Hyperspectral Data Unmixing Algorithm Comparative Analysis Based on Linear Spectral Mixture Model

The mixed pixels of hyper spectral data can be described effectively through linear spectral mixture model. Over the past years, many algorithms have been developed for unsupervised hyper spectral data unmixing, However, there are a lack of effectively compared by using a unified frame for hyper spectral unmixing through quantitative approaches. So, the paper analyze the theory of linear spectral mixture model, and performance of classics unmixing algorithm. By contrast, there is better performance than others for MVSA, VCA and MVC-NMF, MVSA is robustness and effective, the run time of MVC-NMF is long, but its index is better, VCA is excellent algorithm, and its run time is short, The performance of CCA and N-FINDER are bader than the others, so, the use of algorithm accordes to specific circumstances.