While current electromagnetic induction (EMI) sensor technology has shown the ability to detect and localize buried metal objects of a wide range of sizes and at relatively deep depths, the false alarm rate (FAR) is high because the sensor fails to discriminate unexploded ordnance (UXO) from metal objects that pose no risk. Improved discrimination algorithms are needed to reduce the FAR and thus reduce the cost of UXO remediation.
The objective of this project was to investigate signal processing techniques to improve UXO discrimination capability. Two signal and data processing techniques were explored—time decay signature analysis using active EMI time-domain data and spatial EMI data signal processing using a near-field holographic imaging technique.
The project used spatial and wide bandwidth time decay response EMI data gathered at the U.S. Army Blossom Point Test Site to develop the discrimination algorithms. This data set included EMI responses from buried and exposed UXO, UXO simulants, and clutter. The fact that the target is excited with spatially varying magnetic fields can be exploited by taking spatial magnetic data over a target. These different fields tend to excite the different magnetization modes of the target. By combining these spatial and time decay response variations, a library of known target responses was developed for the discrimination algorithm. The three-dimensional holographic imaging algorithm is a back-propagation algorithm that allows the buried object to be localized. The algorithm also provides an estimate of the target’s size.
A time decay signature analysis and parameterization library for a large number of UXO targets was developed. Researchers completed a description of the Levenberg-Marquardt non-linear analysis and the nearfield holographic imaging analysis, as well as a description of holographic processing of UXO target signatures. The Levenberg-Marquardt non-linear analysis, when applied using the algorithms developed to define the time decay and the number of poles necessary to describe that time decay, permits discrimination between UXO and clutter and between different types of UXO; however, initial work with simulated targets illustrated difficulties with near-field holographic imaging techniques. For example, even with the best sampling intervals selected, the holographic code predicted a depth that was usually slightly less than the true depth of the target. The surface field needs to be sampled carefully, based on the characteristics of the target beneath, for optimum depth retrieval.
The goal of the target discrimination algorithms is to reduce the FAR associated with the use of time-domain EMI type sensors. A 50% improvement in the FAR is expected to result in 50% fewer non-UXO targets being excavated. This translates into reduced remediation time and subsequently lower costs. (SEED Project Completed – 2002)