Atmospheric correction is one of the key technologies of water color remote sensing. Chinese land satellites multispectral sensor often only have four bands (blue, green, red and near-infrared) and lack of short-wave infrared band which is needed by atmospheric correction algorithm of turbid water. Therefore, accurate atmospheric correction is very difficult which seriously restrict the application of Chinese land satellites in turbid water. This project will research on blue-green bands ratio characteristics based on the Chinese land satellites multispectral remote sensing image features. We also modify the atmospheric radiative transfer model 6SV to calculate the reflection of sky light. We will construct lookup tables of GF1-WFV and couple blue-green bands ratio and modify 6SV to achieve image own aerosol optical thickness estimation and atmospheric correction for every turbid water pixels. System calibration will be done use surface reflectance spectrum from satellite synchronization experiment to reduce the systematic bias of radiometric calibration and atmospheric correction, and advance the accuracy and stability of atmospheric correction. The atmospheric correction method of this study also could be used in other similar sensor with four bands. However, it need to adjust some parameters, including look-up table and system calibration factor.
大气校正是水色遥感的关键技术之一,国产陆地卫星多光谱传感器通常只有蓝、绿、红和近红外4个波段,缺少浑浊水体大气校正算法所需的短波红外波段,精确大气校正十分困难,制约了国产陆地卫星在内陆浑浊水体中的应用。本项目针对国产陆地卫星多光谱影像特点,研究浑浊水体蓝绿波段光谱特征,修改大气辐射传输模型6SV,计算水面天空光反射,构建面向典型国产陆地卫星(GF1-WFV)的水体-大气辐射传输查找表,建立耦合蓝绿波段比值关系和6SV的浑浊水体区域气溶胶光学厚度反演和大气校正算法,实现基于图像自身的浑浊水体逐像元大气校正。最后利用与卫星同步获取的水面反射率光谱进行系统定标,降低辐射定标和大气校正中的系统误差,提高大气校正的精度和稳定性。本项目构建的面向GF1-WFV影像的浑浊水体大气校正方法也可以应用于其它国产陆地卫星(4波段)影像,不过需要调整一些参数,包括查找表和系统定标系数。
HY1C/D CZI影像拥有蓝、绿、红、近红外四个通道,缺少用于大气校正的水汽通道和短波红外通道,基于图像自身无法反演气溶胶散射,也就无法做到精确大气校正。Sentinel 2 MSI数据有一个水汽通道(第9波段),3个短波红外通道(第10-12波段),可以用于气溶胶反演和精确大气校正。Sentinel 2 MSI的L2A数据为地表反射率产品,在业内具有较高的认可度。且Sentinel 2 MSI 5天的重放周期也和HY1C/D CZI的3天较为相似,可以方便找到前后相邻的影像。如果HY1C/D CZI成像前后各找一景Sentinel 2 MSI L2A图像,分析图像的3*3像元值变化率,可以找到空间均匀的地物;分析两景图像的差异,可以找到随时间变化较小的像元,即为(准)不变地物,二者的交集即为均匀不变地物。而处于两景Sentinel 2 MSI L2A图像之间的HY1C/D CZI图像上均匀不变地物的地表反射率应该约等于两景Sentinel 2 MSI L2A图像的均值。据此可以建立两个传感器数据之间的相对大气校正模型,并用56个采样点数据建立基于实测数据的遥感反射率系统校正模型,实现HY1C/D CZI图像精确大气校正。本研究基于29个采样点对算法进行精度评估,结果表明,4个波段的平均相对误差RE为11.9%-75.5%,均方根误差RMSE为0.0005-0.0012,光谱形状相关系数r和光谱角度距离SAD分别为0.978和0.109。本研究解决了国产4波段多光谱影像大气校正的难题,为其它相似传感器提供了借鉴。另外,水体分布提取是水体大气校正的必要工作,本研究的大气校正必须在正确的水体区域进行,因此水体分布提取对本研究十分重要。本项目发展了基于深度神经网络和基于改进双峰法的地表水体提取方法,大大提高了水体提取的效率和精度。
{{i.achievement_title}}
数据更新时间:2023-05-31
府河-白洋淀硝酸盐来源判定及迁移转化规律
北京市大兴区夏季大气中醛酮类化合物的污染水平、来源及影响
Ordinal space projection learning via neighbor classes representation
胃神经内分泌肿瘤96例临床病理特征及预后分析
长江经济带城市黑臭水体空间分布格局及影响因子
主被动卫星遥感方式结合的河口海岸浑浊水体新型传感器大气校正研究
基于水体光学分类的内陆湖泊水体大气校正研究
内陆水体多角度高光谱图像大气校正研究
浑浊水体中多照明模式下成像优化处理方法研究