Final Report Honours Project: Through-the

Final Report Honours Project: Through-the
Final Report
Honours Project:
Through-the-Wall Imaging Radar
Student:
Thang Bui (1173888)
Partner:
Joseph Rabig (1162140)
Supervisors:
Prof. Douglas Gray
Mr. Richard Drake
October 21, 2011
Executive Summary
A through-wall imaging radar could have many applications where it is important to view
a visually and physically obstructed area. Typical applications for such radar are search and
rescue, police and military operations. The desired purpose of a through-the-wall system is to
estimate or detect the inner content and structure of a room behind the wall, or layout of a
building. Further actions, such as sending personnel inside the obscured area can be followed
accordingly once the result is extracted.
This project was conducted to investigate the feasibility of constructing a Synthetic Aperture
Radar using a Vector Network Analyser as the waveform generator and a pair of horn antennas
being moved to form a Synthetic Aperture, in order to image the objects behind the wall.
The arrangement of the antennas could follow bistatic or pseudo-monostatic configuration. For
each movement of the antennas, the forward gain of the environment was measured. MATLAB
programs were used for focusing implementation to generate the images of the region using
recorded data.
2-D image of the region of interest behind a wall has been successfully generated. The
pseudo-monostatic configuration results in better cross range resolution compared to bistatic
case. Experiments were conducted for various wall materials (pin-up board, plaster wall and
office wall) and the results prove the feasibility of through the wall imaging application using
a Synthetic Aperture Radar. The designed radar system is simple and of low cost, which can
be used for research application. The outcome of this project serves as a foundation for the
implementation of a more complex through-wall radar system.
Acknowledgements
To Prof. Doug Gray and Mr. Richard Drake for your advice and guidance for the project.
Your support and knowledge are invaluable for the completion of this project.
To Joseph Rabig for being part of the project. It has been a pleasure working with you.
To staffs of EEE School for your generous help and support, from administrative works
or part acquirement to construction.
To Prof. Bevan Bates from DSTO for providing the loan of a pair of horn antennas.
Contents
List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
iii
iv
List of Codes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v
Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
1 Introduction
1.1 Aims . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1
1
1.2 Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.3 Background and Significance . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1
1
1.4 Project overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.5 Document structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2
3
2 Synthetic Aperture Radar Imaging
2.1 Power Consideration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4
4
2.2 Range Resolution and Pulse Compression . . . . . . . . . . . . . . . . . . . . . . 5
2.3 Cross-range Resolution and Radar Configurations . . . . . . . . . . . . . . . . . 7
2.4 Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.4.1
2.4.2
Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
Calibration using Electronic Calibration Kit . . . . . . . . . . . . . . . . 11
2.4.3 Calibration using Linear Regression . . . . . . . . . . . . . . . . . . . . . 12
2.5 Focusing Technique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.5.1 Focusing Delay Formulation . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.5.1.1
2.5.1.2
2.5.2
When there is no wall . . . . . . . . . . . . . . . . . . . . . . . 14
When there is a wall . . . . . . . . . . . . . . . . . . . . . . . . 15
Imaging implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.5.2.1 Time domain implementation . . . . . . . . . . . . . . . . . . . 17
2.5.2.2 Frequency domain implementation . . . . . . . . . . . . . . . . 19
2.6 Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3 Experimental Set-up and Initial Experiments
21
3.1 Equipments and Experimental Environment . . . . . . . . . . . . . . . . . . . . 21
3.2 Experiment with Patch Antenna . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
i
3.3 Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4 Imaging Experiments and Results
26
4.1 Experimental Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.2 Imaging of objects without wall . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.2.1 1-object imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.2.2 2-object imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.3 Imaging result for a pin-up board . . . . . . . . . . . . . . . . . . . . . . . . . . 31
4.4 Imaging result for a constructed wall . . . . . . . . . . . . . . . . . . . . . . . . 33
4.4.1
4.4.2
Set up . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
Empty room . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.4.3
4.4.4
4.4.5
No wall . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
One plaster board . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
Two plaster boards . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
4.5 Imaging result for a real office wall . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.6 Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
5 Project Management
45
5.1 Work Plan and Schedule Revisit . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
5.2 Risk Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
5.2.1 Risk occurred . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
5.2.1.1
5.2.1.2
Delay in material procurement and construction . . . . . . . . . 47
Subversion blackout . . . . . . . . . . . . . . . . . . . . . . . . 47
5.2.1.3 Anechoic chamber and VNA unavailable . . . . . . . . . . . . . 48
5.2.1.4 Lack of technical knowledge . . . . . . . . . . . . . . . . . . . . 48
5.3 Budget review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
6 Conclusion
49
6.1 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
6.2 Project Outcomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
Bibliography
52
A MATLAB Code
53
ii
List of Figures
2.1 Stepped-frequency signal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2 Sample reflection measurement from VNA . . . . . . . . . . . . . . . . . . . . .
6
7
2.3 Monostatic configuration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.4 Bistatic configuration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.5 Distance from Horn 1 to Horn 2 via pixel Pi for monostatic and bistatic cases .
8
9
9
2.6 1D experiment set up . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.7 Impulse responses for different antenna separations . . . . . . . . . . . . . . . . 10
2.8 Relative response between two separations . . . . . . . . . . . . . . . . . . . . . 11
2.9 Desired vs. Actual delays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.10 Electronic Calibration kit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.11 Cable response and calibration result using ECal module . . . . . . . . . . . . . 12
2.12 Linear Regression for calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.13 Sample experimental result for linear regression . . . . . . . . . . . . . . . . . . 13
2.14 Calibration technique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.15 Pre- and post-calibration impulse response . . . . . . . . . . . . . . . . . . . . . 14
2.16 2-D imaging system configuration . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.17 2-D imaging system configuration with wall . . . . . . . . . . . . . . . . . . . . 16
2.18 2-D imaging system configuration with wall . . . . . . . . . . . . . . . . . . . . 16
2.19 Imaging technique in time domain . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.20 Imaging technique in frequency domain . . . . . . . . . . . . . . . . . . . . . . . 19
3.1 Synthetic aperture radar and Agilent PNA (VNA) . . . . . . . . . . . . . . . . . 21
3.2 ECal module and DSTO horn antennas . . . . . . . . . . . . . . . . . . . . . . . 22
3.3 Patch antenna and R&S ZVL VNA . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.4 Forward voltage gain for a cross-polarised patch antenna . . . . . . . . . . . . . 23
3.5 Impulse response for a cross-polarised patch antenna . . . . . . . . . . . . . . . 23
3.6 Simulation result for time domain implementation . . . . . . . . . . . . . . . . . 25
3.7 Simulation result for frequency domain implementation . . . . . . . . . . . . . . 25
4.1 Image of one object in the chamber . . . . . . . . . . . . . . . . . . . . . . . . . 28
4.2 Down-range and cross-range profiles at peak location . . . . . . . . . . . . . . . 28
4.3 Image of two objects at same range in the chamber . . . . . . . . . . . . . . . . 29
iii
4.4 Down range profiles at peak location for image of 2 objects . . . . . . . . . . . . 30
4.5 Cross range profile at peak location for image of 2 objects . . . . . . . . . . . . 30
4.6 Imaging result of two objects at different ranges . . . . . . . . . . . . . . . . . . 30
4.7 Result for pin-up board – Pseudo-monostatic . . . . . . . . . . . . . . . . . . . . 32
4.8 Result for pin-up board – Bistatic . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4.9 Experimental set up for constructed wall . . . . . . . . . . . . . . . . . . . . . . 33
4.10 Imaging result for an empty controlled room . . . . . . . . . . . . . . . . . . . . 34
4.11 Imaging result for controlled environment without wall . . . . . . . . . . . . . . 34
4.12 Imaging profiles for pseudo-monostatic configuration . . . . . . . . . . . . . . . 35
4.13 Imaging profiles for bistatic configuration . . . . . . . . . . . . . . . . . . . . . . 35
4.14 Imaging result for one plaster board wall . . . . . . . . . . . . . . . . . . . . . . 36
4.15 Imaging profiles for pseudo-monostatic configuration . . . . . . . . . . . . . . . 37
4.16 Imaging profiles for bistatic configuration . . . . . . . . . . . . . . . . . . . . . . 37
4.17 Imaging result for two plaster boards wall . . . . . . . . . . . . . . . . . . . . . 39
4.18 Imaging profiles for pseudo-monostatic configuration . . . . . . . . . . . . . . . 40
4.19 Imaging profiles for bistatic configuration . . . . . . . . . . . . . . . . . . . . . . 40
4.20 Experimental set-up for a real office wall . . . . . . . . . . . . . . . . . . . . . . 41
4.21 Imaging result for a real office wall . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.22 Combine the results by adding . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4.23 Combine the results by multiplying . . . . . . . . . . . . . . . . . . . . . . . . . 43
5.1 Gantt Chart . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
iv
List of Tables
3.1 Simulation parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.1 Experimental parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
5.1 Project progress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
5.2 Budget review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
v
Listings
A.1 Sample measurement result . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
A.2 Calibration using linear regression . . . . . . . . . . . . . . . . . . . . . . . . . . 53
A.3 Sample image processing program for an empty room . . . . . . . . . . . . . . . 54
vi
Acronyms
DSTO The Defence Science and Technology Organisation. 22, 47, 48
ECal Agilent Electronic Calibration Module. 11, 12, 22
EEE Electrical and Electronic Engineering. 47
ROI Region of Interest. 4, 14, 15, 24, 27, 49
SA Synthetic Aperture. 1, 2, 14, 17, 23, 27, 29, 31, 38, 44, 50
SAR Synthetic Aperture Radar. 1, 2, 24, 27, 48–50
SNR Signal-to-Noise Ratio. 4
SVN Subversion Control. 47
TWIR Through-the-wall Imaging Radar. 1, 2, 4
VNA Vector Network Analyser. 1, 2, 4, 5, 11, 17, 22, 23, 48, 49
vii
Chapter 1
Introduction
1.1
Aims
This project titled Through-the-wall Imaging Radar was conducted to investigate the feasibility
of constructing a Synthetic Aperture Radar (SAR) using a Vector Network Analyser (VNA) as
the waveform generator and a pair of horn antennas being moved to form a Synthetic Aperture (SA),
in order to image the objects behind the wall.
1.2
Scope
The technical scope of this project was to build a simple and low cost imaging system to demonstrate the concept and feasibility. A VNA and a SA were used for conducting experiments and
collecting data. Offline signal processing using MATLAB was employed for imaging purpose.
There was no dedicated electronic hardware used for data collection and generating image,
which reduced the complexity of the technical requirements, but imposed certain constraints
of the capability of the system. This project was to provide foundation research and work for
more complex system to be built in the future.
The scale of the project and the team (two students) was small compared to a real-world
engineering project which decreased the management complexity of the project. There was a
tight set of resource, time and budget and the project outcome is not for commercial use.
1.3
Background and Significance
A Through-the-wall Imaging Radar (TWIR) could have many applications where it is important to view a visually and physically obstructed area. Typical applications for such radar are
search and rescue, police and military operations. The desired purpose of a through-the-wall
system is to estimate or detect the inner content and structure of a room behind the wall, or
layout of a building. Further actions, such as sending personnel inside the obscured area can be
1
followed accordingly once the result is extracted. It can be dangerous going into such situation
without much visual information about the area.
