An automatic unsupervised method based on context-sensitive spectral angle mapper for change detection of remote sensing images

Abstract

This paper proposes an automatic unsupervised method for change detection at pixel level of Landsat-5 TM images based on spectral angle mapper (SAM). In most existing studies, conventional use of SAM does not take into account contextual information of a pixel. The proposed method incorporates spatio-contextual information both at feature and decision level for improved change detection accuracy. First, a similarity image is created using contextsensitive spectral angle mapper, and then it is segmented into two segments changed and unchanged using k-means algorithm to create a change map. The quantitative as well as qualitative comparison of the experiment results shows that the proposed method gives better results than the other existing method.