NASA Technical Reports Server (NTRS) 19790006097: Azimuth correlator for real-time synthetic aperture radar image processing

United States Patent [i9j inj 4,132,989

Arens [45] Jan. 2, 1979

[54] AZIMUTH CORRELATOR FOR REAL-TIME

SYNTHETIC APERTURE RADAR IMAGE
PROCESSING

[76] Inventors: Robert A. Frosch, Administrator of

the National Aeronautics and Space
Administration, with respect to an
invention of Wayne E. Arens,
Pasadena, Calif.

[21] Appl. No.: 843,308

[22] Filed: Oct. 18, 1977

[51] Int.a.2 G01S7/30

[52] U.S. a 343/5 CM; 343/100 CL

[58] Field of Search 343/5 CM, 100 CL

[56] References Cited

U.S. PATENT DOCUMENTS

4.045,795 8/1977 Fletcher et al 343/5 CM

4.084,158 4/1978 Slawsby 343/5 CM

Primary Examiner — Maynard R. Wilbur

Assistant Examiner — Richard E. Berger
Attorney, Agent, or Firm — Monte F. Mott; John R.
Manning; Paul F. McCaul

[57] ABSTRACT

An azimuth correlator architecture is defined wherein a
number of serial range-line buffer memories are cas-
caded such that the output stages of all buffer memories
together form a complete and unique range bin in the
azimuthal dimension at any given time. A range bin is
automatically read out of the last stages of the registers
in parallel on a range line sample-by-sample basis for
subsequent range migration correction and correlation.
Range migration correction is performed on the range
bins by effectively varying the length of a delay register
at the output of each range-line buffer memory. The
corrected range bin output from the delay registers is
then correlated with a Doppler reference function to
form an image element on a real-time basis.

11 Claims, 4 Drawing Figures

SERIAL RANGE
LINE SAMPLES
FROM RANGE CORRELATOR

-ES

RANGE

4,132,989

U.S. Patent Jan. 2, 1979 Sheet 4 of 4

UJ CO
O UJ
Z-J
< Q.

ors

-j ^

uj£:

(O

a:

q:

^ UJ

2 a:
o (T
q: o

U- o

O H
UJ^ UJ

<D

a:

< ^ o
a: S o

z

z2

CO

H

OH

Ulz

O UI

ub UI UJ

q:z5

oSto

UJUl —

z^aiQC

<~03

-itr u-

0. Ui ti-

o:2oo

Q.U- UJ
O UI o
Q QC O

1

AZIMUTH CORRELATOR FOR REAL-TIME
SYNTHETIC APERTURE RADAR IMAGE
PROCESSING

ORIGIN OF INVENTION

The invention described herein was made in the per-
formance of work under a NASA contract and is sub-
ject to the provisions of Section 305 of the National
Aeronautics and Space Act of 1958, Public Law 85-568
(72 Stat. 435; 42 use 2457).

BACKGROUND OF THE INVENTION

This invention relates to a real-time data processor,
and more particularly to a real-time azimuth correlator
for a synthetic aperture radar (SAR) image processor.

Radar imaging using side-looking synthetic aperture
radar techniques is the best known approach for achiev-
ing high-resolution imagery through planetary atmo-
spheric cloud cover. However, if the radar echo data
are not processed into images onboard the spacecraft or
aircraft, very large quantities of raw uncorrelated data
must be sent to Earth for processing. Conversely, if
images are produced onboard in real time, multiple-look
data may be superimposed into single frames and con-
ventioni data-compression algorithms may be applied
to significantly reduce the data volume and rates trans-
mitted to Earth.

During recent years, considerable effort has been
devoted to the development of a digital radar image-
processing capability. Unfortunately, results to date
indicate that the digital data processing required to
produce correlated radar images onboard a spacecraft
or small aircraft is prohibitive based upon cost, com-
plexity, power, size and weight considerations. Since
only limited data reduction, by means of presumming
and time expansion, can be accomplished with the un-
correlated radar echo data, proposed radar mission
requirements to date have implied the need for reliable
high-speed and high-capacity tape recorders for storage
and have imposed potentially severe requirements upon
the telecommunications link and ground data-handling
capabilities.