Due to need for wide range of applications, through-the-wall radar imaging has been a
research interest in recent years. Early research used a simple propagation model and imaging
algorithm while recent advances in propagation modelling and processing power have extended
the capability of TWIR, with more focus on sensor positioning and model-based 3D imaging
[1].
The use of a SA and VNA to form a SAR to image through the wall has been common
for recent research works. In [2], a mobile robot platform was designed and moved to form
the SAR. The ultra-wideband signal was generated by a VNA and two log-periodic antennas
were deployed. The system successfully imaged the objects behind sheetrock or plaster walls.
Another closely related work to our project is described in [3],[4]. The imaging radar presented
has been able to image through a single uniform wall (plywood wall) with known characteristics.
The equipments used for this work were a VNA and one horn antenna used as a transceiver.
The transmitted signal for the papers discussed was a stepped-frequency signal generated by
the VNA. The resolution of the final two dimensional images was approximately 10cm for both
down range and cross range.
Recent attempts to use a SAR to form the approach to produce three-dimensional image
of objects behind the wall can be found in [5], [6] and [7]. However, there is no notable real
experimental result presented, only simulations were conducted to test developed algorithms.
More research are certainly needed for a real implementation of such 3D imaging system.
1.4
Project overview
A simple radar with two horn antennas was built to synthesise an antenna of bigger aperture.
The VNA is utilised as a signal generator and a tool to measure the response of the environment.
Signal processing for collected measurements is implemented in MATLAB to generate the image
of an obstructed region.
The initial stage of the project was to perform necessary research and define RF components.
Horn antennas, VNA and cable were acquired for the project. The mechanical support structure
for the SAR was built. A simulation was carried out to verify initial system design and test
the feasibility. A series of experiments were then conducted to familiarise with the anechoic
chamber and the VNA, and to calibrate the radar.
The next phase of the project was to conduct a series of imaging experiments in controlled
environments with different wall materials and radar configurations. Measurements were taken
for each experiment and offline signal processing to generate images was done using MATLAB.
The analysis was performed to compare theoretical, simulation and experimental results.
The radar configuration for each experiment was bistatic or pseudo-monostatic. Walls being
investigated for the projects were a synthetic brick wall, a pin-up board, a constructed wall
2
(one and two plaster boards) and an office wall.
1.5
Document structure
This document presents a summary for the work conducted for the project. The design challenges for the radar system are discussed in Chapter 2. Initial set up and experiments, together
with simulation result are presented in Chapter 3. Chapter 4 details the experimental set-up,
the results for the experiments in controlled environments and analyses the achieved result.
Future work and direction are briefed in Chapter 5.
3
Chapter 2
Synthetic Aperture Radar Imaging
In this chapter, several design issues for TWIR will be considered including power, resolution,
calibration and image focusing. These are the key challenges during the design phase which set
this project apart from other radar systems.
2.1
Power Consideration
The imaging radar system for this project operates in the same principle with regular radar.
The electromagnetic waves will be transmitted, reflect off the object and come back to the
receiver. The received signal power must be at least bigger than the internal noise level of the
receiver in order to extract useful information about the environment. Due to the spreading
of signal in propagation from the transmitter to the receiver, the power of received signal will
be smaller when the transmitter and receiver are further away or when the objects are further
away from the radar system.
In this project, the transmitter power was limited by the VNA. The VNA acted both as a
signal generator and an equipment to measure the response between the transmitter and the
receiver. There was no dedicated hardware to amplify or control the transmitted signal power.
The range of detection for the radar was the size of Region of Interest (ROI). Typical figures
for this size are the dimension of a testing lab or a real room. The transmitted power therefore
was required to be big enough so that the signal power reflected back from an object in ROI
satisfied certain Signal-to-Noise Ratio (SNR) requirement.
By using the radar equation and assuming the transmitter and the receiver are at the same
location, the relationship between maximum detection range Ro and the transmitted power Pt
for the SNR of 10dB is:
P t Gt Gr λ 2 σ
4
Ro =
10(4π)3 LkTo B
where:
• Gt and Gr are the gains of the transmitting and receiving antennas respectively
4
• λ is the wavelength of the electromagnetic wave
• σ is the radar cross section, which is the effective scattering coefficient of the RF target
• B is the measurement IF bandwidth
• L is the system loss
• k is the Boltzmann constant
• To is the room temperature.
Typical values for the radar system parameter are VNA output power Pt = 1mW , antenna
gain Gt = Gr = 3dB, wavelength λ = 0.1m (f = 3GHz), radar cross section σ = 20cmx20cm,
IF bandwidth B = 10kHz, system loss L = 10dB. These give Ro = 21m which is reasonably
higher than the typical size of a room or a region to image.
When there is a presence of the wall between the radar system and the region to image,
due to the change of the medium as seen by the electromagnetic wave, there will be a loss and
phase change caused by the wall, which are different from them of the air gap. The calculation
above shows there is design space for further spreading or through-wall loss.
2.2
Range Resolution and Pulse Compression
Range resolution of a radar is the smallest distance between two objects in down range such
that both objects are resolvable and distinguishable. For a pulse radar, the system is able to
distinguish two objects when the return pulses from the two objects are separate. Assuming
that pulses have equal width, two receiving pulses can be distinguished when they are one pulse
width apart.
Let τ be the width of a pulse transmitted out by a pulse radar. The down-range resolution
following previous discussion is ∆R = cτ2 where c is the speed of light. Assuming that the pulse
c
where B is the
transmitted out is a sinc pulse, the expression can be rewritten as ∆R = 2B
frequency bandwidth of the pulse. It could be observed that good range resolution requires a
shorter pulse or in other words, a pulse of high bandwidth. The numerical requirement for the
range resolution for this project was ∆R = 10cm, which yielded a requirement for bandwidth
B = 1.5GHz.
However, it is impractical to generate a pulse of such large bandwidth (or extremely short
pulse) which would require high amount of power and dedicated hardware. A technique named
Pulse Compression was used to provide an equivalent result as one single pulse of high bandwidth. This technique forms a basis for the signal generation and measurement of the VNA
used for the project. The stepped frequency signal generated by the VNA could be illustrated in
Figure 2.1. The signal sweeps through specified bandwidth and is composed of multiple pulses
5
over time. The pulses are at discrete frequencies with frequency step between two consecutive
frequencies ∆f .
Frequency
f2
b
b
b
∆f
f1
τ
∆τ
Time
Figure 2.1: Stepped-frequency signal
The signal generated by the VNA is transmitted out to the environment by one horn antenna.
The reflected signal is captured by the receiving horn. The VNA compares the transmitted
and received signals and measures the reflection coefficients of the system. Let S1 (f ) = ST (f )
and S2 (f ) = SR (f ) be the frequency representation of the transmitted and received signals
correspondingly, the measurement performed by the VNA can be expressed as:
S21 (f ) =
S2 (f )
S ∗ (f )SR (f )
= T
S1 (f )
|ST (f )|2
where S21 (f ) is the scattering parameter representing forward voltage gain. The physical meaning of S21 (f ) is the frequency response of the environment. A typical result for scattering
coefficient S21 extracted from the VNA is shown in Figure 2.2.
6
21
|S | (dB)
−40
−50
−60
−70
2.5
3
3.5
Frequency (Hz)
4
9
x 10
21
angle(S ) (rad)
0
−20
−40
−60
−80
2.5
3
3.5
Frequency (Hz)
4
9
x 10
Figure 2.2: Sample reflection measurement from VNA
The number of frequencies and the stepping size of the stepped frequency signal can be
controlled by adjusting related parameters in the VNA. In the attached figure, there are 151
frequencies from 2.5GHz to 4GHz and the step size is 10Mhz. The frequency band used was the
operating frequency band of the horn antennas. Since all the numerical figures and result are
in frequency domain, the measurements at each frequency are the loss (in dB) and the phase
change. The responses for multiple frequencies behave like a response of one single pulse of
1.5GHz bandwidth which allows better down range resolution.
2.3
Cross-range Resolution and Radar Configurations
The quality of the final image is determined by the ability to capture smaller object which can
be assessed based on the down range and cross range resolutions of the image. The criteria used
in [8] and [9] for defining cross range resolution the the half-power or 3dB width of response.
The formula for the two-way angular resolution given the aperture size D of the antenna and
the wavelength λ of the radar pulse is:
θ3dB =
λ
D
Given the range from the radar to an RF target R, the cross range resolution of the image is
∆CR = Rλ
. It could be seen that aperture size D is required to be high to achieve good cross
D
range resolution, since both R and λ are system and antenna parameters.
Typical numerical figures for our imaging radar parameters and requirement were: R = 3m,
λ = 0.1m (f = 3GHz) and ∆CR = 15cm which posed the requirement for the aperture size
D = 2m. Two horn antennas were deployed for the project and the aperture size (surface size)
7
of each horn antenna is 13cm which is much lower than the requirement. Therefore, it had been
decided that a synthetic aperture radar was built and used for our system. The radar would
have a synthetic aperture which is the effective aperture of an antenna linear array formed by
moving the horn antenna. There are two modes to configure the SA array elements which can
yield different results in final imaging: pseudo-monostatic and bistatic configurations.
For pseudo-monostatic configuration, the two horn antenna are put close together such
that the distance between them is small compared to the target range. This distance is kept
unchanged when the horn antennas are moved to form the SA or in other words, both horns
are moved simultaneously. The process of moving two horn antennas is illustrated in Figure
2.3. This configuration is not strictly monostatic since the two horn antennas cannot be put at
the same location due to the mechanical support structure.
Tx
Rx
Tx
Rx
Tx
Rx
Tx
Rx
...
Tx Rx
Figure 2.3: Monostatic configuration
For bistatic configuration, only one horn is moved to form the SA. The transmitting horn is
kept static at the same location. The distance between two antennas can be varied and assumed
to be big compared to the target range. The graphical illustration for this configuration is shown
in Figure 2.4.
8
Rx
Tx
Rx
Tx
Rx
Tx
Tx
Rx
...
Tx
Rx
Figure 2.4: Bistatic configuration
It could be seen from the description above that when a horn antenna is moved, the distance
from it to the other horn antenna, via a pixel of interest is changed. The electromagnetic path
change in pseudo-monostatic case is bigger than it of bistatic case. For comparison, the distance
from transmitting horn to receiving horn via pixel Pi (1m, 3m) for both cases is sketched in
Figure 2.5. For bistatic configuration, am = 0 and bm = 0.2 → 1.6m and for pseudo-monostatic
case, bm = 0.2 → 1.6m and am = bm − 0.2m, where am and bm are the locations of two antennas
in the array. These numbers are chosen only for illustration and are not the real parameters of
the system. It could seen that the pseudo-monostatic configuration provides greater variation
in path length compared to bistatic case. The effect of this difference can result in the difference
in the cross range resolution for the two configurations.
d(am , (1, 3), bm ) (m)
M onostatic
Bistatic
6.27
6.09
0
0.2
1.6
bm (m)
Figure 2.5: Distance from Horn 1 to Horn 2 via pixel Pi for monostatic and bistatic cases
9
2.4
Calibration
2.4.1
Overview
The initial step for generating a 2D image is to extract a correct range profile for each arrangement of the two horn antennas. The range profile is effectively the impulse response of path
of interest. A simple experiment using two directly facing horns (Figure 2.6) was set up for
extracting the range profile and double-checking the design parameters (range and power).
Agilent VNA
Horn 1
Horn 2
r
r
d
r
r
Figure 2.6: 1D experiment set up
The ideal result for the forward voltage gain S21 or S12 is the phase shift corresponding to
the distance between antennas in frequency domain or the spike at this distance in the range
profile. Sample range profiles from collected data are sketched in Figure 2.7.
