Radar Imaging

Margaret Cheney
Department of Mathematical Sciences
Rensselaer Polytechnic Institute
cheney@rpi.edu

September 18, 2005

with thanks to Brett Borden and various web authors for figures
 RADAR = RAdio Detection And Ranging

developed within engineering community

how to transmit high power (physics, engineering)
how to detect signals (physics, engineering, math)
how to interpret and use received signals (math)

mathematically rich
PDE (electromagnetic theory, wave propagation)
harmonic analysis, group theory, microlocal analysis
linear algebra, sampling theory
statistics
scientific computing
coding theory, information theory
 Why make images with radar?

works day or night (unlike optical imaging)

works in all weather

penetrates clouds, smoke
some radars can penetrate foliage, buildings, soil, human tissue

can provide very accurate distance measurements

sensitive to objects whose length scales are cm to m

can measure velocities (changes in range)

 Radar history

1886 Heinrich Hertz confirmed radio wave propagation

1904 Hulsmeyer patented ship collision-avoidance system
1922 ship detection methods at NRL (Taylor & Young, 700MHz)
1930 Hyland used radar to detect aircraft
first US radar research effort, directed by NRL
1930s England and Germany radar programs developed:
Chain Home early warning system (22-50 MHz)
fire control systems
aircraft navigation systems
cavity magnetron to transmit high-power microwaves
1940s establishment of MIT Rad Lab (British + American)
radar for tracking, U-boat detection
 see blip at time 2r/c.
Rudimentary
High imaging
Range-Resolution (HRR) ima
Detection. For a target at distance r,
Real-aperture
see blip at time 2r/c. imaging
High Range-Resolution
Plan (HRR) imagingindicator
position
Real-aperture imaging

Plan position indicator

x2

x2

x1

x1
Synthetic Aperture Radar
(SAR)
 SAR History

1951 SAR invented by Carl Wiley, Goodyear Aircraft Corp.

mid-50s first operational systems, under DoD sponsorship:
U. of Illinois, U. of Michigan, Goodyear Aircraft,
General Electric, Philco, Varian
late 60s NASA sponsorship (unclassified!)
first digital SAR processors
1978 SEASAT-A
1981 beginning of SIR (Shuttle Imaging Radar) series
1990s satellites sent up by many countries
SAR systems sent to Venus, Mars, Titan

8
 Radarsat (Canada)
JERS (Japan)

ERS-1 (Europe) Envisat (Europe)

Venus radar penetrates cloud cover
Venus topography
 AirSAR

CARABAS
UAVs

Lynx SAR
 Applications

military: early warning, tracking, targeting

commercial aviation, navigation, collision-avoidance

land use monitoring, agricultural monitoring, ice patrol,

environmental monitoring

surface topography, crustal change

speed monitoring (police radar)

weather radar: storm monitoring, wind shear warning

search and rescue

medical microwave tomography

Deforestation in Brazil
Ocean waves (texture due to wind)
Oil slicks on the ocean

Sea ice
Ocean internal waves at Gibraltar
 Southern
California
topography
 Glacier flow
via SAR
interferometry
 Outline
1. introduction, history, frequency bands, dB, real-aperture imaging
2. radar systems: stepped-frequency systems, I/Q demodulation
3. 1D scattering by perfect conductor
4. receiver design, matched filtering
5. ambiguity function & its properties
6. range-doppler (unfocused) imaging
7. introduction to 3D scattering
8. ISAR
9. antenna theory
10. spotlight SAR
11. stripmap SAR
 Assumed background

Fourier transform

delta function

(x2 t2 )u(t, x) = 0 has solutions of the form

u(t, x) = f (t x) + g(t + x)
!
Cauchy-Schwartz inequality ( f g #f ##g#)

f = O(g) means f (const.)g

B = 0 B = A and E = 0 E =

E = ( E) 2 E
 Fourier transform
! !
1
F[F ](t) := f (t) = eit F ()d = e2it F ()d
2
!
inverse transform: F () = eit f (t)dt

Properties
"
1. If g(t) = h(t t" )f (t" )dt" , then G() = H()F ().

