RADAR AND SONAR IMAGING - A TUTORIAL INTRODUCTION

H.D. Griffiths

1. Introduction Imaging techniques now form an important part of both radar and sonar signal processing. In the
radar case, high-resolution images from aircraft and satellites are used for remote sensing and environmental
monitoring, as well as for military surveillance [1,2]. Sonar images may be used in surveying and inspection of oil
wellheads, and pipelines and shipwrecks, as well as in mine-hunting [3-51.

Aperture synthesis is one of the most important techniques in radar and sonar imaging. Applied to radar and sonar,
the techniques of aperture synthesis have interesting similarities and differences. Many of the differences are
associated with the much lower velocity of propagation in the sonar case, but also, variations in temperature and
salinity of seawater cause reflection and bending of the ray paths, so sonar propagation has perhaps similarities
with the propagation of HF radio waves in the ionosphere. The purpose of this paper, then, is to provide a tutorial
review of synthetic aperture imaging, attempting to highlight the similarities and differences between the radar and
sonar cases.

g- The principle of aperture synthesis is to store successive echoes obtained from a

where ro is the range at closest approach, which occurs when x = 0. This equation can be expanded:

+ x2 - fl +...)112 . . . (2)
r = ro 1
( 24 8rt

and if the aperture length is restricted such that only the first two terms of the expansion are significant (which will
often, but not always be the case), then:

. . . (3)
The phase history of the sequence of echoes (ignoring the motion of the radar or sonar between transmission and
reception of a pulse) is:

. . . (4)

which is a parabolic function of x (Figure 1b).

It can also be useful to think of the process in terms of Doppler history of the sequence of echoes, which is
obtained by differentiation of the phase history with respect to time:

Professor Griffiths is with the Department of Electronic and Electrical Engineering, University
College London, Torrington Place, London WClE 7E, UK.

Authorized licensed use limited to: The George Washington University. Downloaded on April 20,2010 at 20:40:15 UTC from IEEE Xplore. Restrictions apply.
 target

Figure 1. (a) Synthetic aperture geometry; (b) phase history of the sequence of echoes;
(c) Doppler history of the sequence of echoes.

This Doppler history (Figure IC) is a linear function of x, rather like the linear-FM 'chirp' often used in pulse
compression (for this reason the matched-filtering processing to form the synthetic aperture is often known as
'azimuth compression'). Figure IC also shows the limits kvld of the Doppler history which correspond to the
limits of the footprint of the 'real aperture', of length d, carried by the radar or sonar. Thus the process of aperture
synthesis can be thought of either as focusing of the phase history of the echoes (Figure Ib), or of matched-
filtering of the Doppler history (Figure IC)- and both approaches are useful.

The azimuth (across-track) resolution can be derived in several ways. Perhaps the simplest is to consider the
Doppler bandwidth Afof the echoes from a target of azimuth extent Ax, which from Figure ICis:

Af = 2vdx i.e. Ax = -.r01 Af . (6)

The Doppler resolution of the processing is the reciprocal of the time Ttaken to synthesise the aperture, which is:

The maximum azimuth resolution is thus the value of Ax which corresponds to a Doppler bandwidth of UT:

4 12

Authorized licensed use limited to: The George Washington University. Downloaded on April 20,2010 at 20:40:15 UTC from IEEE Xplore. Restrictions apply.
 i.e. just half the length of the radar antenna or sonar transducer, independent of frequency, range and sensor
velocity. At first sight this may seem curious, since it implies that to achieve high resolution it is necessary to use a
small antenna or transducer, but this is explained when it is realised that this gives a wider beamwidth, and hence
allows a longer synthetic aperture.

3. SamDling In any synthetic aperture system the pulse repetition frequency (PRF) must be sufficiently high to
give adequate sampling of the Doppler at the extremes of the footprint, in other words to avoid grating lobe
responses within the main lobe of the real aperture beam. Thus:

PRF 2 2vld . . . (9)

For a given platform velocity anccross-range resolution, there is a defined minimum PRF.Associated with this is
a maximum unambiguous range R , such that:

R ^ = c . . . (10)
2PRF

or, for a satellite S A R , a maximum unambiguous swath width 9.such that:

W = C
2 PRF sin cp
. . . (11)

where cp is the angle of incidence at the inner edge of the swath.

