1 Remote sensing and SAR radar images processing Physics of radar2 TABLE OF CONTENTS Potentialities of radar Radar trans...
Remote sensing and SAR radar images processing Physics of radar
© copyright CNE
TABLE OF CONTENTS
Potentialities of radar
Radar transmission features
Propagation of radio waves
Radar equation
Surface scattering mechanisms
Volumetric scattering mechanisms
Penetration depth of waves in observed media
© copyright CNE
Potentialities of radar
‘All-weather’ observation system (active system). not sensitive to sun lightening, not sensitive to cloud cover
Other advantages with respect to optics: ranging (simple and accurate geometric modeling), detection capacity (even at medium resolution) Sensitivity to dielectric properties of medium (water content, humidity), and to its roughness the radar response when the moisture and/or when roughness Penetration capabilities
estimation of plant biomass,
observation of buried structures, cartography of subsoils, etc. penetration when the frequency Sensititivity to topography (related to the acquisition geometry)
Sensitivity to geometrical structures with scales of the same order as the wavelength
© copyright CNE
Introduction Imaging radar features
Drawbacks: speckle (difficult visual interpretation)
Sensitive to: roughness relief (slope) humidity metallic and artificial objects
© copyright CNE
Introduction (2/2)
Accessibility With respect to optics:
day/night imaging capacity (x 2) insensitive to cloud cover ( x 5) 10 times more images available Faster information access
Multi-Incidence - Multi-Resolution
With a constellation of 4 SAR Satellites : information access delay shorter than 24h (from decision to interpretation)
© copyright CNE
Radar transmission features
The frequency (carrier frequency + bandwidth)
1
0 .1
10
Ku Ka X
f ( GHz )
30
300
C
100 S
L
3
(cm )
P
0 .3
The propagation direction (Ex: ERS: 23°)
The transmitted power (Ex: ERS: ~ 5 kW pic)
The polarization
vˆ
vˆ
hˆ
hˆ kˆ
Horizontal polarization RADARSAT type
impact on image quality
hˆ
kˆ
nˆ Vertical polarization ERS type
nˆ
hˆ
© copyright CNE
Radar transmission features xˆ electric field E H magnetic field
zˆ
kˆ
energy propagation
yˆ Spatial-temporal variations of the electric field during propagation:
E ( r , t ) E O . exp ( j .( kˆ.r .t )) Configuration of electromagnetic fields in free space:
E ( r , t ) H ( r , t ) kˆ
E ( r , t ), H ( r , t ), kˆ
Propagation of radio waves Maxwell’s equations
form a direct trihedral
© copyright CNE
Radar equation (1/4)
Case of point targets (1/2)
i: incident flux point target Portion of backscattered power i = incident flux = incident power per area unit normal to incident beam:
i Ge .
Pemitted 4 R ²
Ge: Transmitting antenna gain; R: Radar-target distance
Portion of backscattered power:
i
4 R ²
Power received on the receiving antenna:
P received Aeff
i.
² Gr 4
Aeff 4 R ²
Portion of energy sent back by the point target = Radar reflective area (SER )
Effective area of receiving antenna
© copyright CNE
Radar equation
Case of point targets
The radar equation is derived from the transmission-backscattering-reception process:
P received
. Ge Pemitted 4 R ²
backscattering
transmission
P received
Pemitted
Gr ² . 4 R ² 4
. Ge
4
Gr
3
system
Set of terms determined by calibration procedures
reception
² R
4
Target (radar equivalent crosssection) Unit: m²
propagation
© copyright CNE
Radar equation
Case of extended targets
The radar backscattering coefficient (marked σo) represents the average value of the Radar reflective area per area unit (case of an extended target, for example on the scale of a pixel):
d dS
If area is hom*ogeneous:
o
S
o
k
P received Pemitted
Unit: m²/m²
‘ 0 ’ means normalization in relation to an area
σ is expressed in m², σo is expressed in m²/m²
o
Coefficient k is determined by calibration
Representation of 0 on a logarithmic scale: o ( dB ) 10 . log 10 ( o )
Value dynamics
~ -40 dBm²/m²
Unit: dBm²/m²
+10 dBm²/m²
© copyright CNE
Radar equation (4/4)
Case of extended targets (2/2) Behavior and typical values of 0
0 50 dBm²/m²
Point targets: vehicles, ships, etc.
