Advances in Hyperspectral Image Classification: Earth Monitoring with Statistical Learning Methods

TLDR

Hyperspectral images share statistical properties with grayscale images, yet their high dimensionality and scarce labeled data make classification difficult, leading to indeterminacy and complex manifold structures that statistical learning methods aim to address. This tutorial surveys recent advances in hyperspectral remote‑sensing image classification, illustrating key techniques through examples. Recent methods exploit spatial homogeneity, active learning, semi‑supervised manifold learning, invariant feature extraction, and adaptive classifiers to improve classification of unseen but similar scenes.

Abstract

Hyperspectral images show similar statistical properties to natural grayscale or color photographic images. However, the classification of hyperspectral images is more challenging because of the very high dimensionality of the pixels and the small number of labeled examples typically available for learning. These peculiarities lead to particular signal processing problems, mainly characterized by indetermination and complex manifolds. The framework of statistical learning has gained popularity in the last decade. New methods have been presented to account for the spatial homogeneity of images, to include user's interaction via active learning, to take advantage of the manifold structure with semisupervised learning, to extract and encode invariances, or to adapt classifiers and image representations to unseen yet similar scenes. This tutuorial reviews the main advances for hyperspectral remote sensing image classification through illustrative examples.