D UTILITY MAPPING USING ELECTRONICALLY SCANNED ANTENNA ARRAY Egil S. Eide and Jens F. Hjelmstad Department of Telecommunications Norwegian University of Science and Technology, N-79 Trondheim, Norway eide@tele.ntnu.no jens.hjelmstad@tele.ntnu.no ABSTRACT Utility mapping using a single-antenna GPR is a timeconsuming operation especially when large areas are to be covered. The data acquisition can be performed much more efficient by using an electronically scanned antenna array. The stepped-frequency D GPR that is developed at the Norwegian University of Science and Technology uses a meter wide antenna array that consists of transmit/receive antenna pairs. The main application of the system is utility mapping of streets where the underground consists of a jungle of cables and pipes. The data form the meter wide swath is focused into a D image cube using D wavenumber migration. By combining several parallel swaths, it is possible to generate underground maps of the whole street at different depths. The radar has successfully been used for mapping of pipes, cables and old tramlines in Trondheim during. The wide bandwidth ( MHz.6 GHz) gives high enough resolution to map the asphalt thickness and the details of the base layers in addition to the utility lines. In this way, the data from a survey can serve more than one customer. The results from the field tests demonstrate the high user potential of D imaging compared to standard D GPR profiling. Key words: Antenna array, ultra-wideband radar, wavenumber migration, utility mapping. D DATA ACQUISITION USING A SCANNING ANTENNA ARRAY The RadioStar is an ultra-wideband ground-penetrating radar system that was developed at the Norwegian University of Science and Technology (Eide, ). The radar can be programmed to operate in the frequency range from MHz to. GHz using a steppedfrequency (SFCW) waveform. The radar is designed to operate with an antenna array for efficient D data acquisition of GPR data as shown in Figure. The antenna array consists of pairs of transmit/receive bow-tie monopoles mounted on a common ground plane (Eide, ). For low-frequency coverage, the GPR requires antennas of a certain size, but this makes it difficult to space the antennas close enough for proper spatial sampling at the highest frequency. By using a fractal structure consisting of bow-ties with different sizes, we have obtained an array that both gives sufficient low-frequency radiation and fulfils the Nyquist sampling theorem in space. The antenna array is mounted on a lightweight trailer that is pulled by a vehicle along the street being surveyed. In this way, data can be collected on an. x. centimeter grid at a velocity of m/s depending on the integration time required at each antenna element. During operation, the radar uses one T/R antenna pair at the time and records a radar trace at each spatial point. INTRODUCTION In urban areas, excavation work is a very timeconsuming and risky operation because the documentation of existing cables and pipes are very poor or even incorrect. In many areas there is a significant need for subsurface mapping of the utility infrastructure under the streets. Utility mapping using a single-antenna GPR or traditional cable seeking equipment is a time-consuming operation especially when large areas are to be covered. Data acquisition can be performed much more efficient by using an electronically scanned antenna array. Figure : D data acquisition using the electronically scanned antenna array. 9 th International Conference on Ground Penetrating Radar, Santa Barbara, CA, 9 April May,
SIGNAL PROCESSING The received data are processed by the following processing steps. The pre-processor performs filtering, resampling, source deconvolution and amplitude equalization between each antenna elements. This is especially important for the image quality. The radar images are focused using D Stolt migration techniques (Binningsbø, ) (Lopez-Sanhez, Fortuny- Guasch, ). Since the attenuation of electromagnetic waves increases with the frequency, one should expect that the high frequency part of the spectrum have a lower penetration. This can be easily seen in most GPR data where resolution of the deep reflector is always much less than for shallow objects. Equation () contains the radar equation for a lossy medium. Pr Pt GrGtζ rζ tλ σ αr e (π ) R = () Here, G r and G t represent antenna gains for the transmit and receive antennas, while ζ r and ζ t are the transmission losses through the ground surface. Assume that the radar transmits Watt and that the antenna gain is 5dBi for both antennas over the whole frequency range. Assume that the RCS of the target σ =.5 m and that α corresponds to soil attenuation varying between db/m at GHz and 7 db/m at GHz. Figure shows the received power versus frequency for different depths. Note that for depths larger than meter, only the low frequency portion of the spectrum is above the receiver sensitivity. For optimum processing of deep data, one should therefore suppress the high frequency spectrum that otherwise only contributes with noise at large depths. In a stepped-frequency radar the raw data is in the frequency domain, and it is easy to apply a window function that is matched to the expected spectrum at different depths. In practice, one generates two or more sub images using IFFT s with different frequency weighting. Figure shows a D profile which is processed with full bandwidth ( 6 MHz). In Figure the deepest part of the full-bandwidth image is replaced with a low frequency windowed sub-image ( MHz). Note that the composed image has improved signal-to-noise level and reduced resolution at depths below. meter. Received power, dbm - - -6-8 - - R=.m R=.5m R=.m R=.5m - R=.m Radar sensitivity -6 R=.5m -8 6 8 6 8 frequency, MHz Figure : Received spectrum for different depths.... 