Machine Classification of Cloud Particle Types.

Abstract

Classification algorithms have been developed to separate cloud particles into: (1) dendrites, (2) needles, (3) columns, (4) plates, (5) streakers, and (6) a miscellaneous or an unclassifiable class. These algorithms have been incorporated in scheme which when applied to shadow graph images produced by the Knollenberg laser scanning device have demonstrated a capability to classify more accurately than human classifiers and a relative insensitivity to particle orientation. It was concluded that the machine classification developed was greatly superior to manual classification for day to day identification of these particles because the machine classifiers were much faster, less costly, did not suffer from fatigue and were usually more accurate, (i.e. in better agreement with the trainer) then human classifications. (Author)

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Aug 01, 1982
Accession Number
ADA123402

Entities

People

  • Herbert E. Hunter

Tags

Communities of Interest

  • Energy and Power Technologies
  • Human Systems
  • Materials and Manufacturing Processes
  • Space

DTIC Thesaurus Topics

  • Air Force
  • Algorithms
  • Computer Programs
  • Computers
  • Data Analysis
  • Data Sets
  • Detection
  • Detectors
  • Eigenvectors
  • Feature Extraction
  • Information Science
  • Measurement
  • Pattern Recognition
  • Plastic Explosives
  • Recognition
  • Regression Analysis
  • Two Dimensional

Readers

  • Aerosol Science/Aerosol Physics
  • Neural Network Machine Learning.

Technology Areas

  • Directed Energy