Range Profile (note: d = 5.34m)
−4
x 10
Range Profile (note: d = 4.15m)
−4
x 10
10
6
9
5
8
7
4
ho
ho
6
3
5
4
2
3
2
1
1
0
5
10
Range (m)
15
20
0
(a) d = 5.34m
5
10
Range (m)
15
20
(b) d = 4.15m
Figure 2.7: Impulse responses for different antenna separations
It could be seen that the peaks in both range profiles were shifted away from desired locations.
And importantly, the relative shift in peak locations between any two range profiles is equal to
the relative change of the distance between the two horns. By dividing the two responses in
frequency domain, the response representing relative phase shift or time delay can be extracted
as shown in Figure 2.8.
10
Range Profile δd = 5.34 − 4.15 = 1.19m
0.55
0.5
0.45
0.4
0.35
ho
0.3
0.25
0.2
0.15
0.1
0.05
0
1
2
3
4
5
Range (m)
Figure 2.8: Relative response between two separations
Therefore, a reference response of a known distance is needed to correctly retrieve the desired
range profile or in other words, calibration is required. Furthermore, the shift of peak location
above is explained by introducing the system delay to caused by cables, instrument and horn,
which is independent of the antenna separation (Figure 2.9). Therefore, if system delay to is
known, it can be calibrated out by providing equivalent phase shift to the original response.
h(t)
Actual
Desired
d
c
to +
d
c
Delay t
Figure 2.9: Desired vs. Actual delays
2.4.2
Calibration using Electronic Calibration Kit
The Agilent Electronic Calibration Module (ECal) is available for project, which is included
with the VNA (Figure 2.10). It has two ports and acts as the connector between two cables
and the VNA during calibration process.
11
Figure 2.10: Electronic Calibration kit
Since the cable introduces amplitude loss and phase change towards the final response of the
system, the ECal provides an automatic calibration tool which is to remove systematic loss and
delay caused by equipment and cables. The loss and phase change of one N cable is attached
for illustration in Figure 2.11. After the calibration, the VNA remembers the response of the
cables and automatically changes the measurement accordingly to give the correct response of
the system. The delay is removed in the calibrated response or the spike in time domain is
brought closer to correct location.
−3
Amplitude
−3.8
h12
|Scable| (dB)
−3.6
−4
−4.2
−4.4
2.5
3
Original response
x 10
−3.4
3.5
14
12
10
8
6
4
2
1
4
2
3
Frequency (GHz)
Phase
7
8
9
10
−8
x 10
0.025
0.02
−100
h12 cal
angle(Scable) (rad)
0
4
5
6
Delay (s)
Calibrated response
−200
−300
0.015
0.01
0.005
2.5
3
3.5
1
4
Frequency (GHz)
(a) Loss and phase change in the cable
2
3
4
5
6
Delay (s)
7
8
9
10
−8
x 10
(b) Original and calibrated responses
Figure 2.11: Cable response and calibration result using ECal module
2.4.3
Calibration using Linear Regression
When two horn antennas were used as a transmitter and a receiver for SAR configuration, it was
observed that there is remaining unknown delay in the response. It was suggested that further
calibration would be carried out using linear regression analysis which is briefed in Figure 2.12.
This is to reduce systematic and measurement errors and to detect an outlier if there is one.
The location of highest peak in the impulse response was identified for each distance between
two horn antenna. Linear regression using least squares approach with one dependent variable
12
(distance) is fitted to experimental data points. Intersection between best-fit line and y-axis is
the system delay to . The slope of fitting curve is expected to be 1/c where c is the speed of
light in air.
τ
b
b
b
b
to
0
d1
d2
d3
d
d4
Figure 2.12: Linear Regression for calibration
Sample linear regression result using real experimental data is attached in Figure 2.13 for
demonstration. The correlation coefficient is 0.9989 which proves the correctness of linear
model and the assumption on independent system delay. The reciprocal of the slope is (3.44 ×
10−9 )−1 ≈ 2.91 × 108 m/s which is reasonably close to speed of light in air. Linear regression
proved feasible for calibration and was used to calibrate system delay. For each experimental
session, calibration process was repeated to ensure the correct result if there was any change
in the set up of the system.
rd
−8
Delay vs. Separation (May 3 )
x 10
Collect data
Fitted data: delay = 3.44e−009*distance + 2.59e−009
1.6
Delay with no calibration (s)
1.4
1.2
1
0.8
0.6
0.4
Correlation coefficient R = 0.9989
0.2
0
0
0.5
1
1.5
2
2.5
3
Distance between antennas (m)
3.5
4
Figure 2.13: Sample experimental result for linear regression
13
When unknown delay to is found, a simple multipli-
ej2πf to
cation operation is performed on original gain measurement S21 or S12 as in Figure 2.14. The output is the
frequency response of the airgap between two antennas.
The difference between output and input is the increase in phase or the location of peak in time domain
S21
S̃21
Figure 2.14: Calibration technique
representation is brought closer to zero. The sample
result to demonstrate this effect is attached in 2.15.
Range profile without calibration (note: correct d = 0.82m )
Range profile (note: correct d = 0.82m )
0.018
0.016
← d = 1.6
0.016
← d = 0.8
0.014
0.014
0.012
0.012
0.01
hd
hd
0.01
0.008
0.008
0.006
0.006
0.004
0.004
0.002
0.002
0
0
2
4
6
8
0
10
Range (m)
0
2
4
6
8
10
Range (m)
(a) Without calibration
(b) With calibration
Figure 2.15: Pre- and post-calibration impulse response
The distance between two horns for the experiment was 0.82m and the peak of impulse
response was located at 0.8m. This means that the proposed calibration approach performed
well and recovered the correct range profile for the air gap.
2.5
Focusing Technique
2.5.1
Focusing Delay Formulation
2.5.1.1
When there is no wall
The ROI of our imaging system was a single 2D slice of the room at the height of the SA. A
two-dimensional Cartesian coordinate system is fitted into this slide for our imaging purpose,
as seen in Figure 2.16. Axes x and y represent cross-range and down-range respectively. Two
horn antennas are put at (am , 0) and (bm , 0) where subscript m stands for a measurement. It
is noted that there are different measurements for different locations of the antennas. The
inter-element spacing for the synthetic aperture radar is denoted as ∆. Horn 1 can be used as
a transmitter and Horn 2 is a receiver or vice versa.
14
y
b
b
Pixel pi (xi , yi)
am
b
b
...
b
b
bm
∆
b
b
0
Horn 1
b
x
Horn 2
Figure 2.16: 2-D imaging system configuration
The pixel size of the ROI is (δx, δy) such that there is no significant improvement in the result
if the pixel is divided smaller. This figure is important to maintain a reasonable computational
complexity and to reduce processing time. The smaller the size of the pixel is, the more time it
would take to image ROI. The simulation was carried out to select the appropriate pixel size.
The value was chosen to be smaller than simulated range and cross range resolutions.
Consider an arbitrary pixel Pi which has the bottom left corner of coordinates (xi , yi ), the
electromagnetic distance from the transmitter to the receiver via pixel Pi is simply the sum of
two Euclidean distances between Pi and the transmitter, and between Pi and the receiver:
d(am , (xi , yi), bm ) = d(Horn 1 → Pi ) + d(Pi → Horn 2)
=
2.5.1.2
q
(xi − am )2 + yi2 +
q
(xi − bm )2 + yi2
When there is a wall
When there is a wall between the radar and the region of interest, the electromagnetic path
from transmitting horn to receiving horn via a pixel behind the wall is changed. Due to the
change of medium in transmission, there is refraction occurring at the two wall-air surfaces.
The bending effect on the electromagnetic rays caused by refraction is illustrated in Figure
2.17.
15
y
Pixel pi (xi , yi)
yo + t
Wall
t, n
yo
b
am
b
b
...
b
b
b
bm
∆
b
b
b
0
Horn 1
x
Horn 2
Figure 2.17: 2-D imaging system configuration with wall
Let t and n be the thickness of the wall and the refractive index of the wall material correspondingly. Without loss of generality, only one path from one horn to the pixel is considered
as in Figure 2.18. The calculation for the other path can be carried out in a similar manner.
y
Pixel pi (xi , yi)
yi
θ1
yo + t
θ2
Wall
t, n
yo
θ1
b
b
am
b
0
b
b
x1
x2
xi
Horn 1
Figure 2.18: 2-D imaging system configuration with wall
By Snell’s law, the relationship between angles of incidence and refraction is:
sin θ1 = n sin θ2
16
x
By geometry, the following is extracted:
(yi − t) tan θ1 + t tan θ2 = xi − am
By solving the equations above, angles θ1 and θ2 can be found and one-way focusing path
is:
d(am , (xi , yi)) =
nt
y−t
+
cos θ1 cos θ2
However, the symbolic and analytic solution as discussed above is hard to derive. Furthermore, the processing is carried out by computer program. Hence numerical analysis to find the
focusing path is preferred. An equivalent approach to Snell’s law using Fermat’s principle of
least time was proposed and used for this project. The principle states that the electromagnetic
path between any two points is the path of least time. As a result, the problem of finding the
angles as above can be turned into an optimisation problem to find the coordinates of points
where the ray meets the wall surfaces, x1 and x2 . The optimal value for mentioned coordinates,
x̂1 and x̂2 are chosen to minimise the electromagnetic path d, where:
d(am , (xi , yi)) =
q
q
(x̂1 − am )2 + yo2 + n (x̂2 − x̂1 )2 + t2 +
q
(xi − x̂2 )2 + (y − yo − t)2
The focusing path from the pixel (xi , yi) to horn antenna (bm , 0) can be calculated which
gives us the final return focusing delay τ for imaging purpose:
τ=
d(am , (xi , yi )) + d((xi , yi), bm )
c
where c is the speed of light.
2.5.2
Imaging implementation
For all position of horn antennas in both SA configurations, either bistatic or pseudo-monostatic,
measurement for Sam bm can be made using the VNA, where S can be either S21 or S21 and am
and bm are the x-coordinates for horn 1 and horn 2. However, these reflection measurement
only gives the response of whole region of interest. To find imaging value for a pixel of the
region, focusing technique was used, either in Time Domain or Frequency Domain.
2.5.2.1
Time domain implementation
As discussed before, if there is an object at pixel Pi , the peak of the range profile will be located
at the range which is exactly the distance from one horn to the other horn via Pi . This distance
forms an ellipse (its foci are two horns) that the object can stay on, which means exact object
location can not be determined using one range profile. However, if there is an object at pixel
Pi , above discussion will always true for any configuration of the two horns. In other word,
17
different range profiles form different ellipses which theoretically have one intersect at Pi .
The result of calibrating the measurement reflection is the response without system delay.
The inverse Fourier Transform operation performed returns its time domain representation range profile. It is proposed that for each pixel in the region of interest, its imaging value
is the sum of all range bins from different range profiles. Bin location for range profile m is
determined by using:
d(am , (xi , yi), bm
]
nm = [
c/B
where c/B is the smallest quantisation level in range, B is the frequency bandwidth and [·]
denotes rounding operation.