Recent development of charge-coupled device
(CCD) technology offers the potential for considerable
simpliHcation of the complicated digital implementation
of SAR convolution. For instance, a CCD transversal
filter of length N provides N stages of storage while
performing N signal-by-weighting coefficient multipli-
cations each clock period. Since the powerful computa-
tional equivalency of a CCD transversal filter signifi-
cantly alleviates the normal constraints associated with
digiti processing, a CCD SAR image processor offers
a potentially attractive solution to the real-time onboard
SAR signal-processing problem, as described by the
inventor in a U.S. Pat. No. 4,045,795 titled “Charge-
Coupled Device Data Processor for an Airborne Imag-
ing System.’*

It is well recognized that azimuthal resolution in
radar imaging is proportional to the size of the antenna.
The physical size of a real aperture antenna normally
becomes too large to achieve an azimuth resolution
comparable to the range resolution available from typi-
cal radar band-widths. The process wherein a small
antenna is used to simulate a large antenna in order to
achieve a practical azimuth resolution is termed syn-
thetic aperture. Synthetic aperture radar (SAR) is based
upon the fact that there is no difference between a large

2

aperture antenna and a small antenna that successively
occupies all of the positions which are simultaneously
occupied by the larger antenna, provided the data are
successively collected, stored, and subsequently com-
5 bined to simulate the larger antenna.

The problem in SAR data processing is thus collect-
ing and correlating the echo-return pulses in N range
bins, where N is a function of the desired range resolu-
tion. A resulting set of N range-line samples for a given
10 azimuth position is called a range line. For each subse-
quent transmitted pulse, a new range line is generated.
Since the radar physically moves in the time interval
between transmitted pulses, each range line will be at a
different azimuthal position. A number of range lines
15 corresponding to the number of echo returns required
to synthetically simulate the desired real aperture an-
tenna must be stored. The resulting matrix consists of
rows that contain different time delays or ranges, and
columns that provide azimuthal information for a given
20 range. The information required to produce a single
image from the matrix is dispersed throughout the ma-
trix. In the time domain, correlation must be accom-
plished in both the range and azimuth dimensions in
order to convert the dispersed echo data into image
25 elements.

The primary processing functions employed to con-
vert echoes into image elements are sampling, presum-
ming, range correlation and azimuth correlation. The
echo returns are sampled at greater than Nyquist rate
30 and stored. The stored samples are then read out at a
lower rate over the full pulse repetition interval (PRI),
thus resulting in a time expansion of N samples in the
PRI, i.e. a data-rate reduction proportional to the echo
pulse duty cycle.

35 For SAR applications, the PRF is such that the azi-
muth resolution frequently exceeds the range resolu-
tion. On the assumption that the azimuth resolution
does not have to be greater than that in range, echo
pulses may be presummed (range bin 1 added to range
40 bin 1 , range bin 2 added to range bin 2 , etc., over succes-
sive echo pulses). If the azimuth resolution were 6.25
meters, then presumming every four pulses into one
composite pulse would provide a resolution of 25 me-
ters. The result would be a data-rate reduction of 4. It
45 should be noted that the presumming function is not
practically accomplished by direct summation, but must
be achieved by means of filtering to adequately reduce
aliasing effects.

Range correlation of an incoming echo signal from a
50 given target with a replica of the transmitted signal
results in a compressed pulse having a pulse width cor-
responding to the range resolution and a position in
range corresponding to that of the actual target. The
correlated signal pulse width is inversely proportional
55 to the transmitted signal bandwidth. Large bandwidths
yielding high resolution can be accommodated because
pulse compression (correlation) techniques allow the
signal bandwidth to be expanded with negligible sensi-
tivity loss.

