Diapositive 1

download report

Transcript Diapositive 1

VISUPOL version 4
----------------Analysis and visualization of
high resolution polarimetric radar images
Pauline Audenino - CS
1
7th CNES/DLR Workshop on Information extraction and scene understanding for meter resolution images, 28/03-30/03 , 2007
Outline
Context
Visupol in a nutshell
Main functionalities
HR Image example
Theoretical reminders
Visupol innovative features
IHS representation
/4 mode
Polarimetric synthesis
Conclusion and perspectives
2
Context
Future new missions will include sensors with full polarimetric capabilities:
TerraSar X, ALOS, Radarsat 2, etc..
Visupol is a dual tool developed since 2003 by CS for the CNES SI/AR
Altimetry and Radar department, that allows:
 The analysis and visualisation of polarimetric data
 The assessment of new polarimetric concepts (compact polarimetry, target detection, ITS
representation, quaternion analysis)
Main steps overview
End 2003 - initialization of the HMI
2004: v2 - visualization of polarimetric parameters
2005: v3 - compact polarimetry and polarimetric decompositions
2006: v4 - masks, targets analysis and polarimetric synthesis
3
Visupol in a nutshell
Modular, evolutive tool coded in IDL/Envi
User friendly Interface
It allows
 an easy access to the polarimetric images and results
 a wide range of format handling (Jpeg, Tiff, Raw radar)
Available on several types of plateforms:
Sun-Solaris, PC Windows, PC Linux
Visupol availability as a free open source is currently under consideration
Complementarity to Polsarpro hereafter developed
4
Welcome in visupol!
Coherent polarimetry:
All 4 combinations in emission
and reception are available
(HH, HV, VV, VH)
Incoherent polarimetry:
2 combinations are available
(based on the Envisat Asar
Alternating mode)
5
Main Functionalities
Classical Image handling
Histograms
Orientations change
Back up in JPEG, TIFF formats, ..
Image quality Analysis
Azimuth or range Spectra
3D targets visualization
Polarimetric calibration analysis
Targets detection and analysis algorithms:
Internal Hermitian product, simple and multi polarization
Polarimetric analysis
Transform form single look to multi looks
Coherence matrix: H, Alpha Anisotropy representation
Coherent decomposition: Pauli, Kroager, Touzi, Cameron
Compact Polarimetry: /4 mode
Polarimetric signatures
Polarimetric synthesis
6
Nord
Polarimetric
radar Image
RAMSES X band,
01/03/06
HR, metric resolution
Over the Cnes, center of
Toulouse
Uncalibrated data
Colored Composition
Multilook 3x3
HH HV VV
© ONERA 2006
© ONERA/CNES 2006
7
Optical data take
ONERA/IGN/CNES
campaign Pelican
8
The polarization of the electromagnetic waves
vˆ
Magnetic
field

H
electrical
field

E
kˆ
Trajectory of the electrical
field during the
propagation (spiral)
Energy
propagation
hˆ

• In free space, the electric and magnetic fields ( E
Transverse plan of the
electromagnetic wave

and H
)
and the energy propagation direction are positioned in 3
orthogonal directions. The electromagnetic fields are therefore
contained if the le plan which is transverse to the energy
propagation direction.
9
The polarization ellipse
vˆ
Y: Orientation of the major axis
of the polarization ellipse
t: Ellipticity rate
zˆ
t
y
hˆ
Projection of the electrical
field trajectory in the
propagation plan.
Electromagnetic wave
transverse plan
Particular cases:
• The circular polarization (t =1)
• The linear polarization (t = 0)
• Projected in the plan transverse to the
propagation, the extremity of the electrical field
describes an ellipse during its propagation:
this is the polarization ellipse.
10
The polarization of the electromagnetic waves :
The Poincaré Sphere
Representative point of
the polarization (y,t)
Left circular
polarization
zˆ
• At a given polarization state and polarization ellipse, defined
by the orientation and ellipticity angles y et t, is associated
a point at the surface of a unitary sphere, defined by the
azimuth and elevation angles 2 y et 2 t. The associated cartesian
coordinates s1, s2 et s3 are given by the relations:
s3
2.t
s1
s2
yˆ
2.y
xˆ
s1  cos(2.y ).cos(2.t )