Since the result of the summation for each bin is complex, its magnitude is taken for the
final image. The mathematical expression for the imaging value calculation for each pixel is:
I(xi , yi) = |
N
X
ham bm (nm )|
m=1
The block diagram in Figure 2.19 shows the procedure for imaging using time domain data.
ha1 b1
Sa1 b1
IFFT
Sa2 b2
IFFT
ha1 b1 (n1 )
r
r
r
r
ha2 b2
ham bm (nm )
SaN bN
I(xi , yi )
r
haN bN
IFFT
|(·)|
...
r
...
Bin Selection
...
IFFT
ham bm
...
Sam bm
i ,yi ),bm
]
nm = [ d(am ,(x
c/B
ha2 b2 (n2 )
haN bN (nN )
r
r
r
Figure 2.19: Imaging technique in time domain
It is noted that due to the quantisation effect caused by the rounding operation, the ellipse
formed by particular range using one range profile is thicken which could result in a region of
intersection instead of intersecting point when considering multiple range profiles.
18
2.5.2.2
Frequency domain implementation
An equivalent approach in frequency domain is proposed for imaging objects. The distance
between the peak and the origin in a range profile in time domain is equivalent to the phase
difference in frequency domain. The quantitative value for this phase difference is φ = −2πf d/c
where f is a single frequency, d is the distance between peak and origin or the distance between
two horns via an object and c is the speed of light in air. This also means that for each
reflection response, if an increasing phase shift is applied so that it can compensate φ, the
response will theoretically a response of no phase shift (or peak is brought back to time origin).
Therefore, for a pixel Pi , if all compensated responses have phase shifts of zero, they can be
added constructively when there is an object at Pi . If compensated responses have non-zero
phase shifts, sum of response is destructive or there is no object at Pi .
The implementation procedure is described in Figure 2.20. It is noted that the image can
be formed using one single frequency though it will be summed across all operating frequencies
to form a finer image. The final image is formed by taking magnitude value of the complex
image.
ej2πf d(a1 ,(xi ,yi ),b1 )/c
Sa1 b1
ej2πf d(a2 ,(xi ,yi ),b2 )/c
...
Sa2 b2
X
ej2πf d(am ,(xi ,yi ),bm )/c
|(·)|
f
I(xi , yi)
...
Sam bm
ej2πf d(aN ,(xi ,yi ),bN )/c
SaN bN
Figure 2.20: Imaging technique in frequency domain
The mathematical expression for the implementation in frequency domain is the inverse
Fourier Transform of the sum of all shifted responses:
I(xi , yi ) =


N
B X
X
f =0
j2πf d(am ,(xi ,yi ),bm )/c
Sam bm e
m=1
19

j2πf τ /B 
e
τ =0
or
I(xi , yi ) =
B
X
f =0


N
X

j2πf d(am ,(xi ,yi ),bm )/c 
Sam bm e
m=1
The logical flow of the implementation in MATLAB can be described as follow:
for each pixel P(x,y) in the region of interest
initialise the sum So (length of B/delta_f)
for each location of antennas (a,0) and (b,0)
calculate the focusing distance from horn 1 to horn 2, via pixel P
phase shift the response S by the focusing delay to get S’
add S’ to the sum So
sum So across all frequencies
get the amplitude for imaging
2.6
Assumptions
Since the main objective of this project was using simulation and experiment to prove concept
and test the feasibility, the following assumptions were made about the radar system and the
environment to be imaged:
• System was linear and invertible which means multipath reflection, diffraction, dispersion
or near field artifact were insignificant.
• Objects to be imaged were metallic, static and distinct. The resultant images of this
project would provide tool to detect and localise objects behind the wall. Therefore, the
objects should not be RF transparent or either moving. This was to avoid localisation
error and image blurring.
• Wall was uniform and isotropic. The propagation model through the wall was linear and
only valid when the wall was of uniform thickness. The front and back surfaces of the
wall were assumed to be parallel. The wall is assumed to partially allow RF wave pass
through or in other words, electromagnetic shielding wall was not of our interest.
• There was no real-time requirement. The designed imaging radar was not time-critical
which means experimental data was collected and imaging processing was performed
offline. The requirements for software and hardware were therefore much simpler and
achievable within given time frame.
20
Chapter 3
Experimental Set-up and Initial
Experiments
3.1
Equipments and Experimental Environment
There were two horn antennas available to construct a synthetic aperture. The operating
frequency for both horns is from 2.5GHz to 4GHz. To hold and move each or both horn
antennas to form the synthetic aperture, a support structure was designed and built. The
mechanical support provided flexibility in choosing location for each horn (by sliding) while
ensured the whole system was stable for experiment. The width of the support structure was
limited by the width of the anechoic chamber as seen in Figure 3.1(a).
(a) SAR
(b) Agilent PNA
Figure 3.1: Synthetic aperture radar and Agilent PNA (VNA)
The anechoic lab consists of a chamber and an open space and both of which were used
for the imaging experiments. Wall and floor of the chamber are filled with absorptive material
which reduces reflection and limit noise from outside environment. The VNA available in the
lab was the 2-ports Agilent N5230A PNA-L Network Analyser (Figure 3.1(b)) which has a
21
ECal module. This module was used to pre-calibrate system noise and delay of external cables.
Horn antennas were borrowed from The Defence Science and Technology Organisation (DSTO).
To connect the horn antenna to the VNA port and ECal module, two N-type cables (5 metres
each) and several N-to-SMA adaptors were borrowed from EEE School laboratory.
(a) ECal module
(b) Horn antennas
Figure 3.2: ECal module and DSTO horn antennas
3.2
Experiment with Patch Antenna
A cross-polarised patch antenna with dual feeds as seen in Figure 3.3(a) was used to familiarise
with the VNA. Main aim of the experiment was to measure the forward voltage gain S12 (or
S21 ) of the antenna using VNA when input signal was put into one feed and output signal was
taken out of other feed. The expected behaviour of the response is low gain in frequency range
of operation with narrow bandwidth and higher gain for other frequencies. The capabilities
of the VNA (Rohde & Schwarz ZVL model - Figure 3.3(b)) such as scattering measurement,
stepped-frequency signal generation or file saving were tested.
(a) Patch Antenna
(b) Rohde & Schwarz ZVL VNA
Figure 3.3: Patch antenna and R&S ZVL VNA
22
Data collected from VNA was used for offline processing using MATLAB to reproduce
frequency response plot and to find impulse response using Inverse Fourier Transform. Plots
are attached in Figures 3.4 and 3.5.
Frequency response (Magnitude) of the cross−polarised patch antenna
Frequency response (Phase) of the cross−polarised patch antenna
0
0
−10
−10
Phase Response angle(H) (rad)
Magnitude Response |H| (dB)
−5
−15
−20
−25
−30
−35
−20
−30
−40
−50
−60
−40
−70
−45
0.5
1
1.5
Frequency (Hz)
2
2.5
0.5
1
1.5
Frequency (Hz)
9
x 10
(a) Amplitude
2
2.5
9
x 10
(b) Phase
Figure 3.4: Forward voltage gain for a cross-polarised patch antenna
Delay profile of the cross−polarised patch antenna
X: 3.339e−009
Y: 0.1682
0.16
0.14
h = ifft(H)
0.12
0.1
0.08
0.06
0.04
0.02
0
0.2
0.4
0.6
Time (s)
0.8
1
−7
x 10
Figure 3.5: Impulse response for a cross-polarised patch antenna
It could be seen from the result that the operation bandwidth is narrow, with a destructive
resonance at around 1.5GHz. The antenna also has linear phase property which represents
approximate equal delays between two ports of the VNA for the used frequency range. The
impulse response gave the delay τ ≈ 3.34ns or equivalent air gap of 1m. The results proved
the feasibility of using the reflection measurement for 1-D imaging purpose.
3.3
Simulation
To test the proposed SA configurations and processing methods, a MATLAB-based simulator
was built. A reference response with no phase shift was extracted from the real experiment.
23
For each transmitter and receiver location, a simulated reflection measurement was created by
phase-shifting reference response. The shifting amount is corresponding to the electromagnetic
distance from transmitter to receiving antenna via simulated object. The simulated responses
were then used for generating image of region of interest. The variable parameters in the
simulator are: size of region of interest, length and element spacing for SAR and object location.
Other parameters for stepped-frequency signal were taken from real data. The simulation was
run using parameters that are relevant to the real SAR and conducted experiments. Table 3.1
lists values that were used in the simulation of which the result is attached in Figures 3.6 and
3.7.
Parameters
Value
Note
1.5GHz (2.5 → 4GHz)
10MHz
Fixed
Fixed
151
30m
Fixed
Fixed
Pixel size
Size of ROI
SAR aperture
1cmx1cm
1.6mx6m
1.6m
Fixed
Variable
Variable
Array element spacing ∆
Object location
0.04m
(1m,3m)
Variable
Variable
Bandwidth B
Frequency step δf
No. of frequencies
Range of unambiguity c/δf
Table 3.1: Simulation parameters
The following could be observed from the simulation results:
• The frequency domain implementation gave smoother down range and cross range profiles
while the time domain technique suffered from quantisation effect which made the profiles
look discrete.
• The cross range resolution for monostatic configuration was smaller compared to it of
bistatic configuration. This result can be observed in both time domain and frequency
domain implementations: 3dB width of 10cm for monostatic and of 15cm for bistatic
approximately.
• The chosen pixel size (1cmx1cm) is reasonable since it is much smaller than both range
and cross range resolutions.
The simulation result proved the feasibility of using designed system parameters for real
experiments. The frequency domain implementation was used for real imaging experiment due
to its advantage over time domain implementation.
24
monostatic, time domain processing, object is at x = 1.00m y = 3.00m
Range profile at x = 100 cm
monostatic, frequency domain processing, object is at x = 1.00m y = 3.00m
Cross Range profile at y = 301 cm
0
0
−2
20
Range profile at x = 100 cm
Cross Range profile at y = 300 cm
0
0
−2
20
−4
−4
40
−10
−12
100
−10
−8
80
−10
−14
120
−16
140
160
−16
140
−18
100
200
300
Range (cm)
400
500
−15
280
600
(a) Monostatic - 2D image
290 300 310
y [Range (cm)]
−20
50
320
160
100
150
200
x [Cross Range (cm)]
(b) Monostatic - profiles
bistatic, time domain processing, object is at x = 1.00m y = 3.00m
Range profile at x = 100 cm
100
200
300
Range (cm)
400
500
−15
280
600
290 300 310
y [Range (cm)]
−20
80
320
Range profile at x = 100 cm
25
Cross Range profile at y = 300 cm
0
0
−2
20
−4
−4
40
I(x) (dB)
I(y) (dB)
−8
80
−10
−12
100
−10
−8
80
−10
−12
100
−10
−14
120
−6
−5
60
I(y) (dB)
Cross−Range (cm)
−6
−5
60
I(x) (dB)
40
Cross−Range (cm)
90 100 110 120
x [Cross Range (cm)]
(b) Monostatic - profiles
bistatic, frequency domain processing, object is at x = 1.00m y = 3.00m
−2
20
−18
(a) Monostatic - 2D image
Cross Range profile at y = 289 cm
0
0
−14
120
−16
140
160
−10
−12
100
−14
120
I(x) (dB)
80
I(x) (dB)
−8
−6
−5
60
I(y) (dB)
−5
60
Cross−Range (cm)
−6
I(y) (dB)
Cross−Range (cm)
40
−16
140
−18
100
200
300
Range (cm)
400
500
(c) Bistatic - 2D image
600
−15
260
280 300 320
y [Range (cm)]
340
−20
0
50 100 150 200
x [Cross Range (cm)]
(d) Bistatic - profiles
Figure 3.6: Simulation result for time domain implementation
160
−18
100
200
300
Range (cm)
400
500
(c) Bistatic - 2D image
600
−15
280
290 300 310
y [Range (cm)]
320
−20
60
80 100 120 140
x [Cross Range (cm)]
(d) Bistatic - profiles
Figure 3.7: Simulation result for frequency domain implementation
Chapter 4
Imaging Experiments and Results
Once the design phase was completed and simulation gave the expected results, the experiments
to image the objects within a region of interest were carried out. This chapter details the
conducted experimental set up and results for the cases of no wall and with a wall. Various
wall materials were investigated, including pin-up board, contructed wall using plaster boards
and real office wall.
