Polarimetry

Radar Phase
• Phase is a property of a periodic phenomenon.
• The concept of phase is usually applied to the oscillation of
electromagnetic waves
• Phase can be expressed in degrees. One-quarter cycle
represents a phase rotation of 90 degrees; completion of one
complete cycle corresponds to a phase rotation of 360 degrees.
• Waves are considered in-phase, if their origins of phase 0
degrees are perfectly aligned; out-of-phase conditions are met
when phase 0 and 180 degrees are aligned. Precise knowledge
of phase properties in radar signal data is a key element in
interferometric as well as in polarimetric SAR.
 Phase, Amplitude and Frequency

• Period of a wave is the length

from one peak to the next.
• Frequency is the amount of
peaks (or troughs) that pass a
given point in one second
• Amplitude is the height from
centre to peak, or half the
distance from peak-to-peak.
• Phase is the initial angle of the
sinusoidal function at the origin
 Complex SAR data
• Microwave SAR sensors transmit and receive EM
waves. EM waves have amplitude and phase
s = M e-iØ = M(cosØ – isinØ)
• The signal has two components:
sine and cosine components
• I and Q represent the real and imaginary
components respectively, in Cartesian coordinate
system
• Hence, microwave data is termed as complex data
because it has amplitude and phase components
SAR operating modes

Source : Marco Lavalle. Full and Compact Polarimetric Radar Interferometry for Vegetation
Remote Sensing. Signal and Image processing. Universite Rennes 1, 2009.
 Ellipticity Angle
• Also known as axial ratio, the major to minor axis
ratio.
• Defined as the magnitude of the arc tangent of the
ratio of the polarization ellipse’s minor and major
axis.
• If negative, the ellipse rotation is right handed.
• If positive, the ellipse rotation is left handed.
• An ellipticity of zero or infinity corresponds to linear
polarization and an ellipticity of 1 corresponds to
circular polarization.
• This determines the polarization state (linear or
circular) of the propagating EM wave.
• Varies from -45 degrees to 45 degrees.
• If 45 degree, then circular and if 0 degree, then
linear.
 Orientation Angle
• Angle between major semi axis of the
polarization ellipse and the H axis.
• Also known as the tilt angle or azimuth angle.
• This angle indicates the ellipse tilt or orientation
about radar line of sight.
• Orientation angle varies from 0 to 180 degree.
• Used to define the angular position of a target
object.
• Polarization signatures generated from the
scattering matrix can be used to interpret the
scattering properties of the target helping to
understand the roughness, polarization
dependence and scattering mechanisms for
that ice type.
CONVENIENT GRAPHICAL WAY TO DISPLAY THE RECEIVED POWER
AS A FUNCTION OF POLARIZATION

cos   sin    cos  j sin  

A   j sin    uˆ x
 sin   cos    cos  

CO-POLARIZED POWER : CROSS-POLARIZED POWER :

PCO  A S A  S AA PX   A S A  S AA
T 2 2 T 2 2


CO-POLARIZED CROSS-POLARIZED
CORRELATION PHASE: CORRELATION PHASE :
 CO  arg S AA  arg S A A
 
 X  arg S AA  arg S AA 
 Polarimetric Signatures

Signature of co-pol.
= (T11(𝝋) + 2 Re{T12(𝝋)} + T22(𝝋) ) / 2

Signature of cross-pol = T33(𝝋)/2

Figure 9-20. C-band Co-polarization signatures for a) young and
ridged ice and b) for smooth first year ice with different snow cover
properties

Note less peak at VV for ridged ice due to roughness effects and the
reduced polarization dependence in the ice. This is partly due to plates
of thin ice layers in the snow giving rise to differential HH and VV
• Use of polarization signatures for ice type
interpretation can also be demonstrated by
comparing signatures for new, grey, rough grey
and desalinated sea ice types which represent
a time series in terms of evolution and
development.
 Figure 9-21. Co-
polarization C-band
signatures of new,
grey, rough grey,
and desalinated sea
ice types from
scatterometer
measurements.
Note the change in
the peak
polarization
response as the ice
ages from new to
desalinated ice
types

