A hyperspectral classification method and device

Abstract

The invention discloses a hyperspectral classification method and device. The dimensions and data volume of current hyperspectral remote sensing data have increased significantly. However, after data dimensionality reduction, information will be lost. The hyperspectral classification method proposed in this invention overcomes the problem of hyperspectral images in non The disadvantage of label information loss after negative matrix factorization and dimensionality reduction. In addition, the first endmember spectral matrix calculated by the vertex component analysis algorithm is used as the initial value of the constrained non-negative matrix algorithm, which speeds up the operation of the algorithm. Through the classification of machine learning algorithms information to construct a label constraint matrix, which can effectively improve the classification accuracy and classification speed of hyperspectral images, and obtain various endmembers through constrained non-negative matrix decomposition update, which effectively reflects the spatial distribution of various ground objects and solves the current problem The spectral classification method will cause the label information of the hyperspectral image to be lost after the non-negative matrix factorization reduces the dimensionality, which leads to the technical problem of low classification accuracy.

Description

technical field [0001] The invention relates to the technical field of combining remote sensing imaging and machine learning, in particular to a hyperspectral classification method and device. Background technique [0002] In the past, people often carried out remote sensing measurements in wide bands. With the development of science and technology, the emergence of hyperspectral remote sensing images has achieved a breakthrough in the spectral resolution of remote sensing images. People can use many very Narrow electromagnetic bands are used to obtain relevant data from objects of interest, so that substances that were originally undetectable in broadband remote sensing can be detected in hyperspectral remote sensing. [0003] In addition to the two-dimensional plane image, the hyperspectral remote sensing image also includes the spectral dimension, which realizes the integration of the map and spectrum. It can obtain the continuous spectral information of each feature whil...

AI Technical Summary

Problems solved by technology

[0006] The invention provides a hyperspectral classification method and device, which solves the technical problem that the current spectral classification method will cause the label information of hyperspectral images to be lost after non-negative matrix decomposition dimensionality reduction, resulting in low classification accuracy

Images