Classification based on the EMD of hyperspectral curve

Abstract

Hyperspectral remote sensing, as a new remote sensing technology, provides a powerful measure to object identification and classification accurately due to its acuminous ability of spectrum detection. This paper addresses the problem of the classification of hyperspectral remote-sensing images, and studies a novel spectral matching method based on the popular Empirical Mode Decomposition (EMD) technique. EMD is a new data analysis method by which any complicated signal can be decomposed into a set of Intrinsic Mode Functions (IMFs). The IMFs express the tendency of signals at different scales. In this method, spectral curves are decomposed with EMD, the IMFs are extracted and taken as the comparing features for spectral matching and the minimum distance classifier is used to classify ground-objects. The reason to take the IMFs as the comparing components is that the IMFs represent the intrinsic and stable component of a signal. On the other hand, using IMF as comparing element is helpful to remove the effect of noise. Furthermore, to improve the classification accuracy efficiently, similarity measure method and band selection scheme are introduced. The experimental analysis has been carried out by using hyperspectral image acquired by the AVIRIS sensor on the Washington DC Mall. The obtained results confirm the effectiveness of EMD in hyperspectral image classification with respect to conventional classifications.

Ensemble Empirical Mode Decomposition Parameters Optimization for Spectral Distance Measurement in Hyperspectral Remote Sensing Data

“…Furthermore, due to the large amount of large-dimension data processing required, it is not efficient to process hyperspectral images with EEMD [17]. Therefore, parallel processing with a Graphics Processing Unit (GPU) is implemented for EEMD [18,19].…”