论文部分内容阅读
先验知识在遥感反演中起到重要的作用,本文首先根据各向异性平整指数AFX和地表多角度观测数据集建立了BRDF二向性反射分布函数原型库,然后介绍了以BRDF原型作为先验知识的提取多角度数据地表反照率的原型反演算法,为验证算法的精度,将原型反演及全反演算法拟合验证数据子集得到的反照率与MODIS全反演算法拟合完整数据集得到的结果作比较。结果对比表明,观测天顶角在40°以内时,非垂直主平面的观测数据能够代表部分地表的各向异性反射特征,此时原型反演算法的最大平均绝对误差为0.036,相对精度比全反演算法高约5%—10%;当数据位于垂直主平面时,观测信息量严重不足,此时原型反演算法结果相对较稳定,而全反演算法的最大绝对误差高达0.18。综上所述,原型反演算法能够在整个采样空间范围内对BRDF进行约束,对观测数据位置分布的依赖较小,抗噪声能力强,在反演观测信息不足的多角度数据时,结果精度优于全反演算法。
Prior knowledge has played an important role in remote sensing inversion. In this paper, a prototype library of BRDF birefringence distribution function is established based on anisotropic flatness index AFX and surface multi-angle observation datasets. Then, In order to verify the accuracy of the algorithm, the albedo obtained by fitting the prototype inversion and the total inversion algorithm to the validation data subset is fitted to the MODIS full inversion algorithm Data sets obtained for comparison. The results show that observed zenith angle less than 40 °, the observed data from the non-vertical main plane can represent the anisotropy reflection characteristics of some of the earth’s surface. In this case, the maximum average absolute error of the prototype inversion algorithm is 0.036, The inversion algorithm is about 5% -10% higher. When the data is located in the vertical principal plane, the amount of observational information is seriously insufficient. At this time, the results of the prototype inversion algorithm are relatively stable, while the maximum absolute error of the total inversion algorithm is as high as 0.18. In summary, the prototype inversion algorithm can constrain the BRDF within the entire sampling space, has less reliance on the location distribution of the observed data and has strong anti-noise ability. When the multi-angle data with insufficient observational information is inverted, the accuracy Better than the total inversion algorithm.