Civil Engineering Applications of Ground Penetrating Radar - Benedetto & Pajewski

Prévia do material em texto

Springer Transactions in Civil
and Environmental Engineering
Civil Engineering 
Applications 
of Ground 
Penetrating Radar
Andrea Benedetto
Lara Pajewski Editors
Springer Transactions in Civil 
and Environmental Engineering
More information about this series at http://www.springer.com/series/13593
Andrea Benedetto · Lara Pajewski 
Editors
1 3
Civil Engineering 
Applications of Ground 
Penetrating Radar
Editors
Andrea Benedetto
Department of Engineering 
University Roma Tre 
Rome
Italy
ISSN 2363-7633 ISSN 2363-7641 (electronic)
Springer Transactions in Civil and Environmental Engineering
ISBN 978-3-319-04812-3 ISBN 978-3-319-04813-0 (eBook)
DOI 10.1007/978-3-319-04813-0
Library of Congress Control Number: 2015933624
Springer Cham Heidelberg New York Dordrecht London
© Springer International Publishing Switzerland 2015
This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part 
of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, 
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission 
or information storage and retrieval, electronic adaptation, computer software, or by similar 
or dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this 
publication does not imply, even in the absence of a specific statement, that such names are exempt from the 
relevant protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this 
book are believed to be true and accurate at the date of publication. Neither the publisher nor the 
authors or the editors give a warranty, express or implied, with respect to the material contained 
herein or for any errors or omissions that may have been made.
Printed on acid-free paper
Springer International Publishing AG Switzerland is part of Springer Science+Business Media 
(www.springer.com)
Lara Pajewski
Department of Engineering 
University Roma Tre
Rome
Italy
v
Introduction
Ground Penetrating Radar (GPR) is a safe, advanced, and non-invasive sensing 
technique that has several traditional and novel applications, sometimes stand-
ardised by national or international regulations. It can be effectively used for 
subsurface investigation, three-dimensional imaging of composite structures, and 
diagnostics affecting the whole life-cycle of civil engineering works.
The major GPR strengths, on which its success in the civil engineering field 
is based, are related to the non-destructiveness and non-intrusiveness of the sur-
veys, notably lower costs compared to traditional methods, high-speed data acqui-
sition, reliability and representativeness of measurements. The time needed for a 
survey is typically one order of magnitude shorter than with possible alternative 
technology; this results in obvious business benefits of reduced costs, together 
with limited or eliminated charges associated to restricting access of other activi-
ties to the investigated area. GPR provides significant, dense and accurate data; 
the resolution is higher, compared to competing geophysical technologies as seis-
mic, transient electromagnetic, electrical and magnetic approaches. The main per-
formance limitations occur in the presence of high-conductivity materials, such as 
clay or salt-contaminated soils, and in heterogeneous conditions causing compli-
cated electromagnetic-scattering phenomena. Considerable expertise is necessary 
to effectively design and conduct a survey; moreover, the interpretation of radar-
grams is generally non-intuitive, thus specific competences are needed to enable 
measurements to be transformed into clear pictures and engineering decision-
making data. We are confident that, thanks to the improvement and evolution of 
both hardware and software technologies, ground-penetrating radar will become 
an even more efficient, effective, extensively used and less-invasive technique in 
the near future.
Looking at the current interest of the scientific community, technicians and pro-
fessionals all over the world, towards GPR and its civil engineering applications, it 
could seem that the history of this electromagnetic technique is very long. Then, it 
may sound odd that its origin is assumed in the first applications of the radio-wave 
propagation above and along the surface of the Earth that were developed about 60 
years ago. The first use in the field of civil engineering is commonly considered 
Introductionvi
to have taken place in Egypt and it was oriented to identify the water table depth: 
in 1956, El Said implemented a research programme, funded by the Egyptian 
National Research Council, for the geophysical prospection of underground water 
in the deserts. It is interesting to underline that the methodology adopted by El 
Said was essentially the same as is currently used to estimate the thickness of lay-
ers or the burial depth of targets. In particular, he used a continuous-wave trans-
mitter to diffuse electromagnetic energy in the ground through an antenna laid on 
the soil surface. A radio receiver measured the wave reflected by the water table. 
The distance between receiver and transmitter was known. Following procedures 
nowadays still used, he calculated the depth of the table by measuring the time 
delay of the received wave.
The whole history of ground penetrating radar is intertwined with its various 
applications. A significant activity in the field of civil engineering started up in 
the 1960s and has become mature in the 1970s. Mines and underground deposit 
inspections were very frequent. Moreover, in the line of the lunar science mis-
sions, strong efforts were spent to improve new technologies that seemed very 
promising for the subsurface examination. Additional and promising uses were 
observed in archaeology and geology. The electrical characterisation of geological 
materials, as well as the relationships between electrical conductivity and dielec-
tric polarisation, were topics of great interest in the research community. In the 
1980s, the GPR borehole configuration was successfully proposed for the assess-
ment of nuclear waste disposal sites. Starting from these experiences, the borehole 
configuration has become a relevant standard for hydrological studies of porous 
media. Another application, that is now likely the most financed, is mine detec-
tion for security and humanitarian purposes. In recent years, the GPR was pro-
posed and successfully used for the localisation of people buried or trapped under 
snow or debris, aiding rescue activities in disaster scenarios such as avalanches, 
collapsed buildings and earthquakes.
In the civil engineering field, GPR is currently used for inspection, monitoring 
and design purposes. The detection of utilities and buried objects, as well as the 
surveying of road pavements, bridge decks, tunnels, and the measurement of mois-
ture content in natural soils and manmade materials, are the main applications. In 
addition, interesting examples concerning the use of the ground penetrating radar 
in structural, geotechnical and railway engineering have to be mentioned.
This book is a deliverable of the COST (European COoperation in Science 
and Technology) Action TU1208 “Civil Engineering Applications of Ground 
Penetrating Radar.”
COST is the longest-running European (EU) framework supporting cooperation 
among scientists and researchers across Europe; founded in 1971, it has been con-
firmed in Horizon 2020. It contributes to reducing the fragmentation in EU research 
investments, building the European Research Area (ERA) and opening it to coopera-
tion worldwide.It also aims at constituting a “bridge” towards the scientific com-
munities of emerging countries, increasing the mobility of researchers across Europe, 
and fostering the establishment of excellence in various key scientific domains. 
Gender balance and early-stage researchers are strategic priorities of this programme.
Introduction vii
COST does not fund research itself, but provides support for networking activities 
carried out within Actions: these are bottom-up science and technology networks, 
centred around nationally funded research projects, with a 4-year duration and a 
minimum participation of five COST Countries. The Actions are active through 
a range of networking tools, such as meetings, workshops, conferences, training 
schools, short-term scientific missions, and dissemination activities; they are open 
to researchers and experts from universities, public and private research institutions, 
non-governmental organisations, industry, and small and medium-sized enterprises.
The Action TU1208 is running in the “Transport and Urban Development” COST 
domain; it started in April 2013 and focuses on the exchange of scientific-technical 
knowledge and experience of GPR techniques in civil engineering, aiming at promot-
ing throughout Europe a more effective use of this inspection method. The ambitious 
and interdisciplinary project of the COST Action TU1208 is being developed within 
the frame of a unique approach based on the integrated contribution of university 
researchers, software developers, geophysicists, civil and electronic engineers, archae-
ologists, non-destructive testing equipment designers and producers, end users from 
private companies and stakeholders from public agencies. About 300 participants from 
130 institutions in 28 COST Countries (Austria, Belgium, Croatia, Czech Republic, 
Denmark, Estonia, Finland, France, Germany, Greece, Ireland, Italy, Latvia, Malta, 
Macedonia, The Netherlands, Norway, Poland, Portugal, Romania, Serbia, Slovakia, 
Slovenia, Spain, Sweden, Switzerland, Turkey, UK) and a COST Cooperating State 
(Israel) have joined the Action. Partners from COST Near-Neighbour Countries 
(Albania, Armenia, Egypt, Jordan, Russia, Ukraine) and International Partner 
Countries (Australia, Philippines, Rwanda, United States of America) are participat-
ing, too. Applications from further Countries are currently under examination.
