基于Landsat 8的云南松光谱端元选择与评价研究

Evaluation on Spectral Endmember of Pinus yunnanensis Based on Landsat 8

以云南松为研究对象,调查昆明市主城区周围61个样点,选择3块代表性样地,基于Landsat 8影像,采用纯净像元指数(PPI)、连续最大角凸锥(SMACC)和几何顶点的端元提取方法,利用样区1提取的云南松端元波谱对样区2和3进行分类。以外业调查数据提取的平均端元为真值,结合波谱角填图(SAM)分类结果,对比分析不同的端元提取方法。结果表明:研究样区2基于PPI、SMACC和几何顶点端元提取的分类结果整体精度分别为85.00%、35.00%和85.00%;研究样区3基于PPI、SMACC和几何顶点端元提取的分类结果整体精度分别为83.33%、16.67%和75.00%。基于PPI提取的云南松端元平均波谱曲线与真实地表的云南松波谱曲线最为相似,可用于今后基于Landsat8数据的云南松波谱端元提取和混合像元分解。

Sixty-one sampling points of Pinus yunnanensis surrounding the main city of Kunming in Yunnan were investigated. Three representative sample plots were selected. Spectral endmember of P. yunnanensis were ex- tracted based on Landsat 8 using Pixel Purity Index （PPI）, Sequential Maximum Angle Convex Cone （SMACC） and geometric vertex from 1st study plot. Then it has been used to do land cover classification for the 2nd and 3rd study plots. Comparison, analysis and evaluation of different endmember extraction method were done using the av- erage filed data extraction as the true value in combination with Spectral Angle Mapper （SAM） results. Results showed that the 2nd study plot ： the overall accuracy of the classification results based on PPI, SMACC and geomet- ric vertex is 85%, 35.00% and 85.00% respectively. The 3rd study plot ： the overall accuracy of the classification results based on PPI, SMACC and geometric vertex is 83.33%, 16. 67% and 75.00% respectively. The average spectral curve of P. yunnanensis based on PPI endmember extraction, which could be used in the future endmember extraction and spectral mixture analysis based on Landsat 8, is more similar comparing with the ground truth data.