基于非成像高光谱烟叶生化参数估测

Estimation of Tobacco Biochemical Parameters Based on Non-Imaging Hyperspectral

为快速无损地检测出烟叶氯含量,掌握烟草植株的健康状况,控制烟叶的品质,以云南省陆良县大漠古乡烟草种植基地主要栽种的云烟97号为研究对象,通过采集烟叶高光谱数据,并测定其氯含量,利用相关性分析和多元逐步回归或线性拟合法分析了氯含量与原始光谱反射率、反射率的一系列变换形式、高光谱特征变量间的关系。结果表明:烟叶氯含量高光谱反射率特征波长为420 nm;光谱一阶微分特征波长为939 nm;光谱对数特征波长为419~422 nm;光谱对数一阶微分特征波长为372 nm;光谱归一化特征波长为1005 nm;光谱归一化一阶微分特征波长为1694 nm;特征变量λr(红边位置)与烟叶氯含量相关性最好;光谱归一化的逐步回归方程对烟叶氯含量的估测效果最好(R^(2)=0.632,P<0.01)。

In order to quickly and nondestructively detect the chlorine content of tobacco leaves,master the health status of tobacco plants,and control the quality of tobacco leaves,Yunyan No.97,which was mainly planted in the tobacco planting base of Da Mu Gu Township in Luliang County,Yunnan Province,was selected as the research object by collecting the hyperspectral data of tobacco leaves and determining its chlorine content.The correlation analysis and multiple stepwise regression or linear fitting method were used to analyze the relationship between chlorine content and the original spectral reflectance,a series of transformation forms of reflectance and hyperspectral characteristic variables.The results showed that the characteristic wavelength of hyperspectral reflectance of chlorine content in tobacco leaves was 420 nm.The first derivative characteristic wavelength of the spectrum is 939 nm.The spectral logarithmic characteristic wavelength is 419~422 nm;The spectral logarithmic first derivative characteristic wavelength is 372 nm;The normalized spectral characteristic wavelength is 1005nm;The normalized first-order differential characteristic wavelength is 1694nm.The characteristic variableλr(red edge position)had the best correlation with chlorine content in tobacco leaves.The spectral normalization stepwise regression equation was the best for estimating chlorine content in tobacco leaves(R^(2)=0.632,P<0.01).