1. Motivation Introduction Algorithm Validation Conclusion
Iterative Calibration of Relative Platform
Position:
A New Method for Baseline Estimation
Tiangang Yin1 , Emmanuel Christophe1 , Soo Chin Liew1 ,
Sim Heng Ong2
1 C ENTRE FOR R EMOTE I MAGING , S ENSING AND P ROCESSING
2 D EPT. OF E LECTRICAL AND C OMPUTER E NGINEERING ,
N ATIONAL U NIVERSITY OF S INGAPORE
IGARSS 2010, Honolulu

2. Motivation Introduction Algorithm Validation Conclusion
Outline
1 Motivation
2 Introduction
Concept
Baseline Calibration
Expand
3 Algorithm
Coordinate System
Iteration
4 Validation
5 Conclusion
IGARSS 2010, Honolulu

3. Motivation Introduction Algorithm Validation Conclusion
Motivation
We have already know
Baseline precision is significant to the interferometric
accuracy
Precise estimation is required
Idea
Interferometric result can provide information on baseline
Concept can be extended under multiple passes condition,
from baseline to individual sensor position
Iteration and Constraint
IGARSS 2010, Honolulu

4. Motivation Introduction Algorithm Validation Conclusion Concept Baseline Calibration Expand
Outline
1 Motivation
2 Introduction
Concept
Baseline Calibration
Expand
3 Algorithm
Coordinate System
Iteration
4 Validation
5 Conclusion
IGARSS 2010, Honolulu

5. Motivation Introduction Algorithm Validation Conclusion Concept Baseline Calibration Expand
Concept
Baseline Concept
Refer to the relative distance between two sensors
Highlight “relative”
depends on the chosen master image as coordinate origin
build a coordinate system base on master image position,
normally described using “parallel” and “perpendicular”
Initially estimated using orbital information, interpolated
from platform position vector
IGARSS 2010, Honolulu

6. Motivation Introduction Algorithm Validation Conclusion Concept Baseline Calibration Expand
Baseline Error
The root of baseline estimation error is the inaccurate
platform position from orbit data
It can happen on any of the interferometric pair
All the interferograms will be wrong with the same
inaccurate path
IGARSS 2010, Honolulu

7. Motivation Introduction Algorithm Validation Conclusion Concept Baseline Calibration Expand
Geometrical Constraint
The geometric representation of multiple platform positions
can be constructed as polygon(2D) or polyhedron(3D)
Using the orbit estimated baseline, this geometric
representation can be constructed
IGARSS 2010, Honolulu

8. Motivation Introduction Algorithm Validation Conclusion Concept Baseline Calibration Expand
Baseline Calibration
In the past method, error of perpendicular baseline can be
reduced by using GCP or reference DEM
However, the correction is only on the relative distance. No
guarantee for the corrected baseline.
IGARSS 2010, Honolulu

9. Motivation Introduction Algorithm Validation Conclusion Concept Baseline Calibration Expand
Expand
From baseline to relative position
When more information on platform position can be interpreted
from data, global constraint of platform position is needed.
Without constraint, the geometry of platform positions will
break.
IGARSS 2010, Honolulu

10. Motivation Introduction Algorithm Validation Conclusion Concept Baseline Calibration Expand
Expand
Because the problem will become very complicated in 3D
when more passes are used
An iterative optimization method will be provided under
geometry constraint
Global baseline calibration
Detection and quantitative calibration of any pass with
inaccurate orbit information
IGARSS 2010, Honolulu

11. Motivation Introduction Algorithm Validation Conclusion Coordinate System Iteration
Outline
1 Motivation
2 Introduction
Concept
Baseline Calibration
Expand
3 Algorithm
Coordinate System
Iteration
4 Validation
5 Conclusion
IGARSS 2010, Honolulu

12. Motivation Introduction Algorithm Validation Conclusion Coordinate System Iteration
Coordinate System
Requirement
easy to transfer system from one master image to another
error is small enough
TCN (Track, Cross-track and Normal) coordinates is chosen
−P ˆ
n×V
ˆ
n= ˆ
c= ˆ= c×n
t ˆ ˆ (1)
|P| ˆ
|n×V |
IGARSS 2010, Honolulu

