ABSTRACT
Hyperspectral data is increasingly becoming available for practical applications over large areas to better understand, model, and map land characteristics. This chapter provides advanced classification methods and approaches to achieve such a goal. Both supervised and unsupervised classification approaches are presented and discussed using hyperspectral and\or hyperspatial data. The chapter demonstrates how a pixel-based supervised support vector machines (SVMs), a machine learning algorithm, is successfully used in land use\land cover classification using a very small proportion of training data. This is important because, training data over large areas are hard to collect, and resource intensive. Chapter also discusses other machine learning algorithms such as linear and non-linear un-mixing. Such computing on the cloud will allow for fast and accurate crop and vegetation classification using hyperspectral data and help overcome Hughes’ phenomenon.