2. t f (t) = F[iF ](t)

" it
3. (t) = (2) 1
e d

in n dimensions: ! !
1
F[F ](x) := f (x) = eix F ()d F () = eix f (x)dx
(2)n
 Books
B. Borden, Radar Imaging of Airborne Targets, Institute of Physics, 1999.
C. Elachi, Spaceborne Radar Remote Sensing: Applications and Techniques,
IEEE Press, New York, 1987.
W. C. Carrara, R. G. Goodman, R. M. Majewski, Spotlight Synthetic
Aperture Radar: Signal Processing Algorithms, Artech House, Boston, 1995.
G. Franceschetti and R. Lanari, Synthetic Aperture Radar Processing, CRC
Press, New York, 1999.
L.J. Cutrona, Synthetic Aperture Radar, in Radar Handbook, second
edition, ed. M. Skolnik, McGraw-Hill, New York, 1990.
C.V. Jakowatz, D.E. Wahl, P.H. Eichel, D.C. Ghiglia, and P.A. Thompson,
Spotlight-Mode Synthetic Aperture Radar: A Signal Processing Approach,
Kluwer, Boston, 1996.
I.G. Cumming and F.H. Wong, Digital Processing of SAR Data: Algorithms
and Implementation, Artech House, 2005
 Maxwells equations

E = t B (1)
H = J + t D (2)
D = B =0 (3)

E = electric field D = electric displacement

B = magnetic field H = magnetic induction
J = current density = charge density

Constitutive laws in free space

D = #0 E B = 0 H J =0 =0
 (1) + constitutive laws + (2)

E = t B = 0 t H
! "# $ ! "# $
( E) 2 E ! 0 t E
! "# $
0

2 E 0 "0 t2 E = 0
!"#$
1/c20

'

Fourier transform E() = eit E(t)dt
&

2

2 E + 2 E = 0
c
!"#$
k2
Atmospheric Attenuation
 Radar frequency bands

Band Designation Approximate Frequency Range

HF 330 MHz
VHF 30300 MHz
UHF 3001000 MHz
L-band 12 GHz
S-band 24 GHz
C-band 48 GHz
X-band 812 GHz
Ku-band 1218 GHz
K-band 1827 GHz
Ka-band 2740 GHz
mm-wave 40300 GHz
 Decibels
! "
power in
log10 power out = Bel too small
instead use:
2
power in Vin Vin
decibel dB = 10 log10 power out = 10 log10 2 = 20 log10
Vout Vout

power (voltage)2

dB Power ratio
0 dB 1
10 dB 10
20 dB 100
30 dB 1000
 Outline
1. introduction, history, frequency bands, dB, real-aperture imaging
2. radar systems: stepped-frequency systems, I/Q demodulation
3. 1D scattering by perfect conductor
4. receiver design, matched filtering
5. ambiguity function & its properties
6. range-doppler (unfocused) imaging
7. introduction to 3D scattering
8. ISAR
9. antenna theory
10. spotlight SAR
11. stripmap SAR
 Radar systems
1. Stepped-frequency radars (laboratory systems)

transmit receive measure

cos(1 t) RR (1 ) cos(1 t) + RI (1 ) sin(1 t) R(1 )
! "# $ ! "# $
Re(ei1 t ) Re[R(1 )ei1 t ]
Re(ei2 t ) R(2 )ei2 t R(2 )
.. .. ..
. . .
Re(eiN t ) R(N )eiN t R(N )
From the Rs, can synthesize response
& to any waveform
% N
sin (t) = an (n )ein t a()eit d
1
Response would be &
% N
srec (t) = an (n )R(n )ein t a()R()eit d
1
 2. Pulsed radar systems
cos !c t
s(t)
waveform transmitter
generator (amplifier)

p(t) = s(t) cos !c t

antenna

circulator

p (t) = a(t) cos[ !c t +" (t)]

rec
I/Q low-noise
demodulator amplifier (LNA)

correlation
receiver
 I/Q Demodulation
in-phase (I) channel:
prec (t) cos(c t) = a(t) cos((t)
+ c t) cos(c t)

= a(t) 12 cos((t) + 2c t) + cos (t)
$ %& '
filter out
quadrature (Q) channel (90 out of phase):
prec (t) sin(c t) = a(t) cos((t)
+ c t) sin(c t)

= sin((t) + 2c t) + sin (t)
a(t) 12
$ %& '
filter out
I and Q channels together give the analytic signal

srec (t) = a(t)ei(t)

(approximately analytic in upper half-plane, when a(t) is slowly varying,
i.e., in narrowband case)
 Filters

H() transfer function

"
F
# F 1
f (t) F () F ()H() (h f )(t)

