Advanced Imaging Approaches for Detecting Obscured Objects

pancakesbootAI and Robotics

Nov 24, 2013 (3 years and 11 months ago)

53 views


Advanced Imaging Approaches for
Detecting Obscured Objects


Sermsak Jaruwatanadilok


Sumit Roy

Yasuo Kuga

Department of Electrical Engineering, University
of Washington, Seattle, WA

BSI, Bellevue, WA, Feb 26, 2009

Overview


Goal and concepts


Assets and capabilities


Previous and on
-
going work

Goal & Concepts

Concepts:

(1) Waveform design at
transmitters to combat
random media effects

(2) Physics
-
based EM model of
received signals

(3) Signal processing at the
receivers

**Exploit relationship among
(1), (2), and (3)**

GOAL
:

Improve

detection

and

imaging

of

objects

in

obscuring

and

complex

environments

using

electromagnetic

waves

Assets and Capabilities


Analytical formulations


Angular / Frequency correlation functions of surface scattering


Two frequency mutual coherence functions of waves in random media


Numerical simulation tools


Monte Carlo simulations


Scattered waves in the presence of particle scatterings


Full
-
wave simulation tools


FDTD software


COMSOL Multi
-
physics


Experimental tools, equipments and facilities


Array imaging system


MMW systems


Anechoic chamber

I. MMW active imaging of concealed objects

II. MMW passive imaging of concealed objects

III. Microwave imaging using angular/frequency correlation
methods

IV. Time reversal method and time reversal imaging

V. Coherent array imaging


VI. Focused pulse beam imaging

VII. Detection of vehicle and human movement using existing
communication systems


Combined use of the physics
-
based EM modeling and

signal processing

Current and Previous Work Related to BSI

I. MMW Active Imaging of Concealed Objects


Simulated MMW Image Examples


Aperture radius = 30 cm


Distance = 1 m


Cloth material: cotton


Cloth thickness = 8 mm


Plastic explosive (C
-
4)

Optical image

94 GHz simulated image 200 GHz simulated image

Simulated MMW Image Examples


Aperture radius = 30 cm


Distance = 1 m


Cloth material: cotton


Cloth thickness = 1.2 mm


Plastic explosive (C
-
4)

Optical image

94 GHz simulated image 200 GHz simulated image

Multi
-
layer Model


ABCD matrix formulation







4 1 4
4 1 4 4 1 4
//
2
,
////
s s
A B Z Z C D Z
R T
A B Z Z C D Z A B Z Z C D Z
  
 
     
3 3
2 2
3 3
2 2
A B
A B
A B
C D
C D
C D
 
 
 

 
 
 
 
 
 






cos, sin, sin
m m m m m m m m m m m m
A D q h B jZ q h C j q h Z
   
air
cloth
plastic
Human
skin
p

c

h

2 2
2 2
A B
C D
 
 
 
cloth
plastic
Human
skin
3 3
3 3
A B
C D
 
 
 
air
Incident
wave
Incident
wave
h
2
h
3
,
m
m m m m
m
Z q j
q

 
  
Simulated MMW Pulse Imaging

94 GHz

220 GHz

Aperture radius = 30 cm

Distance = 1 m

Cloth material: cotton

Cloth thickness = 1.2 mm

Plastic explosive object

Bandwidth = 10 GHz

II. MMW Passive Imaging of Concealed Objects

[1] R. Appleby, (From previous slide)

[2] National Academies, “Assessment of Millimeter
-
Wave and Terahertz Technology for Detection and Identification of Concealed Ex
plosives and
Weapons,”
http://www.nap.edu/catalog/11826.html
, 2007



2
o o o
8
4
1
3
ambient or sky temperature = 50 ~100 K ou
tdoor, ~295 K indoor
reflection from cloth, medi

reflection from human skin, med
um emission 1 exp
ium to skin,body emission ex
a
m o
b b
T
T
T
T
T T
T
T



    
 




 




5 6 7
target emissibility, body em
reflection from target,
issibility , optic
me
al depth, resolution
dium to target, target emis

sio e p
p
n x
t t
o
o
t b
o
L
D
T T T T




 

 



 
 

Imaging is done by the difference
T
4
6 7
3
5
2
and
T +T
+T

T
+T
+
Cloth
Air gap
or
explosive
Human
skin
Aperture
D
T
1
Target
T
2
Ta
T
3
T
4
T
5
T
8
T
6
T
7
Simulated Passive Imaging Examples

94 GHz

Optical image

OD = 0.123

OD = 0.288

OD = 0.826

OD = 1.9205

Cloth thickness = 1.2 mm

Cloth thickness = 8 mm

Metal object

220 GHz

OD = Optical depth

III. Angular Correlation Function /

Frequency Correlation (ACF / FCF)


Correlation of waves with different angles and
frequencies


Exploit the difference of correlation characteristics
when a target is presence compared to no target


Experimental Studies of ACF/ FCF Memory Line


Strong correlation on ‘memory line’





1 2
19, 20
o o
q q
= =
13
14
15
16
17
18
-4
-3
-2
-1
0
1
2
3
4
Frequency difference (GHz)
Phase of FCF/ACF
Phase (rad)


no explosive
explosive present
Use of Angular and Frequency Correlation Function
(ACF/FCF) for Imaging


beam 1: 92 GHz


96 GHz 10 degree


beam 2: 78 GHz 12 degree


Equivalent to imaging but this shows

presence of particle scattering

#
1
Tx
/
Rx
Angular and
Frequency Correlation
(
ACF
/
FCF
)

1

2

#
2
Tx
/
Rx
Cloth
Air gap
or
explosive
Human
skin
Slope = 5.9 radians / GHz

Slope = 0.39 radians / GHz

shrapnel

IV. Time
-
Reversal Method

Concept of time
-
reversal imaging and focusing

(1)
Send probing signals

(2)
Obtain received signals (targets and surrounding)

(3)
To focus: re
-
transmit time
-
reversed signals


To image: process time
-
reversed signals

Time
-
Reversal Focusing

Snapshots of wave field in random media. (a) Gaussian pulse propagating through

random media, (b) Time
-
reversed pulse back
-
propagated in the random medium.

