Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/224794714

Principles Of Synthetic Aperture Radar

Article  in  Surveys in Geophysics · January 2000


DOI: 10.1023/A:1006790026612 · Source: DLR

CITATIONS READS

76 4,724

1 author:

Richard Bamler
German Aerospace Center (DLR)
405 PUBLICATIONS   13,198 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

So2Sat: Big Data for 4D Global Urban Mapping - 10^16 Bytes from Social Media to Earth Observation Satellites View project

4D City View project

All content following this page was uploaded by Richard Bamler on 18 September 2015.

The user has requested enhancement of the downloaded file.


PRINCIPLES OF SYNTHETIC APERTURE RADAR

R. BAMLER
German Aerospace Research Center (DLR), Remote Sensing Technology Institute (IMF),
Wessling, Germany E-mail: Richard.Bamler@dlr.de

Abstract. Synthetic Aperture Radar (SAR) is an active microwave imaging method. It operates
independently of Sun illumination and cloud coverage. Current spaceborne systems use wavelengths
of 3 to 25 cm and achieve resolutions of 10 to 50 m. The paper attempts to explain the basic SAR
imaging principles using a minimum of mathematics. Emphasis is put on the particular properties of
SAR images that should be understood before interpreting these data.

1. Introduction

A Synthetic Aperture Radar (SAR) image is a 2-D map of the microwave reflectiv-
ity of the imaged scene at resolutions in the order of 0.2 to 100 m. SAR has some
interesting properties not known from optical sensors:
• SAR is an active imaging method, i.e., it provides its own illumination, and,
hence, is independent of sun illumination.
• SAR uses microwaves that penetrate clouds and (partially) canopy, soil and
snow. Typical wavelengths are 3 cm (X-band), 6 cm (C-band), and 24 cm (L-
band).
• SAR uses polarized radiation and, therefore, can exploit polarization signa-
tures of the imaged scatterers for obtaining more information about the scat-
terer’s structure.
• SAR is a coherent imaging method. This is the prerequisite for SAR interfer-
ometry. On the other hand, this makes the radiometric interpretation of SAR
images difficult due to the appearance of speckle.
This paper attempts to explain the principle of SAR image acquisition and
interpretation using a minimum of mathematical equations. Numerical examples
are given for the SAR on the ERS-1/2 satellites whose data are widely distributed
and generally available.

Surveys in Geophysics 21: 147–157, 2000.


© 2001 Kluwer Academic Publishers. Printed in the Netherlands.
148 R. BAMLER

2. Side Looking Radar Imaging Configuration

2.1. T HE RADAR PRINCIPLE

A radar is a microwave or radio frequency distance measuring (= ranging) device.


In the monostatic case, which is of interest here, a transmitter (TX) radiates a short
wave packet of frequency to and duration τp :

g(t) cos(2πf0 t),

where g(t) is the envelope of the pulse and cos(2πf0 t) is the microwave carrier.
After the wave hits an object and part of its energy gets scattered back, the radar
instrument will eventually switch into receive (RX) mode to record the echo (Fig-
ure 1). For a single point-like scatterer a distance (range) R away from the radar
the echo is an attenuated replica of the transmitted wave, delayed by the round-trip
travel time 2R/c:

g(t − 2R/c) cos(2πf0 (t − 2R/c)).

The delay shows up both in the position of the pulse envelope and the phase
of the carrier sinusoid. For ERS the carrier frequency is 5.3 GHz corresponding
to a wavelength of 5.7 cm (C-band). If several objects are present, their echoes
superpose in the received signal. Two objects (e.g., objects #1 and #2 in Figure 1)
separated in range by δR can be distinguished in the radar signal, if

δR ≥ ρR = τp c/2.

Hence, we may call ρR = τp c/2 the range resolution of the radar. Many
radars (including SARs) use high bandwidth phase coded waveforms (e.g., so-
called chirps) instead of short pulses in order to avoid excessively high transmit
peak power. After reception the signals are correlated with a replica of the trans-
mitted waveform, a procedure called range compression. This results in the same
signal as if a short pulse with the bandwidth of the transmitted wave had been used
(if the scatterers did not move). Then the range resolution is ρR = c/(2W ), where
W is the bandwidth.
In natural scenes ensembles of many scatterers (e.g., forests, rough surfaces)
are found more frequently than the discrete point-like objects discussed so far. In
such a case the radar signal is the convolution of the transmitted pulse with the
range profile of the scene’s microwave reflectivity. It is important to understand
how the contributions of all the scatterers superpose; it is the electromagnetic waves
that sum up according to their amplitudes and their relative phases rather than the
envelopes of the pulses. This is true, no matter whether the radar finally detects only
amplitude (incoherent radar) or employs quadrature detection (coherent radar) to
record amplitude and phase (or real and imaginary parts).
PRINCIPLES OF SYNTHETIC APERTURE RADAR 149

Figure 1. Range measurement by a radar.

