线性光谱混合模型影响因子不确定性定量研究  

Study on Uncertainty of Environment Factors Relating to Linear Spectral Mixing Model

作　　者：陈松林[1] 王天星[1] 

机构地区：[1]福建师范大学地理科学学院,福州350007

出　　处：《地球信息科学》2007年第6期49-53,共5页Geo-information Science

基　　金：福建省科技计划项目(2006F5029)

摘　　要：线性光谱混合模型(Linear Spectral Mixing Model,LSMM)是一种像元分解模型,由于其简单和易操作性的特点,在目前亚像元研究中应用颇为广泛。其分离精度受多种因素的影响,但目前对该模型的研究多集中在对模型本身的线性假设评价及端元光谱选取方法上,而忽略了模型应用的环境条件(大气反射、散射、地形起伏等)对模型分解精度的影响等。本文以线性光谱模型提取植被分量为例,探讨环境大气条件、地形因素对模型精度影响的不确定性。研究将数据处理为四个层次,即原始的ASTER数据,利用MODTRAN进行大气校正的数据,经C-地形校正的数据,同时进行了大气校正和地形校正的数据。然后在四个层次上依次提取植被丰度,并将其和NDVI进行线性回归分析,检验植被丰度的分离精度,从而量化大气、地形等因子对LSMM的影响程度。研究结果表明:大气条件、地形因素都会制约LSMM分离精度的提高,特别在有地形起伏的中小空间尺度范围内,地形因子对线性光谱混合模型的影响远大于大气影响。Linear Spectral Mixing Model （LSMM） is one of the pixel unmixing models. LSMM is prevailing presently in sub-pixel applications for it＇s simple and easy operational characteristics. The unmixing accuracy of LSMM is restricted by kinds of factors. However, the research on LSMM is focused on appraisement of linear hypothesis relating to itself and techniques used to select endmembers for the present. The environment conditions of the study area which could sway the unmixing accuracy such as atmospheric reflectance or scatteration and terrain undulation are not previously studied. This paper probes emphatically into the accuracy uncertainty of LSMM from atmospheric condition and terrain undulation by taking unmixing vegetation abundance based on LSMM as an example. Four levels of processing data sets were derived to conduct subsequent unmixing and comparison, namely, the first level which is related to the original ASTER data, the second level related to the data which perform an atmospheric correction using MODTRAN simulations, the third level related to the data which perform a terrain illumination correction equipped with C-correction Method, and the fourth level sequentially, related to the data which were applied to both atmospheric correction and terrain illumination correction. Then the vegetation abundances were extracted from the four processed data sets based on LSMM. The regression analysis between NDVI and vegetation abundance was further conducted to assess the unmixing accuracy which quantitatively measures the atmospheric effect and terrain illumination in the study site. The results indicate that both atmospheric condition and terrain un- dulation could constrain the application of LSMM. Especially, the effective removal or minimization of terrain effects is essential for LSMM applications to moderate or small-scale mountainous areas.

关 键 词：线性光谱混合模型(LSMM) ASTER 地形校正 植被丰度 

分 类 号：P237[天文地球—摄影测量与遥感]