SAR image properties

Thuy Le Toan Centre dEtudes Spatiales de la Biosphre (CESBIO) Toulouse, France Thuy.Letoan@cesbio.cnes.fr

9 September 2010

Lecture D4L1

SAR image properties

Thuy Le Toan

Contents
Physical content of SAR images Geometrical properties Statistical properties of SAR measurements

9 September 2010

Lecture D4L1

SAR image properties

Thuy Le Toan

SAR measurements
Measurements derived from a single SAR image Intensity at single or multiple polarisation Polarimetric measurements (intensity and phase from polarimetric SAR) Measurements derived from multiple SAR images Temporal variation of intensity Interferometric coherence and phase This lecture: intensity measurements

9 September 2010

Lecture D4L1

SAR image properties

Thuy Le Toan

Basic radar measurement

The basic measurement made by a SAR is S (amplitude and phase). This is the complex image.

Main types of images: A is the amplitude image. I = A2 is the intensity image. (the phase of a single image is not exploitable)

9 September 2010

Lecture D4L1

SAR image properties

Thuy Le Toan

The radar cross-section

The radar cross-section (RCS) is defined as

pq = 4 S pq

Ps = 4R Pi
2

[m ]
2

R is the radar-target distance Pi is the incident power, Ps is the power scattered by the target.

9 September 2010

Lecture D4L1

SAR image properties

Thuy Le Toan

The backscattering coefficient

For distributed targets each resolution cell contains many scatterers and the phase varies rapidly with position. The differential backscattering coefficient, o, is

4R 2 Ps = A Pi

[m2/m2]

where A is the area of the illuminated surface over which the phase can be considered constant.
9 September 2010 Lecture D4L1 SAR image properties Thuy Le Toan 6

Scattering mechanisms
Surface scattering Surface scattering

soil, rock

water
Volume scattering Volume-surface scattering

Volume scattering if penetration

snow
Volume scattering

vegetation

Surface scattering

Surface scattering

9 September 2010

Lecture D4L1

SAR image properties

Thuy Le Toan

Scattering mechanisms
The backscattered signal results from: - surface scattering - volume scattering - multiple volume-surface scattering The relative importance of these contributions depend on - surface roughness - dielectric properties of the medium All of these factors depend on the - radar frequency - polarisation - incidence angle
9 September 2010 Lecture D4L1 SAR image properties Thuy Le Toan 8

Surface scattering
Smooth surface Rough surface

r1

r1

The roughness of the surface (wrt to the wavelength) governs the scattering pattern

r2 > r1 medium 2 is wetter than medium 1

Wetter media

r2

r2

The dielectric constant (moisture content) of the medium governs the strength of the backscatter

9 September 2010

Lecture D4L1

SAR image properties

Thuy Le Toan

Effect of surface roughness

Mud

RADARSAT (C band, HH, 45) Quaternary lithology: Bathurst Island, Canada

From : RADARSAT Geology Handbook

Lime stone

Mud fragments (smooth surface) low radar backscatter

9 September 2010 Lecture D4L1

Limestone Higher backscatter because of rougher surface

SAR image properties Thuy Le Toan 10

Angular variation of the backscatter

Very smooth surface

Moderately rough

Rough surface

9 September 2010

Lecture D4L1

SAR image properties

Thuy Le Toan

11

Sensitivity to soil moisture

Irrigated fields: higher backscatter C-band

9 September 2010

Lecture D4L1

SAR image properties

Thuy Le Toan

12

Sensitivity to soil moisture

Experimental results using a ground based scatterometer

(adapted from Le Toan, T. , 1982, "Active microwave signatures of soil and crops: Significant results of three years of experiments", In Proceedings of International Geoscience and Remote Sensing Symposium (IGARSS 82)

9 September 2010

Lecture D4L1

SAR image properties

Thuy Le Toan

13

Effects of roughness and moiture

ERS backscattering coefficient (dB)

The relationship between radar backscatter at C band 23 VV and soil moisture is modulated by surface roughness

Mattia et al., 2000

Volumetric soil moiture content (%)

9 September 2010 Lecture D4L1 SAR image properties Thuy Le Toan 14

The surface roughness

Smooth surface (seed bed): Fractal

Rough surface (ploughed): Single scale

9 September 2010

Lecture D4L1

SAR image properties

Thuy Le Toan

15

Scattering from vegetation

* Order 0 soil * Order 1: simple scattering

soil

soil

(negligible)

soil

0 0 0 0 = soil + veg . + soil veg .

