Assessment of feature extraction techniques for hyperspectral image classification

Using image classification methods to produce thematic maps from hyperspectral data is a challenging image processing task. Feature extraction is an important preprocessing operation to reduce the dimensionality of hyperspectral while preserving most of the information. This research work investigates some of the widely used feature extraction techniques and provides and accuracy analysis by performing experiments on a real dataset. A comparative performance analysis of some of the most important techniques including principle component analysis (PCA), Decision Boundary Feature Extraction (DBFE), and discriminative analysis feature extraction (DAFE) is provided in this work. The classification is carried out using statistical and neural network classifiers. The experimental results shown that DBFE has yielded best accuracy classification among the investigated techniques.