The development and comparison of endmember extraction algorithms using hyperspectral imagery

The mixels in the hypersepectral images not only infl uence the accuracy of target detection and classifi cation, but also greatly hinder the development of quantitative remote sensing. The typical endmember extraction algorithms now available are analyzed and summarized. These algorithms can be classifi ed into two types based on the hypothesis of the existence of the pure pixels: endmember identifi cation algorithm and endmember generation algorithm. Six endmember extraction algorithms, in- cluding N-FINDR, VCA, SGA, OSP, ICE and MVC-NMF, are introduced and compared using experimental data, which further show their advantages and disadvantages. With results of various methods, the future perspective is proposed for further study.