This figure demonstrates the migration of the peak response from HH to

VV related to reduction in DEC of ice surface over this time period.
Polarization response changes once again towards an HH peak as the
ice further evolves, becomes rougher and desalinated.
 FULL-POL-SAR Data

Space-borne Air-borne
TerraSAR-X PiSAR Polarimetric Radar measures
Scattering Matrix

RadarSAT-2

F-SAR
s t
ALOS s E S HH S HV E t
E = H
s = H
= S E
E V S VH S VV E t
V
(2006-2011)

Scattering Matrix
 FULL-POL-SAR Image Analyses

Scattering matrix
HV Basis
= Quad-Pol. data
HH, 2HV, VV
HH HV VV
Pauli Basis Color-Composite

<Average> HH-VV, 2HV, HH+VV

Eigenvalue Entropy, Alpha-angle, Anisotropy

λ1 λ2 λ3

Scattering Power Decomposition

Covariance matrix
double
Coherency matrix Pd
bounce
Pd, Pv, Ps, Pc Ps Pv
surface volume
scattering scattering
 Purpose

Purpose: To decompose different types of scattering contributions

Applications: POLSAR image interpretation, classification, and
scattering parameter inversion .
2 Re Im *
Re Im
Coherency S HH + S VV S +S
HH VV S HH – S VV 2S *
HV S HH + S VV
* 2
Matrix [T]= S HH – S VV S HH + S VV S HH – S VV 2 S HV
*
S HH – S VV
* * 2
2 S HV S HH + S VV 2 S HV S HH – S VV 4 S HV

T11 T12 T13 

[T ]  T21 T22 T23  9
T31 T32 T33 
Scattering Power Decompositions (natural/distributed targets)
 How many independent polarimetric information ?
under the condition of mono-static radar
Coherent j j HV
-j HH
5
S HH S HV e
S relative = VH
j j - j HH j j
VV
- j HH
S VH e S VV e Scattering matrix
S HH + S VV
kP = 1
k p= (1/2) Trace {S[ψ]}
2
S HH – S VV Scattering vector
2S HV

Incoherent Coherency or Covariance

9
matrix
2
Re Im *
Re Im
S HH + S VV S HH + S VV S HH – S VV 2S *
HV S HH + S VV
* 2
S HH – S VV S HH + S VV S HH – S VV 2 S HV
*
S HH – S VV
* * 2
2 S HV S HH + S VV 2 S HV S HH – S VV 4 S HV
 Four Scattering Mechanisms
Total Power (TP)= Surface +Double Bounce+ + +…..
Volume
Scattering Scattering
scattering
Ps Pd Pv

Surf. on the border surface between two different media

DB two surfaces at right angle

Vol. if radar beam penetrates into a medium

Hel. circular pol. power  HH pol. comp. + VV pol. comp. with 900
out of phase
 (1/8) λ separation structures reflect circular pol. urban areas
 FDD
Freeman and Durden 3-Component Scattering Power Decomposition [1]

T   f s [Ts ]  f d [Td ]  f v [Tv ]

1  *   
2
 0

0
    2 0 0
T   f s  0  f v  0 1 0
1
0  f d   * 1
2

    4
 0 0 0  0 0 0 0 0 1
  

with reflection symmetry for natural distributed targets

Total Power (TP)= Surface

+ Double Bounce
Scattering
+ Volume
Scattering scattering
Ps Pd Pv
[1] A. Freeman and S. L. Durden, ``A Three-component scattering model for polarimetric SAR data,``
IEEE TGRS, Vol.36, No. 3, pp. 963-973, May 1998.
Figure 9-24. Total power
images of the Freeman-
Durden decomposition of
C-L, and P-band AIRSAR
data showing surface
versus volume scattering
 Class Assignment To Ice Types
Class Colours Description
ThI / new forming thin ice /
blue
SFYI Smooth first year ice
orange
RFYI / R green Ridged first year ice / Rubble
black
pink
CFYI pastell Compressed First Year Ice
green
white
MYI Multi Year Ice
grey