5.6.8. 5...6 5.8.. 6 8 6 8 5 Figure : Migrated D profile processed with full bandwidth ( 6 MHz).... 5.6.8. 5...6 5.8.. 6 8 6 8 5 Figure : Migrated D profile where SNR from data below. meter is improved by reduced bandwidth processing. SURVEY RESULTS Utility mapping The radar has successfully been used for mapping of pipes, cables and old tramlines in Trondheim during. The radar was operated in the frequency range MHz.6 GHz. Figure 5 shows a map of the street intersection that was surveyed during the summer. Below the asphalt, the rails from a closed down tramline are still lying. 9 th International Conference on Ground Penetrating Radar, Santa Barbara, CA, 9 April May,
Road mapping The wide bandwidth ( MHz.6 GHz) also gives high enough resolution to map the asphalt thickness and the details of the upper base layers in addition to the utility lines. In this way, the data from a survey can serve more than one customer. Figure 9 shows a vertical profile of a test road near Trondheim where various kinds of base materials have been used for the construction. Part of the road was equipped with a number of pressure and deformation transducers for monitoring the road deformations during a one-year period with traffic. Figure 5: Map of the street intersection being surveyed. Figure 6 shows horizontal slices at three different depths while Figure 7 shows a vertical profile passing through the manhole. The tram rails are clearly visible and well focused at a depth of centimeter while a pipe or cable can bee seen at. meter depth. Note that the tram rails have been cut during an earlier excavation of the area. Figure 8 shows a photo of a similar excavation done not far away from the area. The soil in this area consists mostly of clay and wet silt with high attenuation; hence the GPR penetration was about.5 meter. One should expect to see sewer pipes connecting to the manhole in the center of the image, but these pipes are always buried below meter i.e. below the penetration depth CONCLUSIONS The results show that an ultra-wideband GPR can accurately map the shape and location of buried objects like cables and pipes. It is possible to reduce the noise level at large depths by applying a depth dependant frequency window to suppress high-frequency noise at depths where only the low frequencies are above the receiver noise floor. The end user of utility mapping data is mainly interested in knowing only the position and depth of the cables and pipes under the street. Hence, the main future challenge is to develop computer algorithms that automatically can trace and extract the different objects and features from the complex -D image cube. During the last years, much work is done to develop this kind of software (Bernstein et. al, ). ACKNOWLEDGMENTS The work has been funded through a research grant from Witten Technologies Inc., Boston, USA. The authors thank Mr. Jakob Haldorsen, Schlumberger-Doll Research for helpful advises. Mr. Sigve Tjora and Mr. Pål-Aanund Sandnes are also acknowledged for their assistance with the field experiments and data processing.. Figure 8: Photo of an excavation close to the surveyed area. The old tramlines are visible just below the asphalt layer. 9 th International Conference on Ground Penetrating Radar, Santa Barbara, CA, 9 April May,
z = 8 cm 6 8 6 8 z = cm 6 8 6 8 Z = cm 6 8 6 8 Figure 6: Horizontal slices at three different depths.....6.8....6.8. 6 8 6 8 Figure 7: Vertical profile from the street (note that the street has been rebuilt with a convex surface to drain off water. The old street plane can be seen as an interface layer below the asphalt layer. 6 8 6 8 6 8 9 th International Conference on Ground Penetrating Radar, Santa Barbara, CA, 9 April May,
cm Asphalt Crushed rock Leca Material Sand Pressure and deformation sensors m Pressure cell Cable Horizontal slice Figure 9: Asphalt thickness and subbase mapping. The test road was equipped with pressure and deformation cells at various dephts. REFERENCES Bernstein, R., Oristaglio, M., Miller, D.E., and Haldorsen, J.,. Imaging radar maps underground objects in D, IEEE Computer Applications in Power, p.. Binningsbø, J., Eide, E. S., and Hjelmstad, J. F.,. D migration of GPR array-antenna data. In Eighth International Conference on Ground Penetrating Radar, SPIE Vol. 8, pp 59 6. Eide, E.,. Radar Imaging of Small Objects Closely Below the Earth Surface, Doctoral Thesis, Norwegian University of Science and Technology, August. Lopez-Sanhez, J. M., and Fortuny-Guasch, J.,. -D Radar Imaging Using Range Migration Techniques, IEEE Trans. on Antennas & Propagation, Vol. 8, No. 5, pp. 78 77. 9 th International Conference on Ground Penetrating Radar, Santa Barbara, CA, 9 April May,