This project will apply and extend work on advanced electromagnetic induction (EMI) models and sensors (e.g., BUD, TEMTADS, MPV, MetalMapper) for the discrimination of unexploded ordnance (UXO) in cluttered environments as part of the live site demonstrations. Algorithmic and methodological advances for the classification of munitions have been made in several SERDP-sponsored projects led by Sky Research and Dartmouth, in particular MR-1572 on advanced models, MR-1573 on magnetic soils, MR-1592 on subsurface target locations, and MR-1632 on multi-object discrimination. Upon its completion, this project aims to test and deliver flexible, reliable, robust, efficient, and effective advanced EMI models and signal processing algorithms that can be adapted to (virtually) any EMI sensor technology or site condition. Additionally, this project will provide supporting source codes and application programming interface with the dynamic link libraries (DLL), which would allow the algorithms to be accessed from industry standard Geosoft Oasis Montaj software.

This technology is based on the normalized volume/surface discreet, mutually coupled magnetic dipoles model and is applicable for single and multiple target discrimination. Specific techniques that will be used include: (1) robust methods and approaches to estimate the number of targets and to extract target discrimination features from EMI sensor data, in an effort to establish strength and limits of the advanced models; (2) tools and expertise to select robust classification features vectors for a specific live UXO site that can reliably and effectively discriminate hazardous targets of interest (TOI) from nonhazardous items; and (3) advanced models to estimate false negatives and false positives by investigating the sensitivity and specificity of decision making depending on UXO size, type, and material composition.