Identification of buried unexploded ordnance from broadband electromagnetic induction data

A procedure is described for computing range and orientation invariant spectral signatures of buried unexploded ordnance (UXO) from electromagnetic induction (EMI) data. The normalized eigenvalues of the magnetic polarizability tensor that characterizes the target response are used as the orientation-invariant spectral signatures. It is shown that the eigenvalues can be normalized with respect to depth under the assumption that a multiplicative scale factor can be applied at all frequencies. The eigenvalues are derived by measuring the matrix elements of the polarizability tensor from above-ground spatial data and then by diagonalizing this matrix. This method is linear, and does not require a nonlinear parameter search. After normalizing for depth, the eigenvalues derived from an unknown object can then be compared with library eigenvalues using the L2 norm as a goodness-of-fit measure. The procedure is demonstrated using data obtained from cylinders and UXO at different orientations.