Sea Ice Classification Using Synthetic Aperture Radar

This study employs Synthetic Aperture Radar (SAR) imagery from the Marginal Ice Zone Experiment (MIZEX) 1987 to identify an optimal set of statistical descriptors that accurately classify three types of ice (first-year, multiyear, odden) and open water. Two groups of statistics, univariate and textu...

Full description

Bibliographic Details
Main Author: Garcia, Jr, Frank W.
Other Authors: NAVAL POSTGRADUATE SCHOOL MONTEREY CA
Format: Text
Language:English
Published: 1990
Subjects:
Snow
Ice and Permafrost
Active & Passive Radar Detection & Equipment
*SYNTHETIC APERTURE RADAR
INERTIA
STATISTICS
ACCURACY
VARIATIONS
INDEX TERMS
CLASSIFICATION
TEXTURE
DISCRIMINATE ANALYSIS
SEA ICE
OPEN WATER
DISCRIMINATORS
OPTIMIZATION
ICE
*RADAR IMAGES
MARGINAL ICE ZONES
ICE CLASSIFICATION
STATISTICAL ANALYSIS
GRAY SCALE
ODDEN ICE
THESES
Ice
permafrost
Sea ice
Online Access:http://www.dtic.mil/docs/citations/ADA232248
http://oai.dtic.mil/oai/oai?&verb=getRecord&metadataPrefix=html&identifier=ADA232248
Description
Summary:This study employs Synthetic Aperture Radar (SAR) imagery from the Marginal Ice Zone Experiment (MIZEX) 1987 to identify an optimal set of statistical descriptors that accurately classify three types of ice (first-year, multiyear, odden) and open water. Two groups of statistics, univariate and texture, are compared and contrasted with respect to their skill in classifying the ice types and open water. Individual statistical descriptors are subjected to principal component analysis and discriminant analysis. Principal component analysis was of little use in understanding features of each ice and open water group. Discriminant analysis was valuable in identifying which statistics held the most power. When combined, univariate and texture statistics classified the groups with 89.5% accuracy, univariate alone with 86.8% accuracy and texture alone with 75.4% accuracy. Range and inertia were the strongest univariate and texture discriminators with 74.6% and 50.8% accuracy, respectively. Despite the use of a non-calibrated SAR, univariate statistics were able to classify the images with greater accuracy than texture statistics.

Similar Items