Download as pdf or txt
Download as pdf or txt
You are on page 1of 4

TIMIE-FREQUENCY TRANSFORM VS.

FOURIER TFUNSIFORM FOR RADAR IMAGING

Victor C. Chen
Naval Research Laboratory, Washington D .C. 203 75
Shie Qian
DSP Group, National Instruments, Austin, TX 78730

Abstract can be retrieved with a method other than Fourier


transform, such as a time-frequency transform, which
Conventional radar imaging systems use the do not require stationary Doppler spectrum.
Fourier transform for reconstruction of radar images, Due to target motion, the Doppler frequency
To use the Fourier transform adequafely, some shift of a scatterer located at (x,,y,)can be
restrictions must be applied. Due to target's motion approximately expressed as [I]
and maneuvaring, the reconstructed iimage by using
Fourier transform becomes smeared. There:fore, some
sophisticated motion compensation procedures must
be applied to produce clear image. However, the where f is the carrier frequency of the trasmitted
restrictions of Fourier transform can be lifted if a signal, c is the velocity of the wave propagation,
high resolution time-frequency transform can be used v R is the radial velocity of the scatterer, and is
to retrieve the Doppler information. By relplacing the the rotation rate of the scatterer about its rotation
conventional Fourier transform with the ,joint time- center. Froni Eq.(l) we can see that the Doppler
frequency transform, a 2-D range-Dfoppler Fourier frequency shift is also related to the geometric
frame becomes a 3-D time-range-Doppler cube.
With sampling in time, a sequence of clear 2-D location of the scatterer ( x o , y o ) . Thus, another
range-Doppler images can be simply reconstructed scatterer at a drifferent geometric location in the target
without using sophisticated motion compensation. may have a different Doppler frequency shift. From
Eq.( 1) we also notice that the Doppler frequency shift
Introduction is time-varying, even if the rotation rate is a
constant.
Radar image presents a spatial diistribution Based on the returned signal from a single
of an object's reflectivity, and it is mapped onto a point scatterer, the returned signal fii-om the object
range and cross-range plane. Both the synthetic can be represeinted as the integration of the returned
aperture radar (SAR) and the inverse synthetic signals from all scatterers in the object:
aperture radar (ISAR) are imaging radar. SAR refers
to the situation where the radar is moving and the
object is stationary, while ISAR refers to a
geometrically inverse way where ithe object is where r ( x , y ) is the radar range of a scatterer at
moving and the radar is stationary. In BAR, the ( x , y ) . The objective of radar imaging system is to
cross-range resolution is obtained by using Doppler construct the reflectivity finction of the object
spectrum.
The conventional approach to retrieve p(x,y) from the frequency signature S(f). This
Doppler spectrum is the Fourier transform. To use can be done by simply taking inverse Fourier
the Fourier transform adequately, some restrictions transform of the phase compensated frequency
must be applied: the Doppler frequency contents of signature S ( f >exp Ij4r~fR / C} . The process of
the data should not change within the imaging time estimating the object range R, and removing the
duration. If the Doppler contents do ch,ange with
time, the Doppler spectrum by using the Fourier extraneous phase term exp{-j47cfRo / c } is called
transform becomes smeared and, thus, the cross- gross motion compensation. Then, the inverse
range resolution will be degraded. However, the Fourier transfcirm can be applied to construct the
restrictions can be lifted if the Doppler information reflective density function of the object.
':l"f".
390 TFTS'96

p(xt9F!
MWI NC OBJ KTT

e- j 4 x f t 9 C

X Y
CP(X,Y) e- j 2 n ( x $ -

fx =
vfs

2 ftsinQt/c
RPDAR IMAGE

a
Fourier Processing and
Motion Compensation

very
Motion compensation is
important step to achieve a
7y Rawme
fv = 2 ftCQ e d c cutplt
clear radar image. In contrast
with SAR, ISAR utilizes object's

