-
oa Assessment of the Cooperative Source Concept for Single Target Classification using EM63 Metal Detector
- Publisher: European Association of Geoscientists & Engineers
- Source: Conference Proceedings, 24rd EEGS Symposium on the Application of Geophysics to Engineering and Environmental Problems, Apr 2011, cp-247-00101
Abstract
The objective is to assess UXO risk reduction capabilities of a new cooperative source concept. the basic requirement is to develop cost-effective technologies to discriminate munitions of various types from the wide range of non-hazardous items buried in the ground. Classification techniques based on diagnostic target attributes such as principal axis polarizabilities have made significant progress in recent years. However significant limitations remain due to the limited Information content on available sensors. Current development of EMI devices is based on adding transmitters or receivers. Our approach makes use of Geonics EM63 metal detector and a cooperative source. Cooperative source is an object of regular shape and known composition that operates synergistically with a conventional EMI transmitter to illuminate a buried target from a wide range of azimuth positions. the method is based on the mutual inductive coupling between the cooperative source and a buried target after both have been illuminated by the transmitter. Data obtained consists of EMI transients as a function of the cooperative source position. Datasets are dimension-reduced by extracting feature vectors based on the moments of the EMI transients. the feature vectors are loaded into an unsupervised statistical classifier, which in our case is based on a merge self—organizing map (MSOM). Upon output, the primary target is classified as either a target—of—interest (toI), warranting further attention, or else an item not of interest (non—toI) requiring no further attention. Preliminary results show that mutual coupling between cooperative source and target is measurable and strongly dependent on the geometry and orientation of the target, resulting in extraordinarily rich dataset. Unsupervised MSOM learning shows feature vectors clustering from which a-posteriori toI classification can be performed.