高光谱遥感影像地表反射率一体化反演模型研究

Application of Atmospheric correction in Hyperspectral remote sensing is indispensable to retrieve surface reflectance spectra and it is an important precondition for successful application of hyperspectral dataset. In this study, data sources include airborne or spaceborne hypspectral images,such as CASI,HJ-1 HSI or EO-1 Hyperion etc. airborne or spaceboren Hyperspectral, was chosen as a data source. By studying the mechanism of radiative transfer and principles of hyperspectral imaging technology, we built up radiative transfer model, adjacency effect correction model, topographic correction model for diffuse irradiance and retrieval results evaluation method. And then, the method for Aerosol Optical Thickness (AOT) and Column Water Vapor (CWV) retrieval were researched, and corresponding algorithms were developed by which AOT and CWV results were retrieved. AOT and CWV results, combined with predefined models and multi-dimensional look up table, were used to retrieve surface reflectance spectra. Finally, the AOT, CWV and surface reflectance spectra results were validated by in situ measurements based on evaluation method. In this study, the atmospheric parameters and surface reflectance were retrieved from the information provided by the Hyperspectral data itself, avoiding the difficulty to acquire synchronization of atmospheric parameters, and the whole process was based on Hyperspectral data itself and run independently, without any other ancillary information.

对高光谱数据进行大气校正并获取地表反射率光谱曲线是高光谱数据成功应用的重要前提。以CASI/HJ-1 HSI/EO-1 Hyperion等航空或卫星高光谱数据为数据源，通过研究大气辐射传输机理和高光谱成像技术原理，建立辐射传输模型、邻近效应校正模型、漫反射辐射地形校正模型和反演结果评价指标，在此基础上展开气溶胶光学厚度、大气水汽含量反演技术的研究，发展相应的反演算法模型，得到气溶胶光学厚度、大气水汽含量反演结果，借助于辐射传输模型、漫反射辐射地形校正模型、邻近效应校正模型和查找表，反演地表反射率值；根据反演结果评价指标，采用地基同步测量值对气溶胶光学厚度、水汽含量和地表反射率反演结果和精度进行验证与评价。本研究以高光谱数据自身提供的信息进行大气参数和地表反射率反演，避免了大气校正过程中同步大气参数难以获取的困难，整个过程基于影像本身信息、全自动化执行，无需其它辅助信息。

对高光谱数据进行大气校正和地表反射率反演是定量化应用的基础。本项目在研究大气辐射传输模型与理论的基础上，研究了基于高光谱数据的大气气溶胶光学厚度、水汽含量定量反演关键技术，分析了气溶胶、水汽对反射率反演精度的影响，并建立了相应的反演模型；分别利用MODTRAN和6S生成了多维大气参数查找表，并采用多维一次拉格朗日插值方法，即多维线性插值方法实现多维大气参数查找表的插值与计算；最后针对不同下垫面、不同传感器建立了相应地表反射率反演模型，完成高光谱地表反射率一体化反演完整流程。研究结果表明：①对于Hyperion数据：该数据主要是针对城市地表，地物反射率光谱与FLAASH反演结果相比具有较高的一致性，在相同输入参数的条件下，与FLAASH反演结果进行比较，地物反射率光谱的平均相关系数高于0.9，平均光谱角小于0.08弧度，平均均方根误差小于0.02；②对于CASI/SASI数据：该数据主要是针对农田，水汽吸收波段范围之外的地表反射率反演结果与地表实测结果拟合程度较好。在400-2500 nm光谱范围内，反演的地表反射率光谱与地面实测反射率光谱之间的相关系数为：植被R2=0.942，均方根误差RMSE=0.0387，土壤R2=0.786，均方根误差RMSE=0.0406。该结果能满足实际应用需求。本研究的科学意义在于部分解决了高光谱遥感大气校正和地表反射率反演过程中同步大气参数难以获取的问题，提高了地表反射率反演的精度。