Several data-processing techniques for geophysical survey data have been developed for discriminating between unexploded ordnance (UXO) and non-UXO items. Typically the first step of these methods is the recovery of a set of parameters that specify a physics-based model representing the object under interrogation. Electromagnetic induction (EMI) responses depend nonlinearly on the subsurface object’s location and orientation, and inverting for these parameters quickly and accurately is an important part of inversion techniques in use by the UXO community.

The objective of this project was to develop new non-traditional physics-based inverse approaches for determining the location and orientation of buried objects.

The project focused on the fundamental aspects and potential practical applicability to the UXO problem of three new physics-based approaches: (1) pole-series expansions (PSE), (2) energy gradients, referred to as the HAP approach, and (3) use of the physical properties of left-handed media (LHM). All three methods assume that targets’ responses can be approximately reproduced by a set of magnetic dipoles distributed in particular points inside the objects. The algorithms are implemented using a numerical technique called the Normalized Surface Magnetic Source (NSMS) model.

The researchers studied the accuracy with which the methods can estimate an object’s location, orientation, and equivalent magnetic polarizability, their robustness with respect to noise, their computational speed, and their requirements with regard to data quality and quantity.

Several existing data sets collected by the Geophex GEM-3 and the Geonics EM-63 and by next-generation sensors like the TEMTADS array, the MPV-TD and the GEM-3D+ were utilized. The new algorithms were applied to these data sets in order to estimate locations, orientations, and equivalent magnetic polarizability tensors of buried objects; in addition the researchers conducted discrimination studies using the NSMS model to demonstrate the applicability of these techniques.

The researchers determined that all three techniques have the potential to estimate position and orientation at different frequencies. However, when objects’ responses are contaminated with additive noise, the performance of the pole-series-expansion and LHM methods degrade significantly compared to that of the HAP technique. In addition, the computational time required by the PSE and LHM approaches and similar practical considerations prevented the researchers from further testing and developing those two techniques, leaving the HAP as the method of choice for determining buried objects’ locations and orientations.

Determining the location and orientation of buried objects are the most time-consuming, nonlinear inverse problems in UXO discrimination tasks. Once the objects’ locations and orientations are determined, the EMI inverse scattering problem simplifies to determine the physics-based model parameters, such as polarization tensor or amplitudes of the responding sources in the NSMC. Therefore, the development of a fast and accurate physics-based methodology for determining objects’ locations and orientations has the potential to enhance discrimination ability when processing EM data. These improvements in discrimination may lead to fewer excavations of non-UXO geophysical anomalies, thus reducing cleanup costs.