Research Progress of PolSAR Technology in IECAS
Wen Hong, Erxue Chen, Eric Pottier
Institute of Electronics, Chinese Academy of Sciences (IECAS)
National Key Lab of Microwave Imaging Technology (MITL)
POLINSAR (ID. 10609)
1、Overview the research activities from 2014-2015 WP1:land cover /land use and vegetation parameter study WP2:forest type classification and forest parameter WP3:PolSARpro software continued development
2、Study of theory and processing methodology
model-based decomposition for Compact PolInSAR data topography retrieval algorithm based on the polarization-dependent intensity ratio soil moisture change detection model using dual pol InSAR data
3 、Study of system methodology
accurate inversion of multi-baselines based on fusion of SAR and IMU for airborne MB-InSAR
Outline
Overview
Prof. Eric Pottier Prof. Erxue Chen Prof. Wen Hong All participated as lecturers on this advanced training course.
Advanced Training Course on Land and Water Remote Sensing
Overview
Prof. Laurent Ferro-Famil
Dr. Stefano Tebaldini
They gave both lectures and practicals on SAR tomography and applications for one whole week at IECAS.
Advanced Training Course in SAR Tomography and Applications
Overview Young Scientist Exchange
Yin Qiang, Ph.D candidate, Visited ESRIN as Research Fellow from 2014.2-2015.1
Project: Soil Moisture Estimation Based on Interferometric Phase of Multi-Temporal SAR Data ESA Data: Envisat-ASAR & AgriSAR
EUSAR 2014
Trainees & Fellows Meeting 2014
Co-author paper @PolInSAR 2015
Overview Book Translation
S.R. Cloude Published soon in 2015
The new chinese language edition provides a coherent summary of the original book, taking the form of notations.
Notations in Chinese
1、Overview the research activities from 2014-2015 WP1:land cover /land use and vegetation parameter study WP2:forest type classification and forest parameter WP3:PolSARpro software continued development
2、Study of theory and processing methodology
model-based decomposition for Compact PolInSAR data topography retrieval algorithm based on the polarization-dependent intensity ratio soil moisture change detection model using dual pol InSAR data
3 、Study of system methodology
accurate inversion of multi-baselines based on fusion of SAR and IMU for airborne MB-InSAR
Outline
Conventional Freeman-Durden Decomposition Only obtain power of three scattering mechanisms No height information
PolInSAR target decomposition Obtain power Phase center height
We apply the Freeman-Durden decomposition to Compact PolInSAR data to obtain the power as well as phase center height of each scattering mechanism.
PolSAR methodology Background
Surface scattering
Dihedral scattering
Volume scattering
fs, fd, fv are unknown complex coefficients which are to be retrieved
Scattering Model
PolSAR methodology
Compact-PolInSAR decomposition Decomposition of
are real coefficients, we can embed them into the
associated scattering model complex coefficients
Decomposition procedure Step1. Volume power estimation Step 2. phase estimation Step 3. Numerical solution: After two parameters retrieval, the underdetermined equation becomes determined equations which can be solve via numerical method.
PolSAR methodology
Decomposition results Amplitude and phase decomposition
0 50 100 150 200
-20
0
20
Pixel in azimuth
Pow
er (d
B)
Compact-PolInSAR Target Decomposition
0 50 100 150 200
-20
0
20
Pixel in azimuth
Pow
er (d
B)
|β |2Fs - HC pol |α|2Fd - HC pol Fv - HC pol
Fs-VC pol Fd-VC pol Fv-VC pol
0 50 100 150 200
0
2
4
6
8
10
12
Pixel in azimuth
Heig
ht (m
)
Scattering Phase Height
ODD height DBL height VOL height
We can obtain the amplitude and phase height of each scattering mechanism.
In forest area, volume scattering is dominant.
For forest area, the surface scattering comes form the upper layer. For bare surface, the heights of three scattering mechanisms are all located at ground level.
[Guo Shenglong, GRSL 2015]
PolSAR methodology
1、Overview the research activities from 2014-2015 WP1:land cover /land use and vegetation parameter study WP2:forest type classification and forest parameter WP3:PolSARpro software continued development
2、Study of theory and processing methodology
model-based decomposition for Compact PolInSAR data topography retrieval algorithm based on the polarization-dependent intensity ratio soil moisture change detection model using dual pol InSAR data
3 、Study of system methodology
accurate inversion of multi-baselines based on fusion of SAR and IMU for airborne MB-InSAR
Outline
Topography-POA shift model (Lee, 1998)
Polarization ellipse rotation
Topography-POA shift model
A geometrical scheme illustration of incidence plane, flat and tilted surfaces and polarization ellipse rotation
xTop
y
tantansin cos tan
τθη η τ
=−Model:
• We propose a new topography retrieval algorithm based on the polarization-dependent intensity ratio. •The intensity term conforms to both the polarimetric orientation angle shift model and the Lambertian scattering model.