During the first year of activity, the partners worked on highlighting the advan-
tages and limitations of the currently available equipment, surveying procedures, 
and electromagnetic/numerical methods for the interpretation of experimental 
data. Such studies led to a comprehensive assessment of the state of the art in the 
field of the civil engineering applications of GPR, and to the identification of open 
issues and gaps in knowledge and technology. The results of this wide and in-
depth review activity were fruitfully discussed during Action’s meetings and are 
presented in this book.
The organisation of the book reflects the scientific structure of the Action, which 
includes four Working Groups (WGs). In particular, the WG 1 of the COST Action 
TU1208 focuses on the design of innovative GPR equipment, the building of proto-
types, the development of testing and calibration procedures, and the optimisation 
of new systems. The WG 2 deals with surveying of transport infrastructures and 
buildings, sensing of underground utilities and voids, testing of construction materi-
als and estimation of soil water content. The WG 3 is developing accurate and fast 
electromagnetic scattering approaches for the characterisation of complex scenarios, 
inversion and imaging techniques, and data processing algorithms for the elabora-
tion of GPR data collected during civil-engineering surveys. Finally, the WG 4 
focuses on the applications of GPR outside from the civil engineering field, as well 
as on the combination of GPR with other non-destructive testing techniques.
Introductionviii
The book is opened with the first chapter of Part I—“GPR Instrumentation”—
authored by G. Manacorda et al. and entitled “Design of Advanced GPR 
Equipment for Civil Engineering Applications,” where the main issues in design-
ing ground penetrating radar equipment dedicated to civil engineering applications 
are described. A comprehensive review on the commonly available system archi-
tectures along with the main design challenges to build an effective tool are herein 
provided. Overall, the work mostly focuses on three major areas where solutions 
to technical challenges are nowadays more than ever needed, namely, radio-fre-
quency system design, antenna design and data analysis.
The transmitting and receiving antennas are among the most critical parts of a 
ground penetrating radar, performing the essential functions of transferring elec-
tromagnetic energy to the surveyed scenario with the required pattern, bandwidth 
and efficiency, and receiving the energy scattered-reflected by the environment. 
For this reason, Part I is complemented with the chapter by L. Pajewski et al., enti-
tled “Antennas for GPR Systems.” This contribution offers a review on the anten-
nas currently used in GPRs, suggesting ideas for their improvement, and resuming 
the numerical and experimental methods for their electromagnetic characterisation.
Part II of the book is entitled “GPR Surveying of Pavements, Bridges, Tunnels 
and Buildings; Underground Utility and Void Sensing.” It includes five contribu-
tions on this topic.
The chapter authored by J. Stryk et al. is entitled “Innovative Inspection 
Procedures for Effective GPR Surveying of Critical Transport Infrastructures 
(Pavements, Bridges and Tunnels).” This work thoroughly reviews individual 
applications, which are currently in use, and outlines those that are still in the 
phase of research and verification. An overview on issues that need to be dealt 
with GPR is also addressed, to enable the larger applicability of this non-destruc-
tive method in critical transport infrastructures.
The following chapter is entitled “Inspection Procedures for Effective GPR 
Surveying of Buildings,” by V. Pérez-Gracia and M. Solla. It focuses on the main 
achievements in surveying different types of buildings, on the software develop-
ment for enhancing data interpretation and on laboratory studies that can be over-
all relevant for the analyses of complex scenarios. Open issues are also defined as 
a final conclusion, based on the revision of different works.
In their chapter “Inspection Procedures for Effective GPR Sensing and 
Mapping of Underground Utilities and Voids, with a Focus to Urban Areas,” 
C. Plati and X. Dérobert present some studies showing the ground penetrating 
radar performances and limitations in locating and mapping objects such as pipes, 
drums, tanks, cables and underground features or in detecting subsurface voids 
related to subsidence and erosion of ground materials, from single-channel sys-
tems to the potential of multi-channel three-dimensional imaging and integrat-
ing systems. The Authors also discuss the importance of achieving cost-effective 
installations from the deployment of GPR prior to directional drilling for the pre-
vention of damage to existing utilities.
L. Krysinski and J. Hugenschmidt authored the chapter “Effective GPR 
Inspection Procedures for Construction Materials and Structures,” in which a 
Introduction ix
review of approaches related to the assessment of construction details and mate-
rial properties by using ground penetrating radar is presented. The analysis of the 
authors relies on the assessment of electromagnetic properties as a fundamental 
mean for understanding both materials physical properties and as an inherent part 
of any GPR structural study necessary for correcting uncalibrated electromagnetic 
parameters. Major directionsof research along with some benefits and limits of 
different approaches are herein described.
To complete Part II of the book, the chapter by F. Tosti and E. Slob, entitled 
“Determination, by Using GPR, of the Volumetric Water Content in Structures, 
Substructures, Foundations and Soil,” describes the use of several instruments and 
processing techniques for the evaluation of volumetric water content in concrete 
structures and unsaturated soils, at different investigation scales. Strength points 
and main drawbacks of the commonly used approaches for moisture sensing are 
discussed, relative to the most recent research studies on this issue. In addition, 
recently developed methods on this field of application are introduced.
Part III of the book is entitled “Electromagnetic Methods for Near-Field 
Scattering Problems by Buried Structures; Data Processing Techniques;” it 
includes overall four contributions on these issues
This part of the book is opened by the chapter “Methods for the Electromagnetic 
Forward Scattering by Buried Objects,” written by C. Ponti, in which the useful-
ness in using dedicated tools for the solution of forward electromagnetic scatter-
ing by buried objects is outlined, with the main purpose of interpreting the GPR 
responses. A review on the most established approaches in the modelling of 
impulse radar systems, such as Finite-Difference Time Domain or space-time 
integral equations, is developed. Furthermore, the issue of implementing novel 
approaches to approximate the integral equations via series expansions with lower 
computational complexity, when adopting a Method of Moments discretisation, is 
addressed. The spectral-domain Cylindrical Wave Approach is presented.
In the following chapter, entitled “Development of Intrinsic Models for 
Describing Near-field Antenna Effects, Including Antenna-Medium Coupling, 
for Improved Radar Data Processing Using Full-Wave Inversion,” A.P. Tran and 
S. Lambot deal with the proper description of antenna effects on GPR data and 
resume the methods that have been developed for this purpose. Traditional numeri-
cal methods are computationally expansive and often not able to provide an accu-
rate reproduction of real measurements. The Authors thoroughly describe how 
intrinsic modelling approaches, through which radar antennas can be effectively 
described taking into account their fundamental properties, have demonstrated 
great promise for fast and accurate near-field radar antenna modelling in order to 
reliably estimate medium electrical properties.
In the chapter “GPR Imaging via Qualitative and Quantitative Approaches,” 
I. Catapano et al. resume the issue of solving an inverse scattering problem, where 
a set of parameters describing the underground scenario must be retrieved starting 
from samples of the measured electromagnetic field. The authors provide an over-
view of different approaches and algorithms for both quantitative and qualitative 
buried scatterer reconstruction.
Introductionx
N. Economou et al. complete Part III of the book with a chapter on “GPR Data 
Processing Techniques,” wherein the difficulties in automating data analysis are 
mainly addressed. In this regard, after providing the reader with a deep under-
standing of the state of the art and open issues in the field of GPR data processing 
techniques, the authors present an overview on noise suppression, deconvolution, 
migration, attribute analysis and classification techniques with a particular focus 
on data collected during civil engineering surveys.
This book is concluded with Part IV “Different Applications of GPR and Other 
Non-Destructive Testing Technologies in Civil Engineering,” which includes four 
chapters.
The first contribution is entitled “Applications of GPR for Humanitarian 
Assistance and Security,” by X. Núñez-Nieto et al. This chapter reviews a series 
of published works in the frame of the ground penetrating radar applications for 
humanitarian assistance and security, with a special reference to the detection of 
mines and unexploded ordnances. The location of underground spaces and the 
GPR use in rescue operations is also addressed, wherein its contribution in locat-
ing human remains or living victims in disaster areas is always more demanded. 
The authors analyse specific systems, methodologies and processing algorithms 
specifically developed for these applications.