The energy focuses at the original source location.

Geometry of the problem

Focusing improvement in random media
(OD=optical depth)

Time
-
Reversal Imaging



Multistatic data matrix




Time reversal matrix



How to model the time reversal


matrix in the presence of random scattering media


Time reversal imaging


Time reversal MUSIC (multiple signal classification)

1
M
m m m
m




K g g
* * * *
1 1
M M
m m m m m m
m m

  

 
 

T K K g g g g






1 2
,,,
T
m m m m N
G r r G r r G r r

 
 
g
X
X
X
X
X
X
X
Array
SAR
Random
complex
medium
Space
-
time transmitter
-
receiver 7
-
element array with half wavelength spacing is located at, and
a point target is located at and in a random medium. The left figure shows array and image.
Two figures on the right show images (in dB) in the dotted expanded area for OD = 0.1 and 0.5.


Space
-
time time reverse MUSIC images in free space and random complex media at OD = 0.1
and 0.5. (dB scale) Figs. 5 and Fig. 6 show the result for identical physical problems. Note that
space
-
time time reversal MUSIC has superior lateral resolution.

V. Coherent Array (CA) Imaging and Detection
of Object in Random Media

(a) SAR images is formed using backscattering signals.
Received signal is a response of a single transmitter

(b) CA method coherently combines responses from all
receivers and transmitters


Numerical simulations: (a) SAR images (b) CA images

CA method can mitigate effects from random scattering

and clutter, but suffers the reduction in image resolution.

VI. Focused Pulse Beam in Random Scattering Media


Effects from random
scattering media on the
imaging: two
-
frequency
mutual coherence
function


Contribution from target
and media

Focused Beam Imaging

VII. Detection of Vehicles and Human
Movement Using Existing Communication
Systems




Newly Started Project in BSI




Concept


Range
-
Doppler image
using digital correlator


Angle
-
of
-
Arrival using
MUSIC

Source
Passive array
system
Objects
Reference
signal
clutter
clutter
Adaptive
Beamforming
Adaptive
Cancellation
Cross
Correlation
&
Doppler
Processing
Reference
antenna
Array antenna
2
-
D target
imaging
Adaptive clutter
mitigation
Angle of arrival
estimation
DETECTION

SCHEME

Adaptive Cancellation


-

Remove direct signal and clutter from surveillance channels to get true echo signal



-

Adaptive filter uses a lattice predictor structure



Cross Correlation


-
Find Doppler shifts and time
-
delayed echoes of the targets.

Drawbacks:

-
Excessive processing time for long input
signals

-
Decimation technique: discard data at
Doppler frequencies we know targets do not
exist before Fourier Transform

Time Delay


剡湧R


r
1

+ r
2

= 2a

b
2

= a
2
-
c
2

D1

D2

D3

3
2
1
3
)
2
1
(
D
c
td
D
D
c
D
D
D
td







Adaptive Beamforming to Get Angular
Resolution

Array
Processor
Source
s(t)
z(t)
Antenna
Array
y
m
(t)
f(x,t)
MUSIC



1
1
ˆ




e
R
e
MUSIC
H
MUSIC





M
Nz
i
H
i
i
MUSIC
1
1
ˆ
v
v
R
Spatial Subarray Smoothing

For correlated signals:

Results from MUSIC AOA Estimation

-100
-80
-60
-40
-20
0
20
40
60
80
100
-120
-100
-80
-60
-40
-20
0
X: -9.216
Y: 0
Angle of Arrival - Deg
Amplitude - dB
X: 33.42
Y: -16.79
X: 69.32
Y: -7.283
Some Simulation Results

xxxxxxx
1
GHz
Source
Passive
array
Target
1
Target
2
100
m
1
0
0

m
-50
0
50
100
-50
0
50
100
150
x (m)
y (m)
Target - AOA
VIII. Array Imaging Systems


Range


angle imaging


using step CW and angle


of arrival processing

MMW Radar for Imaging


Frequency 30 GHz (to be extended to 100 GHz)


Spotlight images using 2
-
D scan and stepped CW
mode


Doppler images using 2
-
D scan and short pulse

Spotlight image using 2
-
D scan and stepped CW mode

Doppler images using 2
-
D scan and short pulse


With a known vibrating source at 20 Hz (discrimination of an active source)

Resolution

Cross
-
range: ~ 2 degree


(antenna beamwidth)

Down
-
range: ~ 3 cm


5 GHz bandwidth

On
-
going work


Improving modeling of wave propagation in random
scattering media and clutters


Angular / Frequency correlation for detection and imaging of
target


Ultra wide band time reversal imaging and focusing


Detection of vehicles and human movement using existing
communication systems

Future work


Collaborative imaging and detection from several receivers