Linear Spectral Mixture Analysis

Abstract

Linear Spectral Mixture Analysis (LSMA) is a widely used theory in hyperspectral data exploration. It first assumes that for a given finite set of basic material substances a data sample can be modeled as a linear admixture of these substances from which the data sample can be unmixed into their corresponding abundance fractions. In this case, analysis of the data sample can simply be performed on these abundance fractions rather than the sample itself. The commonly used Linear Spectral Unmixing (LSU) is such a technique that realizes LSMA. In other words, it assumes that a data sample is linearly mixed by a finite set of so-called image endmembers. It then unmixes this data sample via finding its abundance fractions of these endmembers for data analysis. So, technically speaking, LSU is a means of implementing LSMA as a data unmixing technique.