Automatic Extraction of Endmember from Hyperspectral Imagery by Iterative Unmixing
高光谱端元自动提取的迭代分解方法

Linear pixel unmixing is a straightforward and efficient approach to the spectral decomposition ofmulti-channel&hy perspectral remotely sensed scenes. Amain drawbackto itsutilization inoperational cases isthatthe endmemberofspectral com ponents can not be retrieved correctly and automatically. Developing unsupervised methods to automatically abstract endmembe is a difficult but significant job. The authors presented an iterative error analysis algorithmto retrieve endmembers and unmixin hyperspectral imagery automatically after obtaining some constraint conditions of selecting endmembers by analyzing error propa gation in linearspectral unmixingmodel, and combinedwiththe propertyof endmemberwhich is cohesive in spatial.The experi mental results showthe algorithm isrobustbytestingvariousthresholds and initial iterative value. Other experiments fortest effi ciency and accuracy of the algorithm by employingAVIRIS and PHI hyperspectral data were also done.