Optimal Wavelet Denoising for High Range Resolution Radar Classification.

Abstract

Noncooperative identification (NCTI) of airborne targets is a top priority for the Air Force and this thesis makes a significant contribution to that area. High range resolution radar (HRR) provides an attractive means to perform NCTI. When measured HRR data is used, classification performance is excellent. However, it is often not feasible to acquire measured HRR signatures for a wide set of targets, thus necessitating the use of synthetically generated HRR data. Classification performance suffers severe degradation when using synthetic data. This thesis suggests that a large portion of HRR signature content is non-discriminatory and that this content is a cause of classifier degradation for the case of synthetic data. A unique wavelet-based denoising methodology is developed which is optimized with respect to classification accuracy. In the case of synthetic data, the denoising method leads to remarkable classification improvements. Classification accuracies are obtained which are comparable to those when training on measured data. This is an unprecedented result. It is also shown that the denoising approach of this thesis leads to superior classification results compared to those obtained with traditional wavelet-based methods.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Mar 01, 1999
Accession Number
ADA361713

Entities

People

  • Brian M. Huether

Organizations

  • Air Force Institute of Technology

Tags

Communities of Interest

  • Air Platforms
  • C4I
  • Energy and Power Technologies
  • Human Systems

DTIC Thesaurus Topics

  • Accuracy
  • Algorithms
  • Bayesian Networks
  • Classification
  • Computational Science
  • Computer Science
  • Computers
  • Data Science
  • Databases
  • Electrical Engineering
  • Feature Extraction
  • Filtration
  • Machine Learning
  • Pattern Recognition
  • Probability
  • Radar
  • Recognition

Fields of Study

  • Engineering

Readers

  • Image Processing and Computer Vision.
  • Neural Network Machine Learning.
  • Sensor Fusion and Tracking Systems.