The following chapter, entitled “Applications of GPR in Association with Other 
Non-Destructive Testing Methods in Surveying of Transport Infrastructures,” writ-
ten by M. Solla et al., reviews a compilation of works in the frame of the appli-
cations of GPR combined to other non-invasive methods in the evaluation of 
transport infrastructures. The authors demonstrate that these integrated approaches 
have significantly benefited the procedures for inspection and they successfully 
solved some of the limitations of traditional methods in monitoring roads and 
pavements, concrete and masonry structures, and tunnels.
The next chapter, entitled “Advanced Electric and Electromagnetic Methods for 
the Characterisation of Soil,” by M. Van Meirvenne, deals with the detailed spa-
tial characterisation of soil properties with different electric and electromagnetic 
methods, which is essential for the management of soil to provide all its functions 
and essential services to the environment. Electrical resistivity sensors, ground 
penetrating radar systems and electromagnetic induction sensors are herein thor-
oughly compared by outlining potential targets of each measurement technique, 
along with advantages and limitations. Despite the strengths of every type of sens-
ing system, it is suggested by the author an increased integration of soil sensors 
into multi-sensor systems enabling their fused processing as a future challenge for 
enhancing the reliability of soil analyses.
The last chapter, entitled “Applications of radar systems in planetary sciences: 
an overview,” by F. Tosti and L. Pajewski, focuses on the remarkable results and 
sophistication of radar systems achieved over the history in several planetary 
explorations, by dividing the treatment according to different planets and celestial 
bodies investigated.
Introduction xi
We would like to thank very much the Authors of all the Chapters, for con-
tributing to this book. We are also sincerely grateful to Springer, to Dr. Pierpaolo 
Riva, Springer Engineering and Applied Sciences Editor, and to the Springer 
Editorial Staff, for giving us the opportunity to publish this book, for their 
patience, suggestions and support, and for the efficient handling of the editorial 
process. Finally, we would like to thank COST for funding the Action TU1208 
“Civil Engineering Applications of Ground Penetrating Radar.”
Andrea Benedetto
Lara Pajewski
xiii
Contents
Part I GPR Instrumentation
Design of Advanced GPR Equipment for Civil Engineering 
Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Guido Manacorda, Raffaele Persico and Howard F. Scott
Antennas for GPR Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
Lara Pajewski, Fabio Tosti and Wolfgang Kusayanagi
Part II GPR Surveying of Pavements, Bridges, Tunnels 
and Buildings; Underground Utility and Void Sensing
Innovative Inspection Procedures for Effective GPR Surveying 
of Critical Transport Infrastructures 
(Pavements, Bridges and Tunnels) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
Josef Stryk, Amir M. Alani, Radek Matula and Karel Pospisil
Inspection Procedures for Effective GPR Surveying of Buildings . . . . . . . 97
Vega Pérez-Gracia and Mercedes Solla
Inspection Procedures for Effective GPR Sensing 
and Mapping of Underground Utilities and Voids,with a Focus to Urban Areas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
Christina Plati and Xavier Dérobert
Effective GPR Inspection Procedures for Construction 
Materials and Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
Lech Krysin´ski and Johannes Hugenschmidt
Contentsxiv
Determination, by Using GPR, of the Volumetric Water 
Content in Structures, Substructures, Foundations and Soil . . . . . . . . . . . 163
Fabio Tosti and Evert Slob
Part III EM Methods for Near-Field Scattering Problems 
by Buried Structures; Data Processing Techniques
Methods for the Electromagnetic Forward Scattering 
by Buried Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
Cristina Ponti
Development of Intrinsic Models for Describing Near-Field 
Antenna Effects, Including Antenna-Medium Coupling, 
for Improved Radar Data Processing Using Full-Wave Inversion . . . . . . . 219
Anh Phuong Tran and Sébastien Lambot
GPR Imaging Via Qualitative and Quantitative Approaches . . . . . . . . . . . 239
Ilaria Catapano, Andrea Randazzo, Evert Slob and Raffaele Solimene
GPR Data Processing Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281
Nikos Economou, Antonis Vafidis, Francesco Benedetto 
and Amir M. Alani
Part IV Different Applications of GPR and Other NDT 
Technologies in CE
Applications of GPR for Humanitarian Assistance and Security . . . . . . . . 301
Xavier Núñez-Nieto, Mercedes Solla and Henrique Lorenzo
Applications of GPR in Association with Other Non-destructive 
Testing Methods in Surveying of Transport Infrastructures . . . . . . . . . . . 327
Mercedes Solla, Henrique Lorenzo, Joaquin Martínez-Sánchez 
and Vega Pérez-Gracia
Advanced Electric and Electromagnetic Methods 
for the Characterization of Soil . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 343
Marc Van Meirvenne
Applications of Radar Systems in Planetary Sciences: An Overview . . . . 361
Fabio Tosti and Lara Pajewski
Part I
GPR Instrumentation
3
Design of Advanced GPR Equipment 
for Civil Engineering Applications
Guido Manacorda, Raffaele Persico and Howard F. Scott
© Springer International Publishing Switzerland 2015 
A. Benedetto and L. Pajewski (eds.), Civil Engineering Applications of Ground 
Penetrating Radar, Springer Transactions in Civil and Environmental Engineering, 
DOI 10.1007/978-3-319-04813-0_1
Abstract This chapter describes the issues to be addressed in the design of Ground 
Penetrating Radar equipment dedicated to civil engineering applications. Radar is 
well known for its ability to detect aircraft, ships, vehicles, birds, rainstorms and 
other above-ground objects. It relies for its operation on the transmission of elec-
tro-magnetic energy, usually in the form of a pulse, and the detection of the small 
amount of energy that is reflected from the target. The round-trip transit time of 
the pulse and its reflection provide range information on the target. The applica-
tion of radar in the detection of buried objects is quite old; there are details of such 
work dating back to 1910, with the first pulsed experiments reported in 1926 when 
the depths of rock strata were determined by time-of-flight methods. The design of 
effective Ground Penetrating Radars requires solutions to technical challenges in 
three major areas:
•	 Radio Frequency system design.
•	 Antenna design.
•	 Data analysis.
Hence, this chapter reviews the commonly available GPR system architectures 
and summarises main design challenges to build an effective tool.
G. Manacorda (*) 
Georadar Division, IDS Ingegneria Dei Sistemi S.p.A., via Enrica Calabresi, 24-Loc. 
Montacchiello, 56121 Pisa, Italy
e-mail: g.manacorda@idscorporation.com
R. Persico 
Institute for Archaeological and Monumental Heritage IBAM-CNR via Monteroni, 
Campus Universitario, 73100 Lecce, Italy
e-mail: r.persico@ibam.cnr.it
H.F. Scott 
OSYS Technology Ltd., Ouseburn Building, Albion Row, Newcastle upon Tyne 
NE6 1LL, UK
e-mail: howard.scott@osys.co.uk
4 G. Manacorda et al.
1 Introduction
The use of Ground Penetrating Radars (GPRs) in civil engineering is well estab-
lished and there are several applications where it is currently utilised; they include 
the location of buried services, the detection voids or cavities, locating steel rein-
forcement in concrete, geotechnical foundation investigations, as well as archaeo-
logical, environmental and hydro-geological surveys.
The main applications of GPR to transportation infrastructures generally 
include measuring the thickness of pavement layers, detecting voids beneath lay-
ers, detecting and locating reinforcing bars, inspecting pavement structure, and 
mapping of the underground utilities.
Ground Penetrating Radars are designed to probe up to a few metres into the 
ground through material that is, usually, non-homogenous and, unlike free-space, 
strongly absorbs radar signals. The frequency range that has been found to be use-
ful for such an application lies within the limits of 100 MHz to 2 GHz.
Requirements for civil engineering applications differ, depending upon the appli-
cation, and each one imposes a particular set of constraints on the design of an effec-
tive GPR. For example, the majority of buried plant is within 1.5–2 m of the ground 
surface, but it may have a wide variation in its size, may be metallic or non-metallic, 
may be in close proximity to other plant and may be buried in any one of a wide 
range of soil types, with implications for large differences in both the absorption and 
the velocity of propagation of electro-magnetic waves, and consequent effects upon 
GPR performance. For this application, the most important performance criterion 
is depth of penetration, with resolution (the ability to distinguish between closely 
space objects), whilst being important, is a secondary consideration.
On the other hand, surveys of concrete or asphalt pavements requires very high 
resolution for accurately measuring the thickness of layers composing roadways or 
the runways; the same applies to the assessment of bridge decks where GPR sig-
nals can be analysed to detect potentially corroded areas.