Figure 9-25. Ice type classification results using data acquired by the
JLP AIRSAR. An entropy based polarimetric decomposition using eight
classes and 12 iterations
• New ice is dominated by surface scattering
where as multi year ice has a large volume
scattering component. The pedestal height
would indicate the degree of roughness. HV
would help identify volume scattering.
Polarimetric classification of scattering
mechanisms could be used to determine areas
of volume scattering.
• Entropy – degree of randomness of scattering process
or degree of statistical disorder (0 to 1)

• Anisotropy – ratio of linear combination of Eigen values.

• Scattering mechanism angle (alpha) 0 to 90

alpha = 0 scattering corresponds to single bounce produced by
rough surface

alpha = 45 volume scattering .

 Cloude and Pottier
decomposition
• This decomposition is based on the three
eigenvalues (λ1, λ2 and λ3 with decreasing
magnitude) of the coherence matrix <T>, and
defines the Entropy H, the anisotropy A, the α
angle and the asymmetry AS as follows

Entropy (H)
Anisotropy (A)

Alpha
Figure 9-23. Entropy (H),
Anisotropy (A), and Alpha
()-Angle for images from
C-band data acquired with
the CV-580 off the North
shore area of PEI in
March, 2001. The bottom
image shows land, while
fast ice is highlighted in the
centre image and rough
first-year ice in the top
image
• Figure 9.23: Smooth ice shows the lowest entropy due
to homogenous scattering where as land has higher
entropy.
• Anisotropy is lower for land and higher for ice providing
the best contrast for land/ice and between ice types.
• Alpha angle is low for smooth ice which along with low
entropy demonstrates surface scattering dominates.
• For rougher first year ice both alpha angle and entropy
are higher which, as demonstrated in the cross
polarization channel, indicates that more volume
scattering is occurring.
• Varying penetration characteristics of multi-
frequency SAR allow further separability
between ice types especially among multi year
ice types and first year ice types.
 Figure 9-22. Classified
image of the north shore of
PEI using CV-580
polarimetric C-band SAR
data using a complex
Wishart classifier with 8
initial classes and 12
iterations
Colour Assignment
Colours Description
blue Smooth, thin Fast Ice Polarimetric data provide enough information
red Fast Ice with rough surface for effective classification of ice types
magenta Rough RYI
cyan Rough, thicker FYI
white Rough, thicker FYI / Land
green Land
dark green Land
black Land
Classes not used for comparison in the
grey
scatterplots (white dark green and black)
 Y4O
4-Component Scattering power decomposition

* *
for general case S S
HH HV ¹ 0, S S
VV HV ¹0

Ps Pv

6 parameters out of 9 are counted in the

expansion
3-component decomposition counts 5 out of 9
[Freeman and Durden Decomposition, FDD]
 General 4-component scattering power decomposition
with Unitary transformation : G4U

unitary dipole
+ extended volume
transformation scattering model dihedral

7 parameters out of 7 (100%) are accounted for the

newly developed decomposition (G4U)
G. Singh, Y. Yamaguchi, et al, “Polarimetric SAR target decomposition with unitary transformation,” 2nd Int’l
POLSAR Workshop in Niigata 2011

G. Singh, Y. Yamaguchi, and S.-E. Park, IEEE Trans. Geosci. Remote Sens., vol. 51, no. 5, pp. 3014-3022,
2013.
1. Extended volume scattering model
 How will dipole and oriented dihedral volume scattering model
effect to decomposition results?
Y4R G4U
vegetation vegetation

Urban area Urban area

Volume scattering caused by the HV Volume scattering caused by the HV

component from vegetation (dipole) component from dipole and oriented
dihedral
Pi-SAR data on 2003/08/20
How to account for all polarimetric information
T11 T12 T13
T = k p k †p = T21 T22 T23
T31 T32 T33
Reduce
9
Increase
more
than 4
Model expansion Measured data
Reduction of independent parameters from 9 to ?

