Unsupervised change detection in remote sensing images using CDI and GIFP-FCM clustering

Abstract

In this paper, an unsupervised change detection method for remote sensing image is proposed. The method takes as input two bi temporal images and outputs a change detection map highlighting changed and unchanged areas. Initially, the method creates a feature vector space by performing the principal component analysis (PCA) of both the images. The first component of PCA of both the images is used to compute the combined difference image. A change map is then created from combined difference image by clustering it into two clusters changed and unchanged using GIFP-FCM clustering technique. The method is applied on Landsat 5 TM images and the results obtained are compared with other existing state of the arts method.