Rapid de-mixing method for noisy hyperspectral image

Abstract

The invention discloses a rapid unmixing method for a noisy hyperspectral image. The method comprises the following steps: extracting a noisy end member spectrum by adopting a VCA algorithm; performing SSA algorithm denoising on the extracted end member spectrum; carrying out abundance estimation by utilizing the denoised end member spectrum according to a least square method. The end member of the original hyperspectral image is directly extracted; denoising is carried out on the end member spectrum; then, abundance estimation is carried out by utilizing the denoised end member; the unmixingof the hyperspectral image is realized. Errors generated in the first step in an existing method are avoided; meanwhile, the problem that the unmixing precision is reduced due to the fact that important information is lost in the original image denoising process is solved. Spectral denoising is conducted on the end members, the data size is small, the problem that the unmixing technology efficiency is low after image denoising is solved, the unmixing precision can be improved, operation is easy, and the processing speed is high.

Description

technical field [0001] The invention relates to the technical field of hyperspectral image unmixing, in particular to a fast unmixing method for noisy hyperspectral images. Background technique [0002] Mixed pixels widely exist in hyperspectral images, which affect the recognition accuracy of ground objects based on hyperspectral images. The existence of noise in images has a greater impact on the sparsity and stability of unmixing results. Therefore, noisy hyperspectral The rapid unmixing of images plays an important role in the fine classification of ground objects in emergency events, and is also of great significance to ground target detection; [0003] To solve the noise problem, researchers at home and abroad have proposed a variety of effective methods. So far, the denoising methods are mainly divided into the following three categories: The first is the denoising method based on the one-dimensional spectral domain. It is regarded as a collection of spectral curves ...

AI Technical Summary

Problems solved by technology

[0005] In view of the above problems, the present invention proposes a fast unmixing method for noisy hyperspectral images. The method of the present invention directly extracts the endmembers of the original hyperspectral image, denoises the endmember spectrum, and then uses the denoised The abundance estimation of the endmembers realizes the unmixing of hyperspectral images. This method avoids the error produced in the first step of the existing method, and also solves the loss of important information in the process of denoising the original image and reduces the unmixing accuracy. In addition, this method denoises the endmember spectrum with a small amount of data, which solves the problem of low efficiency of unmixing technology after image denoising. This method can not only improve the unmixing accuracy, but also has simple operation and fast processing speed

Images