20 dBm²/m²
Urban areas, etc.
0 dBm²/m²
0 dBm ² / m ²
0 0 dBm ² / m ²
-7 dBm²/m² -10 dBm²/m² -15 dBm²/m² -22 dBm²/m²
0 0 dBm ² / m ²
0 0 dBm ² / m ²
Forest
Vegetation
Short grass Noise image limit Concrete, bitumen, etc.
Depends on incidence Depends on frequency
© copyright CNE
Surface scattering mechanisms (3/4) Case of a rough dielectric surface (1/2) medium 1 Medium 2 hom*ogeneous: no volume scattering
medium 2
roughness generates backscattering (part of energy returning to the radar). The dielectric nature produces penetration.
The radar backscattering coefficient o (quantity of energy returning to the radar) depends on: • The surface roughness o ~ f (roughness) . g (r ) • The dielectric permittivity of medium (related to the water content)
o when roughness o when moisture
moisture
hence indetermination:
= Rough dry soil
Wet smooth soil
Indetermination between the moisture and roughness level based on knowledge of 0 alone © copyright CNE
Surface scattering mechanisms (4/4) Case of a rough dielectric surface (2/2)
Quantification of roughness, Rayleigh’s criterion: A surface is not intrinsically smooth or rough from the radar point of view. This concept is meaningful only if referred to wavelength. zˆ
kˆ inc
A
h
B
Rayleigh’s criterion:
• When the phase difference between the 2 reflected waves (at A and B) due to propagation is <
/2, the surface is considered as smooth.
Now: = 2/ = 2/hcos smooth surface if: h < λ/8/cosθ • Δ > π/2 rough surface
Remark: in C-band (l=5.6 cm), condition (1) gives h < 0.8 cm at 23° (ERS-1): all natural surfaces are rough under these observation conditions. © copyright CNE
Volumetric scattering mechanisms Case of the forest 6 1
4 2
3
5
1) Crown scattering
2) Trunk scattering
3) Trunk-soil interaction
4) Attenuated soil scattering
5) Direct soil scattering
6) Trunk-branch interaction 7) Soil-branch interaction
7
Examples of main backscattering mechanisms on the forest Volume backscattering mechanisms generally rely on interaction mechanisms which are highly complex and still not well-known. Main trends: Backscattering coefficient when vegetation volume (biomass) Wavelength penetration when frequency , i.e. when wavelenght
© copyright CNE
Penetration depth of waves in observed media Penetration capabilities of radar waves versus wavelength
L-Band = 23 cm
C-Band = 6 cm
X-Band = 3 cm
1m
6m
20 m SIR-C image Landes Forest, France L-Band, 26° (0HV) High penetration capabilities in canopy. Application: Biomass cartography (CESBIO origin )
© copyright CNE
RADAR SENSITIVITY TO BIO-MASS
o hv (dBm2/m2)
-2
o vv (dBm2/m2)
o
vv (dBm2/m2)
-6 -7 -8 -9
-10 -11
-4
-6
-8
-12 0
33
65
95
130
150
Biomass (tons/ha) L-band, VV-polarisation, 26°
-10 0
33
65
95
130
150
Biomass (tons/ha) C-band, VV-polarisation, 26°
-14 -16
Experimental results show that radar sensitivity to biomass is a complex mechanism depending jointly on frequency and polarisation
-18 -20 -22 -24
33
65
95
130
Biomass (tons/ha) L-band, HV-polarisation, 26°
150
SIRC data, Landes forest, France (origin : CESBIO)
© copyright CNE
Penetration depth of waves in observed media Radar signature differences between X-band (10 GHz) and L-band (1.25 GHz)
X-Band ESAR
L-Band ESAR
• The visibility of a grass runway in the right image demonstrates the volumetric scattering characteristics (thus the penetration characteristics) in L-Band. For the same reason, forest plots are brighter in L-Band. Surface roughness is better reflected in X-band. Also apparent is the rather low image constrast in X-Band as compared to L-Band..
