Classification of Threat and Nonthreat Objects Using the Magnetic Polarizability Tensor and a Large

This article describes the development of a large-scale multicoil arrangement capable of characterizing the magnetic polarizability tensor (MPT) of large threat objects such as firearms. The system has been applied to the measurement of a comprehensive range of weapons made available by the National Firearms Centre of the U.K. For comparison, a number of nonthreat items such as metallic belt buckles, keys, and coins have also been characterized. Clear differences in the magnitude and spectroscopic response of the MPT data for different firearm types and nonthreat items are presented. The application of unsupervised machine-learning (ML) algorithms to MPT data of threat and nonthreat objects enables a better understanding of target object classification. The presented results are encouraging as they demonstrate the ability of the MPT used in combination with the adopted classification algorithms to robustly discriminate between threat and nonthreat objects.