Synthetic Aperture Radar Interferometry Batuhan Osmanoglu USRA - NASA GSFC, Greenbelt, MD Many colleagues in US and Europe Acknowledgements UYGU, TUBITAK, Gebze, 24 June 2014
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Synthetic Aperture Radar
Interferometry
Batuhan Osmanoglu
USRA - NASA GSFC,
Greenbelt, MD
Many colleagues in US and Europe
Acknowledgements
UYGU, TUBITAK, Gebze, 24 June 2014
Radar Remote Sensing Applications
Synthetic Aperture Radar Satellites
Acronyms
SAR - Synthetic Aperture Radar
InSAR - SAR Interferometry
PolInSAR - Polarimetric InSAR
PSI - Persistent Scatterer Interferometry
SBAS - Small Baselines Interferometry
DEM - Digital Elevation Model
DSM - Digital Surface Model
DTM - Digital Terrain Model
Commonly Referred Satellites
TerraSAR-X, TanDEM-X, COSMO-Skymed
ERS, Envisat, Radarsat, Sentinel
JERS, ALOS
Atmospheric Absorption
Radar Wavelength
L-band C-band X-band
~24 cm ~6 cm ~3 cm
Forest
Ice
Dry Soil
��rizontal and vertical polarization
Wavelength and Polarization
C-band (5.7 cm) L-band (24 cm) P-band (68 cm)
L-band HH L-band VV L-band HV
© Jacob Van Zyl
Geometric Effects
Slant range
Image
Geocoded
Image
© F. Sunar
Layover and Foreshortening
ERS-2
Mount
Vesuvius
Ferretti et al. 2007
Envisat
2002-2012
> 8 Tons
Interferometry is child's play!?
Repeat Pass Interferometry
First Pass Second Pass
EcoSAR Mission
© R. Rincon, 5/4/2014
TanDEM-Xsince 2010
DTED Level 2 - 2 m accuracy
theta
dtheta
P_0
1
rho_1
B
alpha
Bperp
Bparbeta=alpha-theta_0
theta_0B
par
Bperp
Bperp’
B’3
23-Pass InSAR
r_1
r_1
P_h
ellipsoid
Radar Images - Istanbul 20041103
© ESA
ASAR
T429
20050112Phase
Amplitude
Phase
Radar Interferometry - Istanbul
InterferogramInterferogram Coherence
-� 0 � 0 1
Real
Imagin
ary
Am
plitu
dePhase Angle
Vector Representation of InSAR data
Persistent Scatterer and Distributed Scatterer
Random Noise
StableDistributedScatterers
R
RealIm
agin
ary
Distributed ScattererLarge and Weak
Imagin
ary
Real
Persistent ScattererSmall and Strong
Bas
elin
eMaster
Acquisition
Slave Acquisition(11 days apart)
B�
B||
BTemporal
Interferometric Baseline
Line of sight
Interferometric Baseline Representations
-120
160
Met
ers
Synthetic Digital Elevation Model
-80
-40
0
40
80
120
Height sensitivity increases with B�
0 m 50 m 100 m 200 m-�
�
Phas
e
Effect of perpendicular baseline
B�
Fri
nge
Rat
e
1 cycle 2 cycles 3 cycles 4 cycles-�
�
Phas
e
Effect of temporal baseline
BTemporal
Nois
e
1 cycle 2 cycles 3 cycles 4 cycles0
1
Coher
ence
Effect of temporal baseline
BTemporal
Coher
ence
Coherence vs. Phase Standard Deviation
InterferogramOrbit Offsets
Scene Offsets4Azimuth Filtering
Coregistration
Resample
4Range Filtering
Interferogram
Flat Earth Removal5Topography Removal
Coherence
Phase Filtering
Unwrap
Geocode
MasterReadfiles
Crop
2DEM offset
3Oversample
1Preprocessing
SlaveReadfiles
Crop
3Oversample
1Preprocessing
Coregistration
Master Slave
Coregistration
Master Slave
Coregistration
Coregistration Error vs. Phase Standard Deviation
Feature Tracking
Gla
cier
Gla
cier
Master(First Pass)
Slave(Second Pass)
©Pritchard & Fielding, UNAVCO SAR Training
Feature Tracking
©Pritchard & Fielding, UNAVCO SAR Training
spacingsearch area
Gla
cier
Gla
cier
Master(First Pass)
Slave(Second Pass)
Feature Tracking
©Pritchard & Fielding, UNAVCO SAR Training
Can detect changes of fractions of a pixel!
