Discrimination Using the Geonics EM63 in a Cued Interrogation Mode at Fort McClellan, AL

Abstract

Data were collected at the former Fort McClellan to test the Geonics EM63 deployed in a cued interrogation mode. Except for one 37mm and several 60mm seed items, all munitions encountered were 75mm or 3.8 inch shrapnel rounds. The EM63 surveys were cued off production mode EM61 data. Polarization tensor models were fit to each surveyed anomaly. Ground truth information was used to train a statistical classifier. For small and medium items, an early time decay rate, equivalent to one that could be obtained with an EM61, provided almost as much discrimination potential as the late time decay rate. The slow rate of data acquisition and the decay curve analysis as good or better than inversion for a physics based model suggest the benefit of the extra information extracted from the EM63 is not justified by the greater costs since some level of discrimination can be obtained without complex inversion and statistical classification, using simpler instrumentation. Better instrumentation and rigorous inversion and statistical classification would still be expected to provide a richer variety of feature vectors to significantly improve discrimination.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Mar 01, 2009
Accession Number
ADA513102

Entities

People

  • Stephen Billings

Tags

Communities of Interest

  • Sensors
  • Weapons Technologies

DTIC Thesaurus Topics

  • Accuracy
  • Computational Science
  • Data Mining
  • Data Processing
  • Detection
  • Detectors
  • Electromagnetic Induction Sensors
  • Explosives
  • Feature Extraction
  • Information Science
  • Kernel Functions
  • Magnetometers
  • Pattern Recognition
  • Supervised Machine Learning
  • Unexploded Ammunition
  • Uxo Detection
  • Warning Systems

Readers

  • Military/Explosive Ordnance Disposal (EOD) Technology
  • Regression Analysis.

Technology Areas

  • AI & ML