This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
2 IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING
angles ranging from 20
◦
to 60
◦
, the scattering mechanism at sea
surface is often predominated by the Bragg resonant scattering
[1], [3]. If small-scale resonant waves are modulated by large-
scale waves, then the two-scale model [10] is often applied.
Although many ocean models were developed to explain some
properties of the ocean backscatter, not a single scattering
model is applicable to all radar frequencies and sea states. Thus,
in this paper, by assuming the perfect ocean scatter, we study
the polarimetric deviation of a scatterer from the Bragg scat-
tering. As is known, any scattering process that affects the
Bragg surface roughness can be imaged with SAR. Oil-slicks
not only damp the ocean capillary and gravity waves (i.e.,
the Bragg waves in certain radar wavelength), but also reduce
the surface tension and friction between the wind and liquid
surface [10]. Therefore, oil-slicks have a low backscattering
feature [11]–[14], i.e., the oil-spill appears as distinguishable
dark patches compared to the ambient sea surface. However,
in SAR imagery, the similar low backscatter feature could also
be presented by other ocean phenomena, known as look-alikes,
such as biogenic films, the low-wind region (LWR), rain affects,
sea ice, and upwellings [11]. I t is crucial to distinguish between
oil-slicks and oil look-alikes when observing oil-spills. In quad-
pol images, the polarimetric signature and pedestal [15], the
Mueller matrix-based filter [16], [17], and the co-polarized
phase difference (CPD) [18] have been investigated for oil-spill
observation and discrimination. However, methods based on the
CP data are rarely found in the literature for the distinction
problem among dark features. In [19], the performance of the
degree of polarization (m) has been studied for both oil-spill
and ship detections in various dual-pol modes, but it did not
refer to the oil look-alikes. In addition, although analysis from
[15]–[17] shows that the biogenic look-alike has polarimetric
signatures similar to the sea surface, biogenic slicks do not fol-
low the typical Bragg scattering and thus should be predicted.
The other class of targets we concern is ships and metallic tar-
gets. Many researchers [1] have demonstrated that polarimetric
information is an important factor for improving the ship detec-
tion performance. Many approaches have been proposed based
on the statistical decision theory with different sea clutter dis-
tributions, in which one popular methodology is the likelihood
ratio test (LRT) detector [20] with Neyman–Pearson criterion.
The LRT detector is to use the scattering matrix directly to
define a decision variable, where the total backscatter intensity
is the main discriminative feature for sea clutter and targets. In
[21], a notch filter was proposed for polarimetric SAR based
on a target polarization space. This method exploits the physi-
cal difference between targets and sea clutter to detect targets
which have different polarimetric behaviors from the sea. In
[22], an approach was proposed from a different perspective
according to the fact that ships are deterministic targets and thus
have higher degree of interlook coherence between images of
different looks than sea clutter, which implies that ships behave
more coherently than the dynamic ocean surface within the
integration time. In this paper, we aim at observing ocean tar-
gets by characterizing their polarimetric features, especially in
compact polarimetry. The scattering process of ships or man-
made metallic targets presented in Pol-SAR imagery of the
ocean is very complicated because it is determined by several
scattering mechanisms, including the single-bounce scattering,
the double-bounce scattering, and multiple reflections between
the ship and sea surface, among which the double-bounce scat-
tering and multiple reflections tend to dominant the backscatter
[1]. Physical scattering mechanisms involved in the sea sur-
face, ships and oil-slicks are quite different from each other.
Thus, analysis based on the physical scattering mechanism of
the ocean is a fundamental way to detect targets in Pol-SAR
images.
The Bragg scattering, which is assumed without the cross-
pol element, is modeled for the rough surface backscatter with
two complex coefficients under a rigorous roughness constraint,
limiting its applicability. In order to widen its application range
and to make it more consistent with the realistic SAR backscat-
tering, an extended Bragg (X-Bragg) model was proposed in
[23] by modeling the rough surface as a reflection symmetric
depolarizer. From this model, two parameters can be derived
[23], [24], i.e., the Bragg alpha and the odd–even bounce scat-
tering coherence. In [25], based on the linear CP mode data,
a method was proposed to invert the X-Bragg parameters by
using the Newton–Raphson iteration technique. In this study,
based on the X-Bragg scattering model and its backscattering
reflection symmetry assumption, we propose a method for sea
surface scattering characterization, and investigate the capabili-
ties of typical CP modes for estimating the X-Bragg parameters.
These parameters are further used to discriminate oil-spills,
look-alikes, and ships from the sea surface. Performances of
the proposed method are evaluated on C-band Pol-SAR data
and results show the promise of the circular transmit CP modes
for maritime surveillance. The organization is given as follows.
In Section II, we summarize the X-Bragg scattering model and
verify the effectiveness of the X-Bragg parameters for ocean
surface description. In Section III, the capabilities of typical CP
modes for estimating the X-Bragg parameters are evaluated and
an equivalent method in CP SAR is proposed with the circular
transmit CP mode. Experimental results and comparisons are
presented in Section IV. Conclusion is given in Section V.
II. X-B
RAGG SCATTERING MODEL FOR SEA SURFACE
DESCRIPTION
The single-look complex scattering matrix, denoted as S =
S
HH
S
HV
S
VH
S
VV
, contains the same information as the Pauli-basis
vector
k
P
. The coherency matrix T can be generated from
the outer product of
k
P
with its conjugate transpose
H
k
p
.In
the case of backscattering from reciprocal scatterers, t he multi-
look coherency matrix T , which is usually used to characterize
the average backscatter mechanism, is shown in (1). We here
present T by the Huynen parameters [26], i.e., A
0
, B
0
, B, C,
D, H, G, E, and F . Each Huynen parameter represents an
element of T . Note that A
0
, B
0
, and F are rotation invariants.
T =
k
p
k
H
p
=
⎡
⎣
T
11
T
12
T
13
T
∗
12
T
22
T
23
T
∗
13
T
∗
23
T
33
⎤
⎦
=
⎡
⎣
2A
0
C − jD H + jG
C + jD B
0
+ BE+ jF
H − jG E − jF B
0
− B
⎤
⎦
(1)