The objectives of this project were to mitigate the effects of discrete metallic clutter from electromagnetic induction (EMI) data and the simultaneous discrimination of multiple anomalies in close proximity with overlapping EMI responses. The project aimed to develop a robust approach for processing high quality data from next generation EMI instruments to mitigate the effects of clutter by isolating their EMI signatures after locating them individually and then discriminate unexploded ordnance (UXO) from non-UXO targets at highly contaminated sites that include challenging terrain, vegetation, and geology using rigorous models that may include interaction effects. Specific objectives included:

  • Develop an N-target estimator able to provide estimates of the number of targets present in the sensor’s field of view along with their locations and orientations, without resorting to computationally expensive optimizations.
  • Formulate robust classifiers that segregate N-targets into UXO and non-UXO.
  •  Discriminate UXO-like targets using rigorous (NSMS, SEA) models that explicitly include coupling between targets if required.

Technical Approach

Three new methods for localizing multiple sources in close proximity using EMI data were developed, tested in lab settings, and applied to field data. These methods were:

  1. A multiple dipole search method based on a Gauss-Newton gradient search algorithm utilizing an analytical Jacobian.
  2. A Joint Diagonalization (JD) method for quickly estimating the number of targets presented in the data.
  3. An Orthonormalized Volume Magnetic Source (ONVMS) method.

Canonical targets of various shapes, sizes, and material parameters have been fabricated. Data acquired from these targets as well as standard UXO targets has been acquired by the TEMTADS and MPV2 instruments in many multitarget configurations. The methods developed under this project have been able to isolate and discriminate up to six targets simultaneously in the case of lab data.


The JD method was able to almost instantaneously provide a good estimate for the number of distinct targets in the EMI data.  After this estimate is obtained, the other methods are used to invert for the parameters of the identified targets. The combined JD ONVMS approach was applied to data acquired from several instruments and from several different live site demonstrations. For example, when applied to the data from Camp Beale, this combined method was able to find all the Targets of Interest (TOI) with few false positives. Additionally, this method found all the small fuze targets in the midst of up to five other pieces of clutter. No other method or group was able to find all of the fuzes.


Using the methods demonstrated in this project, researchers and industry can more confidently leave innocuous scrap in the ground during munitions response projects, saving time and resources.