4.1
Experimental Procedure
For each experiment, the following steps were done in order to generate a 2-D image of the
room:
• Calibrate the system by using Agilent Electronic Calibration Module and linear regression
technique.
• Measure and record physical locations of the wall and the objects in the controlled environment. These figures were used to evaluate the performance of the imaging radar.
• Move the horn antennas to form a Synthetic Aperture. The fashion of moving the horns
was based on the chosen configuration: monostatic or bistatic. For each movement, the
response of the environment was measured and stored into data storage.
• Apply frequency domain implementation of focusing to generate a 2-D image using
recorded data. MATLAB programs were written for this step and Sample code used
can be found in A.
• Analyse the imaging result, including the appearance of objects in the image and range
and cross range profiles at different coordinates. The range and cross range resolutions
were compared against the theory and simulation results.
Objects are paint cans with cross section of diameter of 23cm. The parameters used for the
experiments are listed in Table 4.1.
26
Parameters
Value
Bandwidth B
Frequency step δf
1.5GHz (2.5 → 4GHz)
10MHz
No. of frequencies
Range of unambiguity c/δf
Power at transmitting port
151
30m
10dBm
IF bandwidth
Samples for averaging
10kHz
100
Signal sweeping time
Pixel size
Size of ROI
14.745ms
1cmx1cm
1.6mx6m
SAR aperture D
Array element spacing ∆
Bistatic case
1.6m
0.04m
am = 0m or 0.8m or 1.6m, bm = 0.16 → 1.6m
Pseudo-monostatic case
am = 0 → 1.44m, bm = am + 0.16m
Table 4.1: Experimental parameters
The experiments for no wall case and the pin-up board were conducted in the anechoic
chamber. The experiments for constructed wall were conducted in Final Year Lab. Two offices
in Engineering Maths Building were used for the office wall experiments.
4.2
4.2.1
Imaging of objects without wall
1-object imaging
The images attached in Figure 4.1 are result of processing in frequency domain with two different
SA configurations. The object is correctly located in imaging result for pseudo-monostatic
configuration but not for bistatic case.
27
2−D projection, object is at (70, 375)
20
20
40
40
Cross−Range (cm)
Cross−Range (cm)
2−D projection, object is at (70, 375)
60
80
100
60
80
100
120
120
140
140
160
100
200
300
Range (cm)
400
500
160
600
100
(a) Pseudo-monostatic configuration
200
300
Range (cm)
400
500
600
(b) Bistatic configuration
Figure 4.1: Experimental result for one object at (70cm, 375cm) in the anechoic chamber
By locating the peak’s coordinates of the image in pseudo-monostatic case, the down range
and cross range profiles can be extracted as in Figure 4.2. It is noted that the x direction
represents cross range and y direction represents down range.
Cross Range profile at y = 364cm
0
−5
−5
−10
−10
I(x) (dB)
I(y) (dB)
Range profile at x = 71cm
0
−15
−15
−20
−20
−25
−25
−30
0
100
200
300
y [Range (cm)]
400
−30
10
500
20
(a) Down-range profile
30
40
50
60
70
x [Cross Range (cm)]
80
90
100
110
(b) Cross-range profile
Figure 4.2: Down-range and cross-range profiles at peak location
The physical measurement prior to the experiment gave the location of the object: (70cm,
375cm). However, the location of the peak in the final image is at (71cm, 364cm). If the image
resolution is considered as the two-way 3dB width of the range or cross range profile which is
the distance between -3dB points, Figure 4.2 gives:
Range resolution ∆y ≈ Cross-range resolution ∆x ≈ 10cm
These figures are comparable to or even better than expected results:
∆ytheory =
c
= 10cm
2B
28
and
λ
≈ 20cm for f = 3GHz
D
The real object location falls into 3dB region down from the peak. Experimental results were
as expected and showed that processing in frequency domain was the right approach for the
∆xtheory ≈ y
imaging implementation in the following stages of the project.
4.2.2
2-object imaging
This section presents the result for similar experiment with the presence of two objects at two
different locations. It is noticed that due to the effect of antenna array beam pattern in the
case of same down range (Figure 4.3) or the effect of attenuation in power reflected in the case
of different down ranges (4.6(a)), one object appears stronger in the image compared to the
other. For the experiment of two objects at same cross range (Figure 4.6(b)), one object cannot
be resolved due to the fact that transmitting waves were mostly reflected by the object in front
of it and hence the reflection from this object became insignificant.
2−D projection, objects are at (22,375) and (80, 375)
20
Cross−Range (cm)
40
60
80
100
120
140
160
100
200
300
Range (cm)
400
500
600
Figure 4.3: Experimental result for two objects at (80cm, 375cm) and (22cm, 275cm) in the
anechoic chamber (pseudo-monostatic configuration)
Range and cross range profiles at the local peaks when two objects are at the same range
are attached in Figures 4.5 and 4.4. The range and cross range resolutions of the image can
be computed and they are approximately the same as in the case of one object imaging. The
presence of the wall between the SA and the object is an increase in the level of sidelobes in
the cross range profile.
29
Range profile at x = 15cm
0
−5
−5
−10
−10
I(y) (dB)
I(y) (dB)
Range profile at x = 78cm
0
−15
−15
−20
−20
−25
−25
−30
0
100
200
300
y [Range (cm)]
400
−30
500
(a) Down range profile – object at (80cm, 375cm)
0
100
200
300
y [Range (cm)]
400
500
600
(b) Down range profile – object at (22cm, 375cm)
Figure 4.4: Down range profiles at peak location for image of 2 objects
Cross Range profile at y = 362cm
0
−5
I(x) (dB)
−10
−15
−20
−25
−30
0
20
40
60
80
100
x [Cross Range (cm)]
120
140
160
Figure 4.5: Cross range profile at peak location for image of 2 objects
2−D projection, objects are at (80,303) and (80, 379)
20
20
40
40
Cross−Range (cm)
Cross−Range (cm)
2−D projection, objects are at (24,239) and (70, 369)
60
80
100
60
80
100
120
120
140
140
160
100
200
300
Range (cm)
400
500
160
600
(a) Objects at (24cm, 239cm) and (70cm, 369cm)
100
200
300
Range (cm)
400
500
600
(b) Objects at (80cm, 303cm) and (80cm, 379cm)
Figure 4.6: Imaging results for two objects at different ranges (pseudo-monostatic)
30
4.3
Imaging result for a pin-up board
To synthesise a wall in an controlled environment, a pin-up board with thickness of 1cm was
used for setting up the experiment. A paint can was put at known location in the room and
the VNA was also in the region of imaging. The imaging results for both SA configurations
are attached in Figures 4.7 and 4.8. In each figure, the 2-D image of the region of interest in
linear scale and the range and cross range profiles at the peak location are shown. It is noted
that the pin-up board was not big enough to form a full wall, but the object was arranged so
that it could not be seen by the SA.
The 2-D images for both configurations and their range and cross range profiles were analysed
and the following results are noticed:
• For pseudo-monostatic configuration
– The imaging peak is at (76cm, 239cm) which is relatively close to the real object
location (80cm, 255cm)
– The whole wall is resolved in the final image and its image is thicker than 1cm.
– The range and cross range resolutions are both approximately 10cm.
– The imaging value of the wall is about 5dB lower than it at the object location.
• For bistatic configuration - static horn was put at x = 0cm
– The imaging peak is at (82cm, 248cm) which is relatively close to the real object
location (86cm, 255cm)
– Only part of the wall is resolved in the final image and its image is thicker than 1cm.
The resolved part was directly in front of the static horn. The characteristics of the
radiation pattern of the horn antenna could be used to explain this pattern. Due to
the fall-off in the radiation and reception patterns, the reflection off the wall as seen
by the receiver was only strong when the two horn antennas were spatially close.
– The range resolution is approximately 10cm while the cross range resolution is
roughly 15cm, which is lower compared to the pseudo-monostatic case.
• The blurry region appeared in the 2-D images of both configurations, behind the wall and
the object is the image of the VNA.
31
2−D image, object is (86cm, 255cm), pin board is at y = 135cm
20
20
40
40
Cross−Range (cm)
Cross−Range (cm)
2−D projection, object is (80cm, 255cm), wall is at y = 127 cm
60
80
100
60
80
100
120
120
140
140
160
160
100
200
32
300
Range (cm)
400
500
600
100
200
(a) 2D image
Cross Range profile at y = 239 cm
Range profile at x = 82 cm
−5
−5
−5
−5
−10
−10
−10
−10
−15
I(x) (dB)
0
I(y) (dB)
0
I(x) (dB)
0
−15
−20
−20
−20
−25
−25
−25
−25
0
50
100
150
200
250
y [Range (cm)]
300
350
(b) Range profile
400
450
−30
40
600
−15
−20
−30
500
Cross Range profile at y = 248cm
0
−15
400
(a) 2D image
Range profile at x = 76 cm
I(y) (dB)
300
Range (cm)
50
60
70
80
90
100
x [Cross Range (cm)]
110
120
130
(c) Cross range profile
Figure 4.7: Result for pin-up board – Pseudo-monostatic
−30
0
100
200
300
y [Range (cm)]
400
(b) Range profile
500
−30
0
50
100
x [Cross Range (cm)]
(c) Cross range profile
Figure 4.8: Result for pin-up board – Bistatic
150
4.4
4.4.1
Imaging result for a constructed wall
Set up
A wall using plaster boards and timber studs was constructed for this project. The plaster
boards have thickness of 1cm. The location of the stud in the wall were chosen so that the
constructed wall would be the same as a plaster wall in commercial use.
The following experiments were conducted to study the performance of the imaging radar
with the constructed wall:
• Imaging an open space with no wall and no object.
• Imaging a region with no wall.
• Imaging a region with a single layer wall formed by one plaster board.
• Imaging a region with a double layer wall formed by two plaster boards and a stud frame.
The experimental set-up and controlled environment could be seen in Figure 4.9.
(a) No wall
(b) One plaster board
(c) Stud frame
(d) Two plaster boards
Figure 4.9: Experimental set up for constructed wall
33
4.4.2
Empty room
The result for the imaging experiment for the empty room is shown in Figure 4.10. The SA
configuration used for this experiment was pseudo-monostatic. An unknown repeating pattern
which appeared in the imaging result has not been explained. It could be due to the cluster
around the test area or the internal structure of the room.
2−D image, empty room, log scale
20
20
40
40
Cross−Range (cm)
Cross−Range (cm)
2−D image, empty room
60
80
100
60
80
100
120
120
140
140
160
100
200
300
Range (cm)
400
500
160
600
(a) 2D image - linear
100
200
300
Range (cm)
400
500
600
(b) 2D image - logarithmic
Figure 4.10: Imaging result for an empty controlled room
4.4.3
No wall
The result for the imaging experiment when there were two paint cans in the region to image
is shown in Figure 4.11. The pattern which appeared in the above experiment could be again
observed in the 2-D imaging result. However, the pattern was stronger in bistatic configuration
compared to monostatic case.
