The objective of this paper is to put forward a novel approach to detect the presence of icebergs from Synthetic-Aperture Radar (SAR) imagery obtained from Sentinel-l radar imagery. Strategies like Image Augmentation are used to increase the sample size of the training data. Features are extracted from the image and important features are selected using the Decision Tree algorithm. Important features are used to train different statistical algorithms for classification like Support Vector Machine (SVM) and Convoluted Neural Networks (CNN). A comparative study of performance of different models is used to select the best performing classification algorithm. The best performing model is validated with the test data to obtain accuracy.