Performance characteristics of GPRs are also often affected by ground condi-
tions that may vary rapidly within the area of a radar survey where, for example, 
variations in water content can be crucial and, particularly in urban areas, where 
there could be imported backfill of inconsistent quality. Consequently, it can some-
times be problematic to achieve both adequate penetration of the radar energy and 
good resolution, and some design compromises may have to be accepted.
In addition, a further issue concerns the interpretation of GPR data, which is 
not trivial in many situations; in this respect, the latest developments in GPR are 
oriented towards the design of equipment featuring real-time 3D high resolution 
images of surveyed areas.
Images, such as that shown in Fig. 1 can easily be understood even by an 
unskilled operator; however, this visualisation improvement can be effective only 
if the GPR performs well in terms of signal quality and detection range; in fact, 
if the received signal is too weak, as would be the case in wet, muddy ground, 
enhanced graphics software will solve neither the basic signal problem nor the 
detection performance.
5Design of Advanced GPR Equipment …
Consequently, the design of high performance equipment is a complex but fasci-
nating task for engineers and researchers as it involves a wide range of expertise such 
as electromagnetic wave propagation in media, antenna technology, radar design and 
electronics as well as advanced signalprocessing techniques and computer graphics.
2 The Radio Frequency System
2.1 Introduction
The purpose of the Radio Frequency (RF) system is:
•	 To generate an electrical signal of appropriate power level, frequency range and 
spectral characteristics, and to apply it to the transmit antenna.
•	 To process energy collected by the receive antenna into a form suitable for data 
analysis.
Fig. 1 High resolution GPR image of the archaeological site of Empúries (Spain) (Courtesy of 
Geostudi Astier—Italy)
6 G. Manacorda et al.
The RF energy is usually in the form of a short pulse, a frequency modulated burst 
of electromagnetic energy or discrete frequencies transmitted in a known sequence. 
RF system design and costs are significantly affected by the choice of modulation 
technique. Pulse modulation is, at present, cheaper to implement and is the method 
used in most commercial systems. Performance benefits, particularly increased 
dynamic range, are available from Frequency Modulated Continuous Wave 
(FMCW) and stepped frequency systems, but practical limitations imposed by the 
physics of the process make the benefits difficult, but not impossible, to realise in 
practice. Other modulation techniques are possible, such as noise and pseudoran-
dom coding, but these are very seldom used.
The critical performance parameter of the RF system is dynamic range. Pulse 
modulated systems use sampling receiver techniques where, typically, the dynamic 
range will be 70 dB. Time varying gain is usually applied which can increase the 
dynamic range to 90 dB or more. In addition, averaging may be used (if acquisi-
tion times permit) which can provide a further increase. Frequency modulated sys-
tems, either continuous wave or stepped, can increase dynamic range.
However, when the RF system is connected to practical antennas, internal sys-
tem reflections, ground returns and transmit to receive antenna leakage generate 
time dependent clutter (clutter is the term used for returns identified by the system 
as targets that do not correspond to intended targets or to noise). Unless special 
measures are taken, clutter limits dynamic range regardless of whether time or fre-
quency domain systems are used.
The final stage of the RF system transforms the analogue signal into digital 
form for data analysis and display purposes; this conversion must be executed with 
a suitable sampler resolution that prevents the limiting effects on dynamic range of 
quantization noise.
2.2 Time Domain GPR
Usually, Time Domain GPRs produce transmit signals with the required frequency 
range by using an impulse generator based upon an avalanche transistor. A typi-
cal pulse obtained from such a device is shown below, together with its spectrum 
(note that the signal consists of just a single cycle with a period of, approximately, 
10 ns, as shown in the diagram Fig. 2). Although this is a cost-effective means of 
producing a signal with usable characteristics, the physical mechanism is a ran-
dom process that may produce noise and jitter, which limits the inherent dynamic 
range of the system.
The receivers for such systems are based upon the methods used in high fre-
quency time domain sampling oscilloscopes which also have fundamental limits 
on their dynamic range that rarely exceeds 70 dB.
The top diagram in Fig. 3 depicts a typical GPR system, consisting of an 
impulse source and receiver connected by transmission lines to transmit and receive 
antennas. The system is deployed close to the ground surface, and interactions 
7Design of Advanced GPR Equipment …
occur between the radar and the ground and between its internal components. The 
major interaction paths are marked.
These interactions are extended in time and define what is known as the 
‘Impulse Response’ of the system. It is also known as the ‘Clutter Profile’. This is 
shown in the Fig. 3 bottom diagram as a decaying received signal—with respect 
to time and hence distance from the radar. Reflections from targets buried in the 
ground must be large enough for their peaks to be visible above the clutter profile. 
The system clutter profile is a critical system performance parameter and the radar 
must be designed to minimise its decay time and to make it, as far as possible, 
independent of the electrical properties of the ground.
2.3 Frequency Domain GPR
Frequency domain radar systems have as long a history as their time domain coun-
terparts. For some applications, the advantages of simple Continuous Wave (CW) 
systems is that they avoid the complication of modulation circuitry, have no mini-
mum or maximum range as well as maximising power on the target.
However, because they depend upon the Doppler shift principle they also have 
the disadvantage of only being able to detect moving targets. The main use of such 
systems has been military, where they provide a means of determining the point of 
Fig. 2 Impulse radar pulse and spectrum envelope
8 G. Manacorda et al.
closest approach by guided weapons to their targets so that the warheads may be 
detonated at the correct time.
Being unable to detect targets unless they are moving and producing a Doppler 
shift clearly makes CW radars unsuitable for GPRs. If, however, the source is able 
Fig. 3 Impulse radar GPR Scheme with major signal interaction paths (above) and consequent 
GPR clutter response (below)
9Design of Advanced GPR Equipment …
to produce a range of frequencies continuously varying with time, then it is possi-
ble to detect targets that do not move. Such radars are known as Swept Frequency 
Continuous Wave (SFCW). Usually the signal is generated by a source whose fre-
quency can be controlled by the application of an external DC voltage.
In both the generation and reception of frequency domain signals, the technology 
is very different from that their time domain counterparts, and some aspects of the 
performance of such systems, particularly noise and dynamic range, are superior.
2.3.1 Principles of Frequency Modulated Continuous Wave 
(FMCW) Receivers
FMCW radar systems are able to detect motionless targets because of the time 
delay between the energy originally transmitted and the reception of energy 
reflected from the remote target. Because the frequency of the transmitted signal is 
constantly varying with time, the signal from the target will be different to that cur-
rently being transmitted. If the transmitted frequency changes linearly with time, 
then the frequency difference is a direct measure of the range of the target (Fig. 4).
By choosing the rate of change of frequency appropriate to the target range, the 
difference frequency can be set so that it lies within a range that can be processed 
by audio frequency devices. A typical difference frequency might be 10 kHz, so 
that a narrow-band filter can be used to minimise the noise bandwidth and, hence, 
maximise the dynamic range.
By applying a sample of the transmitted signal and the received signal to a non-
linear receiver, sum and difference frequencies are created. A low pass filter easily 
separates the required difference and sum frequencies.
The diagram in Fig. 5 depicts a simple FMCW radar system. In the GPR appli-
cation illustrated, the same signal interactions are present as in the impulse radar.
Detection is achieved by taking a sample of the transmitted waveform, and 
using it as a Local Oscillator (LO) input to a diode. The much weaker received 
signal is also fed into the diode, which acts as a ‘phase coherent’ detector. This 
Fig. 4 FMCW difference frequency generation
10 G. Manacorda et al.
signal mixing (heterodyning) in the diode creates sum and difference frequencies 
from the LO and received signals. The sum frequency is outside the frequency 
range of the receiver and is thus filtered out, whilst the differencefrequency is pro-
cessed further to yield the required target information. This processing takes the 
form of a transformation of the difference frequency data, by means of a Fourier 
Transform, into the time domain so that range information may be extracted.
FMCW radars offer wider dynamic range, lower noise figures and can radiate 
higher mean powers than the time domain counterpart. In addition, the frequency 
range used is easily tailored to suit the characteristics of the material and targets 
under investigation. If, however, degradation of the system resolution by spectral 
widening of the Intermediate Frequency (IF) is to be avoided, then a high degree 
of the linearity in the variation in frequency as a function of time is required.