Pixel Size: ~10m
Sensitivity: ~1/10 px, ~1m/cycle, ~cm/day
Gla
cier
Gla
cier
Master(First Pass)
Slave(Second Pass)
−800 −600 −400 −200 0 200 400 600 8000
5000
10000
15000
MASTER
−800 −600 −400 −200 0 200 400 600 8000
1
2
filter for master (red is composed)
−800 −600 −400 −200 0 200 400 600 8000
5000
10000
15000
Frequency [Hz]
filtered spectrum for master
−800 −600 −400 −200 0 200 400 600 8000
5000
10000
15000
SLAVE
−800 −600 −400 −200 0 200 400 600 8000
1
2
filter for slave (red is composed)
−800 −600 −400 −200 0 200 400 600 8000
5000
10000
15000
Frequency [Hz]
filtered spectrum for slave
Azimuth Filtering
LP
Master
L0
Slave
pixels
lines
0P00
0 0 0 0 0 0 0 0 0 0 0 00 0 0 0 0 0 0 0 0 0 0 00 0 0 0 0 0 0 0 0 0 0 00 0 0 0 0 0 0 0 0 0 0 00 0 0 0 0 0 0 0 0 0 0 00 0 0 0 0 0 0 0 0 0 0 00 0 0 0 0 0 0 0 0 0 0 00 0 0 0 0 0 0 0 0 0 0 00 0 0 0 0 0 0 0 0 0 0 00 0 0 0 0 0 0 0 0 0 0 00 0 0 0 0 0 0 0 0 0 0 00 0 0 0 0 0 0 0 0 0 0 00 0 0 0 0 0 0 0 0 0 0 00 0 0 0 0 0 0 0 0 0 0 00 0 0 0 0 0 0 0 0 0 0 00 0 0 0 0 0 0 0 0 0 0 00 0 0 0 0 0 0 0 0 0 0 00 0 0 0 0 0 0 0 0 0 0 00 0 0 0 0 0 0 0 0 0 0 00 0 0 0 0 0 0 0 0 0 0 00 0 0 0 0 0 0 0 0 0 0 00 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 000 0 0 0 0 0 0 0 0 0 0 000 0 0 0 0 0 0 0 0 0 0 000 0 0 0 0 0 0 0 0 0 0 000 0 0 0 0 0 0 0 0 0 0 000 0 0 0 0 0 0 0 0 0 0 000 0 0 0 0 0 0 0 0 0 0 000 0 0 0 0 0 0 0 0 0 0 000 0 0 0 0 0 0 0 0 0 0 000 0 0 0 0 0 0 0 0 0 0 000 0 0 0 0 0 0 0 0 0 0 000 0 0 0 0 0 0 0 0 0 0 000 0 0 0 0 0 0 0 0 0 0 00 0 0 0 0 0 0 0 0 0 0 00 0 0 0 0 0 0 0 0 0 0 00 0 0 0 0 0 0 0 0 0 0 00 0 0 0 0 0 0 0 0 0 0 00 0 0 0 0 0 0 0 0 0 0 00 0 0 0 0 0 0 0 0 0 0 00 0 0 0 0 0 0 0 0 0 0 00 0 0 0 0 0 0 0 0 0 0 000 0 0 0 0 0 0 0 0 0 0 00
overlap
Resampling
Interferogram
Reference Phase
Flattened Interferogram
Topographic Phase
Topography Removed
Filtered Phase
Orbit correction
Coherence
Ground Range (Pixels)
Ph
ase
(C
ycle
s)
Absolute Relative Unwrapped
Wrapped�
-�
��
Wrapped and Unwrapped Phase
0 2�Phase
A
A'
2�=28mm Surface Displacement
A A'
2�
Wrapped
Distance
LO
S D
ispla
cem
ent
A A'
Unwrapped
2�
4�
6�
8�
A A'
Unwrapped
�
�
�
�
28
56
84
112[mm]
LO
S D
ispla
cem
ent
Distance
Local vs. Global Unwrapping Methods
Tim is giving me directions to drive from Miami to Tampa:
Miami
Tampa Orlando
Naples
I-95
I-4
I-75
I-7
5
100mi
80mi
120mi 200mi
Global Unwrapping Method
-1 1 0 0 Miami 100-1 0 1 0 Naples 200 0 -1 0 1 Orlando = 120 0 0 -1 1 Tampa 80 1 0 0 0 0
Naples=115miOrlando=185miTampa=250mi
Local Unwrapping Method
Miami=0Naples=100Orlando=200Miami-Naples-Tampa=220Miami-Orlando-Tampa=280
Miami-Tampa=(220+280)/2 =250mi
So how far is Tampa from Miami?
