Feature extraction of hyperspectral images based on SVM optimization of 2D-EMD and GWO

Abstract

Hyperspectral images (HSIs) contain rich spectral information characteristics. Different spectral information can be used to classify different types of ground objects. However, the classification effect is mainly determined by the quality of spectral characteristic information and the performance of the classifier. This paper explores the use of two-dimensional empirical mode decomposition (2D-EMD) to first feature extraction of HSIs, then uses 2D-EMD to carry out adaptive decomposition of each band of hyperspectral data, and optimally extract the features of the sub-band obtained by decomposition. Then, the optimized features are classified in the support vector machine (SVM) recognition classifier optimized by grey wolf optimization (GWO) algorithm to further improve the effect of network recognition and classification. The simulation results show that this scheme can further improve the recognition results of different ground objects in HSIs.