Researh On Method And System For Radiometric Cross-calibration Of China's Optical Satellite Remote Sensing Data On VNIR Bands

In order to better use the domestic satellite imagery in all aspects of national economic development,the quantitative level of remote sensing satellites need to be further improved.Sensor's calibration is the prerequisite to read the remote sensing images correctly and achieve quantitative remote sensing applications.At present,onboard calibrators for most China's domestic optical remote sensing satellites are not perfect,and on-orbit calibration of these satellites depends mainly on site calibration.However,due to cost and experimental conditions,the calibration frequency is low,which makes it difficult to monitor the performance of the sensors in time.Therefore,radiometric cross-calibration with low-cost and high-efficiency has become a research hotspot.The main scientific objects of this thesis and conclusions are summarized as follows:(1)A comprehensive procedure for evaluating radiometric capability of several Chinese optical satellite sensors is proposed.The stability of the surface of the calibration site,the Badain Jaran Desert,has been verified in brightness,spatial homogeneity,altimetry and bidirectional effects,seasonal variation and long-term stability to make sure that the selected calibration area is uniformity and stability to be chosen as the calibration area.Long-term radiometric stability and radiometric accuracy are the two major indicators for radiometric evaluation.The radiometric temporal stability is analyzed by the tendency of long-term top-of-atmosphere(TOA)reflectance variation;the radiometric accuracy is determined by comparing with the TOA reflectance from MODIS after spectrally matching.Three Chinese sensors including HJ-1/CCD,FY-3/VIRR and FY-3/MERSI are evaluated in VNIR bands based on this procedure.The evaluation results show that the existing site calibration frequency is too low to monitor the sensor performance changes timely.Therefore,a more frequent calibration method needs to be developed.(2)A cross-calibration procedure for domestic Moderate and Low-resolution remote sensing satellites in VNIR bands is proposed based on MODIS time series data.The procedure is mainly aimed at Moderate and Low-resolution remote sensing satellites with wide field of view.In this procedure,on the basis of the ‘clean' pixel method,seasonal factor has been taken into account to retrieve the BRDF model and AOD.The Staylor-Suttles model was taken as the BRDF model and two sets of coefficients of BRDF model were produced based on the long time series of MODIS data for summer and winter.The surface reflectance of the sensors to be calibrated is retrieved by the BRDF model and then the TOA reflectance is retrieved by MODTRAN.Four sensors(FY-3A/MERSI,FY-3A/VIRR,FY-3B/MERSI and FY-3B/VIRR)were cross-calibrated based on the procedure.Through compared with site calibration,the cross-calibration result shows that the new cross-calibration procedure can more effectively correct the inconsistency of sensors' radiometric capability than the existing site calibration.(3)A cross-calibration procedure for domestic Moderate and High-resolution remote sensing satellites with wide field of view in VNIR bands is proposed based on the Landsat-8/OLI and ZY-3/TLC data.The procedure improved the existing crosscalibration method based on the Landsat-7/ETM+ and ASTER GDEM data.The new BRDF model was built with the solar illumination and the view geometries from OLI and slope and aspect from DEM produced by ZY-3/TLC.Then the surface reflectance of the images to be calibrated was calculated by the new BRDF model.Compared with the old one,the surface reflectance calculated by the new BRDF has more detail and high accuracy.The TOA radiance was retrieved with the surface reflectance using the atmosphere radiant transfer model and the calibration coefficients were fitted with the TOA radiance and corresponding(DN)values of the image.Two sensors(GF-1/WFV and GF-4/PMS)were cross-calibrated based on the procedure.Through compared with site calibration,the cross-calibration result shows that the new cross-calibration procedure has a higher accuracy(less than 5% error),and can more effectively correct the inconsistency of sensors' radiometric difference among different sensors or different images obtained by the same sensor at different time.(4)A system of cross-calibration is established for domestic remote sensing sensors in VNIR band.According to the resolution and observation characteristics,the existing domestic satellite sensors are classified and the appropriate radiometric crosscalibration procedure is selected for each category.The system covers a variety of domestic optical remote sensing satellite sensor,and also applicable to some foreign sensor with the same type.Based on the data after radiometric normalization produced using this system,a series of quantitative remote sensing products were produced,like 1km 5 day synthetic vegetation index products of China-ASEAN,and 30 m monthly land cover classification in Heihe Basin.These applications illustrates the important role of radiometric cross-calibration in the inversion of surface parameter with multisource remote sensing data.The cross-radiation calibration procedure and system of this thesis are of great significance to monitor the performance of domestic satellite sensors in real time,promote the application of domestic satellite industry and promote the multi-source comprehensive application of domestic satellite and international satellite.