2.2. S IDE - LOOKING / REAL - APERTURE RADAR

A radar as described above provides only a 1-D projection of a scatterer arrange-


ment. In order to get a 2-D map of the microwave reflectivity of the Earth’s surface
a side-looking radar can be used to scan the scene along a flight path or along
an orbit. The basic principle is illustrated in Figure 2. Obviously, two scanning
mechanisms are employed: in range ‘R’ the radar principle is used, i.e., scanning
by the speed of light; in flight direction (or: azimuth ‘x’) the scene is scanned by
the continuously moving antenna beam. Such an imaging configuration requires
the radar to be operated in a repeatedly pulsed fashion. The rate at which pulses
are transmitted and echoes are received is known as the pulse repetition frequency
(PRF). For ERS the PRF is about 1680 Hz and the beam velocity on ground is in
the order of 6700 m/sec, i.e., pulses are transmitted every 4 m.

2.3. A ZIMUTH AND RANGE RESOLUTION

To form an image the received echoes are assembled line-by-line in a raw data
matrix (Figure 3). A single point scatterer (on a ‘black’ background) will contribute
echoes to the raw data matrix as long as it is visible by the radar, i.e., whenever
it is within the antenna beam footprint. Hence, the azimuth resolution of a side-
looking/real-aperture radar (RAR) of azimuth beamwidth β is

ρx,RAR = Rβ.

Considering that an antenna of length L (e.g., 10 m) operated at a wavelength λ


(e.g., 0.057 m) has a far-field beamwidth of β = λ/L it becomes obvious that such
a system cannot give satisfactory azimuth resolution if operated from space. For a
typical value of R = 850 km we find ρx,RAR ≈ 4.8 km.
The range resolution depends on the frequency bandwidth of the transmitted
pulse as discussed before and is usually in the region of 1 to 50 m.
150 R. BAMLER

Figure 2. Side-looking radar imaging geometry.

Figure 3. Formation of an echo raw data signal matrix (top view of Figure 2).
PRINCIPLES OF SYNTHETIC APERTURE RADAR 151

3. Formation of a Synthetic Aperture – The SAR Principle

For an acceptable azimuth resolution of the order of a few meters, a real aperture
system would require unrealistically large antennas of several kilometers. A SAR is
a side-looking radar that records the received echoes coherently, i.e., it detects the
amplitude and phase of the echo signal by quadrature demodulation. This results in
a complex-valued raw data matrix where every sample consists of a real (in-phase)
and imaginary (quadrature) part. In a subsequent compute-intensive digital signal
processing step echo lines of the raw data matrix are phase corrected and summed
up in such a way as to simulate a long antenna. Software or hardware systems to
perform this focusing operation are known as SAR processors.
We have learned that a single scatterer contributes to the raw data matrix over an
azimuth extent of LA = Rβ = Rλ/L which is proportional to range. Clearly, this is
the maximum size of the synthetic aperture that can be formed by the processor. An
antenna of this size, in turn, would have a beamwidth of λ/LA (inversely propor-
tional to range). Considering further that all the phase terms and corrections apply
for the two-way case and, hence, are doubled compared to conventional antenna
theory, we finally obtain the azimuth resolution of a SAR as

ρx = L/2.

Hence, the maximum resolution of a strip-mode SAR is always half of the


physical antenna length and is independent of range and wavelength.
A SAR is equivalent to a real-aperture radar with an antenna of length 2LA as
long as the scatterers in the scene were stationary during data acquisition. This is
true for most land remote sensing applications. If a scatterer is moving while it
is imaged by a SAR, the phase history of this scatterer will change and will not
be correctly compensated in the processor. As a consequence the scatterer will
be defocused and/or displaced in the final image. Prominent examples are the so-
called train-off-the-track effect or the blurring of ocean waves (azimuthal cut-off).