2.3.2 Time Domain to Frequency Domain Transformation
Fourier, in his Théorie analytique de la chaleur (1822), stated that any periodic 
signal can be represented by the sum of an infinite series of sine waves separated 
by the repetition frequency of the time domain signal. The process has become 
known as the Fourier Transform. Further, an inverse process may be applied to the 
Fig. 5 FMCW radar schematic diagram
11Design of Advanced GPR Equipment …
series of sine waves to transform them from the frequency domain into the time 
domain. This is known as an Inverse Fourier Transform.
All time domain signals are represented by a single value that describes their 
amplitude at any instant in time. On the other hand, the frequency domain descrip-
tion of a sinusoid requires both amplitude and phase. The output, therefore, of the 
Fourier Transform requires two channels of information to convey a full descrip-
tion. This can be amplitude and phase, but it can also be the real and imaginary 
parts of a complex number. In order to carry out the Inverse Fourier Transform 
correctly, both the real and imaginary parts of the frequency domain signal must 
be known. The simple schematic of a FMCW system as shown above, cannot sup-
ply the complete information required.
As the repetition frequency decreases, then the frequency difference between 
the sinusoids of the Fourier Transform becomes smaller until, in the limit when 
there is only a single occurrence of the signal, the frequency spectrum becomes 
continuous. The diagram below shows a single occurrence of a real time domain 
signal, together with its Fourier Transform. Note that the time domain signal 
illustrated has a DC component, and the spectrum of the output of the Fourier 
Transform is distributed about zero frequency. With radar pulses, there would be 
no DC component and the spectrum of the Fourier Transform would be distributed 
about the centre frequency of the radar system (Figs. 6 and 7).
To obtain the correct Inverse Fourier Transform, both the real and imaginary 
components of the frequency domain version must be used in the calculation, as 
shown below. The general case is that, in principle, the time domain signal may 
also have real and imaginary parts.
The above has implications for the design of a FMCW radar system, because 
an extra channel of information is needed so that both real and imaginary parts (or 
amplitude and phase) of the frequency domain signal can be captured. To achieve 
this, the Local Oscillator output is divided into two signals which differ in phase 
by 90°, and each is applied to its own mixer. The signal from the receive antenna 
is also divided, with no relative phase shift. These are mixed with the two LO sig-
nals to generate the real and imaginary parts. They are more usually known as the 
‘In-Phase’ (I) and ‘Quadrature’ (Q) components.
Fig. 6 Fourier transform concept
12 G. Manacorda et al.
2.3.3 Stepped Frequency Continuous Wave Systems
Thus far, the microwave source has been described as a voltage controlled swept 
frequency generator where the output frequency is continuously variable; the main 
advantages and disadvantages of such a system have been outlined above.
Another class of microwave sources is available where the frequency can be 
changed in discrete, highly repeatable and stable, steps. In this case, each measure-
ment is made at a constant frequency and, hence, the output of the both the In-Phase 
and Quadrature mixers is a constant voltage. Because both Local Oscillator signals 
are coherent with the received signal, the output of the mixers is proportional to the 
difference in phase between the transmitted and received signals, but the difference 
frequency, as described for the continuous wave source, cannot exist because there 
is no change in frequency. In this sense, it is a CW radar and, technically, is termed 
a homodyne system.
By the application of a command, usually issued by a digital system, the fre-
quency of the microwave source can be changed to a new value and, after allowing 
for the settling time, another CW measurement made where the mixer outputs are 
proportional to the phase difference between the transmit and receive signals at the 
new frequency.
As the microwave source is systematically ‘stepped’, in equal increments, 
through its complete frequency range, the DC voltages from both mixers are dig-
itised and stored. At the completion of the process, the record of the voltages of 
the mixers (which are proportional to change in phase of the received signals, with 
respect to the transmitted signal) may be displayed as a function of frequency 
Fig. 7 FMCW radar with I and Q outputs schematic diagram
13Design of Advanced GPR Equipment …
of the transmitted signal. The result is indistinguishable from the difference fre-
quency that would have been obtained from the continuously swept source. 
However, it is not possible to apply a filter to the data, so the Nyquist requirement 
that the sampling rate provides a range of twice that to the most distant target to be 
detected (whether that target is below or above ground) must be strictly observed.
The advantage gained from the extra complication of a stepped frequency 
source is that the transmitted CW signals are extremely stable and spectrally pure 
to a degree that cannot be obtained from a continuously swept source. Also, the 
accuracy of the frequencies provides an extremely linear sweep, which eliminates 
one of the principal weaknesses of FMCW sources.
3 Ground Penetrating Radar Antennas
The purpose of the antenna system is to illuminate the target and to collect the 
resulting scattered energy from the environment. Two main types of system are used:
•	 Monostatic.
•	 Bistatic.
Monostatic systems use a common radiating element for transmitting and receiv-
ing the signal; most GPRs use bistatic antennas with the transmitter and receiver 
hosted in the same enclosure.
The key requirements for the antennas are:
•	 An ability to radiate a wide range of frequencies, typically over the range 100–
1000 MHz or more.
•	 Linear phase characteristics over the operating frequency range.
•	 Predictable (and, preferably, constant) polarisation characteristics over the oper-
ating frequency range.
The key design objectives for a GPR antenna are to achieve
•	 a high degree of isolation between transmit and receive antennas;
•	 a low return loss from the feed point to minimise ringing and thus clutter generation;
•	 a good immunity from radio frequency interference.
Two general types of antenna have been used so far; dispersive and non-disper-
sive. In dispersive antennas different frequencies are radiated at different times, 
usually with high frequencies being radiated first followed by the low frequencies. 
This type of waveform is sometimes known as a “chirp”. Although some antennas 
are classed as non-dispersive, in fact all antennas are dispersive to some degree. 
Examples of these classes of antennas are:
•	 Dispersive—Spirals (logarithmic, exponential and Archimedean), exponential 
slot, Vivaldi.•	 Non-dispersive—TEM horn, biconical, bow tie, resistive lumped element loaded, 
resistive continuously loaded.
14 G. Manacorda et al.
The design of the antenna system is influenced by the nature of the required tar-
gets. All targets tend to depolarise electromagnetic waves incident upon them, the 
most extreme example of this being long, thin (in terms of wavelength) objects. If 
the long thin object is metallic then the reflected wave’s polarisation is parallel to 
its axis; if it is non-metallic, then the reflected polarisation is orthogonal to its axis 
(Roberts and Daniels 1996).
In the case of radar systems designed to locate pipes and cables, this is a very 
useful property that can, in principle, be used to improve the response to the required 
target. A complete knowledge of the depolarisation characteristics of targets in gen-
eral, particularly how they behave over a range of frequencies, can be a powerful aid 
in target detection and classification. There are many descriptions in the literature of 
antenna configurations specifically designed to extract polarisation information.
3.1 Array of Antennas
Optimum GPR system performance, particularly for high resolution detection of 
shallow objects, is obtained when the whole of the system is designed around a 
specific target type or geometry.
For example, whenever a high density of utilities is expected in several orienta-
tions (e.g. in an inner city road junction), a single antenna GPR must be scanned 
over a dense orthogonal grid (i.e. along two directions at right angles to each 
other) with a step no larger than say 0.5 m; using a larger step might be adequate 
when trying to trace a single pipeline across a field where the approximate entry 
and exit points are known, but is totally inadequate when mapping a complex lay-
out of underground assets.
Thus, when the collection of data requires the execution of a large number of 
profiles, the use of an array of antennas lined up in the transversal direction with 
respect to the direction of movement, and operating simultaneously, will provide 
the most efficient method.
This architecture enables another advantage in respect to the analysis of col-
lected data; in a certain sense, an array GPR implements a scheme very similar to 
the one used by “double threshold detection” radar.
Non-array GPRs are examples of, “single threshold detection radars”, where the 
operator decides whether a target is present on the basis of one “peak detection deci-
sion”; in other words, the output of the receiver is compared to a threshold or bias level.
This bias level, which is dependent upon the sensitivity of the display and the 
human operator’s visual perception, affects the probability of generating a false 
alarm and of missing a genuine target; in other words, when a received signal 
component that has been generated by noise or clutter exceeds the bias level, it can 
be mistaken for a return from a genuine target and there is a “false alarm”. On the 
other hand, when a signal received from a genuine target is interpreted as a noise 
pulse or clutter (which occurs when the signal return is below the bias level), there 
is a “missed detection”.
