Determine the Appropriate Number of Clusters Using the Combination of Fuzzy Algorithms For Hyperspectral Images Clustering

Hyperspectral images clustering, for the appropriate number of clusters, due to their high volume and high noise level is a difficult matter. The proposed method in this paper is a composition of fuzzy clustering algorithms that offers the optimal number of clusters along with the clustered image. The method is based on a hierarchical approach and acts using three fuzzy algorithms FCM, weighted G-K and MCV. Hence, the different shapes of clusters are identified and the sum volume of every cluster is minimum. Each of the clusters according to the proposed condition are separated as step to step, until the appropriate number of clusters are obtained. After implementing the proposed approach on two different hyperspectral images and use of cluster validity parameters, more appropriate values of these parameters is obtained for this method than other fuzzy methods. Finally, a good clustering can be obtained by determining the number of clusters. The results of the cluster assessment such as partition coefficient show that the proposed method, less scattered examples is presented in the clustered image, on the other hand, similarity the samples in each cluster is considerable than other methods.