4. Properties of SAR Images

In order to interpret remote sensing SAR images it is necessary to understand their


particular properties in more detail. Although the term ‘image’ is often associated
with matrices of non-negative real-valued pixels, we will readily extend its ap-
plicability to complex matrices. SAR is a coherent imaging method and the type
of image that contains the most information is complex-valued (in ERS-speak:
SLC for single look complex). Often, we are not interested in the phase of the
image pixels, i.e., we use detected (amplitude or intensity) images. Let the com-
plex image matrix be u[i, k]. Then its modulus |u[i, k]| is referred to as amplitude
and I [i, k] = |u[i, k]|2 as its intensity. Intensity has a straightforward physical
152 R. BAMLER

Figure 4. ERS SAR image (intensity) of a mountain range in the Mojave Desert, CA, USA. Size ∼
=
40 × 40 km. Coordinates: range ↓, azimuth →. Data © ESA.

interpretation in terms of backscattered energy. A typical SAR image is shown in


Figure 4.

4.1. P OINT SCATTERER RESPONSE AND RESOLUTION

Any imaging system has a finite spatial resolution, i.e., all structures in the object,
however small they are, will occur in the final image blurred by a certain amount.
This blurring operation is usually described by a convolution of the (non-blurred)
object with the point response of the system. The point scatterer response of a SAR
system is the image of a very small point-like scatterer. It is similar to a 2-D sinc-
function or some other bell-shaped distribution. Its peak is found at the location of
the point scatterer; its width represents the amount of blurring. It sets the minimum
distance of two objects that can be easily separated in the image. Hence, the width
of the point scatterer response is denoted as the resolution of the SAR image.
PRINCIPLES OF SYNTHETIC APERTURE RADAR 153

Obviously, the point scatterer response is an important function for the sys-
tem description and needs to be measured for quality control purposes. An object
can be considered a point scatterer if the wave scattered back to the SAR is suf-
ficiently isotropic, i.e., it can be approximated by a spherical wave within the
azimuth beamwidth. This requirement is usually met by scatterers much smaller
than the wavelength (like a tiny metallic sphere). However, such a small object
cannot reflect enough energy to be visible in the final image. Therefore so-called
corner reflectors or active transponders are frequently used for the assessment of
point scatterer responses. They behave like point scatterers but reflect several or-
ders of magnitude more energy. The ability of an isolated object to scatter back
microwaves is given by its radar cross section σ and is measured in m2 .
Note that the terms ‘resolution’ and ‘pixel size’ are often confused. Resolution
is a quality measure of an imaging system. Pixel size is the sampling interval in
the final image and can be chosen arbitrarily. In order not to violate Shannon’s
sampling theorem, the pixel size should be smaller than the resolution for complex
images and smaller than half of the resolution for detected images. A slight under-
sampling is often acceptable in detected images. For example ERS images of the
PRI type are sampled at 12.5 m in ground range and azimuth; their resolution is
about 20 to 25 m.

4.2. D ISTRIBUTED ( GAUSSIAN ) SCATTERERS AND BACKSCATTER


COEFFICIENT

The point scatterer response has also another interpretation: if projected on the
ground it is the integration kernel for the imaging process, i.e., it indicates what
a single pixel ‘sees’ in the scene. Considering that the size of such a resolution
element is of the order of e.g., 20 m × 20 m it becomes obvious that in natural
scenes ‘many’ scatterers will contribute to a single pixel value. This is true for
rough surfaces, where every facet of about wavelength size must be considered an
individual scatterer. It is likewise true for vegetation (leaves, branches, etc.). All
the contributions of these scatterers will sum up coherently to give the final pixel
value and the relative phases of the scatterers are essential for the summation result.
These phases in turn depend on the scatterers’ sub-wavelength locations. From that
it should become obvious that natural distributed scatterers can not be described by
their coherent reflectivity. Instead the (dimensionless) normalized radar cross sec-
tion or backscatter coefficient σ 0 is used to describe their mean power reflectivity.
It is the sum of the radar cross sections σn of all scatterers in an area A normalized
by the area:

