Hyperspectral image terminal element extraction method based on Givens rotation

Abstract

The invention particularly relates to a hyperspectral image terminal element extraction method based on Givens rotation. The method comprises using the Givens rotation to obtain the maximum projectionposition to thereby obtain the end-member spectrum, which not only ensures the consistency of the extraction results for the same hyperspectral image, but also only needs to extract the first end-member from the original data to extract all the remaining end-members, and greatly improves the precision of the end-member extraction. The method comprises the steps of 1 inputting the original hyperspectral image data as; 2 estimating that number of terminal element of the original hyperspectral image; 3 carrying out dimensionality reduction on that original hyperspectral data to obtain the dimensionality reduction hyperspectral data; 4 extracting a centroid pixel by using that original hyperspectral data, and obtain a first end element by the centroid pixel; 5 performing QR decomposition of Givens rotation on that first end member to obtain the second end member; 6, carrying out the QR solution of the Givens rotation for p-1 end elements to obtain the p-th end element.

Description

technical field [0001] The invention belongs to the technical field of image processing, and further relates to the field of hyperspectral remote sensing image information extraction, in particular to a hyperspectral image endmember extraction method based on Givens rotation. Background technique [0002] Hyperspectral remote sensing images contain spatial information and spectral information of the captured area at the same time, which has the characteristics of "integration of map and spectrum". Spatial information can provide geometric information of the captured area, and spectral information can be used for practical applications such as classification and identification of real objects in the captured area. Because of this, hyperspectral remote sensing image technology has developed rapidly in recent years and has become a research hotspot. There are more and more studies on its theoretical methods and practical applications. Its technical research mainly focuses on c...

AI Technical Summary

Problems solved by technology

The calculation of the MVHT algorithm is relatively simple, but before the calculation, the first two endmembers must be obtained by using the maximum two-norm method, which reduces the accuracy of endmember extraction to a certain extent.

[0013] Therefore, to sum up, both of the above two methods have their deficiencies, namely: the VCA algorithm has the disadvantage of inconsistent results after multiple runs; the MVHT algorithm needs to obtain the first two endmembers in advance. Affects the endmember extraction accuracy to a great extent

Images