International Society for Industrial Process Tomography

Improving Concealed Metallic Object Characterization using Polarizability Tensors

P.D. Ledger^1*, W.R.B. Lionheart², A.A.S. Amad¹

¹College of Engineering, Swansea University, Swansea, Wales, U.K

²School of Mathematics, The University of Manchester, Manchester, U.K.

^*Email: p.d.ledger@swansea.ac.uk

ABSTRACT

The rapid location and identification of concealed conducting objects is important to a range of security applications and often employs metal detection technology. Our approach is aimed towards improving the location and identification of threats by using a complex symmetric rank 2 magnetic polarizability tensor (MPT) description for the low-cost characterisation of metallic objects. An explicit expression for the computation of the MPT is briefly reviewed and a series of numerical results are presented, which illustrate how its coefficients depend on the object size, shape, material parameters and frequency. The eigenvalues of the real and imaginary parts of the MPT at a finite set of frequencies are chosen for the purpose of identifying different objects and a naïve Bayesian classifier is introduced for classifying objects. The results of cross-validation of the classifier using simulated noisy data are included to demonstrate the success of the proposed classifier.

Keywords Metal detection, Eddy current problem, Magnetic polarizability tensor, Bayesian classification.

Industrial Application General