A review of GPR for landmine detection

The GPR is beginning to be fielded as a sensor for mine detection where its ability against the minimum metal mine often surpasses the ubiquitous metal detector. The US HSTAMIDS and the UK-FRG MINEHOUND hand held detector have been extensively trialled and assessed for both humanitarian and military use. Throughout the world, airborne, vehicle mounted and hand held systems have been extensively researched, developed and trialled. The process has taken over two decades from the early systems devised for Vietnam and the Falkland Islands and has often been fragmented and intermittent. In most soils, GPR can detect mines at greater depths than the metal detector, but in clay or salt-laden soils it does not perform as well. However in some mineralised soils where the metal detector struggles, GPR has a performance advantage. This suggests that it is important that the processing built into a GPR sensor can recognise difficult soil conditions and alert the operator to potential performance degradation. Furthermore GPR has to overcome many potential sources of false alarm due to clutter, which include; large stones, animal burrows, cracks in the soil surface, pooled water in surface and subsurface hollows, tree roots, changes in surface topography and changes in vertical or lateral soil structure. As GPR is now reaching a level of technology readiness it is vital that it is tested in a way that exposes all aspects of performance. Many test procedures developed for metal detector technology are inappropriate for GPR which must take into account not just the mass of metal in the mine but the type of mine, surrogate, inert, etc., the size of mine, the internal structure of mine, the explosive content of mine, the depth of mine, the attitude of mine to the horizontal and the proximity to other targets such as AT mines. The testing of GPR systems should ensure that there is an adequate statistical distribution of AP and AT mines, GPR clutter, ground topography, soil conditions (water content, etc.), operator variance and product batch quality control. The performance bounds of GPR systems can be evaluated but as yet there is not the wealth of understanding of the effects of clutter that is common in conventional radar engineering. This is an important area, which needs to be understood better if predictions of the statistical performance of GPR are to be better quantified. There is still a tendency to evaluate GPR in very benign ground conditions and obtain excellent results only to find that in more realistic conditions the performance is lacking. Typically GPR system can achieve near PDs of 100% and PFA below 0.01% in well manicured test sites but this performance reduces to a PD 50% and PFA of 10% in normal ground conditions (normal that is to the deminer). The most successful developments have been where GPR is used in conjunction with other sensors (primarily the metal detector). The market for humanitarian mine detectors is fragmented and very cost sensitive and will be difficult to develop, but it is to be hoped that parallel developments and procurements for military applications will result in products being used to meet the humanitarian need. © 2006 Springer Science+Business Media, Inc.