IEEE Geoscience and Remote Sensing Magazine - September 2022 - 130

accurately inferred using the ,,m |}
^h three-component
decomposition [26], Bhattacharya expressed the total received
power of partially polarized waves in terms of the
Stokes vectors of the transmitting and receiving waves and,
in 2015, proposed the polarized power fraction Ω considering
both the transmitting and receiving wave properties
|} ,
using m as the DoP parameter, which was widely employed
in ocean applications, such as ship and oil spill detection
[52]-[55]. Bhattacharya and Dey innovatively extended the
meaning of the DoP by proposing Ω and m0
^h and defined it as the ratio of the polarized received
power to the total received power. Also, Bhattacharya proposed
the opposite-sense circular polarization received
power OC and the same-sense CP received power SC representing
odd- and even-bounce scattering, respectively,
based on the analysis of the odd- and even-bounce scattering
properties (odd-bounce, or Bragg, scattering reverses
the transmitting circular polarization, while even-bounce
scattering produces two offset phase reversals, resulting in
strong backscatter from the same circular polarization).
Thus, the surface scattering power and the dihedral scattering
power are denoted as
PO ,== and the
sd
}}
$$ ,CP SC
volume scattering power is Pg1v=-} 0^h [20].
In 2020, Kumar improved the Bhattacharya method by
introducing the circular polarization ratio (CPR) parameter
=
/
)
)
P
.P
ing power is suppressed by the nonpolarization fraction
,
^h1 }),
CPRSCOC
to measure whether surface scattering
(CPR 11 or dihedral scattering (CPR 12 dominates in
Based on this judgment, the nondominant scatterand
the dominant scattering power is obtained by
In
subtracting the nondominant scattering power from P .P
the absence of dominant scattering power (CPR 1= the
scattering model degenerates to the Bhattacharya model
[50]. Dey introduced the Barakat formula into the covariance
matrix in 2021, proposed the 2D Barakat polarization
degree
m ,0 and constructed the second parameter i
to characterize the scattering type information from the
target by using the OC and SC proposed by Bhattacharya,
and the decomposition results are formally similar to the
m-based method [51].
A comparison of the WDT-based methods is provided
in Table 8. It can be seen that the WDT-based method is
essentially a two-parameter approach. The first parameter,
which can be called the DoP parameter, is related to the nature
of the wave and used to decompose the span g0
the sum of a completely polarized component PP
completely depolarized component PUP
into
, where PUP
and a
is assigned
to the volume scattering component. The second
parameter, which can be called the decomposition parameter,
is used to decompose the obtained PP
into the sum of the
surface scattering component and the dihedral scattering
component. Note that g3
appears in the decomposition
parameter of all methods, and its meaning is the average
receiving power of the circular polarization. In the modelbased
method introduced in the following, the sign of g3
is used by many techniques as a discriminator of whether
surface scattering or dihedral scattering dominates in the
residual scattering component with volume scattering removed.
Early studies focused on improving the decomposition
parameter in the residual scattering component by
130
CONSTRUCT SCATTERING MODEL
Except for Guo, who first wrote the scattering model directly
based on the relationship among the covariance elements,
subsequent researchers have built scattering models based
on the relationship between the FP covariance matrix or
the FP coherence matrix and the Stokes vector, which also
takes into account the natural advantages of the Stokes vector
for CP SAR data. The methods of constructing the scattering
model are provided in Figure 13, and the scattering
model is mainly implemented by CC
CTLR ! FP gC ,
,
andgT .
CTLR ! FP
As shown in Figure 13, the scattering model is constructed
according to the following three main ideas. One
is
CTLR ! FP
CC . Analyze the Freeman-Durden model with
the FP covariance matrix:
IEEE GEOSCIENCE AND REMOTE SENSING MAGAZINE SEPTEMBER 2022
CTLR ! FP
, respectively.
Bhattacharya also proposed OC and SC parameters to
characterize the parity of the target polarity variation, and
they are related to the surface scattering component and the
dihedral scattering component, respectively. From the decomposition
results, the WDT-based methods assign all the
completely depolarized components to the volume scattering
components and neglect the possible depolarization effects
caused by surface scattering and dihedral scattering,
suffering from the problem of overestimation of bulk scattering
in urban areas.
MODEL-BASED METHODS
Model decomposition refers to modeling the fundamental
scatterers and then decomposing the covariance matrix into
the sum of the fundamental scatterer models. The method
has obvious physical significance. The model decomposition
theory is oriented to the FP covariance matrix or coherence
matrix. CP is a partial projection of FP. Using the projection
relationship from FP SAR data to CP SAR data, the
Freeman-Durden model form of CP data can be written.
Taking the FP covariance matrix of the Freeman-Durden
model as an example, it contains , , , , , , and ,
ff f aa bb
sd v
with a total of seven model parameters, while the CP covariance
matrix has only four observations, so assumptions
and constraints need to be added to reduce the number
of unknown parameters. Guo first wrote the Freeman-
Durden three-component model for both CTLR and r/4
modes directly based on the correspondence between the
same elements in the FP covariance matrix and CP covariance
matrix, in 2014, and experimental results showed that
the model decomposition is more accurate for CTLR mode
[56]. Subsequent related studies have been carried out on
CTLR mode. Therefore, to facilitate a direct comparison of
model-based decomposition methods, only model decomposition
methods for CTLR mode with transmitted RHCP
are discussed in the following.
**
IEEE Geoscience and Remote Sensing Magazine - September 2022

Table of Contents for the Digital Edition of IEEE Geoscience and Remote Sensing Magazine - September 2022