kt
pitch, roll and yaw motions and
measures the Doppler spectrum to
generate images of the object.
\/ Doppler processing using Fourier
M pllSeS
transforms is adequate only in the
TramttdStepW
Fr" Siwl case that the scatterers remain in
range cells and that their Doppler
frequency shifts are constant
N N during the entire observation time.
'VN
Frwm sicnature
If the scatterers drift out of range
cells or their Doppler frequency
shifts are time-varying, then the
Fig. 1 Illustration of a stepped-frequency
inverse synthetic radar imaging of a moving object. constructed image becomes
blurred. Motion compensation
Fig.1 shows a block diagram of a radar procedures attempt to keep each scatterer within its
imaging system using stepped-frequency waveform range cell and to make constant Doppler frequency
to illustrate the process of the synthetic radar shift for each of them. Thus, the Fourier transform
imaging. The stepped-frequency radar transmits a can be properly applied to construct a clear image of
the object.
sequence of A' bursts. Each burst consists of M
When target is significantly maneuvering,
narrow-band pulses. Within each burst the center
even sophisticated motion compensation is still not
frequency of each succesive pulse f,,, is increased by good enough, and individual scatterers may still drift
a constant frequency step A f . The total bandwidth through range cells. Thus, the Doppler spectrum
of the burst, i.e., M times the frequency may still be time-varying, and the image may still be
step Af ,determines the radar range resolution. The blurred by using Fourier transform.
total number of bursts, N , determines the Doppler However, the restrictions of the Fourier
transform can be lifted if a high resolution time-
or cross-range resolution. To form a radar image,
frequency transform is used. Thus, there is no need
after collecting the returned signals, the M -by-
to eliminate the drift through range cells and no need
Ncomplex data is organized into a two- to keep Doppler frequenqy shift constant for each
diemensional array, which represents the unprocessed scatterer.
spatial frequency signature of the object, Scf) .
The radar processor uses the frequency Time-Varying Doppler Spectrum
signatures as the raw data to do range processing and
Doppler processing. Range processing functions as a From a frequency analysis point of view, the
matched filter for performing pulse compression. complex motion causes the Doppler frequency to be
For the stepped-frequency signals, the range time-varying. Therefore, if we can analyze the time-
processing performs M -point inverse discrete varying Doppler frequency shift very well, the image
Fourier transform (IDFT) for each of the A' received blurring caused by the time-varying Doppler
frequency signatures. Therefore, N range profiles spectrum can be resolved without applying
each containing M range cells can be obtained. By sophisticated motion compensation. On the other
taking discrete Fourier transform (DFT) at each range hand, since the time-varying Doppler frequency shift
cell, a N -point Doppler spectrum can be generated. can also be induced by uncompensated phase errors,
time-varying Doppler processing can also benefit in
Combining the M Doppler spectra at k f range resolving the image blurring due to these.
cells, the -by- N radar image is formed.
TFTS'96 391

The time-varying spectrum can be induced frequency transform, a 2-D range-Doppler Fourier
by time-varying velocity v R ( t )or rotation rate frame becomes a 3-D time-range-Doppler cube. By
sampling in time, a time sequence of 2-D range-
a( t ) . Even if the rotation rate is a constant, the
Doppler images can be viewed [ 3 ].
rotational motion-induced Doppler frequency shift A simulated aircraft in terms of its 3-D
can still be time-varying. Another factor of time-
reflectivity density function p(x,y,z) is used to
varying Doppler spectrum may come from
uncompensated phase errors. due to the object's explore the potential benefit of the time-frequency

Range Range
Fig.2 Images of aircraft from the simulated data with phase errors by using (a) Fourier transform
and (b) time-frequency transform.

