The Research On Radiometric Correction Of Remote Sensing Image Combined With Sentinel-2 Data

With the development of earth observation technology,the application range of remote sensing images is becoming more and more extensive,and the quality requirements of remote sensing image data are correspondingly improved.At present,in the production process of visible remote sensing image products,the quality of radiation correction is difficult to guarantee,resulting in poor repeatability of remote sensing image after correction,which makes the development of subsequent quantitative remote sensing applications more difficult.From the perspective of benchmark thinking,this paper puts forward the idea of adding a radiation correction reference to restrict the calibration results during the radio calibration process.Sampling the surface reflectance data of typical features in the Sentinel-2 image,establishing a radiation correction reference database,and studying the radiation correction method controlled by the radiation correction reference will improve the quality of remote sensing image radiation correction results.Based on the assumption,the following researches have been done:1.Explore the selection method and description model of radiation correction benchmarks for remote sensing images,and propose the construction plan of the radiation correction benchmark database.The Sentinel-2 satellite open-source surface reflectance image with good radiation correction quality and high spatial resolution is selected as the radiation correction reference image for the radiation correction experiment in this paper.The pseudoinvariant feature(PIF)area is used as the reference for radiation correction.Use the surface reflectance data as a reference model for radiation correction and store it in the database.The method and process of establishing the radiation calibration reference database are described.2.Explore the comprehensive radiation correction method under the control of radiation correction benchmark.This paper proposes a comprehensive radiation correction method based on the 6S model of atmospheric correction,terrain C correction and least square linear regression correction of radiation correction reference control.At first,atmospheric correction is used to eliminate or weaken the radiation error caused by atmospheric absorption and scattering.The aerosol optical thickness parameters that have the largest impact on the accuracy of the correction are obtained from Sentinel-2 aerosol optical thickness data.Then,the terrain correction is used to eliminate or weaken the radiation error caused by the terrain fluctuations.In theory,the radiation error can be eliminated after atmospheric correction and terrain correction.But in fact,due to the error in obtaining the calibration model parameters,the radiation error of the image could not be completely eliminated.Therefore,it is necessary to use the radiation correction reference to perform radiation control on the results of the first two steps of correction to ensure the quality of the radiation correction.Radiation correction experiments were carried out on Gaojing-1 remote sensing image which is the product of the first domestic high-resolution commercial satellite.And the accuracy evaluation of the experimental results was carried on.Quantitative evaluation results show that,after correction by the method proposed in this paper,the difference between the measured values and statistical distribution of the surface reflectance of Gaojing-1 image and the radiation correction reference image is reduced,which shows that the radiation correction reference can effectively control the quality of radiation correction.The comprehensive radiation correction method under the control of the radiation correction benchmark studied in this paper provides a feasible solution for improving the radiation quality of remote sensing images.