60 The primary problem of real-time SAR data process-
ing is with azimuth correlation. Signals from a given
target will be received during transit of the SAR
through the desired real aperture. Due to the Doppler
effect, the carrier return from the target will be fre-
65 quency modulated due to the SAR motion through the
desired real aperture. This FM is treated as a chirp
function and is assumed to be a part of the input signal
to the azimuth filter corresponding to the range bin in

4,132,989

3

which the designated target lies. Correlation of this
signal with the expected Doppler chirp function across
all azimuthal target positions relative to the SAR in the
desired real aperture yields a compressed pulse having a
pulse width corresponding to the azimuth resolution 5
and a position in azimuth corresponding to that of the
actual target. More correlation points (i.e., a longer
correlation time) simulates a larger real aperture and
therefore provides a narrower pulse width and im-
proved azimuth resolution. 10

Throughout the image-processing algorithm, it is
desirable to measure both amplitude and phase. This is
best accomplished by I and Q processing wherein the
vector of each echo return is resolved into real (I) and
quadrature (Q) components such that the sum P 15

is proportional to the power of the echo return. In order
to combine successive echo returns, as required for
azimuthal processing, they must be resolved into their
real (I) and quadrature (Q) components. It is the azi-
muth correlation process required for the I and Q sig- 20
nals that gives use to the problem solved by the present
invention. In most spacecraft and some aircraft applica-
tions, azimuth correlation of both the I and Q signals
with the Doppler reference function, requires that both
a linear and quadratic range migration compensation 25
capabOity be provided. Furthermore, a capability for
reprogramming and updating both the range migration
correction and Doppler reference functions in real-time
must be provided to compensate for variations in vehi-
cle flight parameters. 30

SUMMARY OF THE INVENTION

In accordance with the present invention, incoming
samples are transferred tl^ough cascaded range line
memory means, each capable of storing one range-line 35
of serial complex samples. Enough range lines are thus
stored to correspond to the number of points required
to correlate over the real aperture in the azimuth dimen-
sion. The last stage of each range-line memory means is
separated from the last stage of the preceding and subse- 40
quent range-line memory means by exactly one range
line. The outputs of the last stages, taken together,
therefore form a complete and unique range bin in the
azimuthal dimension at any given time. For each com-
plex sample from a range line that is shifted into mem- 45
ory, every sample moves one stage through the entire
cascaded memory array. The net result is that an en-
tirely new set of samples occurs at the output stages of
the range-line memory elements corresponding to a
new azimuthal range bin. The desired azimuthal range 50
bins to be corrected and subsequently correlated are
automatically read out of the memory in parallel on a
range-line sample-by-sample basis. Range migration
compensation is performed on the range bins by reading
the output of the last stage of each range-line memory 55
means into an X-stage register where X is the maximum
number of bins over which migration occurs. In effect,
the X-stage registers as a group are storing parallel
azimuthal range bins. To accomplish compensation, it is
then only necessary to be able to select the information 60
from any location within each register to form a new
range bin for correlation corresponding to the desired
correction curve through suitable taps and in response
to programmable selection codes. The range migration
compensated output from each X-stage delay register is 65
then applied to a corresponding complex multiplier
where it is properly weighted to conform to the appro-
priate Doppler reference function. The multipliers, as a

group, receive Doppler reference function coefficients,
from a microprocessor controller, that may be repro-
grammed and updated in real time. Finally, the outputs
from each complex multiplier are summed resulting in a
correlated image element.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a general block diagram illustrating the
architecture of a time domain synthetic aperture radar
in accordance with the present invention.

FIG. 2 is a schematic diagram illustrating a range-bin
matrix in the azimuth correlator of FIG. 1.

FIG. 3 is a block diagram of a first exemplary em-
bodiment of the azimuth correlator in FIG. 1.

FIG. 4 is a block diagram of a second exemplary
embodiment.

