There are, however, also disadvantages for the combination with electro-optical data:
The side-looking mode makes it difficult to register radar data with nadir images; the
speckle effects caused by the coherent electromagnetic wavelength make interpretation problematic; and the image acquisition process of the active sensor predominantly caused by the roughness and the orientation of the terrain provides completely
different imaging effects. Despite these problems, radar remote sensing may be an
alternative for panchromatic remote sensing data for image fusion if no other images
are available. Consequently, iconic fusion with radar data has been practiced as early
as in the 1980s (Welch and Ehlers, 1988; Harris et al., 1990; Ehlers, 1991).
Preliminary results from the fusion of 1-m-resolution TerraSAR-X and 2.4-mresolution multispectral QuickBird data for the pyramid fields of Giza, however,
showed that even though there was some structural enhancement, an improvement in
spatial resolution was not really visible (Klonus and Ehlers, 2008). This poses the
immediate question as to what extent these two types of data are comparable in terms
of terrain feature interpretation and subsequently to radar-based pan sharpening. It
was also important to know what role the spatial resolution plays in both types of data
and how different resolutions may affect data fusion.
Terrain feature interpretation is difficult to compare between radar and electrooptical systems, given the differences in image acquisition. While the optical sensors
detect the reflected sunlight from objects, mostly in the visible and infrared wavelengths, the radar satellites produce coherent electromagnetic radiation in the microwave region that is then received as they bounce against the objects. Sensor and object
orientation, distance from the object, and relative movement are important factors for
radar remote sensing. Based on the preliminary results, it is evident that spatial
resolution has to be differently defined for EO and radar data. Consequently, we
investigated a number of representative terrain objects of different types and sizes
chosen in a scene in Spain and compared in terms of visibility and identifiability in
both types of images. The study area is located in the North of Spain representing the
region around Santo Domingo de la Calzada. This area was also used as a control site
of the JRC and provided EO remote sensing data as well as a TerraSAR-X scene
(Ehlers et al., 2010). We made use of an IKONOS panchromatic image and a
TerraSAR-X scene—both with a nominal spatial resolution of 1 m. The IKONOS
image was acquired May 30, 2005, and theTerraSAR-X spotlight image on May 3,
2008. The TerraSAR-X image was despeckled with a proprietary algorithm developed at the University of Würzburg in Germany.
A terrain feature was considered visible when it was possible to differentiate it
from the background and was considered identifiable when it could be precisely
recognized on the image. Features that were highly variable in time or across season,
such as agricultural fields, were not considered. Of all the features that were visible in
IKONOS at 1 m ground resolution (being power lines the smallest feature), only
features of size 7–10 m or larger could be seen in TerraSAR-X. This is exemplified by
features such as the main road and a single tree (Figure 2.17 and Table 2.3).
In the case of the dirt road, the low visibility in TerraSAR-X is probably due to the
lack of contrast in the surrounding areas due to only small differences in surface
roughness. Except for the main road, none of the features were identifiable in
FUSION OF ELECTRO-OPTICAL AND RADAR DATA
27
The side-looking mode makes it difficult to register radar data with nadir images; the
speckle effects caused by the coherent electromagnetic wavelength make interpretation problematic; and the image acquisition process of the active sensor predominantly caused by the roughness and the orientation of the terrain provides completely
different imaging effects. Despite these problems, radar remote sensing may be an
alternative for panchromatic remote sensing data for image fusion if no other images
are available. Consequently, iconic fusion with radar data has been practiced as early
as in the 1980s (Welch and Ehlers, 1988; Harris et al., 1990; Ehlers, 1991).
Preliminary results from the fusion of 1-m-resolution TerraSAR-X and 2.4-mresolution multispectral QuickBird data for the pyramid fields of Giza, however,
showed that even though there was some structural enhancement, an improvement in
spatial resolution was not really visible (Klonus and Ehlers, 2008). This poses the
immediate question as to what extent these two types of data are comparable in terms
of terrain feature interpretation and subsequently to radar-based pan sharpening. It
was also important to know what role the spatial resolution plays in both types of data
and how different resolutions may affect data fusion.
Terrain feature interpretation is difficult to compare between radar and electrooptical systems, given the differences in image acquisition. While the optical sensors
detect the reflected sunlight from objects, mostly in the visible and infrared wavelengths, the radar satellites produce coherent electromagnetic radiation in the microwave region that is then received as they bounce against the objects. Sensor and object
orientation, distance from the object, and relative movement are important factors for
radar remote sensing. Based on the preliminary results, it is evident that spatial
resolution has to be differently defined for EO and radar data. Consequently, we
investigated a number of representative terrain objects of different types and sizes
chosen in a scene in Spain and compared in terms of visibility and identifiability in
both types of images. The study area is located in the North of Spain representing the
region around Santo Domingo de la Calzada. This area was also used as a control site
of the JRC and provided EO remote sensing data as well as a TerraSAR-X scene
(Ehlers et al., 2010). We made use of an IKONOS panchromatic image and a
TerraSAR-X scene—both with a nominal spatial resolution of 1 m. The IKONOS
image was acquired May 30, 2005, and theTerraSAR-X spotlight image on May 3,
2008. The TerraSAR-X image was despeckled with a proprietary algorithm developed at the University of Würzburg in Germany.
A terrain feature was considered visible when it was possible to differentiate it
from the background and was considered identifiable when it could be precisely
recognized on the image. Features that were highly variable in time or across season,
such as agricultural fields, were not considered. Of all the features that were visible in
IKONOS at 1 m ground resolution (being power lines the smallest feature), only
features of size 7–10 m or larger could be seen in TerraSAR-X. This is exemplified by
features such as the main road and a single tree (Figure 2.17 and Table 2.3).
In the case of the dirt road, the low visibility in TerraSAR-X is probably due to the
lack of contrast in the surrounding areas due to only small differences in surface
roughness. Except for the main road, none of the features were identifiable in
FUSION OF ELECTRO-OPTICAL AND RADAR DATA
27