Feature-based Detection and Discrimination at DuPont's Lake Success Business Park, Connecticut

Abstract

The objective of this demonstration was to determine if laser-positioned, high-density EM61 data acquired in a moving survey mode could support feature-based discrimination decisions for a canopied site in Bridgeport Connecticut. Inversion results were benchmarked using cued data acquired in a controlled, gridded approach over 40 targets. The site was seeded with inert UXO, ranging in size from 37mm to 105mm projectiles. We inverted the measured EM61 field data assuming a point dipole source and performed classification using a support vector machine classifier. We compared parameter estimates derived from both data sets and evaluated overall discrimination performances. Results indicated that the laser-positioned EM61 data were not of sufficient quality to discriminate UXO from non-UXO. Of the 58 seeded items that were surveyed, 41 were ranked as high confidence ordnance, 10 were ranked as high confidence clutter, and 7 remained unclassified.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Jan 01, 2007
Accession Number
ADA478316

Entities

People

  • Dean A. Keiswetter

Organizations

  • Leidos

Tags

Communities of Interest

  • Human Systems
  • Sensors
  • Weapons Technologies

DTIC Thesaurus Topics

  • Base Closures
  • Data Acquisition
  • Data Analysis
  • Department Of Defense
  • Detection
  • Detectors
  • Explosives
  • Inertial Measurement Units
  • Magnetometers
  • Measurement
  • Munitions
  • Performance Tests
  • Supervised Machine Learning
  • Test And Evaluation
  • Unexploded Ammunition
  • Uxo Detection
  • Warning Systems

Readers

  • Military/Explosive Ordnance Disposal (EOD) Technology
  • Statistical inference.

Technology Areas

  • AI & ML
  • Directed Energy