DESCRIPTION OF PREFERRED
EMBODIMENTS

A block diagram of a SAR image processing system
incorporating the present invention is shown in FIG, 1.
An input rate buffer 10 for SAR data provides the nec-
essary radar echo return samples under control of a
microprocessor controller 12. ITie microprocessor con-
troller also controls the transfer of SAR data to a range
correlator 14. The function of the range correlator is to
provide correlation of an incoming echo signal from a
given target with a replica of the transmitted signal.
This correlation is carried out on echo return samples in
real time on a range-line by range-line basis, each range
line consisting of successive samples 1, 2, 3 ... N. Corre-
spondingly numbered range-line samples from M suc-
cessive range lines form successive range bins 1, 2 ... N
as shown in FIG. 2. Each range bin is then correlated in
an azimuth correlator 16 in real time in accordance with
the present invention to provide an image element out-
put for each range-line sample received. Although ref-
erence will be made hereinafter to range-line samples, it
is to be understood that the samples are complex sam-
ples consisting of both 1 and Q components.

A resulting set of N range-line samples for a given
echo return is called a range line. For each subsequent
transmitted pulse, a new range line is generated. Since
the radar physically moves in the time interval between
transmitted pulses, each range line will be at a different
position in the direction of motion of the aircraft or
spacecraft carrying the radar. A number of range lines
corresponding to the number M of PRF echo returns
required to synthetically simulate the desired real aper-
ture antenna must therefore be stored.

A range bin matrix 18 receives the serial range line
samples from the range correlator, and stores the sam-
ples in the form shown in FIG. 2. A number of range
lines are stored to correspond to the number of points
required to correlate over the real aperture of the SAR
in the azimuth dimension. For example, assume the
number M of range lines is 1000. Furthermore, assume
that integration of 1000 samples yields 6.25m resolution
in azimuth. If the number of samples M were reduced to
250, the resolution would only be 25m, The resolution
desired in azimuth thus dictates the number M of the
matrix.

Before storing a number of range lines in a matrix of
MxN range-line samples, range correlation is carried
out by the real-time range correlator 14. The range-cor-
relation function is carried out by convolving the in-
coming echo signal from a given target with a replica of
the transmitted signal, and the result is a compressed

4,132,989

5

pulse having a pulse width corresponding to that of the
actual target. This real-time range-correlation function
may be carried out by a CCD transversal filter as de-
scribed in the aforesaid U.S. Pat. No. 4,045,795. Assum-
ing the reference function is stable, a fixed tap-weighted 5
transversal filter could be used. A typical implementa-
tion would include four transversal filters on a single
CCD integrated circuit chip to accomodate complex
convolution.

The number N of samples to be stored is determined 10
by the range desired in the imagery. A typical number
might be 1000 range bins. A matrix of MxN samples
must thus be coherently integrated in azimuth to form a
single image line comprised of N azimuth correlated
range bins. However, before integration, the range bins 15
read out for integration must first be corrected for range
migration by a range migration compensator 20. In
addition, each sample of each range bin must be multi-
plied by a proper Doppler reference function using a
multiplier 22 prior to integration by a summer 24. 20

As just noted, for each new range line of N range-line
samples shifted into memory, correlation must be per-
formed over M samples in each of N range bins to pro-
duce a new image line. Unfortunately, corresponding
range-line samples do not stay in the same range bin 25
throughout the entire time a given target remains in the
radar beam. Instead, they migrate over several range
bins, as shown in FIG. 2 for the beginning and end of
each group of serial range line samples. Without range
migration, those samples would be in line, as repre- 30
sented by dots in the memory range bins, but with range
migration, the same samples would be distributed over
range bins as shown by circled X’s. Consequently, in
order to perform correlation for each of N range bins, it
is necessary to select only those samples that lie on the 35
migration curve indicated which has both a linear and a
qu^rature component. In order to do that, more range-
line samples must be stored than the number N required
for the desired range. Typically, 1100 samples may be
stored for N — 1000 to allow for a maximum range 40
migration over 100 range bins.