T11 T12 T13

†
T = kp k p = T21 T22 T23 9 parameters
T31 T32 T33

Possible? Unitary transformation

′
𝑇11 ′
𝑇12 ′
𝑇13 accounted in G4U
′ ′
𝑇′ = 𝑇21 𝑇22 0 source of helicity
′ ′
𝑇31 0 𝑇33
we would like to minimize
2. Unitary Transformation

9  8  7 parameters
 ′ ′ ′
𝑇11 𝑇12 𝑇13
2 steps to transform 𝑇′ ′
= 𝑇21 ′
𝑇22 0
′ ′
𝑇31 0 𝑇33

First step: rotation about the radar line of sight

†
𝑇 𝜃 = 𝑅 𝜃 𝑇 𝑅 𝜃

1 0 0
1 2 Re {𝑇23 }
𝑅(𝜃) = 0 cos 2𝜃 sin 2𝜃 2𝜃 = 𝑡𝑎𝑛 −1
0 − sin 2𝜃 cos 2𝜃 2 𝑇22 − 𝑇33

𝑇23 𝜃 = 𝑗Im{𝑇23 }

No real part in T23

Second step: unitary transformation-change of base

†
𝑇 𝜑 = 𝑈 𝜑 𝑇 𝜃 𝑈 𝜑
1 0 0
1 2 Im {𝑇23 (𝜃)}
𝑈(𝜑) = 0 cos 2𝜑 𝑗 sin 2𝜑 2𝜑 = 𝑡𝑎𝑛 −1
0 𝑗 sin 2𝜑 cos 2𝜑 2 𝑇22 𝜃 − 𝑇33 (𝜃)

𝑇23 𝜑 = Re{𝑇23 (𝜃)}

𝑇23 𝜑 = Re{𝑇23 (𝜃)}= Re 𝑗Im 𝑇23 =0

It is possible to eliminate T23 term

 General 4-component scattering power
decomposition with Unitary transformation
Expansion matrix after unitary transformation
C1 = fs  + fd  = Re C13 + 7 Pv + 1 Pc

30 4

C1 = T11 - T22 + 7 T33 + 1 Pc ≤ 0

8 16
 Equating all elements

′
𝑇33 =

Arrangement of the element relations

Pc and Pv can be determined directly

The remaining equations

3 equations with 4 unknowns

Approximation to solve: Branch condition C0

in covariance formulation
in coherency matrix
expression

C0 > 0
 G4U-
decomposition
algorithm

simplified in terms
of the rotated
coherency matrix
elements
Pd

Ps Pv
HV from dipoles and dihedrals
+
Unitary Transformation
 Azimuth
Niigata University Y40
Y40
6/9

Range
Pi-SAR data on 2003/08/20

Y4R
G4U
Y4R
6/8 7/7
 Azimuth

TerraSAR-X Y4O
FULL-POL-SAR

Range
22 - 04 - 2010 NU

Y4R G4U

Pd

Ps Pv

G4U  PolSARPro v5.0 (ESA)

A few Applications of G4U
 Earthquake/Tsunami Disaster Monitoring

Range

G. Singh, Y. Yamaguchi, W.-M. Boerner and S.-E. Park, Monitoring of the 2011 March 11 Off-
Tohoku 9.0 Earthquake with Super-Tsunami Disaster by Implementing fully polarimetric high
resolution POLSAR Techniques, Proc. of IEEE, vol. 101, no.3, pp. 831-846, March 2013
How can decomposition components help to provide model input?
Shin-Moe-Dake