2−D image, objects are at (140,383) and (63,225), log scale
20
20
40
40
Cross−Range (cm)
Cross−Range (cm)
2−D image, objects are at (140,383) and (63,225), log scale
60
80
100
60
80
100
120
120
140
140
160
100
200
300
Range (cm)
400
500
160
600
(a) 2D image – pseudo monostatic
100
200
300
Range (cm)
400
500
600
(b) 2D image – bistatic
Figure 4.11: Imaging result for controlled environment without wall
The range and cross range profiles of the coordinates of the peaks for both configurations
are shown in Figures 4.12 and 4.13.
34
Cross Range profile at y = 213cm
Cross Range profile at y = 213cm
−5
−5
−5
−5
−10
−10
−10
−10
−15
−15
I(x) (dB)
0
I(y) (dB)
0
−15
−15
−20
−20
−20
−20
−25
−25
−25
−25
−30
0
100
200
300
400
y [Range (cm)]
500
−30
20
600
(a) Range profile - Object 1
30
40
50
60
x [Cross Range (cm)]
70
80
−30
90
(b) Cross range profile - Object 1
Range profile at x = 131 cm
0
100
200
300
400
y [Range (cm)]
500
(a) Range profile - Object 1
Cross Range profile at y = 371cm
−30
600
−5
−5
−5
−10
−10
−10
−10
−15
I(x) (dB)
−5
I(y) (dB)
0
−15
−20
−20
−20
−25
−25
−25
−25
100
200
300
400
y [Range (cm)]
500
(c) Range profile - Object 2
600
−30
0
20
40
60
80
100
x [Cross Range (cm)]
120
140
160
(d) Cross range profile - Object 2
Figure 4.12: Imaging profiles for pseudo-monostatic configuration
−30
0
100
200
300
400
y [Range (cm)]
60
80
100
x [Cross Range (cm)]
120
140
−15
−20
0
40
Cross Range profile at y = 370cm
0
−30
20
Range profile at x = 131 cm
0
−15
0
(b) Cross range profile - Object 1
0
I(x) (dB)
35
I(y) (dB)
Range profile at x = 60 cm
0
I(x) (dB)
I(y) (dB)
Range profile at x = 60 cm
0
500
(c) Range profile - Object 2
600
−30
0
20
40
60
80
100
x [Cross Range (cm)]
120
140
160
(d) Cross range profile - Object 2
Figure 4.13: Imaging profiles for bistatic configuration
The results observed from 2-D images, range and cross range profiles are as follows:
• The physical measurement of the test area gave the location of the paint cans: (140cm,
383cm) and (63cm, 225cm). The locations of the peaks for pseudo-monostatic configuration are (131cm, 371cm) and (60cm, 213cm). The locations of the peaks for bistatic
configuration are (131cm, 370cm) and (60cm, 213cm).
• The range and cross range resolutions at the peaks do match with the results of the
simulation and previous experiments.
• The second object (further from the SA) is just resolvable since it stays closer to the
unknown patterns as discussed above.
4.4.4
One plaster board
The experimental result for when there was a single wall of one plaster board is shown in
Figure 4.14. The down range and cross range profiles at the coordinates of the peaks for both
SA configuration are attached in Figure 4.15 and 4.16.
20
40
40
Cross−Range (cm)
20
60
80
100
60
80
100
120
120
140
140
160
Cross−Range (cm)
2−D image, objects are at (128,333) and (59,203), log scale
100
200
300
Range (cm)
400
500
160
600
200
300
Range (cm)
400
500
(b) 2D image – monostatic – log
2−D image, objects are at (128,333) and (59,203), linear scale
2−D image, objects are at (128,333) and (59,203), log scale
20
20
40
40
60
80
100
80
100
120
140
140
100
200
300
Range (cm)
400
500
160
600
(c) 2D image – bistatic – linear
100
200
300
Range (cm)
400
500
(d) 2D image – bistatic – log
Figure 4.14: Imaging result for one plaster board wall
36
600
60
120
160
100
(a) 2D image – monostatic – linear
Cross−Range (cm)
Cross−Range (cm)
2−D image, objects are at (128,333) and (59,203), linear scale
600
Cross Range profile at y = 196cm
Cross Range profile at y = 196cm
−5
−5
−5
−5
−10
−10
−10
−10
−15
−15
I(x) (dB)
0
I(y) (dB)
0
−15
−15
−20
−20
−20
−20
−25
−25
−25
−25
−30
0
100
200
300
400
y [Range (cm)]
500
−30
600
(a) Range profile - Object 1
0
20
40
60
80
100
x [Cross Range (cm)]
120
140
−30
160
(b) Cross range profile - Object 1
Range profile at x = 127 cm
0
100
200
300
400
y [Range (cm)]
500
(a) Range profile - Object 1
Cross Range profile at y = 327cm
−30
600
−5
−5
−5
−10
−10
−10
−10
−15
I(x) (dB)
−5
I(y) (dB)
0
−15
−20
−20
−20
−25
−25
−25
−25
100
200
300
400
y [Range (cm)]
500
(c) Range profile - Object 2
600
−30
0
20
40
60
80
100
x [Cross Range (cm)]
120
140
160
(d) Cross range profile - Object 2
Figure 4.15: Imaging profiles for pseudo-monostatic configuration
−30
0
100
200
300
400
y [Range (cm)]
60
80
x [Cross Range (cm)]
100
120
−15
−20
0
40
Cross Range profile at y = 327cm
0
−30
20
Range profile at x = 127 cm
0
−15
0
(b) Cross range profile - Object 1
0
I(x) (dB)
37
I(y) (dB)
Range profile at x = 66 cm
0
I(x) (dB)
I(y) (dB)
Range profile at x = 56 cm
0
500
(c) Range profile - Object 2
600
−30
0
20
40
60
80
100
x [Cross Range (cm)]
120
140
160
(d) Cross range profile - Object 2
Figure 4.16: Imaging profiles for bistatic configuration
It is noted that the static horn was located at x = 80cm which was the middle of the SA. It
could be observed from the above figures that:
• The single wall appears in the results for both configurations. However, only part of the
wall which was directly in front of the static horn is resolved in the 2-D image. This result
matches the experimental result with the pin-up board and could again be explained by
using the radiation pattern of the horn antennas.
• The objects were physically located at (128cm, 333cm) and (59cm, 203cm). However,
the peak locations for pseudo-monostatic configuration are (127cm, 327cm) and (56cm,
196cm), and for bistatic configuration are (127cm, 327cm) and (66cm, 196cm). There
is a difference between the real locations and the positions of object as seen in imaging
result. They are relatively close or in other words, the imaging result could provide a
rough estimation of the real location of the objects.
• The resolutions in both range and cross range directions match with previous experimental
results.
• The presence of the wall increases the level of sidelobes in both range and cross range
profiles. The down-range sidelobes of the wall could be seen in the logarithmic 2-D image
in Figure 4.14.
• There is a confusing near field pattern close to the SA for both configurations. The
processing steps were used under an assumption that the target range is of far field as
seen by the SA. However, for the effective aperture size D = 1.6m and the wavelength
λ = 10cm (f = 3GHz), the boundary range between far field and near field is:
Ro = D 2 /λ = 25.6m
which is much higher than the size of the region to image. Near-field artifact or cross-talk
between horn antennas could potentially be the reason for this pattern but it has not
been explained up to this stage of the project.
4.4.5
Two plaster boards
The experimental results for the double wall formed by two plaster boards are shown in Figures
4.17, 4.18 and 4.18. For the bistatic configuration, the static horn was located at x = 80cm
which was the middle of the SA.
The significant aspects observed from the results are as follows:
• The objects stay within 3dB region from the imaging peaks, as discussed for previous
experiments.
38
• Two layers of the wall are resolved in the final image and the studs can be located in the
result of monostatic configuration. The studs are at the imaging holes within the wall as
seen in Figure 4.17(b).
• The presence of the wall introduces an increase in the sidelobe level of both range and
cross range profiles. The resolutions in both axes are comparable to previous simulation
and experimental results.
• Unknown near field pattern again appears in the image for the pseudo-monostatic configuration.
20
40
40
Cross−Range (cm)
20
60
80
100
60
80
100
120
120
140
140
160
Cross−Range (cm)
2−D image, objects are at (120,310) and (55,205), log scale
100
200
300
Range (cm)
400
500
160
600
200
300
Range (cm)
400
500
(b) 2D image – monostatic – log
2−D image, objects are at (120,310) and (55,205), linear scale
2−D image, objects are at (120,310) and (55,205), log scale
20
20
40
40
60
80
100
80
100
120
140
140
100
200
300
Range (cm)
400
500
160
600
(c) 2D image – bistatic – linear
100
200
300
Range (cm)
400
500
(d) 2D image – bistatic – log
Figure 4.17: Imaging result for two plaster boards wall
39
600
60
120
160
100
(a) 2D image – monostatic – linear
Cross−Range (cm)
Cross−Range (cm)
2−D image, objects are at (120,310) and (55,205), linear scale
600
Range profile at x = 50 cm
Cross Range profile at y = 198cm
0
Range profile at x = 50 cm
−2
−4
−4
−4
−4
−6
−6
−6
−6
−8
−8
−8
−8
−10
−10
I(x) (dB)
−2
I(y) (dB)
−2
−10
−10
−12
−12
−12
−12
−14
−14
−14
−14
−16
−16
−16
−16
−18
−18
−18
−18
−20
0
100
200
y [Range (cm)]
−20
300
0
50
100
x [Cross Range (cm)]
−20
150
0
100
200
y [Range (cm)]
(a) Object 1
Range profile at x = 119 cm
−20
300
Range profile at x = 120 cm
−4
−6
−6
−6
−6
−8
−8
−8
−8
−10
I(x) (dB)
−2
−4
I(x) (dB)
−2
−4
I(y) (dB)
40
−2
−4
−10
−12
−12
−12
−14
−14
−14
−14
−16
−16
−16
−16
−18
−18
−18
−18
−20
−20
−20
100
200
300
y [Range (cm)]
400
0
150
−10
−12
0
50
100
x [Cross Range (cm)]
Cross Range profile at y = 309cm
0
0
−2
−10
0
(a) Object 1
Cross Range profile at y = 308cm
0
0
I(y) (dB)
Cross Range profile at y = 197cm
0
0
−2
I(x) (dB)
I(y) (dB)
0
50
100
150
x [Cross Range (cm)]
200
(b) Object 2
Figure 4.18: Imaging profiles for pseudo-monostatic configuration
0
100
200
300
y [Range (cm)]
400
−20
0
50
100
150
x [Cross Range (cm)]
200
(b) Object 2
Figure 4.19: Imaging profiles for bistatic configuration
4.5
Imaging result for a real office wall
An office wall was used for the experiments as an attempt to test the imaging radar for a real
environment. The wall is made of plaster boards but its internal structure was left unknown.
The arrangement of the experiment is shown in Figure 4.20.
(a) SAR and wall
(b) RF targets
Figure 4.20: Experimental set-up for a real office wall
The 2-D images of various configurations of the SA are shown in Figure 4.21. The locations
of the static horn antenna in the experiments are shown in the captions. The result presented
match with the previous experimental results about the wall being partly resolved for bistatic
configuration. The first object (closer to the SA) were clearly resolved in all images but the
second object could not be clearly seen due the noise and clusters. The blurry pattern behind
the second object was assumed to be the image of the second wall of the office since its location
matches with the actual location of the wall.