The range migration effect is due to geometric con-
siderations of the SAR flight path, the rotation of the
surface of the planet, and the curvature of the surface of
the planet. Range migration may be best visualized as a 45
migration of a target in range over a number of PRFs as
the target moves through the real aperture. Range mi-
gration compensation is accomplished through the
compensator 20 for each range bin in the azimuth di-
mension to be correlated with the appropriate Doppler 50
reference. A programmed weighting function corre-
sponding to the Doppler reference is provided for each
range bin sample through the multiplier 22. Both the
compensator 20 and multiplier 22 are under control of
the microprocessor controller 12 which receives the 55
necessary data to compute the range migration compen-
sation and Doppler weighting. The range migration
compensated and Doppler weighted range-bin signals
in the azimuth direction are then integrated in the sum-
mer 24 to provide the desired azimuth output signal on 60
a real-time basis. Each set of integrated range-bin sig-
nals produces a picture element (pixel) for a given
range.

The azimuth-correlation function carried out by the
azimuth correlator 16 utilizes a unique architecture 65
which makes possible real-time SAR image processing
onboard a spacecraft or aircraft. As Just noted, the first
step in achieving this azimuth correlation in real time is

6

to read correlated range lines into the range-bin matrix
18 shown schematically in FIG. 2. Range-bin data from
the range correlator enters a bottom row of range-bin
memory cells at the left and propagates through the
row, left to right, and into the next row, again entering
on the left, and so forth. The serial range-bin data exits
from the last cell on the right of the last row. Enough
range lines are stored to correlate over the real aperture
in the azimuth dimension for the resolution desired.

As further noted with reference to FIG. 2, the last
stage of each range-line memory is separated from the
last stage of the preceding and subsequent range-line
memory by exactly the number of range-line samples in
one range line. TTie outputs of the last stages, taken
together, therefore form a complete and unique range
bin over the real aperture in the azimuthal dimension at
any given time. For each unique range-line sample from
a range line that is shifted into memory, every range-
line sample moves one stage through cascaded buffer
memories. The net result is that an entirely new set of
samples occurs at the output stages of the buffer memo-
ries corresponding to a new azimuthal range bin. The
minimum number of range-bin samples required to pro-
cess an image element are thus automatically and suc-
cessively read out of the memory in parallel for each
range-line sample read in to the memory. The readout
rate is synchronized with the range-line sample rate
resulting in the most efficient transfer of data possible.

Echo-return signals from a given target will be re-
ceived during a number of successive PRTs as the SAR
passes through the desired real aperture. The size of the
real aperture desired thus determines the length of
range-line memories (rows of cascaded memory stages)
required. Azimuth correlation could be carried out by
simply weighting and summing the outputs of the
range-line memories except for the range migration
effect that needs to be compensated.

An output overlay buffer (not shown) may be used to
superimpose single-look images from the output of the
summer 24 to form multi-look image frames which may
be either recorded onto film, used to generate computer
compatible tapes, or the like. How the images are used
is not a part of this invention which relates only to the
real-time image processing of the input signals to the
SAR, and more particularly to the azimuth processing.

Referring now to FIG. 3, an exemplary embodiment
for the azimuth correlator 16 is comprised of range-line
buffer memories (shift registers) 30 connected in cas-
cade to realize the range-bin matrix 18. The range-bin
matrix thus implemented provides for the first step of
real-time azimuth correlation which is accomplished by
reading correlated range lines into the memory so orga-
nized. The second step is for the range bins aligned in
azimuth to be read from the memory in parallel into
variable delay lines 32 which receive range migration
correction control signals. The output range bins of the
variable delay lines are range migration compensated.

The range migration compensated signals are applied
to multipliers 34 and summing network 36 wherein they
are convolved with a Doppler reference function. Due
to the Doppler effect, the carrier return from the target
is frequency modulated as a target passes through the
SAR aperture, first with an increasing frequency to a
point where the target is directly broadside, and then
with a decreasing frequency. This frequency modula-
tion in the form of a chirp function, is treated in essen-
tially the same way as the reference function in the
transversal filter of the range correlator. Thus, for each

4,132,989

7

range bin read from the range-bin matrix, there is pro-
duced an azimuth correlated image element.