In a sonar system this presents a problem; for a typical tow speed of 2 m/s (=4 kts) and a desired cross-range
resolution of 0.1 m, the maximum unambiguous range is only 18.75 m. In a satellite-bome SAR system the same
constraint limits the achievable swath width to approximately 100 km. In an aircraft-bome system it is not a
problem.

There are several techniques that might be used to address this limitation:

(i) Slower speed. This is not practicable for a satellite. For a sonar it may be possible, but the slower speed
exacerbates the problem of motion errors discussed in Section 4.

(ii) Multiple along-track elements or beams. Instead of a single antenna or transducer, a number may be used to 'fill
in' the spatial sampling of the synthetic aperture. This may be regarded as an intermediate stage between a totally
synthetic array and a conventional array. Equally, multiple elevation beams can be used to extend the swath
coverage. In both cases the price paid is the additional complication of multiple antenna elements and receiver
channels [6].

(iii) Wideband techniques. Chatillon et al. [7] have shown how a wide-bandwidth transmission can allow
undersampling in synthetic aperture sonar, relying on the fact that the angular separation of the ambiguous
responses is itself a function of frequency. They discuss and compare three practical implementations in terms of
resolution and computational complexity. However, whilst such a scheme can redistribute the ambiguous energy, it
cannot completely suppress it, so its benefit remains questionable.

(iv) Multiple orthogonal transmissions. If a set of modulation codes can be found such that they individually
possess favourable autocorrelation properties, but whose crosscorrelationproperties are such that they are as nearly
as possible orthogonal, then in principle it is possible to separate which echo belongs to which transmitted pulse.
The simplest such scheme uses two chirps of opposite slope [SI.This gives isolation values of between 20 and 26
dB depending on the time-bandwidth product of the chirps, but is, of course, restricted to two waveforms. Some
further work in this area [9] has identified a promising set of four maximal length biphase codes, and hence a factor
4 increase in maximum unambiguous range, but the performance with distributed targets and over the full range of
Doppler may limit the usefulness of the technique.

1 I3

Authorized licensed use limited to: The George Washington University. Downloaded on April 20,2010 at 20:40:15 UTC from IEEE Xplore. Restrictions apply.
 On balance, the use of multiple along-track elements (or beams) appears to be the best way of suppressing
ambiguous energy, though limited additional use of some of the other techniques may reduce the number of
additional elements and receiver channels required.

4. Autofocus The effect of sensor motion errors is to cause distortion of the image. Motion errors are often
significant with aircraft-borne SAR, particularly at high resolution. With synthetic aperture sonar there are likely to
be problems both with motion error and propagation irregularities [IO]. In principle the motion errors can be
estimated and corrected by inertial sensors, but these are most sensitive to high-frequency errors. There is therefore
considerable interest in estimation and correction of motion errors using the image data themselves - a process
known as autofocus.

Several autofocus algorithms have been proposed and evaluated 111-13]. If an isolated point scatterer can be
identified, this can be tracked through the sequence of echoes and the actual path of the sensor identified (hot spot
tracking). Othenvise, a series of parabolic sensor paths can be tried (cf equation (4)),adopting the one which
gives highest contrast as the best estimate of the true path, and the image focused accordingly. This is known as
contrast optimisation. Another approach [mmulti-lookregisfration)spIits the synthetic aperture into separate Iower-
resolution looks, and co-registers these to estimate the path of the sensor. Both of these work best on an image
with high information content (high spatial frequencies),

A fourth technique (phase gradient autofocus [14,15]) is chimed to work even with images of low contrast. This
works by identifying strong targets in the distorted image as a function of range, employing a circular-shifting
process to centre the targets in azimuth, windowing, and then integrating to estimate the actual parabolic sensor-
target path, relying on the facts that the speckle will average to zero and the actual path is only a weak function of
range.

Further work remains to be done to compare these techniques - and perhaps to devise new ones for motion errors ~

alone (aircraft-borne S A R ) and both motion errors and propagation irregularities (synthetic aperture sonar).

5. Interferometry By utilising either parallel passes of a single sensor (dual-pass) or two receive antennas mounted
on a single platform (single-pass), as shown in Figure 2%it is possible to estimate target height from the measured
phase differences in each pixel of the image, thereby obtaining a three-dimensional image of the target scene, at the
same spatial resolution as the original SAR images [16]. This is now a well-established technique with satellite-
borne SAR, for applications such as topographic mapping, but there is also considerable interest in using the
technique with aircraft-borne SAR, for surveillance applications [ 171, and with synthetic aperture sonar, for
minehunting [IS].

radar I sonar

target
Y

Figure 2. Interferometry: (a) two images are formed, either from two parallel passes of the same sensor (dual-
pass), or from two sensors carried by the same platform (dual-antenna); (b) interferometer geometry.

