THE GPR FOR DETECTION OF BURIED LANDMINES

 

 

The Ground Penetrating Radar (GPR) is one of the most promising technologies for detection and identification of buried landmines. The choice of the central frequency and the bandwidth of the GPR is an important issue, and depends on the particular application, mostly of the kind of object one is looking for: size of the object, electrical properties, depth of the object, properties of the soil... For each application a set of frequency constraints can be developed. Conventional GPR however are mostly designed for geophysical applications and use central frequencies below 1GHz. As landmines are small objects, larger bandwidths are needed for a better depth resolution and detailed echo. Unfortunately, when mines are buried too deep and the frequency is too high, it is possible that we don’t detect anything at all, because of the dramatically increased attenuation of the soil with frequency. So to be sure to detect something, low frequencies are preferred. A solution can be found in the use of a ultra-wideband GPR's in order to benefit from both low and high frequencies. The use of wideband systems involves some technical problems. Critical points are the UWB antenna's and receiver.

 

 

The operating principle of a Ground Penetrating Radar is straightforward. It sends EM waves into the ground and samples the backscattered echoes. An EM wave sent into the ground will be backscattered on any dielectric discontinuity in the ground.

The majority of the GPR’s use an impulse time domain waveform, although recently GPR’s are also developed in the frequency domain. In the time domain radar’s there are two major categories: the amplitude modulated and the UWB GPR. The first one sends a pulse with a carrier frequency. This carrier frequency is modulated by a square envelop. In order to achieve a good depth resolution it is important that the duration of the modulated pulse is as short as possible. In the case of a modulated pulse, the shortest pulse one can send is a mono cycle and that’s what’s usually done in a conventional GPR. The central frequency varies from some MHz up to 1 GHz. The second type of time domain GPR is the Ultra-WideBand GPR. In order to obtain a good depth resolution, one has to use an very large bandwidth. Such need for large bandwidth has led to the development of a UWB radar. The impulse sent by the radar has no carrier. The width of the carrier free impulse is in the order of some 100ps. The shape of the impulse can vary, but typically a Gaussian pulse is used.

As there was no UWB GPR, for the demining application, commercially available at the beginning of the project, we decided to try out existing GPR systems. The systems we tested are the RAMAC 1GHz, the GSSI 1GHz horn antenna , GSSI 2.5GHz horn antenna and the ERA 1GHz and 2GHz. It is not our intention to compare the different GPR systems; we only want to put in evidence the benefit of the UWB approach. The data were analysed and the following conclusions could be drawn:

 

 

LABORATORY UWB GPR SYSTEM

 

To study the advantages and short-comings of an UWB GPR, we decided to develop a laboratory version of such a system. The system components are mainly off-the-shelf laboratory equipment. Only the antennas are developed in RMA. A schematic representation of the UWB GPR system is given in figure 1.

 

Figure 1 : schematic representation of the UWB GPR system

 

The system consists of the following parts: on the transmitting side a Picosecond Pulse Labs step-generator is used. The generated step has an amplitude of 10 Volts, a rise-time of 45 ps and a high waveform purity. This step is than transformed by an impulse-forming-network to an impulse with a maximal amplitude of 2.5 Volts and a FWHM of less than 100 ps. Using an impulse as input signal for the antennas works better than using a step. It limits (in time) the direct coupling between the antennas. The generated impulse is represented in figure 2.

 

Figure 2 : output signal after the impulse-forming-network

 

 

The impulse is fed to a Tx - Rx antenna pair, which are designed for this large bandwidth (the design of these antennas is discussed in the next paragraph). On the receiver side, a 6GHz digitizing oscilloscope is used to measure the backscattered signal. The oscilloscope has an internal delay line, a 14 bit resolution and can average up to 10.000 times, to obtain a higher dynamic range. The oscilloscope data is collected by a computer, using a GPIB bus. The antenna pair is mounted on a xy-table of 2m by 2.5m and 2m high (see figure 3). The xy-table is computer-controlled and has a displacement precision of less than 0.1 mm. In the scanning area of the table, two sandboxes are placed, 1.5m by 1.5m each and 0.8m deep. The first one is filled with sand, the second one with loam. The permittivity of both types of soil is fully characterised in function of frequency and moisture content.