9 September 2010

Lecture D4L1

SAR image properties

Thuy Le Toan

16

Geometric, structural properties

v h k scatterers density soil roughness scatterers size,shape scatterers orientation

Dielectric properties
scatterers water content soil moisture

Vegetation scattering

9 September 2010

Lecture D4L1

SAR image properties

Thuy Le Toan

17

Volume scattering

Single and multiple scattering

9 September 2010

Lecture D4L1

SAR image properties

Thuy Le Toan

18

Polarisation effects: sources of depolarisation

Volume scattering Point scatterer Anisotropic scatterer Multiple scattering

V V -> no depolarisation V V V

H V

-> depolarisation Surface scattering V

H,V -> depolarisation

-> no depolarisation

-> some depolarisation

SAR image properties Thuy Le Toan 19

9 September 2010

Lecture D4L1

Main scattering mechanisms from a cereal canopy

Attenuated ground scattering Dominant at L-band Stem-ground interaction
Dominant for flooded fields

Scattering on leaves, ears

Dominant at X-band

9 September 2010

Lecture D4L1

SAR image properties

Thuy Le Toan

20

Scattering from a rice canopy

HH and VV increases with the plant biomass.The increase is very important (up to 10 dB during the growth season) (Le Toan et al., 1997) HH>VV because of the stronger attenuation of VV by vertical stems (and Fresnel reflection RH > RV )

At C band, HH and VV: the dominant scattering mechanism is the double bounce vegetation-water

Water
9 September 2010 Lecture D4L1 SAR image properties Thuy Le Toan 21

Hongze Lake WideSwath VV 2004 08 18

9 September 2010

Lecture D4L1

SAR image properties

Thuy Le Toan

22

Hongze Lake WideSwath VV 2004 10 08

9 September 2010

Lecture D4L1

SAR image properties

Thuy Le Toan

23

Change based on ratios

1 = 2
Ratio of intensities is equivalent to the difference of the logs. Advantages: Depends only on the relative change in intensity between the images. Unaffected by topography and other multiplicative effects, e.g., calibration.

9 September 2010

Lecture D4L1

SAR image properties

Thuy Le Toan

24

Hongze Lake The ratio image 20040818 / 20041008

9 September 2010

Lecture D4L1

SAR image properties

Thuy Le Toan

25

Hongze Lake WS R : VV_20040818 G : VV_20041008 B : VV_2 / VV_1

9 September 2010

Lecture D4L1

SAR image properties

Thuy Le Toan

26

Hongze (Jiangsu) 2004 09 06

HH

VV

9 September 2010

Lecture D4L1

SAR image properties

Thuy Le Toan

27

Rice mapping using HH/VV

Hongze(Jiangsu) 2004 09 06

9 September 2010

Lecture D4L1

SAR image properties

Thuy Le Toan

28

Forest scattering
4 5

Scatterers contribution
Leaves, Needles

Primary Branches Secondary branches

Higher order branches

1) Direct Crown scattering

4) Multiple trunk-ground 5) Attenuated ground 6) Direct ground scattering

Direct trunk-ground 3) Trunk scattering

2)

Trunk

9 September 2010

Lecture D4L1

SAR image properties

Thuy Le Toan

29

How the trees are seen by radars?

Austrian pine X band

= 3 cm

= 27 cm

L band

= 70 cm

P band

>3m

VHF

The main scatterers in a canopy are the elements having dimension of the order of the wavelength

9 September 2010

Lecture D4L1

SAR image properties

Thuy Le Toan

30

Forest burnt area

ASAR WSM images, Heilongjang province, China

April 2003

October 2003

9 September 2010

Lecture D4L1

SAR image properties

Thuy Le Toan

31

Fire monitoring

04/2003 B&G 10/2003R HH polarization C-band

75 m resolution

Increase in backscatter of burnt forest due to: - weaker attenuation in the canopy due to the consumption of the needles by the fire - exposure of the small twigs and branches with higher - increased backscatter from the
9 September 2010 Lecture D4L1 SAR image properties Thuy Le Toan 32

P HH HV VV

L HH HV VV

ESAR, Remningstorp forest, Sweden

9 September 2010 Lecture D4L1 SAR image properties Thuy Le Toan 33

9 September 2010

Lecture D4L1

SAR image properties

Thuy Le Toan

34

Contents
Physical content of SAR images Geometrical properties Statistical properties of SAR measurements

9 September 2010

Lecture D4L1

SAR image properties

Thuy Le Toan

35

Geometry: Radar vs optics

optical acquisition: angular discrimination

radar acquisition: range discrimination

(From Elachi, 1989)

9 September 2010 Lecture D4L1 SAR image properties Thuy Le Toan 36

Slant range and ground range

G= S.sin ()

9 September 2010

Lecture D4L1

SAR image properties

Thuy Le Toan

37

Geometric distortion

9 September 2010

Lecture D4L1

SAR image properties

Thuy Le Toan

38

Geometric distortion

9 September 2010

Lecture D4L1

SAR image properties

Thuy Le Toan

39

Geometrical artifacts related to the vision in range The foreshortening, layover and shadow effects