PolSAR methodology
"Fake" topographic relief effect 1: Scattering media change border Correction method: Weighted FMG slope revision:
twoweighted a a
g g
( , ) ( 1, ) ( , )
(( , ) ( , 1) ( , ))x yx y x y x y
x y x yω
ρ = ∆ + −∆
+ ∆ + −∆
1 1
0 y00 0
1 1 1 12 20 y0
0 0 0 0
( , ) ( , )( , )
( , ) ( , )
M N
m nM N M N
m n m n
m n m nx y
m n m n
α τω
α τ
− −
= =
− − − −
= = = =
=∑∑
∑∑ ∑∑
The "fake" topographic relief effect induced by geophysical terrain variations is considered and reduced by including an averaged α scattering angle-based weighting coefficient in the second-order residual error equation.
IECAS P-band PolSAR Weighting coefficient DEM, before correction DEM, after correction
Ground range
Azim
uth
Height/m
0
50
100
150
200
250
300
350
400
[Li Yang, TGRS 2015 ]
PolSAR methodology
"Fake" topographic relief effect 2: Bragg resonance Correction method: Weighted FMG slope revision:
The "fake" topographic relief effect induced by Bragg resonance is also considered and reduced by including an wide aperture anisotropy factor-based weighting coefficient in the second-order residual error equation.
IECAS P-band PolSAR Weighting coefficient DEM, before correction DEM, after correction
twoweighted a a
g g2
( , ) ( 1, ) ( , )
( ( , 1) ( ,, ))( )
x y x y x yyx x xy y
ρ
ω
= ∆ + −∆
+ ∆ + −∆
1 1
0 02 1 1 1 1
2 2
0 0 0 0
( , ) ( , )( , )
( , ) ( , )
M N
m nM N M N
m n m n
m n T m nx y
m n T m n
β
β
ωω
ω
− −
= =
− − − −
= = = =
=∑∑
∑∑ ∑∑
PolSAR methodology
Radiometric terrain correction (RTC) RTC method: Flattening Gamma, by David Small
P-band PolSAR, Gen He, Before RTC RCS V.S. LIA P-band PolSAR, Gen He, After RTC RCS V.S. LIA
This method has great potential for PolSAR radiometric terrain correction simply by relying on the DEM extracted by itself. Ref: David Small, Flattening Gamma: Radiometric terrain correction for SAR imagery,
IEEE Transactions on Geoscience and Remote Sensing, vol.49, no.8, p3081-3093, August, 2011.
PolSAR methodology
1、Overview the research activities from 2014-2015 WP1:land cover /land use and vegetation parameter study WP2:forest type classification and forest parameter WP3:PolSARpro software continued development
2、Study of theory and processing methodology
model-based decomposition for Compact PolInSAR data topography retrieval algorithm based on the polarization-dependent intensity ratio soil moisture change detection model using dual pol InSAR data
3 、Study of system methodology
accurate inversion of multi-baselines based on fusion of SAR and IMU for airborne MB-InSAR
Outline
Coherent scattering model for direct ground [Yin Qiang,EUSAR 2014]
• Derivated from coherent scattering model of Treuhaft (2000)
• Normalized interferometric cross-correlation
When the roughness is considered stable in two repeat observations, the changes in soil moisture can be estimated from the phase of interferometric cross-correlation.
( ) ( )( ) ( )
1 2
0 0
1 1 2 2
ˆ ˆ1 1 2 2
24 2 2 4 4ˆ ˆ0 0 0 0 0 ˆ ˆ, ,0
ˆ ˆ
4 cos 2 sin ,0r
t t
i z i rr P p t p t
p E R p E R
A e W d W r e dr k W kπφ α
η η θ θ α α
∗
∞ ∗
−∞
⋅ ⋅
= ⋅ −∫ ∫
roughness moisture stable variant
( ) ( )( ) ( )
( )0 0ˆ ˆ1 2 1 2
2 2 2 21 2ˆ ˆ1 2
ˆ ˆ
ˆ ˆ
t t i zr
t t
t E R t E RA e
t E R t E R
φ α α
α α
∗ ∗⋅ ⋅=
⋅ ⋅
PolSAR methodology
ASAR data processing results
map quick look coherence phase
Track No. t0 t1 t2 t3 t4
444 13/04/2006 18/05/2006 22/06/2006 27/07/2006 31/08/2006
Within all the combination of these five scenes in Track 444, only the interferogram between 18/05/2006 and 22/06/2006 shows high coherence. According to the ground truth data, it is because of either plough activities or rapid vegetation growth. Data from other tracks show similar results.
PolSAR methodology
Due to the different sensitivity of HH and VV InSAR phase on soil moisture change, it is possible to utilize the differential interferogram between HH and VV interferogram.
( ) ( )0 0 0 01 2 1 2
2 2 2 21 2 1 2
i z i zHH r VV rA e A eφ φα α α α
α α α α
∗ ∗ Φ −Φ
Low coherence in agricultural areas constrain the utility of DInSAR technique of time series data, which aims at remove the topography related phase.
