Detecting and discriminating unexploded ordnance (UXO) in the underwater (UW) environment presents additional challenges relative to that in terrestrial environments. In particular, UW sites contain large amounts of environmental clutter, debris, and obstacles, such as pilings, crab pots, anchors, coral, and trash. The high conductivity of sea water severely limits the efficacy of geolocation systems that employ radio frequency (RF) transmission (e.g., global positioning systems [GPS]). Usually UW sites, because of the dynamic nature and unfavorable RF propagation characteristics, are more complex and challenging than terrestrial sites; therefore, even with current technologies it is difficult to obtain the desired centimeter-level position accuracy to support advanced discrimination processing techniques using magnetic and electromagnetic (EM) data. Moreover, positioning in the UW environment relies heavily on the ability to accurately measure the geometry of a towed array or periodic GPS re-acquisition by unmanned UW vehicles – both of which lead to the propagation of positional errors in survey data. Thus, to enhance UW munitions detection and reliable ordnance discrimination and to reduce remediation costs of UW UXO cleanup, new geolocation sensors with centimeter-level position accuracy and sufficient signal-to-noise ratio are needed.

The objective of this project was to explore and develop a new low frequency EM sensor modality for fast and accurate geolocation based on the measured vector magnetic field, which would be sufficiently robust and efficient to be used in the real field for UW UXO detection and discrimination systems tracking and positioning.

Technical Approach

This project concentrated on the fundamental mathematical, physical, computer simulation, and potential practical implementations aspects of the approach for UW geolocation. Namely two techniques using low frequency magnetic field were explored – the first technique was based on the vector magnetic field full tensor gradient estimations at a given point and the second used the non-linear optimization algorithm based on the differential evolution (DE) approach. This project studied the sensitivity of the vector magnetic field gradient estimations using the standard finite different approach. In addition, the three dimensional electromagnetic induction solver was utilized based on the method of auxiliary sources for estimating the noise due to a spherical and spheroidal UXO-like targets. The accuracy with which the system can estimate a transmitter’s location, robustness with respect to noise, and requirements with regard to data quality and quantity were also studied.


The results of this study illustrate that both vector magnetic field full tensor gradient and DE techniques have the potential to provide centimeter-level UW geolocation. However, when the primary magnetic field signals are contaminated with random noise due to UW metallic targets, water conductivity/frequency changes, and transmitter size, the performance of the vector magnetic field full tensor gradient approach degrades significantly compared to that of the non-linear DE optimization technique. In addition, the number of receivers (Rx) required by the vector magnetic field tensor gradient technique and its sensitivity with respect to sensor separations prevented the researchers from further considering this technique for UW geolocation, leaving the non-linear approach that uses only three vector Rx as the technique of choice for tracking the location of UW interrogation sensors with centimeter-level accuracy.


The benefits of this research and follow-on work include high accuracy geolocation for UW systems. Accurately determining UW sensor position will enable high resolution mapping of UW sites, as well as provide information for improved characterization and feature extraction of UW targets.

  • UXO Detection and Classification ,

  • Geolocation ,

  • Differential Evolution (DE) ,

  • Sensors ,

  • Electromagnetic Induction (EMI)