A primary goal of the unexploded ordnance (UXO) research community is to develop technologies that find and locate buried UXO. Geophysical sensors work by detecting and mapping measured quantities that result from changes in properties of the physical material. Thus, the detectors do not really find and locate UXO. Instead, the detectors find and locate changes in the magnetic field or the presence of an anomalous electromagnetic field that could be caused by buried UXO. However, the measured fields are not uniquely associated with UXO and rapidly become complex when signatures from individual objects overlap.

Physics-based modeling and analysis procedures have been shown to aid discrimination of UXO from clutter based on the derived source parameters. Because the models assume spatially discrete target signatures, the accuracy of the inversion results is dependent on the complexity of the data submitted to the fitting routines. In cases where the signatures from multiple targets overlap, the procedures do not work well in determining whether or not they are UXO. Better methods are needed to extract the individual target parameters from a multiple target signature.


The objective of this project was to develop advanced iterative techniques for inverting magnetic and electromagnetic data for situations in which the signatures from two targets overlap.

Demonstration Results

The methods developed were the two-dipole iterative residual algorithm and a simultaneous two-dipole fit algorithm the demonstrators named “double happiness.” On synthetic magnetometer data, the double happiness algorithm performed well, and better than the iterative residual method. On synthetic electromagnetic data, the double happiness algorithm performed better than the iterative residual method, but differences in the assumed sensor (EM61MkI versus EM61MkII) made the direct comparison slightly ambiguous.

On overlapping signature data collected at the Blossom Point, Maryland test facility with an EM61MkII, the iterative residual method performed reasonably well for the 4 cases of target pairs separated by 0.5m, but there was not enough data with real horizontal separations to adequately evaluate the algorithm.

On the synthetic data, both algorithms had some difficulty separating two dipoles that are close in horizontal separation, since the signature from the compact targets will appear like a single dipole. Fortunately, this situation is less important in practice than the situation in which a weaker target lies some distance from a stronger target, but the two signatures overlap and the weaker target is not detected. In such situations, traditional methods will locate the stronger target, but its excavation may miss the weaker one. In this case, the double happiness algorithm may perform more than adequately.

Implementation Issues

The benefits of this approach relate to better estimates of source parameters for those cases where the geophysical signatures overlap. Because classification decisions are based on the estimated source parameters, it is anticipated that more accurate fit parameters will improve the ability to differentiate UXO from non-UXO items in these situations.