Evalution of Machine Learning Methods for Hyperspectral Image Classification

Abstract

Machine learning algorithms are outstanding predictive powerful tools for classification of hypserspectral images. In this paper we summarize the various classification techniques based on machine learning approaches for space borne hypserspectral images. Random Forest (RF), Support Vector Machine (SVM) and a deep learning technique, Convolution Neural Network (CNN) are explored on HySIS images. CNN shows great potential to yield high performance in hypserspectral image classification. 2-D and 3-D CNN techniques provided robust classification results when compared to RF, SVM methods.