摘要
针对无约束最小二乘混合像元分解算法提取地物端元丰度出现的局限性问题,通过野外实地采集的地物光谱数据建立研究区典型的地物波谱库,以Landsat OLI影像作为主要数据源,在经过Gram-Schmidt(GS)影像融合的基础上,利用纯净像元指数(PPI)及基于几何顶点的端元提取技术提取研究区典型地物端元,最后通过完全约束的最小二乘混合像元分解算法完成对研究区典型地物端元丰度的提取。结果较好地解决了无约束最小二乘混合像元分解算法提取的端元丰度信息出现负值的情况,并且提高了典型地物丰度信息提取的精度。完全约束最小二乘混合像元分解算法的RMSE误差均控制在0.174 913左右,在很大程度上提高了混合像元分解精度及实用性。
According tothe fact that there are somelimitations of unconstrained least squares mixed pix- el decomposition algorithm in endmember abundanceextraction. A typical ground objects spectrum database is established through the spectrometer acquired field spectral data, the Landsat OLI image as data source, using pixel purity index（PPI） and geometric vertex end extraction method to extract the typical objects endmember after Gram-Schmidt （GS） image fusion. Finally, applycomplete constrained least squares mixed pixel decomposition algorithm to complete the extraction of typical feature endmember a- bundances. The results showed： The complete constrained least squares mixed pixel decomposition method solve the unconstrained least squares mixed pixel decomposition extraction abundance map appear negative unreasonable situation, and, the constrained least squares method＇ sroot mean square error（RMSE） was controlled at about 0. 174 913, fulfilltheresearchneeds, so thecomplete constrained least squares method- improved the accuracy and utility of mixed pixel decomposition.
出处
《测绘科学》
CSCD
北大核心
2017年第9期143-150,157,共9页
Science of Surveying and Mapping
基金
国家自然科学基金项目(41561048)
云南省中青年学术技术带头人培养项目(2008PY056)
国家测绘地理信息局地理国情监测示范项目"抚仙湖流域生态环境动态监测"测国土函[2014]35号
