Buried Target Detection using a 3-band LWIR Airborne Sensor

Abstract

The comprehensive approach, using both spectral and spatial features, VBGMM for image segmentation to find ROIs, and RVM to classify ROIs that are targets works far better than the windowed RX detector alone in detecting buried targets. Pressure plates are the target best detected, and the 155 rounds are the most difficult to detect. GMMRX as a feature used in the comprehensive approach performs best at higher probability of detection levels, particularly when restricting the search zone only to the road. In a setting where targets are placed within a more heterogeneous background (i.e., off road or side of road), the RX feature would likely miss many more targets than GMMRX, providing a clear advantage in using GMMRX.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Feb 11, 2010
Accession Number
ADA519504

Entities

People

  • Levi Kennedy

Organizations

  • Signal Innovations Group, Inc.

Tags

Communities of Interest

  • Materials and Manufacturing Processes

DTIC Thesaurus Topics

  • Anomaly Detection
  • Bayesian Networks
  • Change Detection
  • Classification
  • Computer Vision
  • Detection
  • Detectors
  • Emissivity
  • False Alarms
  • Flight Paths
  • Image Segmentation
  • Intensity
  • Models
  • Probability
  • Target Classification
  • Target Detection
  • Template Patterns

Fields of Study

  • Physics

Readers

  • Computer Vision.
  • Semiconductor Device Technology