$ $
1 it"
F 1
[H()(Ff )()] (t) = e it
H() e f (t" )dt" d
2 $ %$ &
1 i(tt" )
= e H()d f (t" )dt"
2
' () *
h(tt" )
+
1 || < 1
Example: Low-pass filter. Take H() =
0 otherwise
h(t) = 1 sin 1 t
1 t = sinc
1
1 t
 Outline
1. introduction, history, frequency bands, dB, real-aperture imaging
2. radar systems: stepped-frequency systems, I/Q demodulation
3. 1D scattering by perfect conductor
4. receiver design, matched filtering
5. ambiguity function & its properties
6. range-doppler (unfocused) imaging
7. introduction to 3D scattering
8. ISAR
9. antenna theory
10. spotlight SAR
11. stripmap SAR
 1D Scattering by a fixed perfect conductor at range R
waveform generator sinc (t)
transmitter output:
! "
sinc (t) cos(c t) = Re sinc (t)eic t
:= f (t)
transmitted electromagnetic wave: (1D model)
E in (r, t) = ein f (t x/c) where x = e r
E in is a right-going solution of
1 2 in
x2 E in 2 t E = 0
c

Write total field as E tot = E in + E sc (think f (t x/c) + g(t + x/c))

E tot satisfies
1 2 tot
2 tot
x E 2 t E = 0
c #
#
tot #
E # = 0 conducting B.C.
x=R

1 2 sc
x2 E sc 2 t E = 0
c ! !
!
sc !
!
in !
E ! = E !
x=R x=R

expect E sc (r, t) = esc g(t + x/c) (left-going solution of wave equation)

B.C. esc g(t + R/c) = ein f (t R/c) esc = ein
" #$ %
w
t = w R/c g(w) = f (w 2R/c)

received field at r = 0:
E sc (0, t) = ein f (t 2R/c)

transmit f (t), receive prec (t) = f (t 2R/c) (fixed target)

 1D Scattering by a moving conductor at range R(t)
g(t + R(t)/c) = f (t R(t)/c)
! "# $
w
solve w = t + R(t)/c for t (via Implicit Function Theorem) t = (w)
for pulsed systems: use Taylor series expansion R(t) = R + vt +
R(t)
# $! " w R/c
w = t + (R + vt) /c t= := (w)
1 + v/c
%
g(w) = f (t (R + vt)/c)%t= (w)

) *
1 v/c
= f 1 + v/c (w R/c) R/c

! "# $

Doppler scale factor
 RF field scattered from moving target

For f (t) = s(t) cos(c t),

prec (t) = s((t R/c) R/c) cos[c ((t R/c) R/c)]
! "# $
t(1+)R/c

frequency of cosine = c

v 2v 2v
For << 1, 1 c c c
c c c $
! "#

Doppler shift = D
 I/Q demodulation of signal from moving scatterer

prec (t) cos(c t) = s((t R/c) R/c) cos[c (t (1 + )R/c)] cos(c t)

! "filter
#$
out
% &
1
= s((t R/c) R/c) cos[sum] + cos [c (t (1 + )R/c) c t]
2

I(t) = s((t R/c) R/c) cos c [( 1)t (1 + )R/c)]

Q(t) = s((t R/c) R/c) sin c [( 1)t (1 + )R/c)]

srec (t) = s((t R/c) R/c)eic [(1)t(1+)R/c)]

For v
c << 1 and s slowly varying:

srec (t) s(t 2R/c)eiD (tR/c) e2ic R/c

 Outline
1. introduction, history, frequency bands, dB, real-aperture imaging
2. radar systems: stepped-frequency systems, I/Q demodulation
3. 1D scattering by perfect conductor
4. receiver design, matched filtering
5. ambiguity function & its properties
6. range-doppler (unfocused) imaging
7. introduction to 3D scattering
8. ISAR
9. antenna theory
10. spotlight SAR
11. stripmap SAR
 2. Pulsed radar systems
cos !c t
s(t)
waveform transmitter
generator (amplifier)

p(t) = s(t) cos !c t

antenna

circulator

p (t) = a(t) cos[ !c t +" (t)]

rec
I/Q low-noise
demodulator amplifier (LNA)

correlation
receiver
 Receiver design

For good range resolution, want a short pulse

But a short pulse has little energy hard to detect signal in noise
1
energy density R4 !

signal is swamped by thermal noise in the receiver!

