Computational methods for processing ground penetrating radar data

Abstract

The aim of this work was to investigate signal processing and analysis
techniques for Ground Penetrating Radar (GPR) and its use in civil
engineering and construction industry. GPR is the general term applied to
techniques which employ radio waves, typically in the Mega Hertz and Giga
Hertz range, to map structures and features buried in the ground or in manmade
structures. GPR measurements can suffer from large amount of noise.
This is primarily caused by interference from other radio-wave-emitting
devices (e.g., cell phones, radios, etc.) that are present in the surrounding
area of the GPR system during data collection. In addition to noise, presence
of clutter – reflections from other non-target objects buried underground in
the vicinity of the target can make GPR measurement difficult to understand
and interpret, even for the skilled human, GPR analysts.

This thesis is concerned with the improvements and processes that can
be applied to GPR data in order to enhance target detection and
characterisation process particularly with multivariate signal processing
techniques. Those primarily include Principal Component Analysis (PCA)
and Independent Component Analysis (ICA). Both techniques have been
investigated, implemented and compared regarding their abilities to separate
the target originating signals from the noise and clutter type signals present in the data. Combination of PCA and ICA (SVDPICA) and two-dimensional
PCA (2DPCA) are the specific approaches adopted and further developed in
this work. Ability of those methods to reduce the amount of clutter and
unwanted signals present in GPR data have been investigated and reported
in this thesis, suggesting that their use in automated analysis of GPR images
is a possibility.

Further analysis carried out in this work concentrated on analysing the
performance of developed multivariate signal processing techniques and at
the same time investigating the possibility of identifying and characterising
the features of interest in pre-processed GPR images. The driving idea
behind this part of work was to extract the resonant modes present in the
individual traces of each GPR image and to use properties of those poles to
characterise target. Three related but different methods have been
implemented and applied in this work – Extended Prony, Linear Prediction
Singular Value Decomposition and Matrix Pencil methods. In addition to
these approaches, PCA technique has been used to reduce dimensionality of
extracted traces and to compare signals measured in various experimental
setups. Performance analysis shows that Matrix Pencil offers the best
results.

Date of Award	May 2013
Original language	English
Awarding Institution	University of Portsmouth
Supervisor	Branislav Vuksanovic (Supervisor), Alan Hewitt (Supervisor) & Boris Gremont (Supervisor)