Henning Skriver
2 DATA SET
SAR data from the EMISAR system are used in the project. The Danish airborne EMISAR system is developed at the
Department of Electromagnetic Systems (EMI) at the Technical University of Denmark. The EMISAR system is a
fully polarimetric and interferometric SAR (Christensen et al., 1998). It operates at two frequencies, C-band
(5.3°GHz/5.7°cm wavelength) and L-band (1.25°GHz/24°cm wavelength). The SAR system is flown on a Royal Danish
Air Force Gulfstream G-3 aircraft. The SAR system is normally operated from an altitude of approximately 12.500°m,
the spatial resolution is 2°m by 2°m, the ground range swath is approximately 12°km and typical incidence angles range
from 35 to 60 degrees. The processed data from this system are fully calibrated using an advanced internal calibration
system. In a new research project at EMI a prototype of a new system with a much better spatial resolution is being
developed.
The polarimetric SAR measures the backscattered signal for different combinations of 1) the polarization of the wave
transmitted from the radar antenna, and 2) the polarization of the receiving antenna. If the transmitted signal is
horizontally polarized and the receiving antenna also is horizontally polarized, the resulting signal is called HH (H for
horizontal). Accordingly, the VV signal corresponds to a vertically polarized transmitted signal and a vertically
polarized receiving antenna. If the transmitting and receiving antennas have different polarizations (i.e. H and V, or V
and H, respectively) the resulting signal is said to be cross-polarized, i.e. VH or HV, respectively. The backscattering
coefficient depends on the radar parameters such as frequency, polarization, and incidence angle, and on different
characteristics of the target, such as the geometrical structure and the dielectric properties. The backscattering
coefficient for a bare surface depends on the roughness of the surface, i.e. a smooth surface will not backscatter any
signal toward the radar, whereas a rough surface will scatter the transmitted radar signal in all directions, and part of
that will be backscattered toward the radar. Consequently, a smooth surface (e.g. a lake in calm wind conditions) will
appear dark in the SAR image, whereas a rough surface (e.g. a newly ploughed bare field) will appear bright in the
image. The dielectric properties depend on the composition of the material and the moisture of the target, e.g. a
ploughed field will appear brighter if the soil moisture is larger. The radar s sensitivity to these properties depends, for
instance, on the radar wavelength and the polarization of the radar signal. The latter fact is utilized in the polarimetric
SAR, and the possibility of discrimination between different targets on the ground is improved significantly, when such
data are available.
Polarimetric and interferometric SAR data have been acquired by the EMISAR from 1994 to today for a large number
of scientific applications, such as agricultural crop monitoring, hydrology, monitoring of natural vegetation, studying of
glacier dynamics, dune dynamics, and sea ice monitoring. In connection with the present project, also urban areas have
been covered by EMISAR. Aerial photographs are available for an urban as well as a rural area, together with
appropriate in situ observations. In addition, topographic maps are available in digital form for the two areas. The
complete data set including multitemporal, multifrequency, polarimetric SAR data and aerial photographs for two years
has been geometrically corrected to the same map projection.
3 THEMATIC INFORMATION
Dependent on the size and shape of the objects different categories of methods may be used for the thematic
information extraction. Features like roads, buildings, pylons, rivers, etc. may be enhanced and detected using e.g.
filtering, edge detection, line detection, and point detection methods adapted to the special speckle statistics of SAR
images and utilizing the multi-frequency, multi-polarization, multi-temporal polarimetric and interferometric SAR
information. Using the same range of SAR information, methods for detection of classes like forests, lakes, wetlands,
heath, agricultural areas and urban areas may be applied. In this case, available methods may include traditional
classification schemes, knowledge-based classifiers, classification based on polarimetric decomposition, interferometric
correlation, etc. Some methods and results have been reported but further development is strongly needed including
investigation of the sensitivity to e.g. acquisition time, radar parameters, spatial resolution and environmental
conditions. In addition, the potential of combining SAR images which are acquired at different radar frequencies, at
multiple polarizations, and at different times, as well as of combining polarimetric and interferometric data has to be
investigated further. A very important issue is to study SAR s change detection potential to assist fast, frequent, and
cost-effective updating of existing map information. Another important aspect is to study the possibility to separate
between relevant change detection, e.g. a new building, and irrelevant change detection, e.g. from varying
environmental conditions.
Below, results of the extraction of thematic information from SAR data are described. In view of the available
processing techniques, we define three different approaches for information extraction; namely classification and
308 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part Bl. Amsterdam 2000.
M T ilti badu" Rada s-—--- m MA bx