Study Of Expression And Application Of Remote Sensing Inversion Model Errors

The theories and methods research of multi-stage remote sensing inversion on the basis of prior knowledge is one of the most pressing task in the quantitative remote sensing inversion. Traditional remote sensing inversion were established on some assumptions, which are not always right. This paper devotes to mining the incomplete information of error data as prior knowledge, which can apply to the multi-stage inversion and improve the precision of some model's parameters. Wetake the SAIL model as an example, and the results show that it is feasible.This thesis consists of four chapters. In the first chapter, we summarize thebasis and significance of the selected topic, and the definition of the remote sensing inversion.The second and third chapters are the main parts of this thesis. In the second chapter, we take the difference between the forward simulated BRDF values of Radiosity and SAIL model as error data. Firstly, we test that the error data are not normal distribution, and even the expectation is not zero. Then we analyze the possible type distribution of these error data. Finally, we use the stepwise method to fit the error data and obtain a function for some parameters of the SAIL model. As prior knowledge, the fitting function will be used to inverse these parameters ofthe SAIL model soon.In the third chapter, we focus on the application of the prior knowledge fromthe error data and the genetic algorithms. After establishing the USM matrices, we use the present popular genetic algorithms in the multi-stage inversion of the SAIL model. Through comparing with the multi-stage inversion which did not recommend the error prior knowledge and directional inversion, the results indicate that the superiority of the multi-stage inversion and we can improve the inversion precision after recommending the error prior knowledge. We also give some evaluation of theresults.The last chapter discusses the incompletion of this paper, and puts forwardsome prospects as while, in order to make further research for the process and results of the error.