An unsupervised change detection technique based on Ant colony Optimization

In this paper, an unsupervised technique for change detection in satellite images based on Ant Colony Optimization is presented. Initially, a difference image is computed using normalized neighborhood ratio approach. The cluster centers are then computed using Ant colony based clustering algorithm. The cluster centers obtained from ant based algorithm are used to cluster the difference image into two clusters changed and unchanged. The clustered image is the change map that represents the changed and unchanged areas. The proposed method is compared with some other state of the art methods. Both visual as well as quantative results including percentage correct classification and kappa coefficient verify that the proposed method gives better results.