高光谱组合变换下土壤Cd含量GWR模型反演研究

Inversion of soil Cd content with GWR model under hyper-spectral combined transformation

对高光谱数据进行FD(一阶微分)、SD(二阶微分)、LOG-REC(倒数对数)、CR(连续统去除)处理基础上增加了10种组合变换处理。针对研究区受土壤重金属Cd元素严重胁迫且重金属高度变异程度问题,选取GWR(地理加权回归)模型对高度变异的Cd元素进行建模反演研究,研究结果表明:光谱组合变换中的REC-FD(倒数一阶微分)变换将重金属Cd元素与光谱之间相关性由原始的-0.276提升至-0.493,提升最为显著,其所建立的GWR模型(拟合优度)、MRE(平均相对误差)、RPD(相对分析误差)均比较理想,分别为0.82、29%、2.1。因此,光谱组合变换可以有效提升重金属Cd含量与光谱数据的相关性,GWR模型可以对高度变异的重金属Cd元素进行有效建模反演。

On the basis of conventional FD （first order differential）, SD （second order differential）, LOG REC （reciprocal logarithm）, and CR （continuous system removal） transform processing, ten kinds of combinatorial transform processing are added to deal with the hyper spectral data. Aiming at the severe stress of heavy metal Cd in the study area, the GWR （geographically weighted regression） model is used to model the highly mutated Cd element. Results show that the REC FD （reciprocal first order differential） transformation in spectral combination transformation increases the correlation between heavy metal Cd elements and spectra from the original 0.276 to 0.493. It is the most significant improvement so that the （goodness of fit）, MRE （average relative error）, and RPD （relative analysis error） of the GWR model are ideal, which are 0.82, 29% , and 2. 1, respectively. Therefore, spectral combination transformation can effectively improve the correlation between heavy metal Cd content and spectral data; and the GWR model enables efficient modeling and inversion of highly mutated heavy metal Cd elements.