From: RADARSAT Geology Handbook (RADARSAT International), 1997

9 September 2010

Lecture D4L1

SAR image properties

Thuy Le Toan

40

Contents
Physical content of SAR images Geometrical properties Statistical properties of SAR measurements

9 September 2010

Lecture D4L1

SAR image properties

Thuy Le Toan

41

Initial HH and VV images

HH and VV image after filtering

HH (magenta) and VV (green) images 400 x400 pixels Gaoyou, Jiangsu province, 2004 09 06

9 September 2010

Lecture D4L1

SAR image properties

Thuy Le Toan

42

The physical origine of speckle

Ground scene

Constructive speckle

Destructive speckle

SAR image pixels

Resolution cells are made up of many scatterers with different phases, leading to interference and the noise-like effect known as speckle.

9 September 2010

Lecture D4L1

SAR image properties

Thuy Le Toan

43

Estimating the backscattering coefficient

Given L independent measurements from a uniform distributed target, the MLE of 0 is given by

1 L (k ) I = I L k =1
where the I(k) are individual intensity measurements. N.B. This does not depend on the original form of the data (amplitude, intensity or complex). L is called the number of looks.
9 September 2010 Lecture D4L1 SAR image properties Thuy Le Toan 44

The gamma distribution

1.6 1.4 1.2 1.0 0.8 0.6 0.4 0.2 0.0 0.0 L =1

L = 12 L = 10 L=8 L=6 L=4

Pdf of intensity image for a given backscatter value. The distribution is narrower with an increasing number of looks

PI (I )

0.5

1.0

1.5

2.0

Unit mean Gamma distributions of orders of 1, 4, 6, 8, 10 and 12. The distribution tends to normality as L increases.

9 September 2010

Lecture D4L1

SAR image properties

Thuy Le Toan

45

Equivalent Number of Looks (ENL)

In SAR intensity data, the speckle variance is proportional to the mean intensity squared.

ENL =

(mean )

variance

If ENL is large, the spread of values due to speckle is small.

9 September 2010

Lecture D4L1

SAR image properties

Thuy Le Toan

46

SAR image speckle filtering

To reduce the speckle noise in order to retrieve the useful information content 1. Frequency filtering: spectral filtering during SAR processing (e.g.ASAR PRI images provided to users: 3 Looks) 2. Spatial filtering: local estimation on gliding windows Filters of Lee, Kuan, Frost, MAP widely available 3. Multi channel filtering: applied on multiple images of the same scene: muti polarisation, multi temporal, and multi frequency

9 September 2010

Lecture D4L1

SAR image properties

Thuy Le Toan

47

Multi-image Intensity Filtering

Original Images
I1 I2

. . . . . . .
Filter

IM

Filtered images Purpose of filter: (1) (2)

J1

J2

. . . . . . .

JM

Preserve radiometry unbiased I k ( x, y ) = J k ( x, y ) 1 k M Minimise the variance of constraints in (1)

J k subject to the M

9 September 2010

Lecture D4L1

SAR image properties

Thuy Le Toan

48

HH

Multi-channel speckle filtering

HH

Intensity 3 Looks images Hongze (Jiangsu)

Filtered images using multidate images

9 September 2010

Lecture D4L1

SAR image properties

Thuy Le Toan

49

Multi image filtering

80 60 40 20 0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

Histogram of a homogeneous area before and after filtering using 6 dates HH and VV (12 images) Initial: APP 3 Looks

80 60 40 20 0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

9 September 2010

Lecture D4L1

SAR image properties

Thuy Le Toan

50

Image preprocessing
1. 2. 3. Calibration, to convert the data to standard geophysical measurement units. Geocoding, to allow the data to be referenced to a map and to allow geolocation. Registration, to make sure multiple images can be matched point to point.

There are well-developed methods for each of these operations.

9 September 2010

Lecture D4L1

SAR image properties

Thuy Le Toan

51

date1

date 2

date M

Example of processing chain

Initial images

1
.

2 . M
Calibration Registration .

2 . M

Example of Chain developed using:

Multi image filtering

Calibrated coregistered

Gamma ASAR (Gamma RS) Multi-image filtering (Quegan et al., 2000) Temporal change (Le Toan et al., 1997)

2 . M

Filtered
Spatial filtering Geocoding

2 . M

Analysis , Retrieval Classification

Filtered geocoded

9 September 2010

Lecture D4L1

SAR image properties

Thuy Le Toan

52

Summary
An introduction to SAR measurements and their physical information content , and their geometrical and statistical properties has been given Knowledge of the SAR image properties is essential before to use the SAR image as physical measurements in applications An important topic is the effect of speckle noise which need to be reduced Preprocessing steps are important for quantitative use of SAR images
9 September 2010 Lecture D4L1 SAR image properties Thuy Le Toan 53

9 September 2010

Lecture D4L1

SAR image properties

Thuy Le Toan

54