The whole set-up is controlled by a computer, that commands, via a serial connection, the position of the xy-table scanning head and meanwhile collects the data (A-scan) from the 6GHz oscilloscope. The program that runs on the computer allows the user: to visualise the B-scan, to perform basic signal and image processing and to save the data to disk.

 

Figure 3: set-up of the UWB GPR system.

 

 

In order to reduce the physical size of the antenna, without changing the bandwidth and to approve the directivity, the antenna was filled with a dielectric (). Antenna measurements revealed indeed that the antennas were more directive, the frequency coverage moved towards the lower frequencies but the gain was inferior to the air-filled antennas. In a third stage the antenna plates were etched on a printed circuit board (PCB) and terminated by a 50 Ohm load, by putting two 100 Ohm SMD resistors in parallel between the antenna plate extremities. We also replaced the antenna plates by a set of 41 wires (see figure 4). The distance between the wires is too small to influence the antenna characteristics, but it forces the currents to be radial and it limits the surface of conducting metal. The latter is very important, when using the antennas in combination with a metal detector. Due to the lamination of the antenna plates, the metal detector will be only weakly influenced by the presence of the antennas. Etching the antenna plates also limits the weight of the plates and increases the precision in fabrication.

 

Figure 4 : the bottom- and upper-plate of the

antenna etched on a PCB.

Figure 5 : the Tx-Rx TEM Horn antenna pair

 

The difference between the upper and lower antenna plate, shown in figure 4, functions as a wideband balun. In previous realisations of a TEM Horn, measurements revealed a reflected unbalanced current component on the coax exterior, by the antenna feed. As a consequence, the coax feedline was also acting as an antenna. To avoid this, an ultra-wideband (frequency-independent) balun is needed, to feed the antennas. The principle of this balun is based upon an electrostatic reasoning. The taper in the bottom plate provides a gradual transition from an unbalanced set-up (upper antenna plate on a groundplane), towards a balanced configuration (two symmetrical antenna plates). The coaxial feedline fits into the hole in the bottom plate, and the inner conductor is connected to the upper plate. Measurements of the S11 parameter of the antenna showed the wideband property of the balun and antenna.

Figure 5 represents the final realisation of the dielectric-filled TEM Horn, etched on PCB and terminated on 50W . The dimensions of the TEM horn antennas are small: L=12 cm, =30° and =14.5°, which leads to a physical antenna aperture is 12cm by 6cm. The angles and were calculated so that the antenna matches the 50 W driving cable. The measured characteristic impedance along the antenna is varying progressively from 55W at the antenna feed (the unbalanced part of the balun) towards 65W at the end of the antennas (the balanced part of the antenna).

 

 

 

 

EXPERIMENTAL RESULTS AND DISCUSSION

 

 

 

The UWB GPR system can now be used to acquire B-scans of AP mines buried in the sand- and loambox. As an example three images of B-scans are shown (figure 8-10) and discussed. On the horizontal axis the displacement of the antennas is given in centimetres ; the vertical axis represents the time in ns, which is proportional to the depth. The antennas were moved at 30 cm above the ground. Now signal processing was performed on the B-scans, except an average removal on figure 10.

 

 

 

Figure 8 : PMN mine buried in sand at a depth of 1 cm.

 

The mine appears very clearly on the image and can be distinguished from the air-ground interface, despite of the fact that it is shallow buried. With a conventional GPR system, the echo from the mine would disappear in the echo from the air-ground interface.

 

Figure 9: PMN mine buried in wet loam at a depth of 3 cm.

 

As expected, high frequencies are strongly attenuated, specially in a wet ground. A PMN mine is normally an easy mine to detect and yet with 3 cm of wet loam on top of it, it is hardly visible

One can compensate for this attenuation by a time-varying gain in the receiver part of the system. This is usually done in commercial GPR systems. In an oscilloscope however, this option is not available.

Figure 10 : M35 mine buried in loam at a depth of 1 cm.

 

The M35 is very small APmine (dia. of 6cm). To enhance the image, a clutter reduction was performed by subtracting from each A-scan the average of all the A-scans. The result can also be improved by bringing the antennas closer to the ground. Lowering the antennas means reducing the antenna footprint on the ground, so more energy will be reflected by the small mine. The 3 dB footprint of the antennas at 20 cm off ground is approximately an ellipse with axes 15 cm by 20 cm. Reducing the antenna footprint can also be done by making antennas with a larger antenna aperture. But this means larger and heavier antennas, so a trade-off has to be made.