Pattern Recognition of Images in Atmospheric Turbulence.

Abstract

This report covers the two major activities investigated this year. (1) Preprocessing, feature extraction and classification algorithms have been applied to classify data from the NRL data-base. Invariant moments and normalized Fourier descriptors were the features used in these classification experiments, which show that the Fourier descriptors are better suited to classification of aircraft in turbulent atmosphere. Weak points of the algorithms are discussed and directions for further research are indicated. (2) Post-processing of image data to remove the effects of atmospheric turbulence as also studied. Atmospheric turbulence was simulated, by generation of random phase components, and applied to an idealized model of the drone in the NRL data-base. An algorithm for atmospheric deblurring, known as shift-and-add, was used. The shift-and-add algorithm produced images free of turbulence affects but degraded by blur. Image restoration was applied to the processed images and definite improvements in image quality were obtained.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Jan 01, 1983
Accession Number
ADA146728

Entities

People

  • B. R. Hunt
  • K. West
  • Kyle Morgan

Organizations

  • University of Arizona

Tags

Communities of Interest

  • Air Platforms
  • Autonomy
  • Energy and Power Technologies
  • Human Systems
  • Materials and Manufacturing Processes

DTIC Thesaurus Topics

  • Aircrafts
  • Algorithms
  • Atmospheres
  • Atmospheric Motion
  • Computational Science
  • Databases
  • Engineering
  • Feature Extraction
  • Fourier Analysis
  • Mathematical Analysis
  • Mathematical Models
  • Models
  • Pattern Recognition
  • Power Spectra
  • Preprocessing
  • Simulations
  • Turbulence

Readers

  • Atmospheric Remote Sensing.
  • Computer Vision.
  • Plasma Physics / Magnetohydrodynamics

Technology Areas

  • AI & ML
  • Autonomy