Comparison of NDTI Methodologies to Classify Infrastructure in an Urban Battlespace From Thermal IR Signatures

Abstract

A major challenge of remote infrastructure classification is to distinguish pavements from roofs. We proposed that, due to their differing thermal masses, these elements will heat and cool differently. Therefore, their thermal infrared (IR) signatures will differ throughout a 24-hour cycle. We propose a method to rapidly identify urban infrastructure that will assist the Future Force in dominating the urban battlespace to the same degree that the current force dominates open terrain. We developed a fundamental thermal heating model for these infrastructure elements, and tested the model on spectral data gathered by NASA's Advanced Thermal and Land Applications Sensor (ATLAS) instrument. ATLAS is a 15-channel multispectral scanner that includes both the standard visible and near-IR bands, and adds additional bands in the middle reflective infrared and thermal infrared range (TIR). Image processing techniques were applied to the imagery, focusing on the thermal bands (1015). A new indicator was proposed and tested in this research, the Normalized Difference Thermal Index (NDTI). Development and verification of the thermal model was described in a paper presented at the 25th Army Science Conference (McInerney et. al., 2006). The NDTI is based on the premise that materials have different spectral responses, so that ratioing the significantly distinguishing bands will provide a sensitive and comparable test of thermal character. In analyzing the results, the desire was to have the roofs and pavements appear distinctly separate. Statistical analysis of the NDTI technique was conducted to find the range, mean, and standard deviation of cells that were of either the pavement or roof type. NDTIs from these bands produce well-defined images that accurately identify ground-truth points, as shown in the image of downtown Atlanta.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Dec 01, 2008
Accession Number
ADA503518

Entities

People

  • Jonathan Trovillion
  • Michael J McInerney
  • Robert Lozar

Organizations

  • Engineer Research and Development Center

Tags

Communities of Interest

  • Sensors

DTIC Thesaurus Topics

  • Abstracts
  • Battlespace
  • Data Science
  • Detectors
  • Image Processing
  • Images
  • Information Processing
  • Information Science
  • Infrastructure
  • Materials
  • Pavements
  • Planetary Sciences
  • Remote Sensing
  • Standards
  • Statistical Analysis
  • Statistics
  • Urban Areas

Readers

  • Image Processing and Computer Vision.
  • Systems Analysis and Design
  • Thermal Physics or Thermal Science.