From: http://atlas.op.dlr.de/ne-hf/projects/ESAR/igars96_scheiber.html
© copyright CNE
Penetration depth of waves in observed media Capabilities of low-frequency imaging radars (P-Band)
Centimetric wavelength (2 cm) S-Band
Metric wavelength (290 cm) P-Band
The right image is an example of low-frequency radar imagery acquired in the P-Band (100 MHz). Although of lower image quality compared to the left image, it makes it possible to see underground structures, in this case pipeline segments (VNIIKAN Siberian campaign -1994) © copyright CNE
Left: SIR-C multi-frequency radar image (Nile) (R : CHH, G : LHV, B: LHH). Inverse LUT
Below: Wave penetration in bare soil for different SAR bands as a function of humidity
×
bande L bande C bande X
18 16
From : www.jpl.nasa.gov/radar/sircxasr
penetration ( cm )
14 12 10 8 6 4 2 0 0
10
20
30
40
50
60
Soil humidity ( gr/cm3 ) Left: IR optical image over the same region
RADAR SOIL PENETRATION
© copyright CNE
The speckle noise, consequence of a coherent illumination (1/2) m
pixel
pixel
n 1
n 2
m
e
e
contribution of one scatterer
pixel response
The speckle noise, consequence of a coherent illumination (2/2)
Large radiometry : large noise The speckle noise is a multiplicative noise Low radiometry : low noise
Image SETHI, bande C, 3 m
SAR principle / Image Quality / Processing
© copyright CNE
CONTENT Introduction Reminders: detection radar / antenna scattering Side-Looking Airborne Radar (SLAR) Range processing Synthetic Aperture Radar (SAR) Azimuth processing SAR ambiguities Moving targets Special modes (SAR) Image Quality: Radiometry Image Quality: Geometry Image Quality: localization Processing at CNES: PRISME
© copyright CNE
Reminder: Detection radars target
azimuth
azimuth range
Range
range
Pulses
Radar screen
• The range information comes from the time needed by the pulse to travel way and back 0
t
Pulse transmission chronogram
© copyright CNE
Reminder: Antenna scattering
Wavelength
L
L
Angular aperture (horizontal plane)
Antenna length (horizontal direction)
L'
The larger the antenna, the narrower the aperture (resolution )
Numerical example: L 4m, R 4 km (airborne radar), 3 cm (X band) resolution 30 m
© copyright CNE
SLAR: Side-Looking Airborne Radar (1/9)
s
Azimuth direction
Range direction
Linear displacement of the antenna along the track (aircraft)
Pulses
© copyright CNE
SLAR: Why « Side-Looking » ? (2/9)
Removal of Left/Right Range ambiguity
Left/Right Range ambiguity
rrange
r azimuth 3D representation
© copyright CNE
SLAR (3/9)
Azimuth resolution
s
Prf: Pulse Repetition Frequency
Azimuth direction
L
W
Range direction
H
Chronogram: pulses versus time
Echoes
Transmitted pulse
Numerical example: (airborne example) L=4m W = 5 cm = 20-60° H = 3000 m Swath = 4 km Razi = 25 - 45 m
Azimuth resolution: Rθ, with
L
SLAR azimuth resolution 35m Remark: Azimuth pixel size = S / Prf
Swath
Razi
© copyright CNE
SLAR (4/9)
‘ Ideal ’ range resolution: Case of a Dirac pulse transmission Transmitted pulse (Dirac) ideal time resolution
Sampling of the received echo (with Fs frequency) = sampling in the spatial domain (generation of an image line) c : range (distance) pixel size in 2 Fs the radar geometry, by construction
of an image line c 2 Fs sin
: ground range pixel size
In the case of a Dirac transmission, range resolution = pixel size in range: it depends only on the sampling frequency Fs. This is always true for the range pixel size (by construction), but not for the resolution if the pulse is not a Dirac © copyright CNE
SLAR (5/9)
‘ Real ’ range resolution: case of a pulse transmission of duration (1/2)
Practically, for power budget reason, the pulse duration is . The resulting resolution is dominated by the Factor c as shown in next slide
2
1/ PRF
Pulse duration
distance (or range) resolution:
c / 2
and ground range resolution:
c 2 sin
(Numerical example ERS, 37 s, range resolution 5 km)
© copyright CNE
SLAR (7/9)
Improvement of range resolution: pulse compression
In order to improve distance resolution, the transmitted pulse is frequency modulated (over a bandwidth Bchirp): this can be shown to be equivalent to the transmission of a shorter pulse:
