Baixe o app para aproveitar ainda mais
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.
Compartilhar