irregular motion, fluctuation of the rotation rate, transform iri constructing radar images. The aircraft
inaccuracy in tracking the phase hdstoiry, and other has only a translational motion with a velocity of
variations due to the system and the environment. 120m/s along a track at 130 degree from the radar
Because of the time-varying property of the line of sight. The radar is assumed operating at X-
Doppler spectrum, an efficient waiy to resolve the band, and total 64 stepped frequencies are used in
smeared Fourier spectrum and blurred radar image is each burst. Here, 0.82 sec observation time or 256
to apply a joint time-frequency transform. To samples of the time history series is assumed.
achieve superior resolution and unibiased estimation Therefore, the radar image consists of 64 range-cells
of the instantaneous frequency spectrum, we use the and 256 Doppler frequencies.
time-frequency distribution series (TFDS) to perform If there is a fluctuation in the velocity
time-frequency transform [ 2 ] which is an efficient +
v(t) = v(0) 6V(t)
way to achieve focusing and to resolve the image it can induce an equivalent radial velocity fluctuation
blurring problem caused by time-varying property of as well as an equivalent angular velocity fluctuation.
Doppler spectrum. Even if the fluctuation causes only a maximum
velocity variation of 0.5 m / s with the above
Time-Frequency Transform far Radar mentioned velocity (which results in a maximum
Imaging range fluctuation of 0. lm), the uncompensated phase
error due to this fluctuation can cause the constructed
To achieve superior resolution and unbiased image to be blurred as shown in Fig.2(a), where
estimation of the instantaneous frequency spectrum, range tracking and Doppler tracking has been
the time-frequency properties of joint the-freqiuency applied.
transform are very useful. B!y replacing the
conventional Fourier transform with a joint time-
392 TFTS'96

From time-varying spectrum point of view, [2] S.Qian and D.Chen,''Decomposition of Wigner-
the uncompensated phase error causes the Doppler Ville distribution and time-frequency distribution
spectrum to be time-varying. If we deal with the series", IEEE Trans. on Signal Processing, 42, (1994)
time-varying Doppler spectrum using the pp.2836-2842.
conventional Fourier transform processing, the image [3] V.C.Chen,"Radar ambiguity function, time-
becomes blurred. By replacing the Fourier transform varying matched filter, and optimum wavelet
with a joint time-frequency transform implemented correlator", Optical Engineering, 33(7), (1 994)
by the time-frequency distribution series as illustrated pp.2212-2217.
in Fig. 3 , a single image frame becomes a sequence
of temporal image frames. In other words, the joint
time-frequency transform resolves the single image
frame into a stack of its temporal frame elements.
For each temporal frame element, its range-Doppler Frequency
resolution is higher than the single Fourier frame. Domain
Signatures
Fig.2(b) shows one frame from the image sequence
constructed by the joint time-frequency transform.
We can easily see that by using the joint time-
frequency transform, the time-varying spectrum can
be represented very well and, therefore, the blurred
Fourier image is resolved into a sequence of time-
varying images, which not only has higher resolution,
but also shows the Doppler changing and range
walking from time to time.

Summary

The conventional radar imaging uses the


Fourier transform, which assumes that the contents of Fig.3 Block diagram of ISAR image reconstruction using
Doppler frequencies are not changing during imaging joint time frequency transform.
time. However, in most cases, this assumption is not
true. In many cases, the Doppler frequency shift is
actually time-varying. Therefore, by representing the
time-varying Doppler spectrum with the Fourier
spectrum, the Doppler spectrum becomes smeared.
To represent the-varying spectra, a joint
time-frequency transform is desirable. The joint
time-frequency transform can very well represent the
local time-frequency structures of the radar data. By
replacing the conventional Fourier transform with a
joint time-frequency transform, a 2-D range-Doppler
Fourier frame becomes a 3-D time-range-Doppler
cube. By sampling in t h e , a time sequence of 2-D
range-Doppler images can be viewed. Each
individual time-sampled image from the cube
provides superior resolution.

Reference

[ 11 V.C.Chen,"Reconstruction of inverse synthetic


aperture radar image using adaptive time-frequency
wavelet transform (Invited Paper)", SPIE
Proceedings on Wavelet Applications, 249 1, (1995)
pp.373-386.

You might also like