Fully constrained least-squares based linear unmixing [hyperspectral image classification]

A fully constrained least-squares linear unmixing approach to hyperspectral image classification is presented. It is derived from an unconstrained least-squares based orthogonal subspace projection. It is similar to a method developed by Shimabukuro and Smith (1991) in the least-squares sense, but significantly different from their method in the way of implementing the constraints. Since there is no closed form solution available, an efficient algorithm is developed for finding a fully constrained solution, which can be viewed as a generalization of Shimabukuro and Smith´s method. The effectiveness of this algorithm is demonstrated through computer simulations and real data experiments