Atmospheric and residual baseline errors should also be eliminated from InSAR phase.
New Approach
[Yin Qiang, PolInSAR 2015]
Problems when spaceborne data is applied to the proposed model:
PolSAR methodology
As the incidence angle decreases, the feasible inverted range of moisture changes also decrease very quickly. Compared to the single pol InSAR phase, it has small feasible inverted range.
The proposed dual pol InSAR method has large potential due to the fact the differential phase caused by soil moisture change in realistic soil is much larger.
FEASIBLE inverted
INFEASIBLE inverted
Feasible inverted range of soil moisture change by the proposed model.
Simulation results
PolSAR methodology
1、Overview the research activities from 2014-2015 WP1:land cover /land use and vegetation parameter study WP2:forest type classification and forest parameter WP3:PolSARpro software continued development
2、Study of theory and processing methodology
model-based decomposition for Compact PolInSAR data topography retrieval algorithm based on the polarization-dependent intensity ratio soil moisture change detection model using dual pol InSAR data
3 、Study of system methodology
accurate inversion of multi-baselines based on fusion of SAR and IMU for airborne MB-InSAR
Outline
Multi-baseline interferometric SAR (MB-InSAR) has the advantage of solving the conflict between height sensitivity and interferometric phase aliasing. Mounting multi sensors in one platform usually enable a flexible imaging geometry with large baseline. The flexural deformation often occurs in large baseline because of atmospheric turbulence and mechanical vibration. Such deviation brings phase errors and affects interferometric capability. Therefore, time-varying baseline acquisition is essential for MB-interferometry.
system methodology
One small baseline (called rigid baseline) is constructed firstly, with a high-precision position and orientation system (POS, F.I. Applanix POS/AV610) fixed to it shown in figure. Slave IMUs(inertia measurement unit), are fixed to the other antennas, which form flexible nodes for SAR interferometry.
Phase difference from rigid and flexible baseline is coherent because of observing the same scene. We proposed a model which takes rigid baseline parameters and related interferograms to estimate the flexible baseline.
system methodology
RB InSAR processing
FB InSAR processing
RB parameter optimzation
Slow varying baseline estiamtion
INS DataKalman filter
MOCPOSSlave INS
outputcorrection
MOC
POS measurement compensates the centimeter scale baseline variation. The residual errors is achieved by integration of slow varying baseline estimation and IMU measurement. Slow varing baseline is estimated by related interograms between rigid and flxible baseline and fast varying baseline during short time is measured by IMU. Integration is achieved by a Kalman filter.
Proposed processing frame
system methodology
Simulation results
2 4 6 8 10
9.952
9.954
9.956
9.958
9.96
9.962
9.964
9.966
9.968
time(s)
B x(m)
theoretical value estimation results
0 2 4 6 8 100.869
0.8695
0.87
0.8705
0.871
0.8715
0.872
0.8725
0.873
0.8735
time(s)
B y(m)
We simulated raw signal of the three baseline interferometry system incorporating flexible baseline oscillations. We also simulated INS raw data of three axis accelerometers and three axis gyros corresponding to the movement of sensor.
Flexible baseline estimation is implemented by multi-channel interferometric phase. Theoretical additional movements (blue curves) is composed of slow and fast variations. Slow movement’s reaches 5 millimeters and fast movement oscillation amplitude is 1 millimeters. Estimation results indicated that only slow movements is properly reconstructed and fast information is lost in interferometric phase.
system methodology
0 2 4 6 8 10 12-6
-4
-2
0
2
4
6
8
10x 10
-3
time(s)
Bx m
ovem
ent
inversion resultstheoretical value
0 2 4 6 8 10 125000.8708
5000.871
5000.8712
5000.8714
5000.8716
5000.8718
5000.872
5000.8722
5000.8724
time(s)
B y mov
emen
t(m)
inversion resultstheoretical value
After the INS raw data and baseline estimation result are integrated processed through Kalman filter, high-frequency baseline movement can be constructed and the result can reach a sub-wavelength precision. [Poster: Accurate Inversion of Multi-Baselines Based on Fusion of SAR and IMU for Airborne MB-InSAR]
Simulation results
system methodology
1、Overview the research activities from 2014-2015 WP1:land cover /land use and vegetation parameter study WP2:forest type classification and forest parameter WP3:PolSARpro software continued development
2、Study of theory and processing methodology
model-based decomposition for Compact PolInSAR data topography retrieval algorithm based on the polarization-dependent intensity ratio soil moisture change detection model using dual pol InSAR data
3 、Study of system methodology
accurate inversion of multi-baselines based on fusion of SAR and IMU for airborne MB-InSAR
Outline
Institute of Electronics, Chinese Academy of Sciences (IECAS)
National Key Lab of Microwave Imaging Technology (MITL)