黄河三角洲盐碱土盐分的高光谱定量估测研究

Hypersperpectral Quantitative Estimation of Saline-alkali Soil Salinity in the Yellow River Delta

利用高光谱探测数据,对黄河三角洲盐碱土盐分含量进行定量估测,为土壤盐分含量提供一种快速、实时、准确的估测方法。以黄河三角洲为研究区,在分析盐碱土光谱曲线变化规律的基础上,进行光谱反射率一阶微分和倒数的对数与土壤盐分含量进行相关分析,筛选敏感波长,运用主成分回归分析方法,建立土壤盐分含量估测模型,并对模型进行精度验证。结果表明,一阶微分变换可以扩大样品之间的光谱特征差异,提高了相关性;通过相关分析选出的敏感波长为388 nm、515 nm、962 nm、1164 nm、1623 nm和2150 nm;建立了土壤盐分含量的主成分回归估测模型y=4.389-798x388′-632x515′-151x962′-184x1164′-138x1623′-217x2150′,经精度验证,模型(R2=0.8405,RMSE=1.152)拟合效果较好。利用主成分回归分析方法估测黄河三角洲地区的土壤盐分含量精度较高,为实现土壤盐分的定量估测提供了理论依据和技术支撑。

This paper is to use hyperspectral data to conduct quantitative estimation of saline-alkali soil salinity spectrum on Yellow River Delta, providing a quick and real-time and accurate estimation method for soil salt content. In the studied area of Yellow River Delta, based on the change law analysis of soil spectrum curve, by means of related analysis on the content of soil salt and the transformation of first order differential and the logarithm of reciprocal of spectral reflectance, screening sensitive wavelength, it uses principal component regression analysis method to set up the soil salt estimating model, and verifies the model accuracy. The results show that first order differential transform can expand the difference of spectrum characteristics between samples and improve the correlation; According to the correlation analysis it selects 388 nm, 515 nm, 962 nm, 1164 nm, 1623 nm and 2150 nm as sensitive wavelength; Using principal component regression analysis to establish the soil salt estimating model Y=4.389-798x338＇-632x515＇-151x963＇-184x1164＇-138x1623＇-217x2150＇, by way of the model accuracy verification, the model （R2 = 0.8405, RMSE = 1.152） is good. Principal component regression analysis method is suitable for the estimation soil salt content of the Yellow River Delta, providing theoretical basis and technical support to realize the soil salt quantitative estimation.