鍏抽敭璇�: 绐楀彛鍌呯珛鍙跺彉鎹�, 鍔熺巼璋�, 绾圭悊, 澶氬昂搴﹀垎鍓�, 澶氬眰娆″垎绫�, SPOT 5褰卞儚

Abstract:

The image segmentation is the basis of object\|based image analysis method,which is usually carried out based on the hypothesis of spectral homogeneity.But this hypothesis is not quite applied to the saline lake in arid region which composed of two surface features with high contrast (crystalline salt and brine) in spectrum.Aiming at the shortcoming of segmentation only based heterogeneity of spectrum and shape in object based image analysis methods,the segmentation algorithms based on texture are proposed.In this paper,we firstly extract texture features using window Fourier power spectrum method.Next,multi\|resolution segmentation are carried out on texture image and multi\|spectral image.Finally,we classify polygons produced by segmentation in multi\|level.The paper takes SPOT 5 image acquired in November 2008 of Jilantai Saline Lake and the surrounding areas for example,and extracts saline lake from SPOT 5 image using the method mentioned above.The result shows that the proposed method is capable of effectively extracting features in saline area.
 

Key words: Window Fourier transform, Power spectrum, Texture;Multiresolution segmentation, Hierarchical classification, SPOT5 image