Wetland information extraction of remote sensing imagery based on Markov random field theory

Due to the indistinction of land boundary and the confusion of categories in wetland as well as the big spectral difference of high-resolution remote sensing images, how to segment land boundaries exactly and maintain homogeneity in one category as much as possible are the difficult points of wetland information extraction of remote sensing images. In this paper, Xixi Wetland in Hangzhou is taken as research object and QuickBird high-resolution image as research data. Two approaches for wetland information accurate extraction based on Markov random field (MRF) theory are explored. The experimental results showed that this method has a good effect on exact segmentation of land boundaries and Inhibition of classification noises, and has higher accuracy and speed compared with other MRF methods.