Unwrapping Paths
Following the highest quality path reduces misfit!!
F=Noise Figure, G=System Gain
1 3 5
3 15
1+4+9 = 14
5+8+9 = 22
Friis' Formula (Information Theory):
1 1+3 1+3+5
5 5+3 5+3+1
Unwrapping Errors - 1D
-120
160
Met
ers
Synthetic Digital Elevation Model
-80
-40
0
40
80
120
Unwrapping Errors - 2D
Osmanoglu et al., 2011, Applied OpticsOn the importance of Path for Phase Unwrapping in Synthetic Aperture Radar Interferometry
output.ogv
JessyInk video element
0order number:
-80
0
Phas
e [r
ad]
PDV-Branch Cut 2nd Der. Rel. Fisher's Dist.100
Unwrapping Path
Line Scan Max. Coherence Phase Der. Var.
Misfit Along Path
3-D Unwrapping
Range
Azi
muth
Tim
e
Temporal Baseline
Perp
en
dic
ula
r B
aselin
e
Persistent Scatterer InSAR
Select PSC
Form Network
Modify Network
Estimate
Topography
Atmosphere
Deformation
Expand Network
Persistent Scatterer InSAR - Network Iterations
Persistent Scatterer InSAR - PSP Coherence
Persistent Scatterer InSAR - Multiple Networks
Persistent Scatterer InSAR - Results
Mexico City: Past
Mexico City: Past
Mexico City: Past
InSAR Measurements: Catching spatial variations
InSAR Analysis: Small Scale Changes
Mexico City - Metropolitan Cathedral
Mexico City - Metropolitan Cathedral
Temporal Baseline
Perp
en
dic
ula
r B
aselin
e
Kalman Filter
Developed during the Apollo Program (1968).
Combines observation with a model.
ModelObservation Co
mb
ine
d R
esu
lt
Kalman Filter Unwrapping Assumptions
�(a,r,t)=�topo(a,r)+�defo(a,r,t)+�atmo(a,r,t)+�noise
�topo(a,r): Stable over time. Spatially correlated.
�defo(a,r,�t): Only depends of temporal separation (�t). Correlated in space and time.
�atmo(a,r,�t): Not correlated in time.
�noise: Does not suppress signal.
a
r
Kalman Filter
Prediction
Control
Imagine a stock market forecast...
KF
Price
Price change
Today's Price
Yesterday's Today'sPredicted Price
Tomorrow's
Price
Price change
Each measurement (Price) is accompanied by its uncertainty.
For nonlinear problems "extended" Kalman filters are used.
Pre
dic
tion
Ste
p
x+
a,r|a,r-1
P+
a,r|a,r-1
x+
a,r|a,r+1
P+
a,r|a,r+1
x+
a,r|a-1,r
P+
a,r|a-1,r
x+
a,r|a-1,r-1
P+
a,r|a-1,r-1
x-
a,r
P-
a,rControl
Step
ya,
r,0
ya,
r,1
ya,
r,2
ya,
r,M
x+
a,r|a,r
P+
a,r|a,r
iterate
iterate until updates are below noise level
EKF
...
wra
pped
inte
rfer
ogra
ms
Surf
ace
Rec
onst
ruct
ion
MFC
Sm
ooth
ing S
tep
EKF
iterate until all is unwrapped
SBASEKFM
0
Subsi
den
ce [
mm
/yr]
NSBASM=394.05 M=100.53 M=176.18
PSI EKF300
50
Subsi
den
ce [
mm
/yr]
InSAR: An introduction to Processing and Applications using ISCE and GIAnT
August 4 - 6, 2014
UNAVCO, 6350 Nautilus Drive, Boulder, Colorado
Course will begin at 9am on August 4 and end at 5pm on August 6.
http://www.unavco.org/
GMTSAR
July 21 - 23, 2014
UNAVCO, 6350 Nautilus Drive, Boulder, Colorado
Course will begin at 9 AM on July 21 and end an 12 PM on July 23.
Related Documents
