This paper presents a trendy approach for unmixing of linear hyperspectral images. This method deals with the minimal volume class of the process. The method is SISAL method. This is called as Simplex Identification via Split Augmented Lagrangian method. The linear hyperspectral unmixing is related in finding the hyperspectral vectors which were present in the least possible volume simplex. It is a non-convex optimization problem and it has some convex constrains. The spectral vectors are being forced by the positive constrains which belongs end member signatures of the convex hull which were in turn replaced by the soft constrains. Augmented Lagrangian optimizations in the order of sequences are used to solve this problem. The resultants algorithmic approach is very fast in approach so that the problems will be able to be solved far beyond the present state-of-art algorithms. The concept Simplex Identification via Split Augmented Lagrangian is explained with simulated data.
展开▼
