Applied Mathematics Seminar: Paul Ledger (Keele University)

Applied Mathematics Seminar

Speaker: Paul Ledger (Keele University)

Title: Characterisation and classification of metallic threat objects for security screening Abstract: The location and identification of hidden conducting threat objects using metal detection is an important yet difficult task. Applications include security screening at transport hubs and finding landmines and unexploded ordnance in areas of former conflict.

Based on an asymptotic expansion of the perturbed magnetic field, we have derived an economical object description called a magnetic polarizability tensor (MPT), which is a function of the object’s size, shape, material properties and the exciting frequency. We employ a finite element method, accelerated by a reduced order model and scaling results, to efficiently compute a large dictionary of invariants of MPT spectral signature characterisations of realistic threat and non-threat objects relevant for security screening.

The talk will also discuss the performance of probabilistic and non-probabilistic machine learning classifiers for identifying hidden threat objects learnt from our dictionary.