Use of a spatial expert system shell to develop automated techniques for detection and classification of sea ice in AVHRR imagery

• Other Authors: WILLIAMS, RAYMOND (hasPrincipalInvestigator), WILLIAMS, RAYMOND (processor), Australian Antarctic Data Centre (publisher)
• Format: Dataset
• Language: unknown
• Published: Australian Antarctic Data Centre
• Subjects: oceans; ICE TYPES; EARTH SCIENCE; CRYOSPHERE; SEA ICE; AVHRR; AVHRR > Advanced Very High Resolution Radiometer; SATELLITES; CONTINENT > ANTARCTICA; GEOGRAPHIC REGION > POLAR; Antarctic; Southern Ocean; The Antarctic; Antarc*; Antarctica; Sea ice
• Online Access: https://researchdata.ands.org.au/use-spatial-expert-avhrr-imagery/700340; https://data.aad.gov.au/metadata/records/ASAC_907; http://nla.gov.au/nla.party-617536

The development of an operational sea ice mapping system. This metadata record refers to the development and testing of an prototype system, ICEMAPPER, to interpret NOAA AVHRR imagery on a semi-automatic basis, off the Southern Ocean near to the Antarctic coast. From the abstract of one of the referenced papers: This paper reports work towards the development of a semi-automated technique for creating sea-ice and cloud maps from Advanced Very High Resolution Radiometer (AVHRR) images of the Southern Ocean near to the Antarctic coast. The technique is implemented as a computer-based system which applies a number of classification rules to the five bands of an AVHRR image and classifies each pixel in the image as representing open water, low cloud, high cloud or one of several different sea ice concentration categories. The map produced by the system is then displayed and an experienced sea ice forecaster evaluates the result. If it is deemed satisfactory the map is saved on disk. If not, the expert can alter various parameters within the classification rules to produce a satisfactory map. Experience so far has shown that judicious, but reasonably minor, changes to the rule parameters can produce a satisfactory sea-ice map relatively quickly in most cases. The system is also capable of effectively distinguishing cloudy from clear pixels but it does not accurately distinguish high cloud from low cloud in some of the images. Current work is being undertaken to improve the cloud classification rules.