计量可溯源的高光谱遥感地表反射率反演模型与方法研究

Nowadays the accuracy validation of inversed surface reflectance in hyperspectral remote sensing is based on relative error with reference to in situ measured reflectance. However, the in situ reflectance is merely another measured value of different source and object comparing with the inversed reflectance. The true value of the pixel-scale surface reflectance is unknowable. For quantitative applicaitons of hyperspectral remote sensing, it is needed to validate and express the closeness of agreement between the inversed reflectance and the true value more scientifically..It is proposed, by introducing the concept and analysis methods of measurement uncertainty in metrology, to research on the measurement models and uncertainty evaluation methods for radiative-transfer-model-based inversion methods (including four widely-used inversion chains as system-level geometric correction + terrain correction + atmospheic correction, and system-level geometric correction + atmospheric-terrain correction, based on hypothesis of Lambertian and non-Lambertian surface reflective characteristics respectively) and the empirical line method, so as to validate the closeness of agreement between the inversed reflectance from the existing methods and the true value. It is also proposed to explore the method to get closer to the true value from the aspects of reducing uncertainty of the key factor and reducing the propagation links..It is hoped to construct a framework of data quality analysis of inversed surface reflectance in hyperspectral remote sensing from the perspective of radiometric metrology, and to provide more accurate input to quantitative remote sensing applications.

目前评价高光谱遥感反演地表反射率的精度，多以地面实测反射率为参考来计算相对误差。而实际上地面实测反射率只是与反演反射率来源不同、被测物不同的另一个测量值，像元尺度的地表反射率真值却是不可知的。要实现高光谱遥感的定量化应用，需要更科学地评价和表示反演地表反射率与真值的接近程度。.拟通过引入计量学中 “测量不确定度” 的概念和分析方法，研究基于辐射传输模型反演法（包括地表反射特性分别为朗伯、非朗伯假设的系统级几何校正+地形校正+大气校正、系统级几何校正+大气-地形联合校正共4条常用反演链路）、经验线性反演法的测量模型和不确定度评定方法，评价现有方法反演地表反射率与真值的接近程度；并从减小关键因素不确定度、减少传递环节两方面，探索更接近真值的反演方法。.以期从“辐射度学计量”的角度，搭建高光谱遥感反演地表反射率的数据质量分析框架，为定量遥感应用提供更精准的输入数据。

传统的以地面测量光谱为参考真值的技术框架，不能充分解决“同物异谱”、“同谱异物”的地物光谱多解性和不确定性问题。在遥感反演过程中引入不确定性评估模型，以定量化评定全球和长期系列调查监测结果的质量和可靠性，从而形成可相互比较、可综合利用的“大数据”，成为美欧近几年正在攻关的最新前沿。.本项目针对“高光谱遥感中像元尺度的地表反射率真值不可知，现有的高光谱遥感地表反射率反演方法不能定量评价反演结果与真值接近程度”这一科学问题，完成了基于RT模型反演地表反射率的测量模型、基于辐射传输（RT）模型反演地表反射率的不确定度评定方法、经验线性法反演地表反射率的测量模型及不确定度评定方法、低不确定度的地表反射率反演方法探索及其结果的不确定度评定共4项内容的研究工作，给出了“地表反射率反演结果±指定置信度下的置信区间”的方法，能够在给出地表反射率反演结果的同时，给出指定置信度下的置信区间，定量地评价遥感反演地表反射率与真值的接近程度。.从而实现了比“给出与地面实测光谱相对偏差”更加科学的地表反射率精度评价和表示方法，从“辐射度学计量”的角度初步搭建了高光谱遥感地表反射率数据质量的分析框架，可以支撑下一步开展“同物异谱”、“同谱异物”情况下的地物本征光谱特征提取及识别。