Aspect independent detection and discrimination of concealed metal objects by electromagnetic pulse induction: a modelling approach

The work presented in this thesis describes the research, modelling and experimentation which were carried out so as to explore the use of electromagnetic pulse induction for the detection of nearby or on-body threat items such as handguns and knives. Commercially available finite difference time domain electromagnetic solver software, Vector Fields, was used to simulate the interaction of a low frequency electromagnetic pulse with different metal objects. The ability to discriminate between objects is based on the lifetime of the induced currents in the object, typically around 100 (μs). Lifetimes are different for a different objects, whether they are weapons or benign objects. For example hand grenades, knives, and handguns are clearly threat objects whereas a wrist watch, mobile phone and keys are considered benign. Electromagnetic pulse Induction (EMI) relies on generating a time-changing but spatially uniform magnetic field, which penetrates and encompasses a concealed metallic object. The temporally changing magnetic field induces eddy currents in the conducting object, which subsequently decay by dissipative (i.e. resistive) losses. These currents decay exponentially with time and exhibit a characteristic time constant (lifetime) which depends only upon the size, shape and material composition of the object, whilst the orientation of the object is irrelevant. This aspect independence of temporal current decay rates forms the basis of a potential object detection and identification system. This thesis investigates the possibility of detecting, resolving and identifying multiple objects if they are close together, for example located on an individual. The mathematical analysis used for the investigation implements the generalised pencil of function (GPOF) method. The GPOF algorithm decomposes the signal into a discrete set of complex frequency components; providing the capability to obtain the time constants from data. It was possible to effectively count and identify multiple metallic objects carried in close proximity providing that the objects do not have very similar time constants. The simulation results, which show that multiple objects can be detected, resolved and identified by means of their time constants even when they are close together, are presented.