target cant even be detected, much less imaged
Brilliant solution:
Use (long) coded pulses and mathematical processing

matched filter or correlation receiver
pulse compression
 Matched filter: sketch of derivation

receiver input: r(t) = s(t ) + n(t) ( = demodulator output)

" " # want to find
aei /R4 2R/c noise, assumed white, zero mean
power spectral density N
!
Apply filter (t) = (h r)(t) = h(t t! )r(t! )dt! = s (t) + n (t)
Choose h so that |s ( )/n ( )| is as large as possible.
2"
"! "
! "2
|s ( )|2 h( t )s(t )dt
! !
SN R = max = max !
h E|n ( )|2 h N |h(t)|2 dt
"!
2"
"
! "2
h(t )s(t )dt
! !
= max !
h N |h(t)|2 dt

Cauchy-Schwartz inequality h(t) = s (t)

! ! ! !!
(t) = s (t t)r(t )dt = s (t )r(t + t!! )dt!! correlation
! !
 Pulse compression from matched filtering

Example: the 5-bit Barker code +++-+

+ + + - + correlator output
+ + + - + 1
+ + + - + -1+1=0
+ + + - + 1-1+1=1
+ + + - + 1+1-1-1=0
+ + + - + 1+1+1+1+1=5
 Multiple fixed targets

Two fixed targets: r(t) = 1 s(t 1 ) + 2 s(t 2 ) + n(t)

!
Distribution of fixed targets: r(t) = ( ! )s(t ! )d ! + n(t)
Apply matched filter:
"
(t) = s (t! t)r(t! )dt!
" "
= s (t! t) ( ! )s(t! ! )d ! dt! + noise
" "
= s (t! t)s(t! ! )dt! ( ! )d ! + noise
# $% &
( ! t)
!
(t) = s (t!! + t)s(t!! )dt!! = point spread function for
1D imaging system
High Range-Resolution (HRR) Imaging
 Chirp = Linearly Frequency Modulated (LFM) waveform
s(t) = ei(t) rect(t/tp ) where d
dt (t) = instantaneous frequency
!
1 1/2 < t < 1/2 C
rect(t) =
0 otherwise
d
dt (t) = at (t) = at2"

!+)

!+'

2
s(t) = e iat
rect(t/tp )
!+%

!+#

!!+#

gives rise to a point spread function !!+%

!!+'

!!+)

!"
! " # $ % & ' ( ) * "!

(t) = (1 |t|)sinc(at(1 |t|))

where sinc x = (1/x) sin x.

(see p. 170 in Rihaczek Principles of High Resolution Radar
an upchirp
or work out yourself)
 Matched filter for single moving target

receiver input = demodulator output = r(t) = s(t )eiD (t ) + n(t)

want to find and D .
use a filter bank = set of filters that depend on a parameter :
!
(t, ) = h (t t" )r(t" )dt"

to maximize SNR, choose h (t) = s (t)ei2t

 Matched filter for distribution of moving targets
!! ! 2i ! (t ! )
demodulator output = r(t) = ( , )s(t )e
! !
d ! d !
output of filter bank is
"
2i(tt! )
(t, ) = s (t t)e
!
r(t! )dt!
"
2i(tt! ) ! 2i ! (t! ! ) !
= s (t t)e
!
s(t )e
!
dt ( ! , ! )d ! d !
" "
2i(t ! )
= ! !
( t, )e ( ! , ! )d ! d !

where
! 2it!!
(, ) = s (t + )s(t )e
!! !!
dt!!

(narrowband) radar ambiguity function

point spread function for imaging system
Typically one considers only the magnitude of the ambiguity function.
 Outline
1. introduction, history, frequency bands, dB, real-aperture imaging
2. radar systems: stepped-frequency systems, I/Q demodulation
3. 1D scattering by perfect conductor
4. receiver design, matched filtering
5. ambiguity function & its properties
6. range-doppler (unfocused) imaging
7. introduction to 3D scattering
8. ISAR
9. antenna theory
10. spotlight SAR
11. stripmap SAR
 Properties of the ambiguity function
!
1. |(, )| |(0, 0)| = |s(t)|2 dt = signal energy
= 1 for a normalized signal
!!
2. |(, )|2 d d = 1 (for a normalized signal)
Radar uncertainty principle or conservation of ambiguity volume

3. |(, )| = |(, )|

4. If is the ambiguity function for s, then the ambiguity function a

iat2
for e s(t) satisfies |a (, )| = |(, + a )|

5. The ambiguity function for s(t)eia is the same as that for s(t).

6. The (magnitude of the) ambiguity function for s(t)eit is the same

as that for s(t).
 Resolution and cuts through the ambiguity function

Doppler (frequency) resolution:

!" !
! !
|(0, )| = !! |s(t)| e
2 2it
dt!!

