skip to main content
OSTI.GOV title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: An intercomparison of artificial intelligence approaches for polar scene identification

Abstract

Six advanced very high resolution radiometer local area coverage arctic scenes are classified into 10 classes. These include water, solid sea ice, broken sea ice, snow-covered mountains, land, stratus over ice, stratus over water, cirrus over ice, cumulus over water, and multilayer cloudiness. Eight spectral and textural features are computed. The textural features are based upon the gray level difference vector method. Six different artificial intelligence classifiers are examined: (1) the feed forward back propagation neural network; (2) the probabilistic neural network; (3) the hybrid back propagation neural network; (4) the [open quotes]don't care[close quotes] perceptron network; (5) the [open quotes]don't care[close quotes] back propagation neural network; and (6) a fuzzy logic-based expert system. Accuracies in excess of 95% are obtained for all but the hybrid neural network. The [open quotes]don't care[close quotes] back propagation neural network produces the highest accuracies and also has low CPU requirements. Thin fog/stratus over ice is the class consistently with the lowest accuracy, often misclassified as broken sea ice. Water, land, cirrus over ice, and snow-covered mountains are all classified with high accuracy ([ge]98%). The high accuracy achieved in the present study can be traced to (1) accurate classifiers; (2) an excellent choice formore » the feature vector, and (3) accurate labeling. A sophisticated new interactive visual image classification system is used for the labeling. 33 refs., 8 figs., 7 tabs.« less

Authors:
; ; ; ; ; ;  [1]
  1. (South Dakota School of Mines and Technology, Rapid City (United States))
Publication Date:
OSTI Identifier:
6599667
Alternate Identifier(s):
OSTI ID: 6599667
Resource Type:
Journal Article
Resource Relation:
Journal Name: Journal of Geophysical Research; (United States); Journal Volume: 98:D3
Country of Publication:
United States
Language:
English
Subject:
54 ENVIRONMENTAL SCIENCES; CLOUDS; ACCURACY; CLASSIFICATION; RESOLUTION; ICE; POLAR REGIONS; ARTIFICIAL INTELLIGENCE; COMPARATIVE EVALUATIONS; REMOTE SENSING; WATER; FOG; FUZZY LOGIC; MOUNTAINS; SNOW; ATMOSPHERIC PRECIPITATIONS; CRYOSPHERE; EVALUATION; HYDROGEN COMPOUNDS; MATHEMATICAL LOGIC; OXYGEN COMPOUNDS 540210* -- Environment, Terrestrial-- Basic Studies-- (1990-); 540110; 540310 -- Environment, Aquatic-- Basic Studies-- (1990-)

