Gabor Transforms for Forward Looking Infrared Image Segmentation

Abstract

This research investigated the difficult task of segmenting targets in cluttered FLIR images. This research initiated investigation into classifying the segmented targets based on the same method used for segmentation. The primary means for segmenting and classifying targets was the biologically motivated use of Gabor transforms. This research successfully produced a complete segmentation system based on correlations between FLIR images and non- self similar or modified self similar Gabor functions. Initial investigation into methods for recognition of targets and their detailed sub-structures was accomplished using self similar and modified self similar Gabor transforms correlation coefficients 'stacked' in jets. Nearest neighbor recognition using the jets method produced accuracy of 0.622. Nearest neighbor recognition of targets using jets for input to a k-nearest neighbor network produced final accuracy of 0.6214 (for calculations based on seven nearest neighbors). An optical Gabor transform design is introduced to implement the computation of the Gabor coefficients. Theses.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Dec 01, 1989
Accession Number
ADA215046

Entities

People

  • Kevin W. Ayer

Organizations

  • Air Force Institute of Technology

Tags

Communities of Interest

  • Autonomy
  • Energy and Power Technologies
  • Materials and Manufacturing Processes
  • Sensors
  • Weapons Technologies

DTIC Thesaurus Topics

  • Air Force
  • Artificial Intelligence
  • Computational Science
  • Computer Programs
  • Computers
  • Detection
  • Detectors
  • Gray Scale
  • Image Processing
  • Image Recognition
  • Image Segmentation
  • Operating Systems
  • Pattern Recognition
  • Recognition
  • Signal Processing
  • Three Dimensional
  • Two Dimensional

Readers

  • Neural Network Machine Learning.
  • Sensor Fusion and Tracking Systems.