A robust test for nonlinear mixture detection in hyperspectral images

摘要: This paper studies a pixel by nonlinearity detector for hyperspectral image analysis. The reflectances of linearly mixed pixels are assumed to be linear combination known pure spectral components (endmembers) contaminated additive white Gaussian noise. Nonlinear mixing, however, is not restricted any prescribed nonlinear mixing model. coefficients (abundances) satisfy the physically motivated sum-to-one and positivity constraints. proposed detection strategy considers distance between an observed hyperplane spanned endmembers decide whether that satisfies model (null hypothesis) or results from more general mixture (alternative hypothesis). distribution this derived under two hypotheses. Closed-form expressions then obtained probabilities false alarm as functions test threshold. compared another recently investigated in literature through simulations using synthetic data. It also applied real image.