Randomized Principal Component Analysis for Hyperspectral Image Classification
Mustafa Ustuner
TL;DR
Hyperspectral data are high-dimensional, posing classification challenges. The paper systematically compares PCA and randomized PCA (R-PCA) as dimensionality-reduction steps when classifying with SVM and LightGBM, reducing to 20 and 30 components on Indian Pines and Pavia University. Key findings show that original (unreduced) features generally yield the best performance, with PCA often outperforming R-PCA for SVM, while LightGBM on reduced features can approach original-data performance; the top results are $0.9925$ for Pavia University and $0.9639$ for Indian Pines using LightGBM on original data. The work informs practical classifier design for hyperspectral imagery, highlighting when randomized projections offer limited gains and underscoring strong performance of non-deep-learning, ensemble methods with full-feature inputs.
Abstract
The high-dimensional feature space of the hyperspectral imagery poses major challenges to the processing and analysis of the hyperspectral data sets. In such a case, dimensionality reduction is necessary to decrease the computational complexity. The random projections open up new ways of dimensionality reduction, especially for large data sets. In this paper, the principal component analysis (PCA) and randomized principal component analysis (R-PCA) for the classification of hyperspectral images using support vector machines (SVM) and light gradient boosting machines (LightGBM) have been investigated. In this experimental research, the number of features was reduced to 20 and 30 for classification of two hyperspectral datasets (Indian Pines and Pavia University). The experimental results demonstrated that PCA outperformed R-PCA for SVM for both datasets, but received close accuracy values for LightGBM. The highest classification accuracies were obtained as 0.9925 and 0.9639 by LightGBM with original features for the Pavia University and Indian Pines, respectively.