Method and system for dimensionality reduction of hyperspectral images based on spectral space decomposition and noise component analysis

Abstract

The invention discloses a hyperspectral image dimensionality reduction method and system based on spectral space decomposition and noise component analysis, the method comprising the following steps: S1, acquiring the original hyperspectral image, and calculating the correlation between adjacent bands of the hyperspectral image coefficient and compare it with the threshold value set, if the correlation coefficient is less than the threshold value, it is judged that the band image is a band seriously disturbed by noise, and it is eliminated; The image is decomposed, and similar bands are allocated in a subspace; S3, the weight coefficient of each band image in each subspace is calculated by the method of noise component analysis, and the band is selected according to the weight coefficient to complete the analysis of each subspace Spectral dimensionality reduction. The invention can eliminate redundant information between bands without losing important information, and does not change the physical characteristics of the original bands.

Description

technical field [0001] The invention relates to the technical field of hyperspectral image processing, in particular to a hyperspectral image dimensionality reduction method and system based on spectral space decomposition and noise component analysis. Background technique [0002] Hyperspectral image has the characteristic of "integration of map and spectrum", which can provide the spatial two-dimensional data and spectral data of the scene at the same time, and realize the detection and in-depth analysis of the target scene. Value. However, hyperspectral images contain image information of hundreds or even thousands of spectral bands, and the amount of data is hundreds or thousands of times that of traditional images. [0003] Taking the standard 224 continuous band AVRIS hyperspectral images as an example, the spatial resolution of each band image is 512x614X16bits, and the data volume of such a set of images is about 140Mbits. The huge amount of data puts pressure on t...

AI Technical Summary

Problems solved by technology

[0005] The technical problem to be solved by the present invention is to provide a kind of hyperspectral based on spectral space decomposition and noise component analysis that can effectively reduce the dimension of the spectrum while retaining most of the spectral information in view of the defect that more information is easily lost in the prior art Image Dimensionality Reduction Method and System

Images