1/4

Authorized licensed use limited to: The George Washington University. Downloaded on April 20,2010 at 20:40:15 UTC from IEEE Xplore. Restrictions apply.
 From Figure 2b it is evident that the interferometric phase difference 4 between the two signals corresponding to a
given pixel will be a function of the interferometric baseline B and its orientation a, the wavelength 2 and the
target height h:

These equations are for the the single-pass case; for the dual-pass case the path lengths and hence the phase
differences 4 are doubled.

If the quantities B, a, rl and 1 are known (or can be estimated), then the target height h can be obtained from
the measurement of the interferometric phase difference 4. The phase differences are obtained over the whole
image constitute a fringe pattem or interferogram. The reconstruction process is complicated because $ is only
known modulo-2x, which necessitates a process known as phase unwrapping [191.

For satellite-borne SAR the dual-pass technique has been used exclusively, since it is not feasible to mount two
antennas on a single spacecraft. With aircraft-bome S A R and synthetic aperture sonar the dual-antenna technique is
used. The principal outstanding problem is that the reconstructed target height is very sensitive to motion errors -
particularly roll angle.

6. Where next ? It is interesting to try to predict where the next advances in the subject are likely to be made, and
where cross-fertilization of ideas may occur between the radar and sonar communities. The following areas may
provide some opportunities:

0 The sampling problem discussed in section 3 is more critical in sonar applications. There is greater scope here
for multiple along-track transducers (which are less likely to be feasible with aircraft- or satellite-bome S A R
systems), but multiple beam or coding techniques would be applicable to both.

0 The estimation and compensation of motion errors, both in the context of conventional aperture synthesis and
(more particularly) in interferometry is a key problem common to both radar and sonar domains. Because sonar
and aircraft-borne radar interferometry depend so critically on correction of these errors, it is likely that much
effort will be invested in the search for a solution. In the sonar case the technique may also need to correct for
irregularities in propagation.

0 In interferometry, as well as the motion error problems already mentioned, there are advances to be made in
phase unwrapping algorithms to reconstruct the 3-D target scene.

0 There may be a temptation to expect radar and sonar images to look like optical images. There is no reason why
-
they should the scattering processes are rather different. There is therefore some interesting work to be done
in the context of high-resolution imaging to establish the signatures of particular target types. This is equally
true for interferometry.

0 Nothing has been said so far about ISAR imaging, in which the sensor is stationary and the target motion is
used to obtain the image. The distinction can be made between cooperative ISAR, in which the target motion is
controlled and known, and non-cooperative ISAR, where the target motion parameters have to be estimated.
The technique does not seem so far to have been applied to sonar imaging.

7 . s The work described’ in this paper is the result of several studies. I am grateful to the
European Space Agency, DRA Malvern, DRA Bincleaves, and the Engineering and Physical Sciences Research
Council (grant no. Gm47548 and GRM47753) for their support. I am indebted to Prof. Colin Cowan, Dr
Tahseen Rafik, David Anthony and the late Prof. Roy Griffiths, of Loughborough University; to Dr Chris Baker
and Dr Tony Currie of DRA Malvern; and to Prof. Ralph Benjamin, Prof. Roger Voles, Richard Bullock and
Andrew Wilkinson of University College London, for invaluable advice and discussion.

Authorized licensed use limited to: The George Washington University. Downloaded on April 20,2010 at 20:40:15 UTC from IEEE Xplore. Restrictions apply.
 8. References

1. Ausherman, D.A., Kozma, A., Walker, J.L., Jones, H.M. and Poggio, E.C., 'Developments in radar
imaging', IEEE Trans. Aerospace and Electronics Systems, Vol. AES-20, pp363-400, 1984.

2 . Currie, A.: 'Synthetic Aperture Radar', Electronics and Communication Engineering Journal, Vo1.3, No.4,
~ ~ 1 5 9 - 1 7August
0, 1991.

3. Williams, R.E.: 'Creating an acoustic synthetic aperture in the ocean', J. Acoust. Soc. Am., V01.60, No.1,
July 1976, ~ ~ 6 0 - 7 3 .

