고해상도 원격탐사를 이용한 갯벌 표층 퇴적상 분류 연구

Abstract

In this study, the applicability of an object-based method to the classification of the surface sediment facies in a tidal flat is tested using IKONOS satellite image, a high spatial resolution remote sensing. Object-based classification method has an advantage of incorporating spatial neighbourhood properties into the classification process, and it may accomplish a more accurate result of classification as compared with pixel-based method. We compared the classification result of object-based method with that of pixel-based maximum likelihood method quantitatively, for Hwangdo tidal flat in Cheonsu Bay, Korea. The overall accuracy and kappa value calculated from the result of the object-based method were 90.7 % and 0.86, respectively, which is relatively higher accuracy than the maximum likelihood. It can be concluded that the object-based approach is a effective tool for the surface sediment classification in a tidal flat.