On the effect of nonlinear mixing in hyperspectral images of human settlements

Abstract

In this paper, the nonlinear contribution of human settlements to mixture in hyperspectral images is investigated. Specifically, a method that aims to efficiently evaluate and estimate the extent of urban areas by taking advantage of the results provided by polynomial nonlinear unmixing based on polytope decomposition is proposed. Tests over real images shows how the proposed scheme can actually highlight anthropogenic extents over geometrically complex scenarios.