Concealed Explosives Detection using Swept Millimetre Waves

Smith, Sarah Elizabeth (2012) Concealed Explosives Detection using Swept Millimetre Waves. Doctoral thesis (PhD), Manchester Metropolitan University.

Preview

Available under License Creative Commons Attribution Non-commercial No Derivatives. Download (21MB) | Preview

Abstract

The aim of this project is to develop a system for the stand-o detection (typically ten metres) of concealed body-worn explosives. The system must be capable of detecting a layer of explosive material hidden under clothing and distinguishing explosives from everyday objects. Millimetre wave radar is suitable for this application. Millimetre Waves are suitable because they are not signi cantly attenuated by atmospheric con- ditions and clothing textiles are practically transparent to this radiation. Detection of explosive layers from a few mm in thickness to a few cm thickness is required. A quasi optical focussing element is required to provide su cient antenna directivity to form a narrow, highly directional beam of millimetre waves, which can be directed and scanned over the person being observed. A system of antennae and focussing optics has been modelled and built using designs from nite element analysis (FEA) software. Using the developed system, represen- tative data sets have been acquired using a Vector Network Analyser (VNA) to act as transmitter and receiver, with the data saved for processing at a later time. A novel data analysis algorithm using Matlab has been developed to carry out Fourier Transforms of the data and then perform pattern matching techniques using arti cial neural networks (ANN's). New ways of aligning and sorting data have been found using cross-correlation to order the data by similar data slices and then sorting the data by amplitude to take the strongest 50% of data sets. The signi cant contribution to knowledge of this project will be a system which can be eld tested and which will detect a layer of dielectric at a stando distance, typically of ten metres, and signal processing algorithms which can recognise the di erence 17 between the response of threat and non-threat objects. This has partially been achieved by the development of focussing optics to acquire data sets which have then been aligned by cross-correlation, sorted and then used to train a pattern matching technique using neural networks. This technique has shown good results in di erentiating between a person wearing simulated explosives and a person not carrying simulated explosives. Further work for this project includes acquiring more data sets of everyday objects and training the neural network to distinguish between threat objects and non-threat objects. The operational range also needs increasing using either a larger aperture optical element or a similarly sized Cassegrain antenna. The system needs adapting for real time use with the data processing techniques developed in Matlab. The VNA is operated over a band of 14 to 40 GHz, future work includes moving to a stand-alone transmitter and receiver operating at w-band (75 to 110 GHz).