P
σn
σ =
0 n
.
A
154 R. BAMLER

4.3. S PECKLE

The backscatter coefficient σ 0 of a distributed object is a measure for the expecta-


tion value of backscattered power. Arbitrarily many scatterer arrangements within a
resolution cell may lead to the same value of σ 0 . A SAR, on the other hand, images
the coherent superposition of individual scatterer responses. As an example let us
consider that we want to image a pebble beach of a certain σ 0 by a SAR. If we re-
arranged the pebbles, σ 0 would not change. The SAR pixel value, however, would
react sensitively to the new phase values of the individual scatterer contributions
and may take on an arbitrary value. Hence, SAR images of distributed objects,
exhibit a noise-like structure, known as speckle, also from other fields of coherent
imaging (e.g., lasers).
If there is a sufficiently high number of sub-scatterers in every resolution ele-
ment and none of theses scatterers dominates the others, the complex pixel values
u[i, k] can be described as a circular Gaussian process, whose real and imaginary
parts are zero-mean, mutually uncorrelated Gaussian random variables. The ex-
pectation value of the pixel intensity is proportional to the backscatter coefficient:
E{I [I, k]} = E{u[I, k]2 } ∝ σ 0 .

The speckle probability distributions of these Gaussian scatterers are well known.
An interesting property of speckle fields is that their standard deviation is equal
to their mean (expectation) value, i.e., if we blame speckle to be noise, then a
(single look) SAR image has a signal-to-noise ratio of 1 (or: 0 dB)! Often speckle
is reduced at the expense of resolution by averaging over L independent pixels,
where L is historically referred to √
as the ‘number of looks’. The signal-to-(speckle)
noise ratio of an L-look image is L.

4.4. G EOMETRY

The mapping of the 3-D real-world coordinates (x, y, z)T into a 2-D SAR image
with coordinates (x, R)T is different from the optical central perspective. Using
Figure 5 we find the relationship between range R and y, z:
p
R = (y0 + y)2 + (H − z)2 .
This type of geometry leads to distortions of terrain in the image not known
from optics. According to Figure 6 we can distinguish three effects:
• Foreshortening describes the fact that terrain tilted towards the sensor appears
squeezed in the SAR image, while slopes tilted away from the sensor get
stretched (see also Figure 4).
• If the terrain slope exceeds the incidence angle of the SAR the mapping from
ground range y to slant range R gets inverted and a lay-over of different areas
occurs.
• Radar shadow can be found behind steep mountains at shallow incidence angles.
PRINCIPLES OF SYNTHETIC APERTURE RADAR 155

Figure 5. Mapping of the real world into radar coordinates (flight path perpendicular into plane).

Figure 6. SAR image geometry effects caused by radar mapping (a simple conical mountain on
flatground is assumed).

4.5. R ADIOMETRY

We have defined the backscatter coefficient σ 0 as the sum of radar cross sections
per unit ground area. For real-world scatterer configurations σ 0 is in general a
function of the local incidence angle: σ 0 (θi ). The brightness of the scatterer in
the SAR image depends additionally on the size of the range resolution element
projected on ground; the larger the resolution cell, the more scatterers contribute
to a pixel value. Hence, the radiometric quantity seen in a SAR image (save from
speckle) is the radar brightness, defined as
β 0 (θi ) = σ 0 (θi )/ sin θi .
In order to recover σ 0 from a SAR image the influence of the local incidence
angle (i.e., SAR look angle + local terrain slope) must be eliminated. This requires
156 R. BAMLER

a digital elevation model of the imaged scene. In images of rugged terrain (as the
one in Figure 4) local slope variations, i.e., topography, contribute more to the
visual appearance of the image than spatial variations of the backscatter coefficient.
A third radiometric quantity, γ 0 , is often used. It is the backscatter coefficient
as measured by scatterometers and is related to σ 0 via

γ 0 (θi ) = σ 0 (θi )/ cos θi .

Note that for Lambertian scatterers (e.g., rain forest at C- or X-band) whose
σ 0 (θi ) well follows a cosine law, γ 0 is almost independent of the incidence angle.