41
2−D log image, objects are at (86,189) and (143,313)
20
20
40
40
Cross−Range (cm)
Cross−Range (cm)
2−D log image, objects are at (86,189) and (143,313)
60
80
100
60
80
100
120
120
140
140
160
50
100
150
200
250
Range (cm)
300
350
160
400
(a) Bistatic (static horn at 0cm)
50
40
40
Cross−Range (cm)
Cross−Range (cm)
20
60
80
100
140
300
350
400
100
140
200
250
Range (cm)
350
80
120
150
300
60
120
100
200
250
Range (cm)
2−D logarithmic image, objects are at (86,189) and (143,313)
20
50
150
(b) Bistatic (static horn at 80cm)
2−D log image, objects are at (86,189) and (143,313)
160
100
160
400
(c) Bistatic (static horn at 162cm)
0
50
100
150
200
250
Range (cm)
300
350
400
(d) Pseudo-monostatic
Figure 4.21: Imaging result for a real office wall
In an attempt to combine various experimental results using different configurations, the
image of the region of interest was extracted as seen in Figures 4.22 and 4.23. The individual
2-D images of the region using the bistatic configurations as seen above were combined and
superimposed by adding or multiplying. These operations are the equivalence of the crosscorrelation operation between the images. By doing so, the quality of the images increased
which can be reflected in both 2-D images and range and cross range profiles. The range and
cross range resolutions of the images formed by combining multiple images are much better
compared to the original individual images.
42
bistatic combination, objects are at (86,189) and (143,313)
20
20
40
40
Cross−Range (cm)
Cross−Range (cm)
bistatic combination, objects are at (86,189) and (143,313)
60
80
100
60
80
100
120
120
140
140
160
160
50
100
150
43
200
250
Range (cm)
300
350
400
50
100
150
(a) 2D image
Cross Range profile at y = 185cm
Range profile at x = 85 cm
0
−2
−6
400
Cross Range profile at y = 185cm
0
0
−10
−10
−20
−20
−30
−30
I(y) (dB)
I(x) (dB)
−10
−15
−12
I(x) (dB)
−10
−8
I(y) (dB)
350
−5
−4
−40
−40
−50
−50
−60
−60
−70
−70
−20
−14
−16
−25
−18
−20
300
(a) 2D image
Range profile at x = 85 cm
0
200
250
Range (cm)
0
50
100
150
200
250
y [Range (cm)]
300
(b) Range profile
350
400
−30
0
20
40
60
80
100
x [Cross Range (cm)]
120
140
(c) Cross range profile
Figure 4.22: Combine the results by adding
160
−80
20
40
60
80
100 120 140
y [Range (cm)]
160
180
(b) Range profile
200
220
−80
70
75
80
85
90
95
x [Cross Range (cm)]
100
105
(c) Cross range profile
Figure 4.23: Combine the results by multiplying
110
4.6
Remarks
Various experiments using different wall materials were conducted and their results were analysed. The significant results from the experiments can be summarised as follows:
• Objects were resolved at correct locations in the final image.
• Down range and cross range resolutions were comparable to theoretical value and simulation result. The down range resolution for both SA configurations was 10cm. The cross
range resolution for the pseudo-monostatic configuration was approximately 10cm while
its value for the bistatic configuration was 15cm.
• The wall was partly resolved in the result of the experiments using bistatic configuration
for the SA. Full wall appeared in the result for pseudo-monostatic case.
• The processing technique and propagation model performed well for single and double
layer walls.
• There was a confusing near-field pattern close to the SA.
44
Chapter 5
Project Management
This chapter will give a brief look at some management aspects of the project, which include
schedule management, risk mitigation and budget review.
5.1
Work Plan and Schedule Revisit
The proposed work has been successfully completed as planned. There is no significant change
to the Gantt chart during the execution phase of the project. Previous versions of the project
schedule could be found in [10],[11]. There was a delay in the initial imaging experiments in
Semester A due to the late arrival of the horn antennas. The schedule was adjusted accordingly
and time delay was compensated by the work during the mid-year break. One optional task
was completed which is conducting the experiment with various wall materials. The up-to-date
Gantt chart of the project is shown in Figure 5.1.
Table 5.1 shows the status of planned tasks up to this stage of the project. All experiments
in the chamber and real environment were conducted and met the schedule.
The tasks were allocated to team member based on the availability and the experience of
members. Weekly informal and formal meetings were held for planning and reporting progress.
More on initial group role allocation and auditing plan for this project can be found in [10].
Due to the main focus of the second stage of the project was software and signal processing,
task and group roles were slightly modified. The experiments were conducted by the whole
team and tasks for software development and result analysis were equally divided between
team members.
45
46
Figure 5.1: Gantt Chart
Planned Task
Status
Proposal
System design
Complete
Complete
System simulation
1-D imaging
2-D imaging in anechoic chamber
Complete
Complete
Complete
2-D imaging in controlled environments
System testing and validation
Complete
Complete
Experiment with various wall materials
3-D imaging
Complete
Incomplete
Table 5.1: Project progress
5.2
Risk Management
5.2.1
Risk occurred
5.2.1.1
Delay in material procurement and construction
In Semester A, horn antennas were chosen for the configuration of SA and it was decided that
two horns would be borrowed from DSTO. However, there was a short delay in getting the
antennas delivered which resulted in an adjustment in the project schedule.
In Semester B, a wall formed of plaster boards was built in an attempt to test the system
with a real wall. There was a long delay (about one week) for the plaster boards to be ordered
and delivered to campus. Due to various reasons, the wall construction using these plaster
boards was also delayed for about one week. However, the delays were quickly compensated
since the project was ahead of schedule.
5.2.1.2
Subversion blackout
Since the project dealt with large amount of data and programs, it was decided in planning phase
that Electrical and Electronic Engineering (EEE) School’s Subversion Control (SVN) repository would be used for managing and storing data. While it provided flexibility, SVN posed a
great risk in the case that it was corrupted or data was lost or inaccessible. It did happen twice
over two weekends (end of week 10 and 11, Semester A) when there was unexpected power
outage in Engineering South building, where the SVN server is housed. The processing of data
was therefore terminated until server was manually rebooted in the following Mondays. The
solution for the following stage was to have working copy stored offline to be accessed if needed.
47
5.2.1.3
Anechoic chamber and VNA unavailable
Since the access of the anechoic chamber and the VNA is managed by booking, it was booked
out and could not be accessed for some particular days. This caused the delay for conducting
experiments which can only be done when both lab and instrument are available. This risk did
cause a short delay for the project, but did not affect the overall progress.
5.2.1.4
Lack of technical knowledge
Due to the inexperience of team members with reflection measurement calibration and radar
imaging technique, handling related challenges was time-consuming. This caused the project to
stay behind schedule most of the time in semester A. However, when the problem was solved,
delay could quickly be compensated and tasks were back on track.
5.3
Budget review
Table 5.2 lists all up-to-date expenses for this project. All numerical figures are finalised and
there will be no further purchases being made for this project. The real figure is slightly
lower than proposed cost since most RF components and instrument were borrowed and not
externally purchased.
Item
Cost
Note
SAR mounting
N-type cable x 2
-
Built by workshop
Borrowed from phased array lab
VNA
Horn antennas
-
Borrowed from the anechoic chamber
Borrowed from DSTO
SMA male - N female adaptor x 4
RF targets
Wall construction
$83.06
Borrowed from EEE labs
Borrowed from the workshop
Material bought from Bunning
Exhibition printing
$10
Estimate cost
Total Cost
Proposed budget
$290
$93.06
Real cost
$93.06
Remaining budget
$500 - $93.06 = $406.94
Table 5.2: Budget review
48
Chapter 6
Conclusion
6.1
Future Work
The work done for this project has laid down the foundation so that extension or future work
can be performed:
1. Automatic 2D imaging
The experiments were manually conducted by sliding the horn antennas across the SAR.
The data from each measurement is acquired and stored for offline processing. The
process for getting image of ROI is consequently slow, although it does not take long
to generate the image once data is collected. The followings could be used to enable
automatic imaging using the same system and configuration:
• Stepper motors to quickly and automatically move the horn antennas along the SAR.
• Live data acquisition and signal processing by either remote controlling the VNA
by MATLAB using Local Area Network (LAN) or General Purpose Interface Bus
(GPIB), or using dedicated hardware and signal processing unit.
2. Setting up theory for near-field Synthetic Aperture Radar imaging
The experimental work has proved the difficulty of using far-field theory to explain beamforming artifacts in the final image. The theoretical background is needed to completely
explain the experiment result, which includes the disturbance in the image close to the
synthetic aperture or the presence of repeating pattern in the image of the empty room.
3. 3D imaging
In practical situation, it is not often useful to only see one 2-D slide of the scene. The
radar being able to generate a 3D image of objects behind the wall would be ideal for
many reasons. The following work can be done to achieve such requirement:
49
• Vary the height of current SAR to capture multiple 2D slides and reconstruct 3D
scene using 2D results.
• Change the focusing delay model to deal with third dimension (elevation direction)
and redesign mechanical structure for a two dimensional SA.
6.2
Project Outcomes
2-D image of the region of interest behind a wall has been successfully generated. The pseudomonostatic configuration results in better cross range resolution compared to bistatic case.
Experiments were conducted for various wall materials (pin-up board, plaster wall and office
wall) and the results prove the feasibility of through the wall imaging application using a
Synthetic Aperture Radar. The designed radar system is simple and of low cost, which can
be used for research application. The outcome of this project serves as a foundation for the
implementation of a more complex through-wall radar system.
50
Bibliography
[1] E. J. Baranoski, “Through-wall imaging: Historical perspective and future directions,”
Journal of the Franklin Institute, vol. 345, no. 6, pp. 556 – 569, 2008.
[2] A. Braga and C. Gentile, “An ultra-wideband radar system for through-the-wall imaging
using a mobile robot,” in Communications, 2009. ICC ’09. IEEE International Conference
on, pp. 1 –6, june 2009.
[3] F. Ahmad, M. Amin, and S. Kassam, “Synthetic aperture beamformer for imaging through
a dielectric wall,” Aerospace and Electronic Systems, IEEE Transactions on, vol. 41,
pp. 271 – 283, jan. 2005.
[4] M. Amin and F. Ahmad, “Wideband synthetic aperture beamforming for through-thewall imaging [lecture notes],” Signal Processing Magazine, IEEE, vol. 25, pp. 110 –113,
july 2008.
[5] S. Kidera, T. Sakamoto, and T. Sato, “A high-resolution 3-d imaging algorithm with
linear array antennas for uwb pulse radar systems,” in Antennas and Propagation Society
International Symposium 2006, IEEE, pp. 1057 –1060, july 2006.
[6] J. Zhu, Y. Hong, and L. Tao, “3d imaging algorithm and implement for through-wall
synthetic aperture radar,” in Circuits and Systems, 2009. MWSCAS ’09. 52nd IEEE International Midwest Symposium on, pp. 561 –564, aug. 2009.
[7] F. Ahmad, Y. Zhang, and M. Amin, “Three-dimensional wideband beamforming for imaging through a single wall,” Geoscience and Remote Sensing Letters, IEEE, vol. 5, pp. 176
–179, april 2008.
