Department of Computer Science and Electrical Engineering

Land Mine Identification Using Microwave Resonance Signatures

David Kinnon

Abstract:
It has been proposed that plastic land mines can be identified based on their physical properties using Ground Penetrating Radar (GPR). GPR distinguishes landmines from extraneous subterranean clutter by virtue of their dielectric constant. Classification patterns can be formulated by identifying the Complex Resonance Signatures (CNRs) of the targets. CNRs are comprised of the real and complex parts of the respective resonance frequencies.
The backscattering from targets of particular length-to-diameter ratio will resonate at specific frequencies. This resonance is unique to that particular length-to-diameter ratio. The larger the length-to-diameter ratio, the larger the exhibition of resonance. Hence long-thin targets can be more easily identified that circular targets such as land mines.
Experiments involving thin copper wires of lengths 13, 14 and 15cm were conducted using a Network Analyser - anechoic chamber set-up. The wires were oriented parallel to the polarisation of the antennas. It was observed that the backscattering contained complex resonances at wavelengths that could be factored, using integer multiplication, into the length of the wires.
The wires were then tested with azimuths of 45 degrees and 90 degrees and elevations of 0 degrees and 45 degrees to test the polarisation and orientation independence of CNRs respectively. It was observed that the resonance frequencies remained constant as expected. The magnitudes of the complex resonances (residues) decreased.
Auto-regressive spectral estimation techniques, based on the original and least squares Prony algorithms, were applied to the transient response of the backscatter to determine the CNRs. It was observed that the least squares technique provided a very high degree of accuracy. The original Prony algorithm proved fallible in the presence of noise.
Experiments involving surrogate land mines were conducted. Analysis showed no evidence of CNRs. The feasibility of plastic and mine detection and classification using CNRs, based on the results published in this dissertation, is thus questionable. However, CNRs of unexploded ordnance is worthy of further investigation.

Complete thesis:
thesis.pdf

Additional material:
conference.PDF

About the Author