4.6. PARAMETERS INFLUENCING RADAR BRIGHTNESS

For radiometric interpretation of SAR images it is important to know how differ-


ent parameters influence the radar brightness observed in the image. Besides the
properties of the scatterers in the scene also sensor parameters like wavelength,
polarization, look angle, and resolution determine the pixel brightness, e.g., longer
wavelengths penetrate canopy better than shorter ones. Most natural surfaces ap-
pear rougher in X-band than in L-band. Cross-polarization is sensitive to volume
scattering, while a 180 degrees phase shift between horizontal and vertical polariz-
ation is an indicator for double-bounce scattering (e.g., soil-trunk). High resolution
SAR images allow for textural structure analysis in forested areas while a low
resolution SAR may blur texture information.
The dominant scene parameters that influence radar brightness are:
• Surface roughness
• Local incidence angle (local terrain slopes, geomorphology)
• Scatterer density (e.g., bio-mass, leaf density)
• 3-D distribution of scatterers (canopy thickness, trunks, buildings)
• Dielectric constant (scattering material, soil moisture, vegetation status)

References

Allan, T.D.: 1983, ‘Satellite microwave remote sensing’, in T.D. Allan (ed.), Ellis Horwood Series in
Marine Science, Ellis Horwood Limited, Chichester, England.
Bamler, R.: 1992, ‘A comparison of range-Doppler and wavenumber domain SAR focusing
algorithms’, IEEE Transactions on Geoscience and Remote Sensing 30(4), 706–713.
Bamler, R. and Schättler, B.: 1993, ‘SAR data acquisition and image formation’, in G. Schreier (ed.),
SAR Geocoding: Data and Systems, Wichmann, Karlsruhe, pp. 53–102.
Barber, B.C.: 1985, ‘Theory of digital imaging from orbital synthetic-aperture radar’, International
Journal of Remote Sensing 6(7), 1009–1057.
Brown, W.M.: 1967, ‘Synthetic aperture radar’, IEEE Transactions on Aerospace and Electronic
Systems, AES-3(2), 217–229.
Brown, W.M. and Porcello, L.J.: 1969, ‘An introduction to synthetic aperture radar’, IEEE Spectrum,
September, 52–62.
PRINCIPLES OF SYNTHETIC APERTURE RADAR 157

Cafforio, C., Prati, C. and Rocca, F.: 1991, ‘Full resolution focusing of Seasat SAR images in the
frequency-wavenumber domain’, International Journal of Remote Sensing 12(3), 491–510.
Curlander, J.: 1982, ‘Location of spaceborne SAR imagery’, IEEE Transactions on Geoscience and
Remote Sensing GE-20(3), 359–364.
Curlander, J.C. and McDonough, R.N.: 1991, Synthetic Aperture Radar: Systems and Signal
Processing, John Wiley & Sons, New York.
Elachi, C.: 1988, Spaceborne Radar Remote Sensing: Applications and Techniques, IEEE Press, New
York, NY.
Elachi, C.: 1991, ‘Spaceborne imaging radars’, International Journal of Imaging Systems and
Technology 3, 167–185.
Gough, P.T. and Hawkins, D.W.: 1997, ‘Unified framework for modern synthetic aperture imaging
algorithms’, International Journal of Imaging Systems and Technology 8, 343–358.
Harger, R.O.: 1970, Synthetic Aperture Radar Systems: Theory and Design, Academic Press, New
York.
Lee, J.-S., Hoppel, K.W., Mango, S.A. and Miller, A.R.: 1994, ‘Intensity and phase statistics of
multilook polarimetric and interferometric SAR imagery’, IEEE Transactions on Geoscience
and Remote Sensing 32(5), 1017–1028.
Li, F.K. and Johnson, W.T.K.: 1983, ‘Ambiguities in spaceborne synthetic aperture radar systems’,
IEEE Transactions on Aerospace and Electronic Systems AES-19(3), 389–397.
McDonough, R.N., Raff, B.E. and Kerr, J.L.: 1985, ‘Image formation from spaceborne synthetic
aperture radar signals’, Johns Hopkins APL Technical Digest 6(4), 300–312.
Raney, R.K.: 1982a, ‘Processing synthetic aperture radar data’, International Journal of Remote
Sensing 3(3), 243–257.
Schmullius, C.C. and Evans, D.L.: 1997, ‘Synthetic aperture radar (SAR) frequency and polarization
requirements for applications in ecology, geology, hydrology, and oceanography: a tabular status
quo after SIR-C/X-SAR’, International Journal of Remote Sensing 18(13), 2713–2722.
Tomiyasu, K.: 1978, ‘Tutorial review of synthetic-aperture radar (SAR) with applications to imaging
of the ocean surface’, Proceedings of the IEEE 66(5), 563–583.
Wu, C., Liu, K.Y. and Jin, M.: 1982, ‘Modelling and a correlation algorithm for spaceborne SAR
signals’, IEEE Transactions on Aerospace and Electronic Systems AES-18(5), 563–574.
View publication stats

You might also like