Dimensionality Reduction on Hyperspectral Data Set

Hyperspectral image data is a continuous succession of hundreds of narrow spectral bands (each 10 nm wide) collected over visible to infrared regions of the electromagnetic spectrum. These image data sets hold useful information in the form of a data cube, which can be noisy and redundant. Redundancy in these data sets can lead to sparseness, which is the cause of curse of dimensionality. Dimensionality reduction techniques resolve the problem of curse of dimensionality in the hyperspectral data set by extracting meaningful and compact information. In this paper, an attempt has been made to minimize the dimension of the hyperspectral data set by employing Principal component analysis (PCA), Singular value decomposition (SVD), and Independent component analysis (ICA) techniques for the extraction of spatial features. The extracted spatial and spectral features are classified using the Support Vector Machine (SVM) classifier. The classification result shows that with the increase in the number of dimensionalities, classification accuracy starts to decrease.