13. Motivation Introduction Algorithm Validation Conclusion Coordinate System Iteration
Transfer Equation: Bji −Bij
Is it valid?
Assumption can be made that all of the platform have the
same direction of V
Image pixels within one range row will share the same
baseline TCN coordinates
| Bij · c |2 + | Bij · ˆ |2
ˆ t
∆θ = arctan (2)
Ai + R
Ai : the platform altitude of image i (691.65 km for ALOS)
R: the radius of the earth (6378.1 km)
IGARSS 2010, Honolulu

14. Motivation Introduction Algorithm Validation Conclusion Coordinate System Iteration
System Error
The baseline component along ˆ is very small
t
ˆ
Therefore, for baseline of 1 km along c , the axis error is
0.0081 ◦
ˆ
the baseline error is Bij · c × tan ∆θ 14 cm for this system
Conclude: TCN coordinates system will be considered at
corresponding point between all passes
Bji −Bij (3)
IGARSS 2010, Honolulu

15. Motivation Introduction Algorithm Validation Conclusion Coordinate System Iteration
Iteration: Starting Point
K + 1 passes over same area
Differential interferogram and baseline is generated for all
combinations
Processed with both baseline vector and baseline
changing rate
Initialization:
Bji = −Bij ˙ ˙
Bji = −Bij (4)
IGARSS 2010, Honolulu

16. Motivation Introduction Algorithm Validation Conclusion Coordinate System Iteration
Iteration Steps
Take one pass as master image, calculate the baseline
error to be corrected
IGARSS 2010, Honolulu

17. Motivation Introduction Algorithm Validation Conclusion Coordinate System Iteration
Iteration Steps
Take one pass as master image, calculate the baseline
error to be corrected
(n) 1
Average the result: ∆Pi = K × j=i ∆Bij
IGARSS 2010, Honolulu

18. Motivation Introduction Algorithm Validation Conclusion Coordinate System Iteration
Iteration Steps
Take one pass as master image, calculate the baseline
error to be corrected
(n) 1
Average the result: ∆Pi = K × j=i ∆Bij
(n)
Update all the baseline vectors: Bij = Bij + ∆Pi
1 (n)
A weight coefficient n can be added before ∆Pi to slow down the convergence
IGARSS 2010, Honolulu

19. Motivation Introduction Algorithm Validation Conclusion Coordinate System Iteration
Iteration Steps
Take one pass as master image, calculate the baseline
error to be corrected
(n) 1
Average the result: ∆Pi = K × j=i ∆Bij
(n)
Update all the baseline vectors: Bij = Bij + ∆Pi
1 (n)
A weight coefficient n can be added before ∆Pi to slow down the convergence
Update the reversed baseline Bji
IGARSS 2010, Honolulu

20. Motivation Introduction Algorithm Validation Conclusion Coordinate System Iteration
Iteration Steps
Take one pass as master image, calculate the baseline
error to be corrected
(n) 1
Average the result: ∆Pi = K × j=i ∆Bij
(n)
Update all the baseline vectors: Bij = Bij + ∆Pi
1 (n)
A weight coefficient n can be added before ∆Pi to slow down the convergence
Update the reversed baseline Bji
Change another master image and go back to first step,
until all of the images have been taken once as master
image
IGARSS 2010, Honolulu

21. Motivation Introduction Algorithm Validation Conclusion Coordinate System Iteration
Iteration Steps
Take one pass as master image, calculate the baseline
error to be corrected
(n) 1
Average the result: ∆Pi = K × j=i ∆Bij
(n)
Update all the baseline vectors: Bij = Bij + ∆Pi
1 (n)
A weight coefficient n can be added before ∆Pi to slow down the convergence
Update the reversed baseline Bji
Change another master image and go back to first step,
until all of the images have been taken once as master
image
Calculate the total displacement of all platform:
(n)
∆P (n) = K +1 | ∆Pi |
i=1
IGARSS 2010, Honolulu

22. Motivation Introduction Algorithm Validation Conclusion Coordinate System Iteration
Iteration Steps
Take one pass as master image, calculate the baseline
error to be corrected
(n) 1
Average the result: ∆Pi = K × j=i ∆Bij
(n)
Update all the baseline vectors: Bij = Bij + ∆Pi
1 (n)
A weight coefficient n can be added before ∆Pi to slow down the convergence
Update the reversed baseline Bji
Change another master image and go back to first step,
until all of the images have been taken once as master
image
Calculate the total displacement of all platform:
(n)
∆P (n) = K +1 | ∆Pi |
i=1
Iteration n finished, Take n = n + 1 and restart
IGARSS 2010, Honolulu

