An optimality based unsupervised method for hierarchical land cover classification

This study proposes an optimal hierarchical clustering methodology which can be used for land cover classification. In each step of the proposed method, optimal cluster number is determined by comparing the validity indices of clustered data for given cluster numbers. Optimal cluster number is chosen as the number optimizing the used cluster validity index. Then, each of the clusters is splitted again into optimal number of subclusters by using the same validation index. Because the proposed method is applied in a hierarchical manner; the method clusters the data in to a hierarchical structure. Optimal hierarchical clustering method is presented and applied to land cover classification from LANDSAT data.