Accession Number : ADA299088
Title : Semi-Automated Mesoscale Analysis System, Version 1.2. Evaluation Test Results.
Descriptive Note : Final rept.,
Corporate Author : NAVAL RESEARCH LAB STENNIS SPACE CENTER MS
Personal Author(s) : Lybanon, Matthew ; Peckinpaugh, Sarah H.
PDF Url : ADA299088
Report Date : 28 SEP 1995
Pagination or Media Count : 20
Abstract : The Naval Research Laboratory's Remote Sensing Applications Branch has developed the Semi-Automated Mesoscale Analysis System (SAMAS), a prototype image analysis system for satellite images of the Gulf Stream region. Previous tests demonstrated that SAMAS shows rudimentary skill in automatically locating the Gulf Stream and its associated eddies. Since those tests, SAMAS has been updated in several ways. This report describes the performance of SAMAS Version 1.2, a newer version that incorporates some new techniques, as well as improvements to some old ones. SAMAS 1.2 was tested by applying it to a set of 22 warmest-pixel composite multichannel sea surface temperature (MCSST) images covering February through June 1993, which were also analyzed by a human expert. The human analyses were regarded as "truth" for the purposes of the statistical analysis of the test results. The human analyses are subjective combinations of analyses from the Naval Oceanographic Office's Warfighting Support Center and information from inspection of the composite images. The analyses, both human and automated, include North Wall positions and eddy definitions, center position coordinates, and sizes. The statistical analysis consists of tabulations and accuracy measures. For eddies, the tabulations compare the number of eddies found by SAMAS versus the number found by the analyst, and the locations and sizes of the eddies. Accuracy measures include position errors and size fractional errors, their first- and second-order statistics (means and standard deviations), as well as minimum and maximum values. The Gulf Stream statistical results were obtained using the same software that was used to evaluate the Gulfcast and Data Assimilation Research and Transition (DART) numerical models.
Descriptors : *IMAGE PROCESSING, *PROTOTYPES, *GULF STREAM, *COMPOSITE IMAGES, TEST AND EVALUATION, COMPUTER PROGRAMS, MATHEMATICAL MODELS, POSITION(LOCATION), SKILLS, SIZES(DIMENSIONS), HUMANS, ACCURACY, EDDIES(FLUID MECHANICS), ERROR ANALYSIS, COORDINATES, IMAGES, ERRORS, REMOTE DETECTORS, NAVAL RESEARCH, ARTIFICIAL INTELLIGENCE, EXPERT SYSTEMS, STATISTICAL ANALYSIS, NORTH(DIRECTION), MEAN, ASSIMILATION, STANDARD DEVIATION, LONGITUDE, SCIENTIFIC SATELLITES.
Subject Categories : Physical and Dynamic Oceanography
Distribution Statement : APPROVED FOR PUBLIC RELEASE