干旱区绿洲典型地物MESMA模拟分解与验证

混合像元作为遥感信息的不确定性,一直是定量遥感科学研究的核心领域之一,干旱区由于下垫面均匀、气象条件单一等先天条件,已成为定量遥感产品真实性检验的理想场所。本文以塔里木盆地北缘的库车河绿洲为研究区,首先,针对不同地物类型分别采用不同方法进行地物端元提取;然后,以端元均方根EAR(Endmember Aver-age RMSE,EAR)和最小平均波谱角(Minimum Average Spectral Angle,MASA)值来选取最优端元;最后,用多端元光谱混合分析(Multiple Endmember Spectral Mixture Analysis,MESMA)模型进行光谱混合分解,并对结果作了精度评价与比较分析。结果表明:MESMA模型能有效提高像元内基本组分丰度信息精度,从而为典型地物高精度提取提供了科学方法。

Research on Pixel Unmixing of Typical Surface Features in Oasis Based on the MESMA Model

Soil salinization is an important worldwide environmental problem, especially in arid and semi-arid regions. Quantitative remote sensing provides accurate and up-to-date information of spatio-temporal dynamics of salinity. Studies at both home and overseas showed that it is hard to acquire reliable information of salinity by using a single wavelength in visible, near infrared (NIR), thermal or microwave domain, especially in a sophisticat-ed surface conditions such as agricultural fields, while the reported methods inherited many limitations in practical applications including time-lag effect, being too complex to calculate, being excessively dependent on meteorological observations and field measurements etc. Therefore, developing of simple, effective and operational methods for the satellite estimation of surface salinity, especially vegetation cover is of great interest for both researchers in remote sensing community and policy makers for the sustainable development of eco-environments. Mixed pixels as an important aspects in remote sensing information uncertainty, it always been one of the core areas of quantitative remote sensing science research. With the congenital features, i.e., a homogeneous underlying surface and simple meteorological conditions and so on, arid area has become an ideal place for quantitative remote sensing products testing. In this article we took the Kuqa River, in north of Tarim Basin oasis as the study field. At first we selected spectral mixture analysis models, according to different feature types we extracted the endmembers in different ways. Depending on EAR (Endmember Average RMSE) and MASA (Minimum Average Spectral Angle) we chose the optimal endmembers. At last we carried out pixel unmixing by using the MESMA (Multiple End-member Spectral Mixture Analysis) model, then we finished the accuracy evaluation and comparative analysis. The results showed that MESMA model can effectively improve the information accuracy of a pixel. Therefore, it provided a scientific basis for the high-precision information extraction for typical oasis surface features.