Comparative Study of Machine Learning Algorithms for Glacier facies Classification

Classification of different glacier facies are the important factors which determine the rate of retreat/advance of a glacier. In this paper, the potential of using freely available Sentinel-1 radar images for glacier zones mapping in the Chandra basin of Himachal Pradesh is examined. The three images of three different months (January, May & September) used and different supervised algorithms namely MD, MLC and SVM were implemented. Google earth imagery and RGB stacked image was used to create training dataset which is used to train the model. Post processing of the classification results overall accuracy have been calculated. It is found that the significantly better accuracy was achieved through support vector machines i.e. more than 90%.