Digital Signal Processing Reference
In-Depth Information
Fig. 10.6
Real polarimetric
image in the Pauli basis
misclassified when using the incorrect euclidean definition of the mean. This calls
strongly for an application of the geometry of information on a real scene, where
a lot of polarimetric mechanisms can be encountered and where the opportunities
for misclassification abound. This is why the algorithms used in the simulations in
Sect.
10.3
have been applied on a data set acquired by the ONERA RAMSES system
in the region of Brétigny, France. This image, presented on Fig.
10.6
has a resolution
of approximately 1.3 meters in both range and azimuth and has been acquired in
X-band with a 30
◦
incidence angle.
10.4.1 Classification Results
Wishart distance measure
The image on Fig.
10.6
is represented in the Pauli basis: blue for the
+
HH
VV
−
component, red for the
component. This
decomposition allows for a very quick assessment of the polarimetric mechanisms in
an image. Indeed, a strong
HH
VV
and green for the
HV
HH
+
VV
component is characteristic of scattering over
a surface, a strong
HH
−
VV
component indicates double-bounce scattering and
a strong
component indicates volume scattering, which is generally present in
forested areas. However, color mixtures are very subject to interpretation and should
be handled carefully.
In this image, a lot of polarimetric diversity is present. Two large buildings can
be observed in the left part of the image, one in purple and white, the other in green.
Smaller buildings and several fields are scattered among the image. A forested area
is present on the right side. On the lower right side, four corner reflectors have a
strong signal in red.
HV
