Geology Reference
In-Depth Information
(a)
(b)
σ
0
vv
σ
0
hh
(c)
(d)
Griffith
Island
OW
Smooth FY ice
Thin ice
Rough FY ice
γ
co
N
Figure 10.6
ENVISAT ASAR image of an area surrounding Cornwallis Island, Nunavut, Canada, in 25 April 2004. All
images are enhanced using Gaussian stretch. The composite image in (d) is generated with RGB channels correspond-
ing to
0
0
, and
γ
co
, respectively. Ice types marked in (d) were verified in the field. [© 2004 European Space
Agency; adapted from
Geldsetzer and Yackel
, 2009].
,
hh
vv
co‐polarization parameters and the simple thresholding
technique.
Polarimetric SAR data are the most promising source
for ice type classification. That is because of the availa-
bility of several parameters and matrices that describe
the backscatter from individual pixels (section 7.6.2.3).
Two major approaches can be pursued in ice classifica-
tion. They can be used separately or combined. In the
first approach, multipolarization parameters can be used
in any standard classification scheme such as the super-
vised Bayesian maximum‐likelihood or the unsuper-
vised ISODATA algorithm. The parameters can be the
direct linear polarization backscatter components
(
derived from the covariance or coherence matrices as listed
in equations (7.97) and (7.99
)
.
In the second approach the entire matrix representa-
tion of polarimetric SAR signal, namely the covariance
or coherence matrix, is utilized (i.e., not just selected
decomposition parameters). This approach should be
better because it utilizes the full suite of the polarimetric
information in the received signal to derive the polarimet-
ric parameters
. Lee et al.
[1994] showed that the inclusion
of the phase in the classifier's parameters improved the
classification significantly (the classification in that study
was of an image of an urban area). The two approaches
have been explored in several studies. A brief account
of each approach is introduced in the remainder of this
section.
, , ) or any set derived from equations (7.81)-
(7.86). It can also be any decomposition parameters
0
0
and
0
hh
vv
hv
