Surface and Buried Mine Detection with Variance-Based Multispectral Data Fusion

Abstract

The detection of surface or buried landmines is complicated by effects of occlusion due to overlying vegetation and confusion of the mine spectral response due to overlying obscurants. Such effects often cause mine detection systems that are strictly model-based to fail in field practice due to brittleness resulting from lack of input coverage. University of Florida has assisted Frontier Technology, Inc. in analyzing, developing, and implementing prototype algorithms and software for buried land mine detection using infrared imagery. This technique detects statistical differences between target and background regions and produces and estimate of target probability at a given location and, where possible, an estimate of target identity. University of Florida has assisted FTI in analyzing its Tabular Nearest Neighbor Encoding paradigm that has been highly successful in detecting small targets imagery and other signatures. TNE has been applied to mine detection problems to investigate its utility in producing increased probability of detection and decreased rate of false alarms. This technique has been found to be useful in a wide variety of military and commercial applications.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Nov 10, 2000
Accession Number
ADA387243

Entities

People

  • Gary Key
  • Mark Schmalz

Tags

Communities of Interest

  • Air Platforms
  • Materials and Manufacturing Processes
  • Sensors
  • Weapons Technologies

DTIC Thesaurus Topics

  • Algorithms
  • Automated Target Recognition
  • Coding
  • Data Fusion
  • Detection
  • Detectors
  • False Alarms
  • Land Mines
  • Multispectral
  • Notation
  • Pattern Recognition
  • Performance Tests
  • Probability
  • Recognition
  • Statistics
  • Target Recognition
  • Warning Systems

Readers

  • Computer Vision.
  • Regression Analysis.
  • Sensor Fusion and Tracking Systems.