Highly Contaminated Uxo Sites: Combination of GPR and Emi for Discrimination of Clustered Scatterers

The most fundamental electromagnetic limitation on discrimination of subsurface unexploded<br>ordnance (UXO) during cleanup operations is that one must use quite low frequencies to penetrate the<br>ground. Operating between some 10’s of Hz and some 100’s kHz, electromagnetic induction (EMI)<br>sensor signals are sensitive to many aspects of target shape and composition. In that band, the signals do<br>not suffer the scattering and absorption loss problems that challenge ground penetrating radar (GPR).<br>However, EMI transmitted wavelengths are many, many orders of magnitude greater than the size of<br>targets of interest. This means that distinct targets cannot readily be picked out by timing the arrival of<br>echoes or by noting the direction they are coming from, as for wave phenomena. Clustered targets will<br>respond simultaneously and their signals overlap. This is a particularly important problem because most<br>UXO cleanup sites contain much metallic clutter. The number of targets and their locations are hard to<br>tell from EMI data only.<br>Our full-polarimetric UWB GPR operates between some 10’s of MHz and about 800 MHz, i.e. at<br>a low enough frequency to penetrate the soil, minimize scattering losses, and elicit essential target<br>resonances, but necessarily too low to form precise target images. What GPR can often do, however, is<br>time the arrival of target echoes from distinct targets, even when they are clustered, and feed into EMI<br>processing some crucial information on number of targets, approximate locations, and other geometrical<br>data. Altogether, EMI signal optimization constrained by GPR data produces separate EMI signature<br>patterns for each item, indicating whether the object is UXO-like or not.<br>Traditional fast EMI forward modeling contains too many free parameters, which is a serious<br>challenge to inversion algorithms, especially for multiple targets. In this paper we propose a three step<br>approach for UXO discrimination: (1) preliminary screening with GPR information to identify or rule<br>out obvious UXO candidates; (2) Analyze EMI data with simple dipole model, using GPR information<br>as prior information. The results are again used to identify or rule out obvious UXO candidates; (3) For<br>cases where final decisions can not be made in step one and two, a pattern matching approach is<br>employed to identify each candidate UXO, using the first two step results as prior information. Study on<br>examples illustrates how this three step approach may help improve UXO discrimination and<br>identification.