[8] W. M. Brown and L. J. Porcello, “An introduction to synthetic-aperture radar,” Spectrum,
IEEE, vol. 6, pp. 52 –62, sept. 1969.
[9] A. Currie, “Synthetic aperture radar,” Electronics Communication Engineering Journal,
vol. 3, pp. 159 –170, aug 1991.
[10] T. Bui and J. Rabig, “Stage 1 progress report - project: Through-the-wall imaging radar,”
tech. rep., Adelaide University, 2011.
51
[11] T. Bui, “Stage 2 progress report - project: Through-the-wall imaging radar,” tech. rep.,
Adelaide University, 2011.
52
Appendix A
MATLAB Code
This appendix lists out sample MATLAB programs used to process collected data and generate
the 2-D image. It also includes a segment of the typical measurement result collected using the
VNA.
Listing A.1: Sample measurement result
! Agilent Technologies , N5230A , US43500294 , A . 0 4 . 2 5
! Agilent N5230A : A . 0 4 . 2 5
! Date : Thursday , September 1 5 , 2011 1 5 : 4 8 : 1 3
! Correction : S11 ( Full 2 Port SOLT , 1 , 2 ) S21 ( Full 2 Port SOLT , 1 , 2 ) S12 ( Full 2 Port SOLT , 1 , 2 ) ←֓
S22 ( Full 2 Port SOLT , 1 , 2 )
! S2P File : M e a s u r e m e n t s : S11 , S21 , S12 , S22 :
# Hz S dB
R 50
250000000 0 −7.430793 e+000 −1.600484 e+002 −3.742124 e+001 1 . 1 1 5 3 3 1 e+002 −3.746042 e+001 ←֓
1 . 1 1 9 3 8 5 e+002 −6.741900 e+000 1 . 7 4 1 2 6 2 e+002
251000000 0 −9.430097 e+000 −1.599856 e+002 −3.682345 e+001 9 . 0 1 9 6 6 2 e+001 −3.680711 e+001 ←֓
8 . 9 8 2 1 2 1 e+001 −5.338504 e+000 −1.762050 e+002
252000000 0 −7.807305 e+000 −1.694729 e+002 −3.502031 e+001 7 . 3 9 1 9 7 4 e+001 −3.509335 e+001 ←֓
7 . 4 5 1 7 5 1 e+001 −4.510386 e+000 −1.626250 e+002
253000000 0 −8.758544 e+000 −1.584523 e+002 −3.319084 e+001 5 . 3 4 9 7 7 5 e+001 −3.321054 e+001 ←֓
5 . 3 5 3 4 8 2 e+001 −4.070266 e+000 −1.505258 e+002
254000000 0 −9.448528 e+000 −1.741781 e+002 −3.262580 e+001 3 . 5 2 5 9 7 2 e+001 −3.275987 e+001 ←֓
3 . 5 7 3 0 8 6 e+001 −4.045329 e+000 −1.358550 e+002
255000000 0 −8.199015 e+000 −1.650800 e+002 −3.151055 e+001 1 . 0 1 4 8 2 8 e+001 −3.151619 e+001 ←֓
1 . 0 6 9 7 3 4 e+001 −4.165150 e+000 −1.226444 e+002
256000000 0 −1.054701 e+001 −1.651589 e+002 −3.112403 e+001 −6.580334 e+000 −3.128778 e+001 ←֓
−6.677758 e+000 −4.720467 e+000 −1.096675 e+002
257000000 0 −8.813895 e+000 −1.712190 e+002 −3.048989 e+001 −3.088564 e+001 −3.044874 e+001 ←֓
−3.038239 e+001 −5.575526 e+000 −9.557133 e+001
258000000 0 −9.635010 e+000 −1.571554 e+002 −3.061926 e+001 −5.116904 e+001 −3.077199 e+001 ←֓
−5.148961 e+001 −7.355277 e+000 −8.485722 e+001
259000000 0 −1.014102 e+001 −1.730694 e+002 −3.013548 e+001 −6.606699 e+001 −3.009271 e+001 ←֓
−6.575909 e+001 −1.008304 e+001 −7.305842 e+001
260000000 0 −8.658289 e+000 −1.612389 e+002 −2.969870 e+001 −9.283371 e+001 −2.979784 e+001 ←֓
−9.288857 e+001 −1.628804 e+001 −6.644978 e+001
...
# The remaining was d e l i b e r a t e l y deleted !
53
Listing A.2: Calibration using linear regression
f u n c t i o n p = calibrati o n ( name , d )
% d i s a v e c t o r o f d i s t a n c e s between a n ten n a s
% name i s th e s t a r t i n g s t r i n g o f th e c a l i b r a t i o n data f i l e s
% p i s a 1 x2 v e c t o r c o n t a i n i n g th e system d e l a y p ( 2 ) and th e s p e ed o f
% wave i n th e a i r g a p 1/p ( 1 )
len = l e n g t h ( d ) ;
delays = z e r o s ( 1 , len ) ;
f o r i =1: len
fname = strcat ( name , num2str ( d ( i ) ) , ' . s2p ' ) ;
% open f i l e
fid = f o p e n ( fname , ' r ' ) ;
% r ea d data
data = textscan ( fid , '%f%f%f%f%f%f%f%f%f ' , ' H e a d e r l i n e s ' , 6 ) ;
f c l o s e ( fid ) ;
% e x t r a c t data
freq = data { 1 } ;
S21_magdB = data { 4 } ;
S 2 1 _ p h a s e D e g = data { 5 } ;
S21 = db2mag ( S21_magdB ) . ∗ exp (1 i ∗ S 2 1 _ p h a s e D e g ∗ p i / 1 8 0 ) ;
h_d = i f f t ( S21 ) ;
fs = max( freq ) − min ( freq ) ;
time = ( 1 : l e n g t h ( h_d ) ) / fs ;
h_abs = abs ( h_d ) ;
[ pks , locs ] = findpeaks ( h_abs , ' s o r t s t r ' , ' d es cen d ' ) ;
delays ( i ) = time ( locs ( 1 ) ) ;
end
f i g u r e , p l o t ( d / 1 0 0 , delays , ' ∗ r ' ) ;
xlim ( [ 0 . 9 ∗ min ( d ) /100 1 . 1 ∗ max( d ) / 1 0 0 ] )
ylim ( [ 0 , 1 . 1 ∗ max( delays ) ] ) , h o l d on , g r i d on ;
x l a b e l ( ' D i s t a n c e between a n ten n a s (m) ' )
y l a b e l ( ' Delay with no c a l i b r a t i o n ( s ) ' )
t i t l e ( ' Delay v s . S e p a r a t i o n ( September 26 th ) ' )
p = p o l y f i t ( d / 1 0 0 , delays , 1 ) ;
delays_t r ue = p ( 1 ) ∗ d /100 + p ( 2 ) ;
p l o t ( d / 1 0 0 , delays_true , '−ob ' )
str1 = s p r i n t f ( ' F i t t e d data : d e l a y = %.2d∗ d i s t a n c e + %.2d ' , p ( 1 ) , p ( 2 ) ) ;
l e g e n d ( ' C o l l e c t e d data ' , str1 , ' l o c a t i o n ' , ' South ' )
% regression coefficient
r = c o r r c o e f ( d / 1 0 0 , delays ) ;
t e x t ( 1 , 0 . 2 5 e −8 , s p r i n t f ( ' C o r r e l a t i o n c o e f f i c i e n t R = %.4 f ' , r ( 2 , 1 ) ) )
return
Listing A.3: Sample image processing program for an empty room
clear
close
all ;
all ;
54
clc ;
%% p er f o r m c a l i b r a t i o n
d = [ 5 3 93 137 1 8 6 ] ;
p1 = calibratio n ( ' S c a l ' , d ) ;
d = (0:4:144) ;
d_len = l e n g t h ( d ) ;
c = 3 e8 ;
N = 151;
freq = z e r o s ( 1 , N ) ;
H = z e r o s ( d_len , N ) ;
h = z e r o s ( d_len , N ) ;
f o r i = 1 : d_len
fname = strcat ( ' S e 2 ' , num2str ( d ( i ) ) , ' . s2p ' ) ;
% open f i l e
fid = f o p e n ( fname , ' r ' ) ;
% r ea d data
data = textscan ( fid , '%f %f %f %f %f %f %f %f %f ' , ' H e a d e r l i n e s ' , 6 ) ;
f c l o s e ( fid ) ;
% e x t r a c t data
freq = data { 1 } ;
S21_magdB = data { 4 } ;
S 2 1 _ p h a s e D e g = data { 5 } ;
S21 = db2mag ( S21_magdB ) . ∗ exp (1 i ∗ S 2 1 _ p h a s e D e g ∗ p i / 1 8 0 ) ;
S21_cal = S21 . ∗ exp (1 i ∗2∗ p i ∗ freq ∗ p1 ( 2 ) ) ;
h ( i , : ) = i f f t ( S21_cal ) ;
H ( i , : ) = S21_cal ;
fs = max( freq ) − min ( freq ) ;
end
%% 2D i m a g i n g
% d i m en s i o n o f a r e a i n i n t e r e s t
width = 1 . 6 ;
len = 6 ;
w = 1 : width ∗ 1 0 0 ;
l = 1 : len ∗ 1 0 0 ;
N_w = l e n g t h ( w ) ;
N_l = l e n g t h ( l ) ;
%% Frequency domain t e c h n i q u e
pixels_f r eq = z e r o s ( N_w , N_l , N ) ;
f o r i = 1 : N_w ;
f o r j = 1 : N_l
d i s p ( [ '@ row ' num2str ( i ) ' c o l ' num2str ( j ) ] ) ;
% d i s t a n c e i s th e d i s t a n c e from horn1 to horn 2 v i a p i x e l ( i , j )
distance = 1/100 ∗ ( s q r t ( ( i−d −16) .ˆ2+ j ˆ 2 ) + s q r t ( ( i−d ) .ˆ2+ j ˆ 2 ) ) ;
sum1 = z e r o s ( 1 , N ) ;
f o r k = 1 : d_len
sum1 = sum1 + H ( k , : ) . ∗ exp (1 i ∗2∗ p i ∗ freq . ' ∗ distance ( k ) / c ) ;
end
pixels_fre q ( i , j , : ) = sum1 ;
end
end
I = abs ( sum ( pixels_fr e q ( ) , 3 ) ) ;
%% p l o t th e r e s u l t
55
% linear scale
f i g u r e , i m a g es c ( I ) , colormap ( j e t )
y l a b e l ( ' Cross−Range (cm) ' ) , x l a b e l ( ' Range (cm) ' )
t i t l e ( '2−D image , empty room ' )
% log scal e
f i g u r e , i m a g es c ( mag2db ( I /max(max( I ) ) ) ) , colormap ( j e t )
y l a b e l ( ' Cross−Range (cm) ' ) , x l a b e l ( ' Range (cm) ' )
t i t l e ( '2−D image , empty room , l o g s c a l e ' )
% r a n g e a t x = 80
f i g u r e , p l o t ( mag2db ( I ( 8 0 , : ) /max( I ( 8 0 , : ) ) ) , ' L i n ew i d th ' , 2 ) ;
x l a b e l ( ' y [ Range (cm) ] ' ) ; y l a b e l ( ' I ( y ) (dB) ' )
t i t l e ( ' Range p r o f i l e a t x = 80 cm ' )
g r i d on ; ylim ([ −30 0 ] )
56
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement