Study On Sampling Method And Uncertainty Of Surface Reflectance Based On Simulated Annealing

Surface reflectance products are the basis of optical remote sensing products,and their accuracy restricts the application and development of surface reflectance products and their derivatives.Authenticity test is a process to verify the quality and accuracy of remote sensing products.Obtaining the relative truth value of product pixels through ground spectral sampling and comparing the pixel value of remote sensing products with it is an important way to test the authenticity of surface reflectance products.The ground spectrum sampling is required to be completed within one hour before and after the satellite transit.Due to the limitations of work efficiency and technical means,only a few points can be observed within the specified time.Using a few point data to represent the surface reflectance of a large area will lead to low sampling accuracy and difficult to ensure the reliability of the measured relative truth.Solving this contradiction is the difficulty of the accuracy inspection of surface reflectance products.When the number of sample points is small,simulated annealing sampling can improve the sampling accuracy by optimizing the distribution position of sampling points,but there is less quantitative evaluation of sampling accuracy in the existing research.During the authenticity inspection of surface reflectance products,it is unknown how effective simulated annealing optimized sampling is and whether it can reduce the uncertainty of scale conversion in the authenticity inspection process,and whether it can be applied to different types of products,This research has not been reported.Therefore,it is of great theoretical significance and practical value to carry out sampling research on the authenticity test of different types of surface reflectance products and find the best sampling method.In order to determine the best sampling method for different types of surface reflectance products,the study area is divided into Gobi study area and vegetation coverage study area according to the type of surface features,and 150m and 3km sites are set in each study area to carry out the authenticity inspection and sampling research of surface reflectance products with different types of surface features and different resolution;Based on the multi temporal 10m resolution sentry-2 multispectral satellite data of each site,the sampling accuracy of simulated annealing sampling method and system sampling method is quantitatively evaluated and compared;Combined with the parameters such as spatial registration error and reference image error that affect the scale conversion uncertainty of authenticity test,the influence of the accuracy of sampling method on the scale conversion uncertainty is estimated based on sentry ? data and high-resolution UAV data,so as to determine the best sampling method;Based on the analysis results of scale conversion uncertainty,the authenticity of surface reflectance products of FY-2 and landsat8 satellites is tested in this paper.Research findings:(1)In the study areas of different ground feature types,simulated annealing sampling can improve the sampling accuracy.Among them,the accuracy of simulated annealing sampling in the vegetation coverage study area is significantly higher than that of systematic sampling,but there is little difference between the accuracy of the two sampling methods in the Gobi study area.(2)In the vegetation cover study area,the sampling accuracy has a great impact on the scale conversion uncertainty of product authenticity test with different resolutions.The use of simulated annealing sampling can effectively reduce the scale conversion uncertainty;In the Gobi study area,the sampling accuracy has little effect on the scale conversion uncertainty,and the scale conversion uncertainty of different sampling methods is close.(3)For the authenticity test of Landsat surface reflectance products,the best sampling method to obtain the true value of Gobi pixel is 5-point systematic sampling method,and the uncertainty of scale conversion is 0.46%.The best sampling scheme of vegetation pixel is 5-point simulated annealing sampling method,and the uncertainty of scale conversion is 3.85%;For the authenticity test of meteorological satellite surface reflectance products,the best sampling method for obtaining the true value of Gobi pixels is the 5-point systematic sampling method,and the uncertainty of scale conversion is 0.98%.The best sampling scheme of vegetation pixels is the 9-point simulated annealing sampling method,and the uncertainty of scale conversion is 6.79%.(4)The authenticity test results of the surface reflectance products of FY-2 and landsat-8 satellites show that there is an error in the surface reflectance of FY-2 in the visible band,and the surface reflectance products of landsat-8 are relatively accurate,and the accuracy of the visible band is high.