996, 1997, and 1998.
Figures 8 a) to c) show a section of three ERS-2 images taken in May, June and J uly 1996. Furthermore, a Landsat TM
data set of April 1996 was available (Fig. 7 a) and Fig. 8 d)). First, the ERS data was subjected to a principle component
analysis (Fig. 7 b)). Using the first three principle components the data was clustered into eight classes. The classification
results (Fig. 8e)) show a high degree of fragmentation due the speckle effect. Then, agricultural field segments were
extracted from the Landsat data. As in the first example, inside of the field segments a majority filter was applied to
the landuse classification results. The results of this step are shown in Figure 8f). A comparison with the original ERS
classification (Fig. 8e)) reveals that segment boundaries resulting from the ERS data should have been added to the
segment boundaries stemming from the Landsat TM data before applying the majority filter.
Figure 7: a) Landsat TM image (red: band 4, green: band 3, blue: band 7), b) first principal components of multitemporal
(23 dates) ERS-1/2 SAR data.
5 CONCLUSIONS
An approach to scene interpretation using multisensor fusion was introduced. It is suggested to extract geometric details
from high-resolution panchromatic optical imagery, material properties from multispectral optical imagery, details of
object development in time from SAR data, and to solve remaining ambiguities with the help of three dimensional data.
This work will be continued to further integrate the various methods of object extraction and classification into a Bayesian
network framework using uncertainty information provided by the individual methods.
ACKNOWLEDGMENTS
The authors thank E. Aigner, F. Kurz, and R. Ludwig for providing some of the data sets.
REFERENCES
Bruzzone, L. and Serpico, S. B., 1997. An Iterative Technique for the Detection of Land-Cover Transitions in Multitem-
poral Remote-Sensing Images. IEEE Transactions on Geoscience and Remote Sensing 35(4), pp. 858-867.
394 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000.