15Design of Advanced GPR Equipment …
Missed detection and false alarm rates are subject to trade-off; it means that 
the number of false alarms (missed detection) may be decreased (increased) as the 
bias level is raised, and vice versa; therefore bias level is a primary parameter in 
the radar’s design.
For solving problems related to the human operator, “double threshold detec-
tion” radars have been introduced. Such systems impose detection criteria 
whereby there must be several occurrences of the threshold of detection being 
exceeded in a defined period of time before a target is confirmed; therefore, the 
performance of these equipment are less vulnerable to the sensitivity of human 
operators.
Analogously, in GPR operations, the use of antenna arrays improves the detec-
tion of buried utilities because elongated targets (e.g. pipes or reinforcement bars 
in concrete) produce echoes (in the form of hyperbolae) in the same position in 
most (ideally in all) of the data windows; therefore, the operator can easily dis-
tinguish this category of target from those that are concentrated (e.g. a stone), and 
performance, in terms of probability of detection, is improved.
A further benefit of this array based architecture is the possibility of using 
advanced signal processing techniques; data collected with the array are geometri-
cally coordinated and can be stored as a three-dimensional data set that may be 
displayed as slices in vertical or horizontal planes. These are effectively micro-
wave images, and image-processing techniques may be applied to enhance wanted 
features; in this respect, as the target sought has identifiable properties (a pipe is 
long and thin), then this may be taken into account in the processing by the appli-
cation of appropriate filters, such as line finding algorithms.
Finally, the latest developments in GPR are tending towards providing the capa-
bility of generating real-time 3D high resolution images of surveyed areas. This 
objective is achievable by having a complete, very dense, coverage of the surveyed 
area, in the form of a 3D data set.
This could be achieved, albeit less efficiently, by performing hundreds of 2D 
profiles, very close to each other, with a single antenna GPR. However, since this 
data collection procedure is very time consuming, several dense arrays have been 
developed with the capability of producing 3D data volumes with a single scan.
4 Data Processing and Analysis
The processing of GPR data, with its several aspects both theoretical and practical, 
could form a “chapter” in its own right. At first sight, it appears that the mathemat-
ical ill-posedness of the problem (Colton and Kress 1992) makes it impossible to 
retrieve any details of the buried scenario. This should not make us unnecessarily 
pessimistic. However, this means that, whether the exploited processing algorithm, 
there will be a finite resolution and a finite quantity of information extractable 
from the data. Consequently, it is illusory to think that gathering an indefinitely 
growing number of data will achieve an increasingly precise image, and it is even 
16 G. Manacorda et al.
harmful to think of prolonging indefinitely the processing in order to improve the 
achieved results. This over-processing is the equivalent of squeezing a lemon in an 
attempt to extract more juice than it contains.
That being said, several categories of processing can be identified, depending 
upon whether a 1D, 2D or (more rarely) a 3D approach is taken and depending 
upon whether the problem is treated by a linear or (more rarely) nonlinear 
method. Within each approach, the configuration of the antennas has also to be 
taken into account (Persico et al. 2005), as well as the height of the measurement 
line (Persico 2006), the frequency band (Sala and Linford 2012) and the a priori 
information available in relationship with the case history at hand. In particular, 
if reliable a priori information is available, that is if the nature of underground 
targets are already presumed, and only some of their geometrical features are 
looked for, then a forward modelling technique may be exploited (Daniels 2004; 
Utsi 2012).
4.1 One-Dimensional Processing
Proceeding in order of increasing complexity, and taking for granted the prelimi-
nary step of the zero timing, 1-D processing is, in general, a technique where GPR 
traces (gathered at any fixed measurement point and presented either in time or 
frequency domains), are processed independently from each other. This can be 
done if a one-dimensional scenario is scanned so that a 1-D model of the propaga-
tion can be exploited (Persicoand Soldovieri 2004; Pieraccini et al. 2006), or the 
interest is in equalizing the deformation of the signal when it is reflected by the 
targets of interest.
In particular, it can be the case that incident waves may characteristically be 
scattered by some targets, allowing them to be recognised and, possibly, distin-
guished from other objects that may be geometrically similar, but of a different 
nature. This procedure is called deconvolution (Daniels 2004; Jol 2009), and is 
sometimes exploited where such an identification is particularly important, as e.g. 
in de-mining operations (Daniels 2004).
The predominant characteristic of GPR data is that it diminishes in amplitude 
as a function of time. This is caused by the geometrical spreading of the radi-
ated (and scattered) energy plus the electrical losses usually present in the soil. 
These effects combine to make the echoes from the deepest targets possibly much 
weaker than those from shallower targets. By applying an increase in gain versus 
time along the received signal, it is possible to compensate for this attenuation and 
make those targets visible.
It should be noted that varying the gain as a function of depth is a non-sta-
tionary processing step, which can cause a spurious enlargement of the spectrum 
of the traces, causing a deterioration of the image. This enlargement of the band 
is theoretically easily explained by the fact that a variable gain is equivalent to a 
multiplication of the time domain trace by a monotonically increasing function. 
17Design of Advanced GPR Equipment …
In the frequency domain, this is equivalent to the convolution product of the two 
spectra that, in general, leads to an enlargement of the band.
This effect can, usually, satisfactorily be mitigated by means of a further 1-D 
processing procedure consisting of filtering the traces to limit their spectra to be 
within the original band. This can be done via software making use of an ideal 
filter (the causality requirement is not essential because the GPR post-processing 
is not a real time operation) or an algorithm imitating a physically realisable filter, 
as e.g. a Butterworth filter (Di Lorenzo 2013). Any choice has its pros and cons. In 
particular, an ideal filtering process can, theoretically, anticipate the depth of the 
target whereas a “real filter” might add some further distortion in the band (in any 
case, both of these effects are usually negligible).
Variable gain and 1-D filtering procedures are usually contained in the routines 
available in commercial codes for GPR data processing, as e.g. the Reflexw 
(Sandmeier 2003) or the GPRslice (Goodman and Piro 2013). It is worth noting 
that variable gain and 1-D filtering are also usually exploited within the processing 
chain performed in 2-D or 3-D contexts. In other words, 1-D processing steps can 
be (and in practice are) mixed with 2-D or 3-D processing steps. To summarise with 
a final formula the 1-D filtering, let Tˆ(ω) represent the spectrum of a trace T(t).1
A 1-D filter is the multiplication of Tˆ(ω) by an established filtering function 
H(ω), so to achieve a filtered spectrum TˆF(ω) given by:
4.2 Two-Dimensional Processing
The most common forms of 2-D processing are spatial filtering and migration. The 
difference between 1-D and 2-D filtering is that, in the second case, several traces 
are treated and combined together in some manner to produce a comprehensive 
representation known as a B-scan. The most common 2-D spatial filters are imple-
mented by multiplying the spectrum of the data by a 2-D filtering function.
More precisely, labelling the 2-D Fourier transform of the datum as d(x, t) 
with respect to the measurement abscissa and to the time as ˆˆD(k,ω), the 
most common 2-D filter is implemented by multiplying this spectrum by an 
 established filtering function H(k,ω) to produce the filtered spectrum of the data 
ˆˆ
DF(k,ω), given by
Included in this category is the FK filter, which aims to reject the effect of possible 
reflection from targets in air (Chan and Stewart 1994) and the Background 
1 The data may be gathered either with a pulsed or a swept frequency GPR, in the latter case T(t) 
is meant as an equivalent trace in the “synthetic” time domain.
(4.1)TˆF(ω) = H(ω)Tˆ(ω).
(4.2)ˆˆDF(k,ω) = H(k,ω) ˆˆD(k,ω)
18 G. Manacorda et al.
Removal (BKG), that is aimed to remove background clutter caused by quasi-hor-
izontal reflectors that may mask targets of interest (Persico and Soldovieri 2008).
In particular, the quasi-horizontal “disturbing” reflections can be generated 
from either actual physical discontinuities (such as the air-soil interface, and also, 
possibly, any layered structure of the underground scenario) or from an apparent 
constant target, physically generated (e.g.) by a vehicle that moves the antennas 
or by ringing (Daniels 2004) of the antennas. The background removal can be 
performed on all of the traces or on a limited subset centred on the current trace 
(which is called a moving average). In theoretical terms, it can be shown that the 
first case is a particular instance of the second.
