Liam A Marsh

KNN classification of metallic targets using the magnetic polarizability tensor

Walk-through metal detectors are used at check points for preventing personnel and passengers from carrying threatening metallic objects, such as knives and guns, into a secure area. These systems are capable of detecting small metallic items, such as handcuff keys and blades, but are unable to distinguish accurately between threatening objects and innocuous items. This paper studies the extent to which a K-nearest-neighbour classifier can distinguish various kinds of metallic objects, such as knives, shoe shanks, belts and containers. The classifier uses features extracted from the magnetic polarizability tensor, which represents the electromagnetic properties of the object. The tests include distinguishing threatening objects from innocuous ones, classifying a set of objects into 13 classes, and distinguishing between several similar objects within an object class. A walk-through metal detection system is used as source for the test data, which consist of 835 scans and 67 objects. The results presented show a typical success rate of over 95% for recognizing threats, and over 85% for correct classification. In addition, we have shown that the system is capable of distinguishing between similar objects reliably. Overall, the method shows promise for the field of security screening and suggests the need for further research.

Non-contact multi-frequency magnetic induction spectroscopy system for industrial-scale bio-impedance measurement

Biological tissues have a complex impedance, or bio-impedance, profile which changes with respect to frequency. This is caused by dispersion mechanisms which govern how the electromagnetic field interacts with the tissue at the cellular and molecular level. Measuring the bio-impedance spectra of a biological sample can potentially provide insight into the samples properties and its cellular structure. This has obvious applications in the medical, pharmaceutical and food-based industrial domains. However, measuring the bio-impedance spectra non-destructively and in a way which is practical at an industrial scale presents substantial challenges. The low conductivity of the sample requires a highly sensitive instrument, while the demands of industrial-scale operation require a fast high-throughput sensor of rugged design. In this paper, we describe a multi-frequency magnetic induction spectroscopy (MIS) system suitable for industrial-scale, non-contact, spectroscopic bio-impedance measurement over a bandwidth of 156kHz-2.5MHz. The system sensitivity and performance are investigated using calibration and known reference samples. It is shown to yield rapid and consistently sensitive results with good long-term stability. The system is then used to obtain conductivity spectra of a number of biological test samples, including yeast suspensions of varying concentration and a range of agricultural produce, such as apples, pears, nectarines, kiwis, potatoes, oranges and tomatoes.

Measurement system for determining the magnetic polarizability tensor of small metal targets

This paper presents an apparatus to measure the spectroscopic magnetic response of small metallic objects and deduce the magnetic polarizability tensor. The measured transimpedances of a .222 Remington rifle cartridge and titanium cube are compared to simulated results and are found to match well providing verification of the method. The eigenvalues of the two objects are calculated and discussed highlighting the potential discriminatory aspect. The results support the proposed use of the eigenvalue spectra to provide subsurface classification and discrimination between landmines and clutter.

Magnetic Polarizability Tensor Spectroscopy for Low Metal Anti-Personnel Mine Surrogates

The magnetic dipole polarizability tensor is an object-specific property possessing information about the size, shape, and material. This information could be used by electromagnetic induction sensors typically used for demining operations to discriminate between buried mines and clutter, reducing false alarm rates, and improving demining throughput and safety. This paper presents a methodology capable of obtaining the spectroscopic tensors of small metallic objects and low metal anti-personnel mine surrogates. The experimental results are validated against simulated and analytical solutions to ensure that the obtained tensor truly represents the absolute object tensor. Absolute tensors for a number of typical clutter items and mine surrogates are presented, with significant variance observed between those of mine and clutter.

Determination of material and geometric properties of metallic objects using the magnetic polarisability tensor

A walk-through metal detector system has been used for measuring the magnetic polarisability tensor for a variety of metallic objects. We propose a method for classifying objects by their metallic composition using features of the tensor. Furthermore, we investigate the potential of using the tensor representation as an indication geometric properties of the object. The method used is shown to be accurate for classification of material composition. Furthermore, the results suggest that it is possible to use the tensor to distinguish between similar objects of different sizes in limited scenarios. These findings demonstrate the potential for this method, but also suggest the need for further studies.