Range
23.1°
2009/06/10
Volcano Eruption
on March 13, 2011

2011/03/16
Pd

Ps Pv

GNR 639
 Ambiguity for vegetation
monitoring---
can be resolved with G4U

Y4R and G4U RGB over the part

of ALOS-PALSAR data
of May 14, 2010 in Rajasthan area
Y4R
G4U RGB

Range
(G. Singh, JAXA PI Report, May 2013)

Pd

Ps Pv
Forest monitoring
DLR /F-SAR S-band Image: June 8, 2010

Range
G4U

Kaufbeuren, Germany
Fig. G4U RGB image for the fully polarimetric acquisition of ALOS-PASAR on
April 23, 2009, over the part of Chinnur, India (scene ID: ALPSRP172950370; scene
center latitude/ longitude: 18.920N / 79.800E): Forest cover in green. Azimuth
direction is from left to right and range direction is from top to bottom.

(G. Singh, JAXA PI Report, May 2013)

 Range
Full-POLSAR images
over snow : G4U RGB

Mt. Iwate
March 28, 2007

After 46 days

May 13, 2007

 Range
May 25, 2010 June 06, 2010

Snow line delineation

 1

175
0.6
Scattering Power

Surface(Ps) Double Bounce (Pd)

0.5
Volume (Pv) Helix (Pc)
0.4

0.3

0.2

0.1

0
1 22 43 64 85 106 127 148 169
Distance (in Pixels)
 G4Usnow – an Inversion Algorithm Development
1. 2 main scattering contributions come from snowpack
2. Double bounce is negligible in wet snow covered area
3. Helix scattering power can be generated from the sloped terrain
4. Shape of snow particle becomes roundedG4U volume scattering models
will be modified/developed for snowpack.

Pd
PV
Ps

Pc
Generalized Volume and Surface scattering
Parameters
1 𝛽∗ 0 𝛼2 𝛼 0
†
𝑈 𝜑 𝑇 𝜃 𝑈 𝜑 = 𝑈 𝜑 𝑓𝑠 𝛽 𝛽2 0 + 𝑓𝑑 𝛼 ∗ 1 0
0 0 0 0 0 0

2
𝛾 0 0
1 0 0 0
0 𝑓
0 + 𝑐 0 †
+𝑓𝑣 2 1 ±𝑗 𝑈 𝜑
2 0 ∓𝑗 1
1
0 0
2
1 0 0
1 2 Im {𝑇23 (𝜃)}
*** 𝑈(𝜑) = 0 cos 2𝜑 𝑗 sin 2𝜑 with 2𝜑 = 𝑡𝑎𝑛−1
2 𝑇22 𝜃 −𝑇33 (𝜃)
0 𝑗 sin 2𝜑 cos 2𝜑

***G. Singh, Y. Yamaguchi, and S.-E. Park, IEEE Trans.

Geosci. Remote Sens., vol. 51, no. 5, pp. 3014-3022, 2013.
Generalized Volume Scattering Parameter

𝛾𝐻𝐻 + 𝛾𝑉𝑉 2 𝑇11 𝜃 𝑇12 𝜃 + 𝑇13 𝜃 2

2
𝛾 =
𝛾𝐻𝐻 − 𝛾𝑉𝑉 2 = 2 𝑇33 𝜃 − 𝑓𝑐 – 2 𝑇 𝜃 − 𝑓 𝑇22 𝜃 − 𝑇33 𝜃
33 𝑐

Generalized Surface Scattering Parameter

α𝐻𝐻 − α𝑉𝑉 𝑇12 (𝜃) + 𝑇13 (𝜃)

𝛽 = =
α𝐻𝐻 + αVV 𝑇11 (𝜃) − 𝑓𝑣 𝛾 2

Inversion f (|εs|, 𝜃𝑖 /𝜃𝑟 ) Wetness Validation*

*C-band