s2  sin (2.y ).cos(2.t )
s  sin (2.t )
 3
s1, s2 et s3 follow the relation:
Linear
polarizations
Right circular
polarization
s12  s22  s32  s02  1
s0, s1, s2 et s3 are the wave Stokes parameters.
11
The linear polarizations
vˆ
vˆ
hˆ
hˆ
kˆ
Horizontal polarization
(Ex: RADARSAT HH)
kˆ
Vertical polarization
(Ex: ERS-1/2 VV)
• In the case of a linear polarization, the electrical field is characterized by a fixed orientation (doesn’t
change during propagation). This is the mode selected by the conventional SAR systems, with polarization
uniqueness, such as RADARSAT (horizontal polarization) or ERS-1/2 (vertical polarization).
12
Example of different polarizations
Single polarization image HH
Full polarimetric image HH, HV, VV
Site of Ulan-Ude, Russia, 1994 Image SIR-C (L band, 24°)
13
The wave depolarization
Incident wave
(linear polarization)
Backscattered
wave
Polarized wave
( adequate for
polarimetric analysis )
Depolarized wave
polarimetric noise.
‘ pseudo-random ’ environment
The polarimetric analysis is relevant only if the polarized part of the backscattered
wave is predominant:
The computation of the coherence matrix allows to quantify the two parts of the waves
 Its decomposition leads to the definition of polarimetric parameters: for example the
Cloude Pottier H, Alpha, Anisotropy description
14
Synoptic vizualisation of the polarimetric information: IHS
Fusion of
(SPAN)
f HH , f HV , f VV
Hue
ˆ
Saturation
Intensity
Projection on a
Intensity / Hue / Saturation (IHS)
color system, driven by
polarimetric discriminators
1 H
The color value is driven
by the value of
ˆ
Local colorization locale of the image
Low entropy high saturation,
High entropy  ‘black and white’ image
15
Ground truth
© ONERA/CNES 2006
16
Target identification
Trihedral F
Trihedral B
Trihedral E,D,A
Luneberg C
Dihedral I,J,K
S
3 twingos
Luneberg G Doris
© ONERA 2006
N
17
IHS representation
Odd number of
reflexions (trihedral)
Trihedral B
Luneberg C
Dipoles (~ linear
structures)
Dihedral
double reflexion
(~ buildings,
dihedral)
Luneberg G Doris
18
The /4 mode
19
Polarization synthesis
It aims at generating a virtual radar image similar to the one that would
have been obtained with the emitted and received chosen polarizations.
From the existing channels, you can emulate the polarimetric radar
images derived by any polarization combinations in emission and
reception
Applications:
Contrast enhancement between areas of different intensities
Improvement of the sensibility toward one specific phenomenon (urban,
forest, etc)
 Within VISUPOL
Static: an image of polarization synthesis related to the selected polarization
states is created
Dynamic: a sequence of images of polarization synthesis are generated
along a specified path
20
Polarization synthesis
21
22
Dynamic polarization synthesis
23
Dynamic
Polarization
synthesis
Interface
24
Dynamic polarization synthesis
Copolarization
(same polarizations in
emission and reception)
Initial state : psi=0. tau=0.
Path: isolatitude (equator)
25
Dynamic polarization synthesis
Crosspolarization
(orthogonal polarizations
in emission and reception)
Initial emission state : psi=0. tau=0.
Initial reception state : psi=/2 tau=0.
Path: isolongitude (meridian H)
26
Dynamic polarization synthesis
Copolarisation – General path
Departing point : psi= -60. tau= 34.
Arrival point:
psi= 5. tau=-18.
27
Complementarity between POLSARPRO
(©ESA)
and VISUPOL
(©CNES)
Education
Free Open source
Current Airborne
and Spaceborne
input formats
Bitmap results
viewer
Complete classical polarimetric data processing
Not an open source. But an
Objective is to insert some tools
in Orfeo Tool Box
Engineering
Quick visual results on full
resolution extracts
Current Airborne
and Spaceborne
input formats
In addition to standard analysis,
focus on polarimetric advanced concepts
(polarisation synthesis viewer, target detection and analysis, compact polarimetry,
28
quaternion analysis, …)
Conclusion and perspectives
Visupol is an easy to use, original toolkit dedicated to the
Analysis and visualization polarimetric data
Assessment of the relevancy of new polarimetric concepts
 Further development:
Colored dynamic polarization synthesis
MLC calibration
Quaternions analysis
Compact polarimetry for the P band data….
 Open discussion: Visupol, a free software ?
Step by step, beginning with the polarization synthesis module
Within the OTB ?
29