It is important to outline that any “therapy” on the signal has disadvantages. In 
particular, it is impossible to filter out completely the whole of the undesired parts 
of the image and leave intact the useful parts. Therefore, the nature and the spatial 
band of the filtering have to be determined by the data, and by the expertise of the 
human operator. The background removal on all of the traces may generate “new” 
horizontal artefacts (paradoxically, in a sense) when, in particular, the top of a 
strong reflector makes the level of the average trace significantly different from the 
level of most traces at some points along the depth axis.
Indeed, more complex 2-D filtering processes have been proposed, as e.g. the 
eigenimages (Kim et al. 2007), based on the singular value decomposition (SVD) 
(Bertero and Boccacci 1998) of the matrix of the data, that is the function d(x, t) 
(we can gather only a finite number of data, which means that, in the end, there 
is a sampled version of d(x, t) representing a continuous function). In particular, 
the matrix d(x, t) can be decomposed along component matrixes (called eigenim-
ages) by means of its SVD, and the subsequent eigenimages are weighted by the 
associated singular values, forming a decreasing sequence, which makes them 
more affected by the noise. The filtering, in this case, consists of retaining the first 
eigenimage or a few eigenimages corresponding to the highest singular values.
Beyond 2-D filtering, there are 2-D focusing procedures, which constrain, 
within certain limits, the “signature” of the target so that it is within its actual geo-
metric size (more precisely the radar cross-section of the target below the scan tra-
jectory. The most exploited focusing algorithm is 2-D migration, operating either 
in the frequency domain (Stolt 1978) or the time domain (Schneider 1978). The 
migration is an approximated closed (integral) form solution for the shape of the 
buried targets. Several approaches can be followed to describe it, and the object 
function associated with the buried target can be given by the contrast of dielectric 
permittivity (Persico 2014) or by some equivalent electric field in the medium of 
propagation (Stolt 1978; Schneider 1978). Whatever the object function, there are 
several forms of the 2-D migration formulas. Reported here are two of the most 
common, labelling as O
(
x′, z′
)
 the object function.
In particular, denoting again with ˆˆDF(k,ω) the filtered (and possibly “enhanced” 
by some variable gain) data, a 2-D migration formula in frequency domain is given by
(4.3)O
�
x′, z′
� =
+∞ˆ
−∞
+∞ˆ
−∞
ˆˆ
DF(k,ω) exp
�
jkx′
�
exp

jz′
�
4ω2
v2
− k2.

dkdω19Design of Advanced GPR Equipment …
and a 2-D migration formula in time domain is given by
where dF(x, t) is the space-time domain representation of the filtered datum 
 corresponding to 
ˆˆ
DF(k,ω) in the frequency and wave number domain.
In Eqs. (4.3) and (4.4) v is the propagation velocity of the electromagnetic waves 
in the soil and, to enhance clarity, some inessential constants have been omitted 
before the integral. In both cases, measurements at the air soil interface have been 
implicitly assumed. Equation (4.3) is also known as the 2-D Stolt’s migration for-
mula, whereas Eq. (4.4) is also known as 2-D Kirchoff’s migration formula. Details 
on their derivation, starting from the Maxwell’s equations, are available in (Persico 
2014). What is important to emphasise here is that the migration is, in any case, an 
approximate procedure based on a linear approximation of the scattering phenome-
non (Chew 1995) [which is actually intrinsically nonlinear (Colton and Kress 1992)] 
and on the assumption that targets’ depths are large in terms of wavelength (Persico 
2014). Another important point is that the migration (at least in its basic forms) is a 
procedure that neglects the losses of the soil (Lesselier and Duchene 1996).
However, the theory underlying the migration algorithms [that are also col-
lectively known as diffraction tomography (DT) (Lesselier and Duchene 1996; 
Meincke 2001)] provides not only a solution of the problem, but also allows an 
analysis that can estimate the resolution limits and the required sampling rate in the 
spatial and frequency (or alternatively time) domain. In particular, the essence of 
DT is the identification of an algebraic relationship between the spectrum of the 
data and that of the object function, which allows estimations to be made of the 
available resolution and the required rate of sampling data in the spatial and fre-
quency domains. The available resolution for shallow targets is of the order of �s
2
 
(�s being the centre frequency wavelength in the soil), and degrades as the depth 
increases. Several more detailed formulas describe, in different ways, this decay 
(Jol 2009; Persico 2014; Sheriff 1980). The vertical resolution for shallow targets 
is expected to be of the order of v
B
, B being the available frequency band and v is 
the propagation velocity of the electromagnetic waves in the soil. Usually, the Ultra 
Wide Band (UWB) antennas exploited for GPR prospecting have a bandwidth of 
the same order as the centre frequency fc, so that 
v
B
≈ v
fc
= �s.
Also, with regard to the vertical resolution, because the soil attenuates the higher 
frequencies more than the lower frequencies (Daniels 2004; Jol 2009), the “received 
bandwidth” reduces as the depth of the target increases (Sala and Linford 2012), 
and so some degradation vs. the depth of the vertical resolution is expected too.
Based on the DT, moreover, the expected required spatial step for the data is, of 
the order of, �s min
4
 (�s min being the minimum expected wavelength in the soil).
Invariably, the GPR signal never consists of a single harmonic, and the spatial 
steps may be redundantly narrow with respect to the lowest harmonic components 
(4.4)
O
(
x′, z′
) = ∂
∂z′
+∞ˆ
−∞
dx
+∞ˆ
2
√
(x−x′)2+z′2
v
dF(x, t)√
t2 − 4
[
(x−x′)2+z′2
]
v2
dt
20 G. Manacorda et al.
of the signal. With a pulsed system, the spatial stepping of the signal cannot be 
varied to make it larger for the lower harmonics, and for a stepped frequency sys-
tem it is impractical. If, however, it is necessary, for computational efficiency in 
particular cases, the redundancy of the spatial stepping necessary for gathering 
the data can be reduced in the processing phase through some suitable “harmless” 
decimation of the data.
The reasoning outlined above demonstrates that spatial sampling is not inde-
pendent of the time variability of the signal. This is predicted by Maxwell’s equa-
tions, where the spatial and time (or frequency) variabilities are coupled. The 
necessary frequency step (for swept frequency systems) is of the order of v
2D
, 
where D is the maximum expected penetration depth of the signal.
Sometimes, D is implicitly understood as the maximum depth of interest, but 
caution should be exercised. For example, if the range of depth interest is, say, one 
metre, as it may be in the case of asphalt monitoring, it should be borne in mind that 
radiation penetrates to greater depths. In some situations, deeper targets may cause 
disturbances within the depth range of interest if the frequency step is too large.
Finally, for a pulsed system, a time step of the order of 1
B
 is necessary. All of the 
required sampling rates, either in space, frequency or time, are essentially deter-
mined by the anti-aliasing requirement. More details are available, e.g., in (Persico 
2014).
It is a common practise suitably to combine (possibly with some interpola-
tion) several processed B-scans together to obtain a “parallelepiped” containing 
information representing a volume of the sub-surface. This allows visualisation 
of the buried scenario relative to a series of planes parallel to the three coordi-
nate Cartesian planes (i.e. the planes with equations, x = 0, y = 0 and z = 0, 
respectively). With some additional effort, if needed in particular cases, an image 
of the underground scenario along an oblique plane or even a curve surface can 
also be produced. The images at fixed depth (i.e. the images on the planes at 
constant z, also called depth-slices or time-slices) are often particularly useful, 
because they provide a map of the buried infrastructure in planes parallel to the 
air soil interface. This can be important for the mapping of the buried services 
or for archaeological investigations (Conyers 2004). Three dimensional perspec-
tive representations can also be produced from the reconstructed parallelepiped of 
the sub-surface volume. However, these involve the choice of some on-off energy 
thresholds in order establish the inner and the outer regions of the targets.
