论文部分内容阅读
以矿区复垦农田土壤为研究对象,利用实验室获取的土壤重金属元素砷(As)、锌(Zn)、铜(Cu)、铬(Cr)和铅(Pb)的含量与土壤可见近红外高光谱数据建立重金属元素含量的定量估算模型。为了保证模型预测的精度和稳定性,首先,对原始光谱数据进行平滑处理,并进行光谱变换,即:一阶导数,标准正态变量变换及连续统去除变换;然后,通过相关性分析提取不同变换光谱的特征波段;最后,将最小二乘支持向量机与传统的多元线性回归和偏最小二乘回归方法的结果相比较。研究表明:(1)以不同变换光谱数据建立反演模型均有较好的稳定性并达到一定精度,其中以最小二乘支持向量机方法优于偏最小二乘回归优于多元线性回归模型(除少数几个情况外);(2)从不同光谱变换数据中提取的光谱特征对反演模型结果有一定影响,其中以连续统去除和标准正态变量变换建模结果较好,一阶导数变换稍差。因此,利用高光谱遥感技术来定量估算土壤重金属含量是可行的,而且,必要的光谱预处理对提高估算模型的精度很有帮助。
Taking the reclaimed farmland soils in the mining area as the research object, the contents of arsenic (As), zinc (Zn), copper (Cu), chromium (Cr) and lead (Pb) Quantitative Estimation Model of Heavy Metal Elements Content Based on Spectral Data. In order to ensure the accuracy and stability of the model prediction, the original spectral data is smoothed and the spectral transformation is performed first, that is, the first derivative, the normal normal transformation and the continuous unification transformation; then, the correlation analysis is used to extract the differences Finally, the least square support vector machine is compared with the traditional method of multiple linear regression and partial least squares regression. The results show that: (1) The inversion model established with different spectral data has better stability and accuracy, and the least square support vector machine is better than the PLS regression model in multiple linear regression model ( Except for a few cases). (2) The spectral characteristics extracted from different spectral transformation data have some influence on the inversion model results. The modeling results are obtained by the continuum removal and standard normal transformation. The first derivative Transform slightly worse. Therefore, it is feasible to use hyperspectral remote sensing to quantitatively estimate heavy metal content in soil, and the necessary spectral pretreatment is helpful to improve the accuracy of the estimated model.