Modulation bandwidth:
Bchirp
comp 1 / B chirp
comp
equivalence
1/ PRF 1/ PRF 0
t
Numerical example ERS, 37 s, Bchirp=15.5 MHz comp=64 ns
c 2.B chirp
Achieved range resolution (slant range):
Re sdist
Achieved range resolution (ground range):
Resdist _ sol
c 2.B chirp . sin(i )
i
/ sin(i)
© copyright CNE
SLAR (8/9)
Pixel size vs. Resolution in range
comp
Numerical example: ERS resolution
Compressed Pulse duration
Fs 18.96 MHz t1
t0
t2
time
Pixel slant _ range 7 ,9 m
Pixel ground
_ range
26 to 18 m
swath
B 1
c 2 Fs
comp
15.5 MHz
Resslant _ range 9.7 m Pixel size
c 2 Fs sin
The pixel size is defined by the sampling frequency Fs The range resolution is defined by the modulation Bandwidth Bchirp
Resground _ range 22 to 32 m The pixel size is generally “built”slightly smaller than the resolution: FsBchirp
© copyright CNE
Synthetic Aperture Radar (SAR) Principle
Pulse transmission
v
Azimuth direction
Range direction
(1/12)
The antenna progression along the orbit allows to observe each given point at different times
Resolution improvement in the azimuth direction
© copyright CNE
SAR Principle (2/12) Signal processing in azimuth: principle (1/2)
illumination duration
L illumination duration v
v
Synthetic Aperture
azi
Equivalence
The moving small antenna is equivalent to a long fixed antenna (size , directivity , resolution )
SAR T
T'
S azi Coherent adding of successively received echoes
v
T
T'
Resolution gain in the azimuth direction (Ex: ERS: 5 km 5 m)
The compression rate Na equals the number of coherently added echoes (complex addition). It is
the resolution gain in the azimuth direction
© copyright CNE
SAR Principle (4/12) Signal processing in azimuth: Doppler analysis
(1/5)
The range variations between a target and the sensor produce a linear Doppler effect of the transmitted pulse
(quadratic distance&phase variations with time linear frequency variations with time in a frequency band: Doppler Bandwidth)
Fd > 0
Fd = 0
Fd < 0
© copyright CNE
SAR Principle (5/12) Signal processing in azimuth: Doppler analysis illuminati on duration : T int
Doppler frequency
v
L
Instantaneous phase azi
Total Doppler bandwidth
R
T Position origine des temps
(2/5)
dop
d 1 2 dt
L
2
0 2 f
dop
B
dop
T
int
2
where:
R ² v ² t ²
1/ 2
v² t R
v ² T int 2 R R
B dop 2
T' S azi R R
f
1 L
v
v L
v L spatial resolution Bdop 2
Target-antenna range variations during the illumination time produce a Doppler effect, resulting in spreading the backscattered energy over a bandwidthB dop 2 v © copyright CNE
L
SAR Principle (6/12)
Signal processing in azimuth: Doppler analysis (3/5)
v forward look central look
Frequency spectrum in azimuth (antenna pattern modulation)
backward look
S ( f ) azi
f
f
dop
B T
t B int
/2
T
T
int
/2
dop
dop
f Doppler excursion versus time (case of a zero Doppler centroïd)
dop
2
v² t R
int
© copyright CNE
Image quality: geometry (1/4) radar versus optics
radar acquisition: range discrimination of the space: A’,B’,C’
(From Elachi, 1989)
optical acquisition: angular discrimination of the space: A”,B”,C”
© copyright CNE
Image quality: geometry (2/4) geometrical artifacts related to the vision in range The foreshortening effect Radar discrimination capacity
radar
(From Elachi, 1989)
• ‘shortening’ of slopes facing the radar • ‘stretching’ of slopes oppositely oriented to the radar
© copyright CNE
Image quality: geometry (3/4) geometrical artifacts related to the vision in range The layover effect
Look direction
Radar trajectory
A shadow
A’
B B’
The point A (top) is projected before B (base) in the direction of the radar pass
Layover effect on airborne image “Sethi” Tour Eiffel, Paris, C band (resolution: 3m)
© copyright CNE
Image quality: geometry (4/4) geometrical artifacts related to the vision in range example of foreshortening, layover and shadows
Standard beam position 1: acquired Feb.12, 1996
From: RADARSAT Geology Handbook (RADARSAT International), 1997
© copyright CNE
ENVISAT MERIS Not quite the same
geometry…!! Where is Spain?Where is the North? Where did the satellite pass????
© copyright CNE