中高分辨率多源光学遥感图像辐射归一化模型与方法研究

With the increasing of medium and high resolution satellites at home and abroad, it is possible to conduct surface dynamic monitoring with a medium-high resolution scale in the region and even in the world. However, the difference of radiometric performance caused by sensors themselves and imaging conditions present a great challenge to the comprehensive utilization of these data. To solve these problems in this project, we analyze in depth analyze the radiometric difference and distribution among multi-source satellite data, which is caused by factors such as spectral response, imaging geometry, surface topography and non-Lambertian, and solve problem of difference of the calibration accuracy and inconsistency of the radiometric references; we analyze the distribution of multi angle observations provided by multi-source satellites, and construct the bidirectional model with the medium-high resolution to describe the mountainous surface; we study the varying dependence of the surface reflectance derived from different remote sensing data with types of ground suface and the imaging angle, and establish the normalized model for ground surface reflectance pixel by pixel, thus to minimize the difference of spatial resolution, spectral response, imaging angle and imaging time among different satellite data. This project is eventually expected to form a set of radiometric normalization models and methods for the multi-resource remote sensing data with medium-high spatial resolution independent of sensor’s characteristics, to radiometric normalize the multi-source satellite data in two levels of satellite apparent reflectance and ground surface reflectance. This project has a prominent academic innovation and application prospect.

随着国内外中高分辨率卫星的日益增多，开展区域乃至全球中高分辨率地表动态监测已逐渐成为可能，但不同遥感器自身及成像条件造成的辐射差异为综合利用这些数据提出了很大挑战。本项目针对这一问题，深入分析中高分辨率多源卫星数据因光谱响应、成像几何、地形起伏、地表非朗伯性等因素造成辐射差异和分布规律，建立高精度的交叉定标模型与方法，解决不同卫星定标精度和辐射参照基准不一致问题；深入分析多源卫星提供的多角度观测分布规律，构建山区地表的中高分辨率二向性模型；研究不同遥感数据反演的地表反射率随地物类型、成像角度等因素变化规律，建立逐像元的地表反射率归一化模型，降低地表反射率数据因遥感器空间分辨率、光谱响应、成像角度、成像时间造成的差异。本项目最终形成一套独立于遥感器特性的中高分辨率卫星数据辐射归一化模型与方法，从星上表观反射率和地表反射率两个层面实现多源遥感数据辐射归一化，具有较高学术创新和应用价值。

本项目主要围绕多源遥感数据定标与大气校正的关键问题开展研究，解决辐射基准之间的差异，根据执行之间研究方向的相关性，主要研究内容分为5个方面：中高分辨率遥感器高精度交叉定标；快速大气校正查找表构建方面；多种大气校正算法对比与精度评估；地形起伏区域的辐射校正模型研究与反射率计算；考虑像元尺度下的地表反射率归一化模型。.在中高分辨率遥感器高精度交叉定标方面，以GF-4中分辨率静止轨道卫星光学传感器为例，开展了极轨卫星与静止卫星高精度交叉定标研究，对GF-4/PMS遥感器定标系数进行计算和精度评估。在多种大气校正算法对比与精度评估方面，融合多种算法实现多源国产卫星数据自动化地表反射率生产，解决了商业软件和国内已有产品系统针对性不强、无法自动化处理的难题。系统开展了HJ-1和GF1/2的地表反射率精度验证工作，总体误差约为TM/ETM+产品二倍左右。在地形起伏区域的辐射校正模型研究与反射率计算方面，提出一种基于辐射传输的地形辐射校正模型结合严控误差源的地形辐射校正方法。在考虑像元尺度下的地表反射率归一化模型方面，本研究考虑不同遥感器光谱响应以及地物光谱差异，提出一种逐像元辐射归一化方法，可以有效提高多源遥感器归一化精度，为多源中高分辨率遥感图像高精度辐射归一化提供新的思路。在快速大气校正查找表构建方面，在多节点高性能集群计算环境下设计了一套高效运行流程和查找表构建方法，通过合理分割查找表计算任务、多个计算节点任务调度、以二进制方式存储计算结果等来解决辐射传输模型计算时间长和存储空间占用巨大的问题。.