Localized soft classification for super‐resolution mapping of the shoreline

Abstract

The Malaysian shoreline is dynamic and constantly changing in location. Although the shoreline may be mapped accurately from fine spatial resolution imagery, this is an impractical approach for use over large areas. An alternative approach using coarse spatial resolution satellite sensor imagery is to fit a shoreline boundary at sub‐pixel scale. This paper evaluates the use of soft classification and super‐resolution mapping techniques to accurately map the shoreline. A localized soft classification approach was used to provide an accurate prediction of the thematic composition of each image pixel. This involves the use of training statistics derived locally rather than globally in the classification. Using the derived class proportion information the shoreline boundary was determined within the pixels using super‐resolution techniques. Results show that by using a localized approach in the prediction of the pixel's thematic class composition, the accuracy of shoreline prediction was increased. Notably, the use of the localized approach resulted in the shoreline with an rms error of <1.51 m, smaller than the rms error of 2.13 m derived from the use of the global approach.