Frequency (Doppler) resolution is determined by amplitude.

For good Doppler resolution, want |s(t)| 1.
Range resolution:
!" !
! !
|(, 0)| = !! |S(2)| e
2 2i
d !!
#
where S() = eit s(t)dt.
Range resolution (for a fixed target) is determined by bandwidth.
 Example: Range resolution with a CW pulse

baseband signal is s(t) = rect(t/tp ) tp = time duration of pulse

ambiguity function is
!" #$ % " #&$
| | $ | | $
|(, )| = 1 tp $sinc tp 1 tp $ for | | < tp
0 otherwise
Range resolution is obtained from
!" | |
#
|(, 0)| = 1 tp for | | < tp
0 otherwise
whose first null is at pn = tp .
 ambiguity
function
for CW pulse

N. Levanon,
Radar Principles,
Wiley 1988
 Example: Range resolution with a chirp
iat2
For s(t) = rect(t/tp )e
the ambiguity function is
!" #$ % " # &$
| | $ | | $
|(, )| = 1 tp $sinc tp 1 tp ( + a ) $ for | | < tp
0 otherwise
Range resolution is obtained from
!" #$ % " # &$
| | $ | | $
|(, 0)| = 1 tp $sinc tp 1 tp a $ for | | < tp
0 otherwise
The first null is at pn = 1
atp = 1
B where B = bandwidth
Phase modulation improves range resolution by a factor of
pn,CW tp
pulse compression ratio = = = tp B !
pn,chirp (1/B) '()*

time-bandwidth product
ambiguity
function
for chirp
 A train of high-range-resolution (HRR) pulses
Doppler shift can be found by change in phase of successive returns
Suppose target travels as R(t) = R0 + vt; write Rn = R(nT )
1. transmit pn (t) = s(t)eic t
receive rn (t) = pn (t 2Rn /c)eiD (t2Rn /c)
2. demodulate: sn (t) = s(t 2Rn /c)eiD (tRn /c) e2ic Rn /c
! #
3. correlate: n ( ) = s (t )sn (t# )dt# =
! # iD (t! Rn /c) 2ic Rn /c #
s (t )s(t 2Rn /c)e
#
e dt
4. at peak, = 2Rn /c:
! 2 iD (t! Rn /c) # 2ic Rn /c
n (2Rn /c) = |s(t 2Rn /c)| e
#
dt e
" #$ %
(0,D )

5. phase difference between successive pulses:

2c [R0 + v(n + 1)T ]/c 2c [R0 + vnT ]/c = 2c v/c = D
6. note blind speeds when 2c v/c = 2(integer)
 ambiguity
function for
a train of pulses

pulse repetition
frequency gives
rise to delay
ambiguities
 Outline
1. introduction, history, frequency bands, dB, real-aperture imaging
2. radar systems: stepped-frequency systems, I/Q demodulation
3. 1D scattering by perfect conductor
4. receiver design, matched filtering
5. ambiguity function & its properties
6. range-doppler (unfocused) imaging
7. introduction to 3D scattering
8. ISAR
9. antenna theory
10. spotlight SAR
11. stripmap SAR
 Range-Doppler Imaging v
!
v cos !

(x,y)
r

Stationary radar, rotating 2D object !

x

If radar is at (0, R), scatterer at (x, y):

range is R + y

if rotation rate is , then

|v| = r vLOS = vy = |v| cos = r! cos
"# $
x
D 2vLOS 2x
recall Doppler shift is = =
c c c
As the object rotates, x and y change (scatterer moves out of
resolution cell)
blurring
Need 3D scattering model that incorporates target motion
 Moving radar imaging a stationary planar scene

delay range scatterer lies on a constant-range sphere

scatterer on plane lies on a constant-range circle

Doppler shift line-of-sight relative velocity

scatterer lies on the iso-Doppler cone vLOS = R v = const
scatterer on plane lies on iso-Doppler hyperbola

does not account for change in radar position as measurements are

taken (scatterers migrate through resolution cell)
get an unfocused image
Need a 3D scattering model that incorporates changes in sensor position