Umberto Spagnolini
Dip. Elettronica e Informazione
Politecnico di Milano

I. Introduction

Ground penetrating radar (GPR) is a short-range pulse system for remote sensing which measures short-pulse electromagnetic (EM) reflections due to variations of the electrical properties of the investigated medium. Twenty years ago, GPR was considered as just an interesting tool with only a few applications, and now it has evolved into a commercially available technology.

GPR represents a simple and fast way of providing information for ground mapping. The increasing number of applications of non-intrusive analysis has been the key of the success of GPR in the past few years. The currently available commercial equipment is low-cost, portable (or, at least, small enough to be mounted on a vehicle) and equipped with a user-frendly PC interface. In addition, simple processing algorithms allow a continuous display of all resulting profiles, which make the interpretation of the results more intuitive in most of the cases. All applications that use monostatic or bistatic systems are mainly devoted to detection and/or estimation of objects and interfaces beneath the ground surface or, in general, located in a visually opaque structure.

II. Characteristics of GPR

GPR technology is mainly application-oriented. Pulse width varies from 100 ns (low resolution medium-range applications – within 10-40 m) to 0.5 ns (very-short range applications – below 0.5 m). In order to obtain a reasonable resolution, the bandwidth of the overall radar system must be at least one octave (ideally one decade), with a linear phase response. The need of reducing linear radiation distortion limits the choice of the antenna. Commercial GPR systems often use TEM horn and bow-tie antennas. Apart from some specific applications (airborne radar), the antenna is close to the ground in most GPR systems. As a consequence, the performance of the system depends on how accurately the interaction between the antenna and the ground has been modeled. Ground attenuation and reflections at the interface between layers are some of the most important factors that affect the radar signal-range. Radar penetration and resolution tend to be reduced by the EM attenuation, with a range that goes from a few meters in conductive media, to 50 m at most for low-conductivity (below 1 mS/m) materials (sand, gravel, rock and fresh water). High-power pulses are essential in order to increase the ratio between signal and clutter (or noise). For this reason, instead of using a pulse radar, one may use swept-FM or step-frequency transmitters. Even though frequency synthesizers are still costlier than pulse radars, they seem to overcome some of the limitations of pulse radars which makes them interesting for future developments. In order to improve the overall performance of GPR systems, research needs to focus on all signal processing aspects. This will allow us to leap from a qualitative level to the quantitative one.

III. Signal Processing

Most of the basic signal processing tools such as filtering, waveform averaging, time-varying gain, clutter and ringing reduction by optimum filtering (matched and/or Wiener filter) are available in most commercial GPRs and are enough to fulfill most of the requirements. The detection of voids, pipes, buried objects and interfaces rely on the spatial transient given by the change in
permittivity. To correlate echoes’ time delay with depth and measure the target’s location, the value of permittivity must be known. The estimation of the permittivity profile can be carried out in bistatic experiments using the synthetic aperture methods to correlate echoes’ time delay with local wavefront curvature. In monostatic experiments the permittivity estimation exploits simultaneously amplitude and time delay of echoes but assumes a lossless medium. In addition, there is a wide range of other possibilities that are mainly based on travel times, such as travel time tomographic methods. Single channel monostatic and bistatic images are distorted; unfocused images of subsurface structure are due to the apparent position shift associated with dipping interfaces and to diffractions from small objects and edges. A focused and correctly positioned target can be obtained by using any of the migration algorithms developed for seismic data. Kinematic similarities between GPR and seismic wave propagation can be exploited by adapting most of the seismic processing to be routinely used in GPR processing (the main change required is rescaling). Geometrical optics and nondispersive propagation should hold for kinematical processing; this occurs in the frequency range currently used in GPRs when conductivity is below 10 mS/m. Even if reliable quantitative analysis using global inversion algorithms and tomographic methods seems to be limited to some simple cases, increasing computing power will provide the necessary framework for their routine use. Their reliability needs to be fully assessed in GPR applications.

IV. Applications

The areas of application for GPR are diverse. The non-destructive tests with GPR are used in civil engineering for void detection, prediction of concrete deterioration from variations of its permittivity by the presence of moisture and chloride, pavement profiling for programmed road and bridge deck maintenance, reinforcing bar location, subgrade deterioration in railroad and airport runways. The objects’ detection and classification by their extent and permittivity contrast with the overburden is crucial in environmental engineering. It is useful for mapping hazardous wastes and buried contaminant containers, imaging and monitoring subsurface contaminants (e.g., gasoline and other hydrocarbon fuels). Stratigraphic and bedrock mapping are essential in geothecnical, archaeological, and hydrogeological applications, in site characterization, in mining planning (e.g., borehole profiling), in tunnel excavations, ice thickness profiling, and permafrost mapping. GPR has been proven to be useful for detection of metallic and nonmetallic buried mines. Pipe and shallow object detection are widely used to avoid their damage in excavation. Persons buried by avalanches can also be detected with GPR. Further applications of GPR to investigate an area (3D mapping) instead of a line are now becoming more feasible.