GF-1星WFV相机的快速大气校正

高分一号卫星(GF-1)WFV相机是中国新型高分辨率传感器,为了更好地进行定量应用,需完成高精度大气校正,但需要解决数量大,辅助数据不足等关键问题。针对WFV相机构建了快速大气校正模型,(1)采用交叉定标方法借助Landsat 8数据完成辐射定标;(2)从WFV相机的辅助数据出发,计算得到太阳天顶角、观测天顶角等辅助信息;(3)考虑不同海拔大气分子散射的不同,完成基于海拔数据的分子散射校正;(4)采用深蓝算法,从第一波段(蓝光)反演得到气溶胶信息;(5)计算每个像元的大气校正参数,进而获取地表反射率,完成大气校正。在此基础上,利用IDL语言建立相应的大气校正模块,以过境华北地区的3景WFV数据为例进行大气校正实验。结果表明,模型能够快速完成大气校正,并能较好的去除大气分子与气溶胶影响,较好地还原植被、裸土等典型地表类型的光谱反射曲线,校正后的NDVI更好地反映了各地物的特征。

Quickly atmospheric correction for GF-1 WFV cameras

Four Wide-Field-Viewing (WFV) cameras are taken onboard the GF-1 satellite, which is a newly launched earth-observing satellite from China. The satellite is employed to monitor land use, environmental parameters, and agriculture, among others. However, a highaccuracy Atmospheric Correction (AC) algorithm is imperative to process the GF-1 WFV data for quantitative applications. The key problems in the AC of WFV cameras include the following:large amount of data, lack of auxiliary data, and aerosol and molecular variations. In the paper, an AC algorithm for GF-1 WFV data is introduced. Based on radiance transfer theory, the fast AC algorithm for WFV data was established as follows:(1) The radiometric calibration was completed in four seasons by cross-calibration method using Landsat 8 data. The apparent reflectance in all four bands of the WFV camera was received at the solar zenith angle, and the solar irradiance was obtained at the top of atmosphere.(2) The sun and viewing zenith angles were calculated at 1km resolution with the use of the auxiliary WFV data, including projection information, satellite passed time, view zenith angle at nadir, and pixel position.(3) Rayleigh scattering was corrected for each pixel in an image with the use of altitude data and the second simulation of the satellite signal in the solar spectrum (6S) in the same view geometry.(4) Aerosol Optical Depth (AOD) was derived from the apparent reflectance in the blue band by the deep blue algorithm at 10 km resolution with the use of MODIS 8-day surface reflectance product.(5) In every 10 km×10 km block of WFV image, the retrieved AOD was inputted into the6S, and the three atmospheric parameters were determined. Then, from the apparent reflectance in the four bands, the surface reflectance in the four bands was retrieved using the atmospheric parameters in every block. After all the blocks were processed, the AC of the WFV image was completed.The AC module for GF-1 WFV data was developed using interactive data language and our AC algorithm. Three GF-1 WFV images over North China Plain acquired on September 27, 2013, March 13, 2015 and May 25, 2015 were selected to conduct the experiments that will verify the performance of our AC algorithm and module. The results show that our algorithm has significantly removed the atmospheric influences, including molecular and aerosol scattering and absorption. However, if the aerosol layer is thick, the influence of the atmosphere cannot be completely removed. From these images, we select three typical surfaces for further study, including vegetation, soil, and urban. Then, the reflectance after AC is compared with that before AC. The reflectance after AC was close to the spectrum of these surfaces, and the corrected normalized difference vegetation index reflects the character of the typical surface.In this paper, a new AC algorithm based on the aerosol retrieved from the deep blue algorithm was built for GF-1 WFV data. The scattering and absorption of molecules and aerosols in the GF-1 WFV data were well corrected using the proposed algorithm, which also allowed for the rapid acquisition of surface reflectance. However, our algorithm may still be improved in terms of robustness against high-concentration aerosol, such as haze, and against adjacency effect over non uniform surface.