齐长城土壤电导率的高光谱定量估测模型

Hyperspectral Quantitative Estimation Models of Soil Conductivity of the Qi Great Wall

建立夯土齐长城黄岛段土壤电导率高光谱估测模型。由采集的夯土齐长城黄岛段的土壤样本提取光谱数据,利用SG平滑和光谱微分技术,通过相关系数法筛选敏感波长,并以敏感波长作为自变量建立土壤电导率的高光谱定量估测模型,对比分析所建立的主成分回归、支持向量机和随机森林模型的精度,选择最优模型并验证。结果表明:839 nm、975 nm、1 279 nm和1 284 nm为敏感波长,经过对比分析所建立的模型,以随机森林模型为最优估测模型。随机森林模型能较好地估测夯土齐长城黄岛段的土壤电导率。

The hyperspectral estimation model of soil conductivity in Huangdao section of the Qi Great Wall of rammed earth was established. The spectral data were extracted from the soil samples collected in Huangdao section of the Qi Great Wall of rammed earth. The sensitive wavelength were selected by the correlation coefficient method with by SG smoothing and spectral differentiation technique. The hyperspectral quantitative estimation model of soil conductivity was established with the sensitive wavelength as the independent variable. The accuracy of principal component regression, support vector machine and random forest model was established by comparison analysis. The optimal model was selected and verified. The result demonstrated that 839 nm, 975 nm, 1 279 nm and 1 284 nm were sensitive wavelength. The models were established by comparative analysis. The random forest model was the best estimation model. The random forest model can estimate the soil electrical conductivity of Huangdao section of the Qi Great Wall of rammed earth better.