The choice of threshold is intrinsically linked to the mechanism of human 
vision. Indeed, when we observe the scenario of any room, what we usually distin-
guish are the external surfaces of the objects present that, in most but not all 
cases,2 appear to be “sharp”. With the reconstruction of buried targets there are 
two problems, namely: (i) even more so than in air, buried targets may not be 
intrinsically sharp, but diffuse because of smoothing mechanism, such as in the 
case of ingress of water, the spreading of polluting substances from leaking pipes, 
2 E.g. let us think of an unequally dense, or ‘patchy’ fog.
21Design of Advanced GPR Equipment …
or targets broken into discrete pieces of varying sizes. (ii) even if the target is 
sharp, its reconstruction, in general, may not be because of the spatial filtering 
properties of the scattering operator.
In these circumstances it is, in most cases, unavoidable that the appearance of the 
reconstruction is dependent on the chosen threshold level, which introduces some 
degree of arbitrariness, which must be assessed by human expertise. An examination 
of the perspective reconstruction at several threshold levels, in order to test heuristi-
cally the sensitivity to this parameter, is well advised (Leucci et al. 2011).
These three dimensional representations, as well as horizontal slices, are some-
times improperly referred to as “3-D” processing. In fact, it is more exact to say that 
it is pseudo-3-D processing based on an amalgamation of several 2-D solutions.
4.3 Three-Dimensional Processing
Authentic 3-D processing must be based Maxwell’s equations without imposing any 
invariance direction. This is computationally expensive and is, therefore,carried out 
more rarely. Adopting a linear approximation of the scattering phenomenon is com-
monly done also in 3-D processing, and also in 3-D it is possible to determine DT 
relationships (Persico 2014) and migration formulas (Stolt 1978; Schneider 1978; 
Persico 2014). In particular, expressions for 3-D migration formulas, in frequency 
and time domains, are provided in Eqs. (4.5) and (4.6), respectively.
Symbols in Eqs. (4.5) and (4.6) are direct extensions of the homologous symbols 
in Eqs. (4.3) and (4.4). In general, there is a plane for the measurements instead 
of a line, so that the spectrum of the data is a triple Fourier transform with respect 
to two spatial variables and time. Similarly to Eqs. (4.3)–(4.6), in Eqs. (4.5)–(4.6) 
it is implicit that the measurements are taken from the air-soil interface. The data 
sampling and resolution limits available in this framework are of the same order 
as the homologous 2-D version (Persico 2014). However, in comparison to a 2-D 
model, a 3-D version is able to provide a visualisation with respect to the required 
transect, i.e. the theoretical maximum allowed spacing between adjacent B-scans. 
(4.5)
O
�
x′, y′, z′
� =
+∞ˆ
−∞
+∞ˆ
−∞
+∞ˆ
−∞
ˆˆˆ
DF(k1, k2,ω)
× exp �jk1x′� exp �jk2y′� exp

jz′
�
4ω2
v2
− k2.

dk1 dk2 dω
(4.6)O
(
x′, y′, z′
) = ∂
∂z′
+∞ˆ
−∞
+∞ˆ
−∞
dF
(
x,y,
2
√
(x−x′)2+(y−y′)2+z′2
v
)
√
(x − x′)2 + (y − y′)2 + z′2
dx dy.
22 G. Manacorda et al.
This is expected to be of the order of �s min
4
, i.e. the same as the maximum allowed 
spatial step of the data along any single B-scan. In other words, in order not to lose 
information, and avoid spatial aliasing phenomena, the measurement points should 
form a square grid on the air soil interface. From a practical point of view, the 
 theoretically required transect is an important result. In fact, while there is usually 
no problem in guaranteeing a spatial step of �s min
4
 along a single B-scan (the GPR 
is in most cases equipped with an odometer that allows, in most cases, an even 
 narrower spatial sampling rate), to gather data with a transect of �s min
4
 is impractical 
with a single bi-static antenna system. To quantify, if the relative dielectric permit-
tivity of the soil is 4 and the maximum usable frequency is 500 MHz (which are, on 
average, two “favourable” hypotheses that relax the requirement) then the transect 
should not be larger than 7.5 cm. For practical reasons, transects used in the “real 
world” are normally of the order of 50 cm.
As already explained in par. 3.1, the development of GPR systems equipped 
with antenna arrays offer significant advantages because they allow the automatic 
gathering of several parallel B-scans at reciprocal distances of few centimetres, 
which allows the saving of time, and better spatial sampling of the datum.
4.4 Final Remarks on GPR Processing
In this overview, the most commonly used processing techniques have been 
described. There are, however, other focusing processes that complement the 
migration algorithm, namely tomographic inversions, both linear (Persico et al. 
2005; Persico 2006; Meincke 2001) and nonlinear (Qin and Cakoni 2011). Linear 
inversion algorithms are usually based on the SVD of the linearized scattering 
operator. The linearization can be achieved in several ways, depending of the spe-
cific assumption adopted for the case at hand. The most applied linearization’s 
are the Born (Chew 1995), Rytov (Devaney 1981) and Kirchoff approximations 
(Liseno et al. 2004).
Whatever the adopted method, in general, a linear inversion is theoreti-
cally more refined than a migration and, in particular, it does not require it to be 
assumed that soils are lossless and that targets are electrically deep. On the other 
hand, it is computationally much more cumbersome than a migration because 
the SVD of the relevant linear operator is, in general, not known in a close form 
and has to be implemented numerically. In general, with a “medium power” com-
puter, it is possible to invert an investigation domain that is a few wavelengths 
(computed at the centre frequency) in extent, both in the abscissa and the depth 
directions, whereas the length of a B-scan can be of the order of hundreds of 
wavelengths. This problem can be addressed by combining a sequence of sepa-
rately reconstructed investigation domains. However, optimising the combina-
tion is in itself a problem that, if not correctly treated, can affect the quality of the 
achievable result (Persico and Sala 2014).
23Design of Advanced GPR Equipment …
A further niche aspect of processing GPR data is the possible application of 
tomographic inversion using nonlinear processing. This method is theoretically 
more realistic than a linear inversion, because the physical nature of the scatter-
ing process is also nonlinear. However, the computational effort is considerable, 
and nonlinear inversions are, usually, applied only in particular cases (and most 
often for the purpose of studying the algorithm rather than solving a practical 
problem). Computation times usually restrict investigations to domain sizes of the 
order of a very few square wavelengths (Qin and Cakoni 2011). Finally, any non-
linear inversion has to consider the problem of false solutions caused by the local 
minima generated by, possibly, minimising a non-convex cost-function (Persico 
et al. 2002).
This section closes with a reference to forward modelling, which consists of 
postulating a scenario that provides artificial data similar to those recorded. By 
adjusting the parameters of the simulation, the result is driven to be progressively 
more similar to the real measured data. This approach can be useful in particular 
cases when the data and/or the available a priori information allow a reasonably 
precise hypothesis on the nature, size and disposition of any buried anomalies to 
be constructed to provide the basis for the forward modelling process. There are 
several available codes for this, the best known of which is, probably, GPRmax 
(Giannopoulos 2003). However, forward modelling cannot be exploited as a gen-
eral method, because (as stated at the beginning of this section), the problem at 
hand is ill-posed, which means that it is possible for two different buried scenarios 
to provide similar data.
5 Advanced Equipment for Civil Engineering Applications
5.1 Introduction
Ground-penetrating radar is one of a number of non-destructive inspection 
methods that is gaining acceptance in a wide range of applications in the Civil 
Engineering sector.
It has become a viable means of detecting and mapping buried infrastructure, 
locating reinforcement bars (rebars), voids and cracks in concrete, inspecting 
foundations, measuring asphalt layers’ thickness in roadways and airport runways, 
evaluating the fouling of ballast in rail-tracks, etc.
Given the diverse nature of all these applications, requirements for building 
effective equipment are very different from case to case and a detailed, exhaustive 
description for all of these is not within the scope of this chapter; thus, the follow-
ing paragraphs describe three examples of advanced GPRs
1. a dense array system designed for mapping underground assets;
2. a sparse array, high resolution system for the evaluation of bridge decks;
3. a continuous wave, reconfigurable GPR.
24 G. Manacorda et al.
5.2 Dense Array GPR
One of the main objectives in developing any new GPR system is to reduce the 
time needed to collect data while maintaining the performance in terms of targets’ 
detection; this is a key point in the design and is the origin of the 3-D GPR con-
cept, i.e. a system which is able to detects all targets within the surveyed swath, 
irrespective of their layout.