23. Motivation Introduction Algorithm Validation Conclusion
Outline
1 Motivation
2 Introduction
Concept
Baseline Calibration
Expand
3 Algorithm
Coordinate System
Iteration
4 Validation
5 Conclusion
IGARSS 2010, Honolulu

24. Motivation Introduction Algorithm Validation Conclusion
Data Over Singapore
8 passes of PALSAR over the Singapore between
December 2006 and September 2009 are used
SRTM is used as reference DEM
GAMMA software is used for the interferograms
Python used for programming
Starting Point:
IGARSS 2010, Honolulu

25. Motivation Introduction Algorithm Validation Conclusion
Results:Relative Position Iteration
h a n g e Vi h a n g e Vi
XC e XC e
F- w F- w
PD
PD
er
er
!
!
W
W
O
O
N
N
y
y
bu
bu
to
to
k
k
lic
lic
C
C
w
w
m
m
w w
w
w
o
o
.d o .c .d o .c
c u -tr a c k c u -tr a c k
250
Relative Normal Corrdinate(m)
Before iteration
Before iteration
After iteration
After iteration 160
200
20070623
159.5
150
20070923
Relative Normal Corrdinate(m)
159
100
−95 −94.5 −94 −93.5 −93
Relative Cross−Track Coordinate(m)
50
(b) for 20070923
0
20090928
Relative Normal Corrdinate(m)
16 Before iteration
20081226 20090628
After iteration
15
−50
20061221
14
20090210
13
−100 20081110
12
11
−150
−200 −100 0 100 200 300
72 74 76 78
Relative Cross−Track Coordinate(m)
Relative Cross−Track Coordinate(m)
(a) Global Relative Position Iteration (c) for 20090928
IGARSS 2010, Honolulu

26. Motivation Introduction Algorithm Validation Conclusion
Results:Displacement plotting without weight
coefficient
20
Total Displacement ∆P(n)
18
20081226
16 20061221
20070923
Displacement ∆P(m) 14 20090928
The total 12
20090210
20070623
displacement 10
20081110
20090628
∆P (n) 8
converges 6
4
2
0
0 1 2 3 4 5 6 7 8 9 10
Interation Number n
IGARSS 2010, Honolulu

27. Motivation Introduction Algorithm Validation Conclusion
Results:Displacement plotting with weight coefficient
20
Total Displacement ∆P(n)
The 18
20081226
convergence 16 20061221
20070923
is slower but 14 20090928
Displacement ∆P(m)
20090210
result in a 12 20070623
20081110
smaller value 10
20090628
8
Speed can 6
neither be too 4
slow nor too 2
fast 0
0 1 2 3 4 5 6 7 8 9 10
Interation Number n
IGARSS 2010, Honolulu

28. Motivation Introduction Algorithm Validation Conclusion
Results:Differential interferogram after calibration
Before iteration
16
After iteration
Relative Normal Corrdinate(m)
15
14
13
12
11
71 72 73 74 75 76 77 78 79
Relative Cross−Track Coordinate(m)
IGARSS 2010, Honolulu

29. Motivation Introduction Algorithm Validation Conclusion
Outline
1 Motivation
2 Introduction
Concept
Baseline Calibration
Expand
3 Algorithm
Coordinate System
Iteration
4 Validation
5 Conclusion
IGARSS 2010, Honolulu

30. Motivation Introduction Algorithm Validation Conclusion
Conclusion
IGARSS 2010, Honolulu

31. Motivation Introduction Algorithm Validation Conclusion
Conclusion
Concept
Satellite platform position can be relatively calibrated from
multiple interferograms
IGARSS 2010, Honolulu

32. Motivation Introduction Algorithm Validation Conclusion
Conclusion
Concept
Satellite platform position can be relatively calibrated from
multiple interferograms
Result
The SAR passes which gives inaccurate platform position
are successfully detected and calibrated
IGARSS 2010, Honolulu

33. Motivation Introduction Algorithm Validation Conclusion
Conclusion
Concept
Satellite platform position can be relatively calibrated from
multiple interferograms
Result
The SAR passes which gives inaccurate platform position
are successfully detected and calibrated
Disadvantage
Platform position can only be calibrated along
perpendicular baseline
IGARSS 2010, Honolulu

34. Motivation Introduction Algorithm Validation Conclusion
Conclusion
Possible Application
Orbit refinement for SAR
Baseline problem for deformation monitoring, like
earthquake
IGARSS 2010, Honolulu