CAJ | 학술논문

大气校正是遥感数据定量化应用的关键步骤，国产环境与灾害监测预报小卫星星座（简称 HJ-1） CCD相机数据在进行大气校正时，面对数据量大、观测角度变化大、气溶胶多变等各种问题。本文针对HJ-1 CCD相机的波段特征，利用Hsu等提出的深蓝算法借助地表反射率库反演得到气溶胶光学厚度，结合辅助角度数据计算观测几何，利用核驱动模型完成了BRDF校正，提出了针对HJ-1 CCD数据的在植被、裸土等地表类型通用的大气校正算法；在数据处理中，采用查找表和双线性插值的方法分块计算HJ-1 CCD图像的大气参数，利用IDL语言的矩阵运算大大加快了大气校正速度，实现了快速大气校正。以2012年7月3日过境中国华北平原的1景CCD数据进行了算法实验，结果表明，该算法较好的校正了大气对地表反射率的影响，能够在8 min内快速完成1景数据量约为1 G的CCD数据的大气校正，校正后的土壤和植被等典型地物的反射率更接近其光谱特征；同时，将校正结果与同期过境的M ODIS地表反射率产品进行了比对，由于分辨率的较高，HJ-1的结果细节表现要优于MODIS ，而二者获得的典型地物反射率相关性较好（相关系数大于0.9）。误差分析表明，气溶胶类型的误判会对近红外波段地表反射率带来较大的误差，在0.05左右，而地表反射率库0.02的误差，会对红波段和绿波段的大气校正结果带来0.01左右的误差。
대기교정시요감수거정양화응용적관건보취，국산배경여재해감측예보소위성성좌（간칭 HJ-1） CCD상궤수거재진행대기교정시，면대수거량대、관측각도변화대、기용효다변등각충문제。본문침대HJ-1 CCD상궤적파단특정，이용Hsu등제출적심람산법차조지표반사솔고반연득도기용효광학후도，결합보조각도수거계산관측궤하，이용핵구동모형완성료BRDF교정，제출료침대HJ-1 CCD수거적재식피、라토등지표류형통용적대기교정산법；재수거처리중，채용사조표화쌍선성삽치적방법분괴계산HJ-1 CCD도상적대기삼수，이용IDL어언적구진운산대대가쾌료대기교정속도，실현료쾌속대기교정。이2012년7월3일과경중국화북평원적1경CCD수거진행료산법실험，결과표명，해산법교호적교정료대기대지표반사솔적영향，능구재8 min내쾌속완성1경수거량약위1 G적CCD수거적대기교정，교정후적토양화식피등전형지물적반사솔경접근기광보특정；동시，장교정결과여동기과경적M ODIS지표반사솔산품진행료비대，유우분변솔적교고，HJ-1적결과세절표현요우우MODIS ，이이자획득적전형지물반사솔상관성교호（상관계수대우0.9）。오차분석표명，기용효류형적오판회대근홍외파단지표반사솔대래교대적오차，재0.05좌우，이지표반사솔고0.02적오차，회대홍파단화록파단적대기교정결과대래0.01좌우적오차。
In the present ,for the characteristic of HJ-1 CCD camera ,after receiving aerosol optical depth (AOD) from deep blue algorithm which was developed by Hsu et al .assisted by MODerate-resolution imaging spectroradiometer (MODIS) surface re-flectance database ,bidirectional reflectance distribution function (BRDF) correction with Kernel-Driven Model ,and the calcula-tion of viewing geometry with auxiliary data ,a new atmospheric correction method of HJ-1 CCD was developed which can be used over vegetation ,soil and so on .And ,when the CCD data is processed to correct atmospheric influence ,with look up table (LUT) and bilinear interpolation ,atmospheric correction of HJ-1 CCD is completed quickly by grid calculation of atmospheric parameters and matrix operations of interface define language (IDL) .The experiment over China North Plain on July 3rd ,2012 shows that by our method ,the atmospheric influence was corrected well and quickly (one CCD image of 1 GB can be corrected in eight minutes) ,and the reflectance after correction over vegetation and soil was close to the spectrum of vegetation and soil .The comparison with MODIS reflectance product shows that for the advantage of high resolution ,the corrected reflectance image of HJ-1 is finer than that of MODIS ,and the correlation coefficient of the reflectance over typical surface is greater than 0.9 .Error analysis shows that the recognition error of aerosol type leads to 0.05 absolute error of surface reflectance in near infrared band , which is larger than that in visual bands ,and the 0.02 error of reflectance database leads to 0.01 absolute error of surface reflec-tance of atmospheric correction in green and red bands .