Citation Formats

Tovinkere, V.R., Penaloza, M., Logar, A., Lee, J., Weger, R.C., Berendes, T.A., and Welch, R.M. An intercomparison of artificial intelligence approaches for polar scene identification. United States: N. p., 1993. Web. doi:10.1029/92JD02599.
Tovinkere, V.R., Penaloza, M., Logar, A., Lee, J., Weger, R.C., Berendes, T.A., & Welch, R.M. An intercomparison of artificial intelligence approaches for polar scene identification. United States. doi:10.1029/92JD02599.
Tovinkere, V.R., Penaloza, M., Logar, A., Lee, J., Weger, R.C., Berendes, T.A., and Welch, R.M. Sat . "An intercomparison of artificial intelligence approaches for polar scene identification". United States. doi:10.1029/92JD02599.
@article{osti_6599667,
title = {An intercomparison of artificial intelligence approaches for polar scene identification},
author = {Tovinkere, V.R. and Penaloza, M. and Logar, A. and Lee, J. and Weger, R.C. and Berendes, T.A. and Welch, R.M.},
abstractNote = {Six advanced very high resolution radiometer local area coverage arctic scenes are classified into 10 classes. These include water, solid sea ice, broken sea ice, snow-covered mountains, land, stratus over ice, stratus over water, cirrus over ice, cumulus over water, and multilayer cloudiness. Eight spectral and textural features are computed. The textural features are based upon the gray level difference vector method. Six different artificial intelligence classifiers are examined: (1) the feed forward back propagation neural network; (2) the probabilistic neural network; (3) the hybrid back propagation neural network; (4) the [open quotes]don't care[close quotes] perceptron network; (5) the [open quotes]don't care[close quotes] back propagation neural network; and (6) a fuzzy logic-based expert system. Accuracies in excess of 95% are obtained for all but the hybrid neural network. The [open quotes]don't care[close quotes] back propagation neural network produces the highest accuracies and also has low CPU requirements. Thin fog/stratus over ice is the class consistently with the lowest accuracy, often misclassified as broken sea ice. Water, land, cirrus over ice, and snow-covered mountains are all classified with high accuracy ([ge]98%). The high accuracy achieved in the present study can be traced to (1) accurate classifiers; (2) an excellent choice for the feature vector, and (3) accurate labeling. A sophisticated new interactive visual image classification system is used for the labeling. 33 refs., 8 figs., 7 tabs.},
doi = {10.1029/92JD02599},
journal = {Journal of Geophysical Research; (United States)},
number = ,
volume = 98:D3,
place = {United States},
year = {Sat Mar 20 00:00:00 EST 1993},
month = {Sat Mar 20 00:00:00 EST 1993}
}
  • A comprehensive survey of computer-based systems that apply artificial intelligence methods to detect and identify component faults in nuclear power plants is presented. Classification criteria are established that categorize artificial intelligence diagnostic systems according to the types of computing approaches used (e.g., computing tools, computer languages, and shell and simulation programs), the types of methodologies employed (e.g., types of knowledge, reasoning and inference mechanisms, and diagnostic approach), and the scope of the system. The major issues of process diagnostics and computer-based diagnostic systems are identified and cross-correlated with the various categories used for classification. Ninety-five publications are reviewed.
  • An intercomparison study of low-level {sup 237}Np determination in artificial urine samples has been carried out. The purpose of this study was to rind the {open_quotes}optimal{close_quotes} method presently available for use in a routine in-vitro radiobioassay program for occupationally exposed workers. Four synthetic urine samples with differing {sup 237}Np concentrations were prepared: (1) 3 mBq kg{sup -1} of {sup 237}Np; (2) 3 mBq kg{sup -1} of {sup 237}Np with natural uranium, {sup 239}Pu and {sup 241}Am as interferences; (3) 50 mBq kg{sup -1} of {sup 237}Np; and (4) a matrix blank. The solutions were submitted to 10 alpha-particle and 10more » inductively coupled plasma-mass spectrometry (ICP-MS) laboratories of which six and four laboratories, respectively, returned results. Two laboratories performed assays using both techniques. The radiochemical method of choice used {sup 239}Np as the tracer, which utilized a combination of coprecipitation and anion-exchange separation. The best results obtained by ICP-MS were comparable with but not better than the most accurate results obtained by alpha-particle spectrometry. Alpha-particle spectrometry measurements overall gave consistently better agreement with known values. 15 refs., 1 fig., 7 tabs.« less
  • An intercomparison study of low-level {sup 237}Np determination in artificial urine samples has been carried out. The purpose of this study was to find the present {open_quotes}optimal{close_quotes} method available for use in a routine in vitro radiobioassay program for occupationally exposed workers. Four synthetic urine samples with differing {sup 237}np concentrations were prepared: (1) mBq kg{sup -1} of {sup 237}Np; (2) 3 mBq kg{sup -1} of {sup 237}Np with natural uranium, {sup 239}Pu and {sup 241}Am as interferences; (3) 50 mBq kg{sup -1} of {sup 237}Np; and (4) a matrix blank. The solutions were submitted to 10 alpha-particle and 10more » inductively-coupled plasma-mass spectrometry (ICP-MS) laboratories of which six and four laboratories, respectively, returned results. Two laboratories performed assays using both techniques. The radiochemical method of choice used {sup 239}Np as the tracer which utilized a combination of coprecipitation and anion-exchange separation. The best results obtained by ICP-MS were comparable with the most accurate results obtained by alpha-particle spectrometry but not any better. Alpha-particle spectrometry measurements overall gave consistently better agreement with the known values.« less
  • The author discusses the application of artificial intelligence in the office, showing how it has been used in over 100 companies throughout the world.