High spectrum image classifying method based on reference diagram mutual information wave band selection and EMD (Empirical Mode Decomposition)

Abstract

The invention discloses a high spectrum image classifying method based on reference diagram mutual information wave band selection and EMD (Empirical Mode Decomposition), and relates to high spectrum image classifying methods in the field of remote sensing. The high spectrum image classifying method comprises the steps of: step I, selecting a high spectrum image wave band by using a reference diagram mutual information wave band selection-based algorithm; step II, carrying out empirical mode decomposition and image reconstruction on a selected high spectrum image wave band subset; and step III, classifying reconstructed high spectrum images by using an SVM (Support Vector Machine) classifier to obtain a classified result. As the high spectrum image wave band subset with high separability is selected by using the reference diagram mutual information wave band selection, the influence of redundant information of the high spectrum image is solved, the number of support vectors required in a classifying process is reduced, the classifying speed is accelerated, feature extraction and feature reconstruction are carried out on the high spectrum image wave band subset by using the EMD, the influence of high spectrum image noises is overcome, and the classifying precision is improved.

Description

technical field [0001] The invention relates to a hyperspectral image classification method in the field of remote sensing, in particular to a hyperspectral image classification method based on reference map mutual information band selection and empirical mode decomposition. Background technique [0002] Hyperspectral remote sensing images have high spectral resolution and can provide almost continuous spectral curves of surface objects for each pixel. Therefore, hyperspectral remote sensing can retrieve land details. At present, hyperspectral images have been widely used. Since the hyperspectral imaging spectrometer is easily affected by atmospheric molecular scattering and absorption, aerosol scattering and absorption, surface scattering, terrain, etc. during the transmission of radiation energy, the spectral shape of the hyperspectral data will be distorted, thereby introducing various noises. . In addition, hyperspectral sensors observe the surface by simultaneously s...

AI Technical Summary

Problems solved by technology

In actual engineering, most hyperspectral data obtained by satellite remote sensing and aircraft remote sensing do not have real ground object reference maps, so the hyperspectral image classification algorithm based on band selection and empirical mode decomposition has great limitations.

Images