4. Lee, H.E.: 'Extension of synthetic aperture radar (SAR) techniques to undersea applications', IEEE J. Oceanic
Engineering, Vol.QE-4, No.2, April 1979, pp60-65.

5 Cutrona, L.J.: 'Comparison of sonar system performance achievable using synthetic aperture techniques with
the performance achievable by more conventional means', J. Acoust. Soc. Am., Vo1.58, No.2, August 1975,
~~336-348.

6. Currie, A. and Brown, M.A., 'Wide-swath SAR', IEE Proc. Pt.F., Vol.139, No.2, April 1992, pp122-135.

7. Chatillon, J., Bouhier, M.E. and Zakharia, M.E.: 'Synthetic aperture sonar for seabed imaging: relative merits
of narrow-band and wideband approaches', lEEE J. Oceanic Engineering, Vol.QE-17, No. 1, January 1992,
~~95-105,

8. Giuli, D., Fossi, M. and Facheris, L.: 'Radar target scattering matrix measurement through orthogonal
signals', IEE Proc. Pt.F, Vol.140, No.4, August 1993, pp233-242.

9. Griffiths, H.D., Normant, E. and Wilkinson, A., 'Adaptive SAR beamforming network'; ESA Contract
Report ESTEC 6553/89mrUIW, l a p p . , ESA Scientific and Technical Publications Branch, ESTEC, Postbus
299, Noordwijk, The Netherlands, April 1992.

10. Checketts, D.G. and Smith, B.: 'Analysis of the effects of platform motion errors upon synthetic aperture
sonar', Proc. Inst. Acousfics, voI.18, Pt.3, ppf37-145, 1986.

1 1 Finley, I.P. and Wood, J.W.: 'An investigation of synthetic aperture radar autofocus', RSRE memorandum
~

No. 3790. April 1895. Ministry of Defence, RSRE Malvern, Worcs, UK.

-
12. Blake, A.P.: 'Autofocus techniques: multilook registration and contrast optimisation a comparison', RSRE
memorandum No. 4626. Sept. 1992. Ministry of Defence, RSRE Malvern, Worcs, UK.

13. Blacknell, D. and Quegan, S.: 'Motion compensation of airbome synthetic aperture radars using autofocus',
GEC Journal of Research, Vo1.7, No.3, pp168-182, 1990.

14. Eichel, P., Ghiglia, D. and Jakowatz, C., 'Speckle processing method for synthetic-aperture radar phase
correction', Oprics Letters, VoI.14, No.20, ppl-3, January I989

15. Wahl, A., Eichel, P.H., Ghiglia, D.C. and Jakowatz, J.R., 'Phase gradient autofocus - a robust tool for high
resolution SAR phase correction', IEEE Trans. Aerospace and Electronic Systems, Voi.30, No.3, pp827-
834, July 1994.

16. Goldstein, R.M., Zebker, H.A. and Wemer, C.L., 'Satellite radar interferometry: two-dimensional phase
unwrapping'; Radio Science, Vo1.23, No.4, pp713-720, July-August 1988.

17. Cume, A., Baker, C.J., Bullock, R., Griffiths, H.D., Edwards, R. and Griffiths, H.D., 'High resolution 3-D
SAR imaging'; Proc. RADARP.5 Conference, Washington; IEEE Publ. No.95CH357 1-0, pp468-472, 8-1 1
May 1995.

Authorized licensed use limited to: The George Washington University. Downloaded on April 20,2010 at 20:40:15 UTC from IEEE Xplore. Restrictions apply.
 18. Griffiths, H.D., Shafeeu, H., Griffiths, J.W.R., Rafik, T.A., Meng, Z. and Cowan, C.F.N.: 'Mine detection
using interferometric synthetic aperture sonar'; Proc. Undersea Defence Technology '95,Cannes, pp229-234;
Nexus Information Technology, 4-6 July 1995.

19. Griffiths,H.D. and Wilkinson, A.J.: 'Improvements in phase unwrapping algorithms for interferometric
SAR'; L'Onde Electrique, Vo1.74, No.3, pp46-52, Mai-Juin 1994; also published in proceedings of
Colloque International sur le Radar, Paris, pp342-349; Sociktk des Electriciens et des Electroniciens, Paris,
3-6 May 1994.

1 /I

Authorized licensed use limited to: The George Washington University. Downloaded on April 20,2010 at 20:40:15 UTC from IEEE Xplore. Restrictions apply.