Estimating magnetic polarizability tensor of buried metallic targets for land mine clearance

This paper addresses the problem of identifying metallic objects in buried landmines and discriminating them from clutter using low frequency electromagnetic induction (EMI) techniques. From dipolar fields, the magnetic polarizability tensor extracted from the target response can be used as a basis for identification. Here, a deterministic nonlinear optimization method is presented to estimate target polarizability matrix and location by fitting a dipole model to EMI data collected above target in a least squares sense. Using finite element simulated data with added synthetic low frequency noise (10 dB SNR), results show initial guess misestimating target position with few centimeters in the transversal (x, y) plane can be corrected very close to the true location. The method is also able to estimate the polarizability tensor to within 12 % error of the true tensor. Keywords-component; Electromagnetic Induction, landmines, UXO, Magnetic polarizability, nonlinear inverse problems.

Spectroscopic identification of anti-personnel mine surrogates from planar sensor measurements

The electromagnetic response of metallic targets is known to vary as a function of frequency. In this paper we demonstrate the ability to measure these frequency-dependent variations for buried metallic targets from stand-off sensor measurements, with a focus on its potential application to humanitarian demining. A planar measurement system is presented which is capable of measuring the trans-impedance between a transmit-receive coil pair at five simultaneous frequencies. These measurements have been calibrated against a known ferrite target to yield absolute spectroscopic responses. Three targets are presented; two surrogate anti-personnel landmines and a bullet casing. These measurements are compared with previously recorded spectra using a separate measurement system, showing an agreement in the range of 15 to 25%.

Design of electromagnetic sensor arrays optimised for inversion of the magnetic polarisability tensor

This paper presents a method for the simulation of sensitivity maps from an array of coils. Some of the criteria necessary for designing a coil array capable of inversion of the magnetic polarisability are examined, and sensitivity maps are analysed with this in mind. The summarised sensitivity map for a single optimised array is presented, as well as the results of noise testing of an inversion algorithm using simulated measurements. Finally, a method for the construction and testing of such an array is presented.

Improving reliability for classification of metallic objects using a WTMD portal

In this paper, a walk-through metal detection (WTMD) portal is used for classification of metallic objects. The classification is based on the inversion of the magnetic polarisability tensor (tensor) of the object. The nature of bias and noise components in the tensor are examined by using real walk-through data, and consequently, a novel classifier is introduced. Furthermore, a novel method for detecting poorly inverted tensors is presented, enabling self-diagnostics for the WTMD portal. Based on the results, the novel methods increase the accuracy of metal object classification and have the potential to improve the reliability of a WTMD system.

Evaluation of the thin-skin approximation boundary element method for electromagnetic induction scattering problems

A conductive object subject to an applied varying magnetic field will emit a secondary magnetic field or scattered field. The scattered field is dependent on the geometry of the object and its material properties (conductivity, permeability, etc). If we can calculate the scattered field for a given geometry (the scattering problem), we can infer the material properties from the detected electromagnetic response. Our motivation is the production of induction based classifiers for object and material classification. Applications include sorting of high value scrap metal and identifying UXO from clutter in landmine clearance. To this end, we require methods of solving the scattering problem quickly and accurately. In this paper, we evaluate the thin-skin approximation boundary element method. The method offers a particularly compact formulation of the scattering problem which is quick to solve. We compare this method to the more established finite element method. We find that larger objects at higher frequencies and conductivities appear to give good agreement between the two methods. However, the agreement breaks down for smaller objects even when the frequency or conductivity is relatively high for typical induction based sensing. This is especially true when the object has a complex geometry. This imposes limitations on the practical usefulness of this approach.