基于集合经验模态分解的土壤重金属Cd定量反演

针对高光谱定量反演土壤重金属含量的模型精度不足的问题，本文从时频空间的角度引入时频分析法——集合经验模态分解（EEMD）。采用EEMD法分解土壤高光谱，获得不同频率的本征模态（IMF）分量，通过分析IMF分量与重金属含量的相关性，提取特征光谱，构建EEMD-SVM定量反演模型。研究结果表明，通过EEMD法分解土壤光谱，可有效地提取土壤光谱中的微弱信息；构建EEMD-SVM模型可较好地反演土壤重金属Cd含量，模型的决定系数R²为0.920 3，明显高于基于一阶微分处理光谱数据后构建的SVM模型的R²（0.786 6）。即说明在土壤重金属定量反演领域，EEMD可作为一种新的光谱处理方法。

Quantitative inversion of soil heavy metal Cd based on ensemble empirical mode decomposition

In view of the lack of accuracy of the model for hyperspectral quantitative inversion of soil heavy metal content, this paper introduces the time-frequency analysis method seting ensemble empirical mode decomposition(EEMD) from the perspective of time-frequency space, and explores the application of EEMD method in hyperspectral quantitative inversion of soil heavy metal content.The EEMD method is used to decompose soil hyperspectral to obtain intrinsic mode function (IMF) components with different frequencies. By analyzing the correlation between IMF components and heavy metal content, characteristic spectra are extracted to construct an EEMD-SVM quantitative inversion model.The results show that the EEMD method can effectively extract the weak information in the soil spectrum.The determination coefficient R² of the EEMD-SVM model is 0.920 3, which is significantly higher than that of the SVM model based on the first-order differential processing of spectral data (0.786 6).It is explained that EEMD can be used as a new method of spectral treatment in the field of quantitative introspection of soil heavy metals.