The manner in which the variable delay lines 32,
comprised of shift registers 38 and gating networks 39,
are used to effect range migration compensation will 5
now be described. Since the outputs of the last stages of
the range line memories form a complete and unique
range bin in the azimuthal dimension at any given time,
and since every sample moves one stage through the
entire cascaded memory array for each unique sample 10
from the range correlator shifted into memory, the
result is that an entirely new set of range-bin samples
occurs at the output of the range line memories for
every range-line sample interval. If there were no range
migration, these azimuthal range bins could be corre- 15
lated directly. But with range migration, a particular
target may appear in successive range lines in different
range bins, as indicated by the solid line curve of FIG.

2 which represents the range position of a given target
in the successive range lines. Therefore to effect range 20
migration compensation, the shift registers 38 are
tapped at successive range-line sample intervals, and the
particular taps used across the array of delays are se-
lected by range migration compensation control from
the microprocessor controller 12.

If the variable delays are implemented as CCD serial
shift registers, as disclosed in the aforesaid patent, the
uncompensated azimuthally aligned range-bin signals
move one stage through the shift registers during every
range-line sample interval. Thus the output of the last 30
stage of each range-line memory 30 is read into a shift
register 38 having a number of stages corresponding to
the maximum number of bins over which migration can
occur. In effect, the shift registers taken as a group store
uncompensated azimuthally aligned range bins. To ac- 35
complish compensation, it is only necessary to select the
appropriate range-bin signals from any location within
each register to form a new array of compensated azi-
muthally aligned range-bin signals for correlation corre-
sponding to the range migration curve. The net effect is 40
the same as being able to slide every range line in the
range-bin matrix (range line memories) in either direc-
tion through range bins until the desired samples are
aligned for correlation. This may be implemented by
selectively enabling output gates 39 which couple 45
stages of the variable delays (shift registers) into the
multipliers 34. There the Doppler weighting is applied
by the reference function coefficients used for azimuth
correlation. The Doppler reference coefficients are
computed and provided by the microprocessor control- 50
ler. In that manner range migration compensated azi-
muthal range bins are automatically correlated to pro-
duce a new image element during every range line sam-
ple interval in real time.

To effectively interpolate more precise compensa- 55
tion, the range migration curve may be accommodated
by effectively resampling the range-bin signals at the
correct phase (time location of the range line within a
range bin). A system for this resampling technique is
shown in FIG. 4. Components common to the system of 60
FIG. 3 are identified by the same reference numerals. If
the range correlator output in the form of range line
samples (which is exactly what is in each shift register
38) is convolved with a digitally sampled (sin x/x) func-
tion, the original range correlated signal will be repro- 65
duced, and by shifting the phase of the (sin x/x) func-
tion, the original range correlated signal is sampled at
points different in range by the amount of phase shift.

8

Thus the same range correlated signal can be produced
following convolution, only it will be sampled at differ-
ent points corresponding to the phase shift of the (sin
x/x) function. As schematically illustrated in FIG, 4, a
range migration compensated sample within a range bin
may be accomplished to fit the range migration com-
pensation curve precisely. The microprocessor control-
ler positions the (sin x/x) function for each shift register
38 location to accomodate the range migration curve by
computing (sin x/x) function values for each output tap
(stage) of every shift register and applying the com-
puted (sin x/x) function values to a bank 40 of multipli-
ers. The outputs of the multipliers are summed in net-
works 42 and then applied to multipliers 34. This entire
function could be achieved with a single CCD transver-
sal filter wherein the (sin x/x) function is the reference.
The outputs of each multiplier 34 are then summed to
form an image element for each new range line sample
from the range correlator.

Although particular embodiments of the invention
have been described and illustrated herein, it is recog-
nized that modifications and variations may readily
occur to those skilled in the art. It is therefore intended
that the claims be interpreted to cover such modifica-
25 tions and variations.

