The objective of this project was to test and evaluate procedures for target detection and classification in the context of the Environmental Security Technology Certification Program (ESTCP) Classification Pilot Program Live Site Demonstrations. The procedures were developed in Strategic Environmental Research and Development (SERDP) projects MR-1658, MR-1711 and MR-2100. Demonstrations were performed using data from ESTCP demonstration projects at the Southwestern Proving Ground, the former Fort Bliss, the former Camp Hale, and the former Camp Beale.
Detection is based on applying the standard dipole inversion model as a filter over the entire survey site. Locations where the model fits the measured data well are target locations. Classification decisions are based on two parameters easily calculated from magnetic polarizabilities of unknown targets and targets of interest. The parameters are measures of the mismatch between the strength and the shape of the respective polarizability curves.
Results for the Southwestern Proving Ground demonstration were excellent. At the stop-dig threshold all TOI were correctly identified and 18 clutter items had been marked for digging, leaving 474 clutter items (96.3%). The performance for the Fort Bliss and Camp Hale demonstrations was significantly poorer due to inadequate training data. Had the TOI library included examples of the difficult targets, the performance would have approached that demonstrated with the Southwestern Proving Ground data.
The basic problem is adequate training data and a complete TOI library. Performance at Fort Bliss would have been significantly improved had the target library included an example of one of the legacy 37mm rounds found at the site. This is not the only issue, however. Some munitions items (e.g. the 40mm illumination round found at Camp Hale) have polarizabilities which are very similar to those of munitions debris such as fins and fuzes. If such items are present, then the goal of significant reduction in clutter digs while identifying all TOI can be compromised.