Hybrid Optical/Digital Architecture for Distortion Invariant Pattern Recognition

Abstract

This research investigated optical techniques for pattern recognition. An optical joint transform correlator was implemented using a magneto-optic spatial light modulator, and a charge coupled device (CCD) camera and frame grabber under personal computer (PC) control. A hybrid optical/digital architecture that could potentially perform position, scale, and rotation invariant pattern recognition using a computer generated hologram (CGH) was also implemented. The joint transform correlator was tested using forward looking infrared (FLIR) imagery containing tactical targets, and gave very good results. New techniques for binarizing the FLIR inputs and the fringe pattern of the joint transform were discovered. The input binarization used both scene average and a localized energy normalization technique for binarization. This resulted in reduced scene background, while retaining target detail. The fringe binarization technique subtracted the Fourier transform of the scene from the joint transform, and binarized on the average difference. This new technique was a significant improvement over recent published designs. Keywords: Artificial intelligence, Theses. (aw)

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Dec 01, 1989
Accession Number
ADA215628

Entities

People

  • John D. Cline

Organizations

  • Air Force Institute of Technology

Tags

Communities of Interest

  • Air Platforms
  • Energy and Power Technologies
  • Ground and Sea Platforms
  • Space

DTIC Thesaurus Topics

  • Air Force
  • Artificial Intelligence
  • Charge Coupled Devices
  • Computer Programming
  • Computer Programs
  • Computers
  • Detection
  • Detectors
  • Engineering
  • Information Processing
  • Modulators
  • Optical Correlators
  • Optical Modulators
  • Optics
  • Pattern Recognition
  • Recognition
  • Test And Evaluation

Fields of Study

  • Physics

Readers

  • Image Processing and Computer Vision.

Technology Areas

  • AI & ML