What is claimed is:

1. In a synthetic aperture radar system, an azimuth
correlator for real time image processing of range cor-
related range line samples from echo return signals, the
combination comprising:

a plurality of serial range-line buffer memories cas-
caded so that the output stage of one feeds into the
input stage of another, and the output stages of all
buffer memories together form a complete and
unique range bin, whereby a range bin is automati-
cally read out on a range line sample-by-sample
basis,

a plurality of variable delay means, one for each
range-line buffer memory, for receiving successive
range bins and in each controllably delaying indi-
vidual range bin samples a selected number of
range line sample periods to effect range migration
correction across a range bin,
multiplying means, connected to receive the selec-
tively delayed outputs from said variable delay
means, for correlating a Doppler reference func-
tion with each successive range-migration cor-
rected range bin to form an image element on a real
time basis, and

means for summing the Doppler reference correlated
range bin samples.

2. The combination of claim 1 wherein each of said
serial range-line buffer memories is comprised of a pre-
determined number of complex sample memory stages
greater than the number, N, of range bins to be pro-
cessed for the number N of image elements at successive
ranges, the predetermined number of complex sample
memory stages being greater than the number N by a
number sufficient to accommodate the maximum range
migration to be corrected.

3. The combination of claim 2 wherein said variable
delay means is comprised of shift registers each having
a number of complex sample memory stages necessary
to accommodate the maximum range migration to be
corrected, and gating means connected to the stages of
each shift register for selectively coupling the output of
any stage to said correlation means for range-migration
correlation.

4,132,989

9

4. The combination of claim 3 wherein said Doppler
reference function is a controlled variable.

5. The combination of claim 2 wherein said variable
delay means is comprised of shift registers each having

a number of complex sample stages necessary to accom- 5
modate the maximum range migration to be corrected
and a separate means for collectively correlating the
outputs of all stages of each shift register with a phase
variable (sin x/x) function to selectively couple a resam-
pled range bin sample output, which is interpolated to 10
within a fraction of the range line sample period, to said
means for Doppler reference function correlation.

6. The combination of claim 5 wherein said Doppler
reference function is a controlled variable.

7. The combination of claim 1 wherein said Doppler 15
reference function is a controlled variable.

8. In a synthetic aperture radar system, an azimuth
correlator for real time image processing of range cor-
related range line samples from echo return signals, the
combination comprising:

a plurality of serial range-line buffer memories cas-
caded so that the output stage of one feeds into the
input stage of another, and the output stages of all
buffer memories together form a complete and
unique range bin, whereby a range bin is automati- 25
cally read out on a range line sample-by-sample
basis,

correlation means connected by coupling means to
receive outputs from said range-line buffer memo-
ries for correlating a controllably variable Doppler 30
reference function with each successive range bin
to form an image element on a real time basis, and

means for summing the Doppler reference correlated
range bin samples.

10

9. Ah azimuth correlator for real-time synthetic aper-
ture radar image processing under control of a radar
controller, comprising

a plurality of cascaded range-line memory means,
each having a plurality of stages for storing one
range line of serial complex samples, whereby the
outputs of all range line memory means together
form a complete and unique range bin in the azi-
muthal dimension at any given time, and for each
complex sample from a range line that is shifted
into memory, every sample moves one stage
through the plurality of range-line memory means,
means connected to the output stages of said plurality
of range-line memory means for performing range
migration compensation under control of said
radar controller, and

means connected to said range migration compensa-
tion means for correlating range-migration cor-
rected range bin samples with a Doppler reference
20 function under control of said radar controller.

10. The combination of claim 9 wherein said range
migration compensation means is comprised of a plural-
ity of shift registers, each having the same number of
stages, and means under control of said radar controller
for gating range line samples from selected stages of
said shift registers.

11. The combination of claim 10 wherein said range
migration compensation means is comprised of a plural-
ity of shift registers, each having the same number of
stages, and means under control of said radar controller
for correlating a selectively phase shifted (sin x/x) func-
tions with the outputs of said shift register stages,
thereby to provide range migration compensation.

« « « * ♦

40

45

50

55

60

65