Land cover classification using super-vised and unsupervised learning techniques

The aim of this study is to propose a suitable methodology for accurate Land cover classification in the Joshimath district, India. The study proposes K-mean clustering algorithm approach for accurate mapping of land cover. We tried several combinations of parameters and opted for the one which gave the best classification results. The results were also compared with current state of art machine learning algorithms, artificial neural network, maximum likelihood classifier and iso-cluster algorithm. Accuracy assessment was performed by means of confusion matrix. It was observed that the proposed approach gave the highest classification accuracy (93.5%) with value of kappa coefficient 0.91. While the lowest accuracy (77.8%) was achieved by iso-cluster algorithm.