Investigation of the random forest framework for classification of hyperspectral data

TLDR

Statistical classification of hyperspectral data is difficult because of high dimensionality, mixed classes, and limited labeled data, leading to unstable classifiers with poor generalization. This study evaluates two random‑forest–based approaches within a binary hierarchical multiclassifier system to enhance generalization when training data are scarce. The proposed method employs bagging of training samples and adaptive random‑subspace feature selection in a binary hierarchical classifier, selecting the number of features at each node according to the available training data, and is compared to a CART‑based random forest. Experiments on NASA AVIRIS and Hyperion datasets show that both methods outperform the original best‑basis BHC and a random‑subspace BHC, achieving superior classification accuracy.

Abstract

Statistical classification of byperspectral data is challenging because the inputs are high in dimension and represent multiple classes that are sometimes quite mixed, while the amount and quality of ground truth in the form of labeled data is typically limited. The resulting classifiers are often unstable and have poor generalization. This work investigates two approaches based on the concept of random forests of classifiers implemented within a binary hierarchical multiclassifier system, with the goal of achieving improved generalization of the classifier in analysis of hyperspectral data, particularly when the quantity of training data is limited. A new classifier is proposed that incorporates bagging of training samples and adaptive random subspace feature selection within a binary hierarchical classifier (BHC), such that the number of features that is selected at each node of the tree is dependent on the quantity of associated training data. Results are compared to a random forest implementation based on the framework of classification and regression trees. For both methods, classification results obtained from experiments on data acquired by the National Aeronautics and Space Administration (NASA) Airborne Visible/Infrared Imaging Spectrometer instrument over the Kennedy Space Center, Florida, and by Hyperion on the NASA Earth Observing 1 satellite over the Okavango Delta of Botswana are superior to those from the original best basis BHC algorithm and a random subspace extension of the BHC.