Classification of unexploded ordnance

In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Date Abstract ii Abstract In this thesis I investigate methods for discriminating between unexploded ordnances (UXOs) and clutter items (e.g: shrapnel, geology). I first describe a numerical forward model, the method of auxiliary sources (MAS), which can be used to model the magnetic and electromagnetic response of a conductive, permeable body. I use this model to validate the connection between the parameters of approximate forward models and target properties (i.e target shape). I also examine how model parameters can be estimated from observed data using inversion. I then describe algorithms for discriminating between UXO and clutter. In the statistical classification framework, model parameters are basis vectors within a multi-dimensional feature space. I prioritize features based upon their ability to separate UXO and clutter using canonical analysis. I describe two approaches for partitioning the feature space: modelling the underlying distributions from which the observed feature data are drawn, or directly defining a decision boundary. A suite of statistical classifiers are then applied to magnetics data acquired at three field sites. Finally, I propose an algorithm for selecting a classifier as target excavation proceeds.