Due to the nature of imaged objects or scenes, proper segmentation and classification of image contents may be very difficult. In some cases, ancillary data or other additional information which can be used to assist a classification may be available. One type of imagery which has such additional information available can be found in satellite remote sensing of the polar oceans.; Many automated techniques have been employed in the classification of satellite synthetic aperture radar (SAR) imagery of sea ice, but few have been attempted on imagery of the polar oceans under summer melt conditions. Summer melt conditions cause classification based upon backscatter intensity to become invalid, as the backscatters of open water, thin ice, first-year ice, and multi-year ice overlap to a large degree. This instability makes proper segmentation and automated classification of image contents extremely difficult.; There are many factors which are important in the analysis of summer ice imagery; melt stage, wind speed, and surface temperature are among them. My thesis shows that using temporally-accumulated information and historical information as additional data sources, fused with ancillary temperature and wind data, can result in an automated interpretation of summer marginal ice zone imagery.
展开▼
