一种采用稀疏促进的高光谱影像端元提取方法  

An Endmember Extraction Method for Hyperspectral Imagery Based on Sparsity Promoting

作　　者：余岸竹[1] 陈伟[1] 张驰[1] 冯伍法[1] 

机构地区：[1]信息工程大学测绘学院,河南郑州450052

出　　处：《测绘科学技术学报》2011年第4期288-291,共4页Journal of Geomatics Science and Technology

摘　　要：利用稀疏促进原理以及高光谱影像端元提取传统算法,结合线性光谱混合模型,提出了一种采用稀疏促进的高光谱影像端元提取方法。该方法不需要预先对端元数量进行估计,也不需要假设影像中存在纯像元。利用模拟数据以及真实高光谱影像对提出方法、ICE算法和NMF算法进行了对比实验分析。实验结果表明:提出方法能稳定地从影像中提取端元并同时估计出端元数目,精度也优于其他两者。An endmember extraction method for hyperspectral imagery was proposed, which was based on the theories of sparsity-promoting, traditional endmember extraction algorithms and the linear spectral mixture model. This method doesn＇t need a prior estimation for the quantity of endmembers or the existance of pure pixels in hyperspectral images. Some experiments were carried out with simulative and real hyperspectral images, and the results among the sparsity-promoting based method, ICE and NMF were compared and analyzed. The result showed that the sparsity-promoting based method could extract the endmembers and find their quan- tity steadily, and this method was more accurate than ICE and NMF.

关 键 词：高光谱影像 稀疏促进 线性混合模型 端元提取 混合像元分解 

分 类 号：P237[天文地球—摄影测量与遥感]