Key words : Demining, Ultra-Wideband Ground penetrating radar, Time domain modelling

Ground Penetrating Radar (GPR) designates the family of radar systems that image the sub-surface. GPR is considered as a promising technology for the detection of AP mines. The potential for detecting non-metallic objects in the ground makes the GPR complementary to a metal detector. At the moment, the use of commercially available GPR in real demining operations is negligible, mainly due to three drawbacks. The first drawback concerns the antennas. In most cases, GPR antennas have a low directivity and therefore perform best when they are in contact with the ground, which is for safety reasons unacceptable for the application. The second and third drawback is the limited depth and cross-range resolution. Therefore, conventional GPR can have difficulties in discriminating the target echoes from the air-ground interface and from other mine-like targets. For the application of demining, more resolution is needed, which means a larger bandwidth of the system.

The thesis reports on the possible use of the ultra-wide band (UWB) GPR as demining sensor. In a first step a pair of dielectric-filled TEM horns is developed. The demining application imposes specific requirements to the GPR antennas. The most important are the mobility of the antennas and the bandwidth. The antennas have to be small, light-weighted and directive, so that they can be used off-ground. The design of the TEM horns is based on the wire model. In this model the antenna plates are replaced by a set of wires. The model permits to simulate accurately the antenna pattern as well as the antenna impedance. In order to reduce the physical size of the antennas and to improve the directivity, the TEM horns are filled with a dielectric. It is shown that the dielectric-filled TEM horns are capable of radiating and receiving very fast transient pulses (< 300 ps). In a second part of the work the antennas are integrated in an instrumentation GPR, using mainly off-the-shelf equipment. The system uses frequencies between 1 and 5 GHz, with a central frequency of 3 GHz. The upper cut-off frequency seems to be a good compromise between the resolution of the system and the attenuation of the higher frequencies in the ground. The whole system is modelled in the time domain by considering it as a cascade of linear responses, resulting in a time domain equivalence of the radar range equation. The time domain radar range equation allows us to calculate the received voltage as a function of time at the receiver in terms of the radar -, ground - and target characteristics. The time domain model is used e.g. to optimise the offset angle for the GPR antennas and to calculate the range performance of the UWB GPR system. It is shown that the moisture content of the soil limits drastically the range performance of the UWB system. A key element in the time domain modelling of the radar is the description of the antennas by means of the normalised IR, totally characterising the antennas in the time domain. Another important term in the time domain radar range equation is the IR of the lossy ground. In this work we propose an analytical expression of the impulse response, modelling the propagation in the ground. A last main contribution was made in the domain of 3D signal processing. Migration is a technique to focus reflections in the recorded data into their true position and physical shape. Most of the existing migration techniques do not take into account the characteristics of the acquisition system and the ground. As we dispose of a good time domain description of our UWB GPR, we propose a novel migration method that integrates the time domain model of the UWB GPR in the migration scheme. We calculate by forward modelling a synthetic 3D point spread function of the UWB GPR, i.e. a synthetic 3D scan of a small point scatterer. This 3D point spread function, containing system characteristics like the waveform of the excitation source, the combined antenna footprint and the IR of the antennas, is then used to deconvolve the recorded data. Results of this migration method on real data obtained by the UWB GPR system show that the migration method is able to reconstruct the top contour of small targets like AP mines, in some cases even with the correct dimensions. The method is also capable of migrating oblique targets into their true position. The migration scheme is not computational intensive and can easily be implemented in real time.

From the experience we have obtained by working with UWB GPR for mine detection, we learned that the difficulty of detecting small objects in an inhomogeneous background is often underestimated. The UWB GPR will most likely never be used as a stand-alone mine detector, but always in combination with other sensors. An advantage of the UWB GPR over a conventional GPR is the ability of detecting shallow buried objects and the gain in cross-range resolution. We recommend to record 3D data. After the adapted image processing, the 3D images have enough resolution for extracting the shape of the object. This additional information on location and shape of the target will reduce the number of false alarms and thereby speed up the mine clearance.