Spectral unmixing scheme based on abundance sparse and endmember orthogonal constraint NMF(Non-negative Matrix Factorization)

Abstract

A spectral unmixing scheme based on abundance sparse and endmember orthogonal constraint NMF(Non-negative Matrix Factorization) is an algorithm of a spectral image decomposition field. A classical NMFtarget function is a non-convex function, and a constraint condition needs to be added into the target function for solving the problems. The characteristics of spectral images are combined to put forward the spectral image unmixing algorithm which combines endmember orthogonality with abundance sparse constrains NMF on the basis of a linear spectral mixture model. The endmember orthogonal constraint guarantees independence among spectral endmembers, meanwhile, the abundance sparsity fully utilizes the sparsity of spectral data, two constraint conditions are introduced into the target function, a least square method is adopted to obtain the iteration rules of an endmember matrix and an abundance matrix, and a final result is obtained through the setting of an iterative termination condition. Through the experiments of simulation data and truthful data, the effectiveness of the algorithm is verified.

Description

Technical field [0001] The invention relates to a hyperspectral remote sensing image processing technology, in particular to a hyperspectral unmixing scheme based on abundance sparseness and endmember orthogonality constraint NMF, and belongs to the field of hyperspectral remote sensing image unmixing. Background technique [0002] Due to the limited spatial resolution of spectral imagers and the complex diversity of ground features, some pixels in hyperspectral images often contain multiple substances, which are end members. These pixels containing other end members are called mixed pixels. The development of hyperspectral remote sensing provides a new way to solve the problem of mixed pixels. Since the imaging spectrometer obtains the information of each pixel, and the spectral information reflects the physical and chemical characteristics of the ground objects, it can The spectrum is decomposed to study the end members contained therein and the degree of mixing involved in eac...

AI Technical Summary

Problems solved by technology

[0003] The purpose of the present invention is to solve the problem that due to the non-convexity of the classic NMF objective function, it is easy to fall into the local extremum problem in the process of finding the optimal solution, thus proposing a method based on endmember orthogonality and abundance sparsity constraints As the condition, combined with NMF to decompose the mixed pixels, it is called the hyperspectral image unmixing scheme based on abundance sparse and endmember orthogonal constraints nonnegative matrix factorization (abundance sparse and endmember orthogonal constraints nonnegative matrix factorization), referred to as SONMF

Images