基于偏差原则和正则化方法反演晴空地表BRDF和反照率
Retrieval of Clear-sky Land Surface BRDF and Albedo Based on Discrepancy Principle and Regularization Method
-
摘要: 提出一种反演晴空地表反照率快速稳健的新算法——混合算法。该方法首先利用双参数模型函数确定正则参数的初始值,然后由基于Morozov偏差原则的高阶收敛算法确定正则参数,继而通过Tikhonov正则化手段来反演BRDF模型。从POLDER-3/PARASOL BRDF数据库中任意挑选不同覆盖的地表像元测量数据,与MODIS全反演法结果作比较,对比较结果给予了讨论;最后选择天津市地区的卫星图像进行反演实验,并就反演结果给出了误差分析和算法模型的评价。Abstract: A fast and robust novel algorithm is presented to retrieve land surface albedo over clear-sky areas,which is called hybrid algorithm.In this method,the initial guess value of regularization parameter was obtained by using the two parameter model function and then defined by a high order convergence algorithm.And the inversion of the kernel-driven BRDF model was implemented by Tikhonov regularization techniques.Measured data of surface targets with different cover types were chosen from POLDER-3/PARASOL BRDF datasets and then retrieved by the proposed algorithm.Comparison of results via the proposed method with that by MODIS full inversion was analyzed and discussed.At last,the retrieval experiments with MODIS satellite images over the regions of Tianjin were carried out,both error analysis and the evaluation of selected model and algorithm based on the retrieval results were also presented.
下载:
