Retrieving Urban Land Cover By Linear Spectral Unmixing Based On Bayesian Model

This research work concentrates on the concept of hyperspectral unmixing, which is the disintegration of pixel spectra received by spectral sensors into a group of elemental spectra, or endmember spectral signatures, as well as their corresponding abundance fractions. numerous unmixing algorithms and software tools have been developed for Hyperspectral images of different spectral and spatial resolution. This study presents an application of hyperspectral unmixing method based on Bayesian theorem on a real dataset. Linear mixing model decomposed each pixel of the hyperspectral image as a linear combination of pure endmember spectra. Posterior distribution of abundances and endmember limits under a hierarchical Bayesian model estimated unknown endmember spectra in a homogenized manner. This model assumes unite prior distributions for these parameters, accounts for physically meaningful, the positivity condition requires all abundances to be positive and as a way of describing for the composition of a mixed pixel, the full additivity constraint needs. Implementation of the Gibbs sampler extend the proportions on a lower dimensional simples as well as the expectation of any measurable functional of the abundance parameters, related to the posterior distribution, can be determined efficiently. This general method can be applied to contain extra conditions.