风云卫星中分辨率遥感数据几何定位误差分析  

Systematic Analysis of Geometric Performance of Fengyun-3C MERSI Satellite Data Using Image Chip Matching Method

作　　者：范锦龙[1] 张晔萍[1] 李昌宝 许文波[2] 刘少杰 薛飞 覃志豪[3] Fan Jinlong;Zhang Yeping;Li Changbao;Xu Wenbo;Liu Shaojie;Xue Feil,;Qin Zhihao(National Satellite Meteorological Center,Beijing 100081,China;University of Electronic Science and Technology of China,Chendu 611731,China;Institute of Agricultural Resources and Regional Planning,Chinese Academy of Agricultural Sciences,Beijing 100081,China)

机构地区：[1]国家卫星气象中心,北京100081 [2]电子科技大学,四川成都611731 [3]中国农业科学院农业资源与农业区划研究所,北京100081

出　　处：《遥感技术与应用》2018年第4期621-627,共7页Remote Sensing Technology and Application

基　　金：国家重点研发计划(2016YFA0600301;2016YFA0600302);国家自然科学基金项目(41271429);欧盟FP7SIGMA603719

摘　　要：几何定位误差是影响卫星遥感数据应用的基础性问题。提出了一种适用于量化分析FYMERSI数据几何定位精度的方法(90%同名控制点误差CE90、平面几何误差RMSEH、经向几何误差RMSEX和纬向几何误差RMSEY 4个指标来确定几何定位精度):首先通过图像自动匹配技术对单幅图像进行控制点检查,然后对大批量的数据进行系统性几何定位误差分析。华北地区的典型研究表明,FY-MERSI数据经系统性纠正后,其几何定位精度存在不均一性,有些数据的几何定位误差可低至CE90=1.41像元,RMSE-H=0.97像元,RMSEX=0.74像元和RMSEY=0.64像元,而有些数据的几何定位误差则高达CE90=14.87像元,RMSEH=13.0像元,RMSEX=9.45像元,RMSEY=8.93像元。2013年10月5日至2015年5月30日所有数据的分析表明,FY3CMERSI数据经在系统性几何纠正后,CE90为5.97像元,RMSEH为4.94像元,RMSEX为3.29像元,RMSEY为3.28像元。该方法可以减少人工参与,大幅减轻目视确定几何定位必需的控制点的工作量,提高几何定位工作效率。因此,该方法将有助于量化分析FY-MERSI数据的几何定位误差,为FY-MERSI数据进行几何精校正提供技术基础。Geometrically corrected satellite image with the high accuracy is the key in support of the satellite data applications.However,the geometric errors of the data from the MEdium Resolution Spectrum Imager onboard the second generation of polar orbiting Chinese Meteorological satellite after the systematically geometric correction is still high.This paper came up with an approach to quantify the geometric errors and made the further analysis of these errors,aiming at the next step for the further geometric correction.The approach takes 4 indices to quantify the geometric errors：CE90 indicates the radius of the circle errors for the 90%control points;RMSEH indicates the horizontal errors on the image;RMSEX and RMSEY indicate the longitude and latitude errors,respectively.The image chip matching technology is applied to identify the control points on the image.The errors for the single image and time series of images were further calculated using abovementioned 4 indices.The case study in North China Plain shows that the geometric errors for various images are uneven and no uniform systematical geometric error found.Sometimes the geometric error may be small as CE90 1.41 pixels,RMSE H 0.97 pixels,RMSEX 0.74 pixels and RMSEY 0.64 pixels but for example April 29,2014,the geometric may be large as CE 90 14.87 pixels,RMSEH 13.0 pixels,RMSEX 9.45 pixels,RMSEY 8.93 pixels.The averaged geometric errors for the period of October 5,2013 to May 30,2015 were as CE90 5.97 pixels,RMSEH 4.94 pixels,RMSEX 3.29 pixels and RMSEY 3.28 pixels.A dramatic decrease of the geometric errors was observed since the beginning of the year 2015 due to the update of the systematical preprocessing of the satellite data at the end of 2014.This update improved the geometric errors with the decrease from approximate 15 pixels to below 8 pixels.This proposed approach has obvious advantages in terms of efficiency in comparison with manually identifying control points since it adopts image chip matching technology.

关 键 词：风云卫星 中分辨率数据 几何纠正 误差分析 图像自动匹配 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置]