基于最优光谱指数的大豆叶片叶绿素含量反演模型研究  

Study on Inversion Model of Chlorophyll Content in Soybean Leaf Based on Optimal Spectral Indices

作　　者：刘爽 于海业[1] 张郡赫 周海根 孔丽娟 张蕾[1] 党敬民 隋媛媛[1] LIU Shuang;YU Hai-ye;ZHANG Jun-he;ZHOU Hai-gen;KONG Li-juan;ZHANG Lei;DANG Jing-min;SUI Yuan-yuan(School of Biological and Agricultural Engineering,Jilin University,Changchun 130022,China)

机构地区：[1]吉林大学生物与农业工程学院,吉林长春130022

出　　处：《光谱学与光谱分析》2021年第6期1912-1919,共8页Spectroscopy and Spectral Analysis

基　　金：国家自然科学基金青年科学基金项目(31801259,32001418);吉林省科技发展计划项目(20200402015NC)资助。

摘　　要：叶绿素含量的准确获取及预测可为作物种植的精准化管理提供理论依据。利用最优光谱指数建立大豆叶绿素含量反演模型,以大豆花芽分化期叶片为研究对象,获取高光谱和叶绿素含量数据。首先构建了7种与叶绿素含量相关的典型光谱指数,分别为比值指数(RI)、差值指数(DI)、归一化差值植被指数(NDVI)、修正简单比值指数(mSR)、修正归一化差值指数(mNDI)、土壤调节植被指数(SAVI)和三角形植被指数(TVI),并对原始高光谱进行一阶微分(FD)处理,随后分别对原始和一阶微分高光谱在全光谱波长范围内两两组合所有波长,进行14个光谱指数的计算。再采用相关矩阵法进行最优光谱指数的提取,将所有波长组合计算出的光谱指数与叶绿素含量进行相关性分析,以相关系数最大值为指标,提取出14组最优的波长组合,并进行对应光谱指数值的计算作为最优光谱指数。最后将最优光谱指数划分为3组模型输入变量,分别与偏最小二乘回归(PLS)、最小二乘支持向量机回归(LSSVM)和套索算法LASSO回归3种方法组合建模并对比分析,以决定系数R_(c)^(2),R_(p)^(2)和均方根误差RMSEC,RMSEP作为模型评价指标,最终优选出精度最高的大豆叶片绿素含量反演模型。结果表明:14组最优光谱指数波长组合分别为RI(728,727),DI(735,732),NDVI(728,727),mSR(728,727),mNDI(728,727),SAVI(728,727),TVI(1007,708),FDRI(727,708),FDDI(727,788),FDNDVI(726,705),FDmSR(726,705),FDmNDI(726,705),FDSAVI(727,788)和FDTVI(760,698),相关系数最大值r_(max)均大于0.8。建立最优模型的方法为输入变量为一阶微分光谱指数(组合2)与LSSVM组合的建模方法,所建模型的R_(c)^(2)=0.7518,R_(p)^(2)=0.8360,RMSEC=1.3612,RMSEP=1.2204,表明模型精度较高,可为大面积监测大豆的生长状态提供参考。The accurate acquisition and prediction of chlorophyll content can provide a theoretical basis for precise management of crop planting.Optimal spectral index was used to establish the soybean chlorophyll content inversion model in this paper.The hyperspectral and chlorophyll content data of soybean flower bud differentiation were obtained.Firstly,seven typical spectral indices related to chlorophyll content were constructed,namely ratio index(RI),difference index(DI),normalized difference vegetation index(NDVI),modified simple ratio index(mSR),modified normalized difference index(mNDI),soil-adjusted vegetation index(SAVI)and triangular vegetation index(TVI),respectively.First derivative(FD)processing was performed on the original hyper spectrum,and then the original and first derivative hyper spectrum are combined with all wavelengths in the full spectrum wavelength range to calculate 14 spectral indices.Then use the correlation matrix method to select the optimal spectral index.The correlation analysis was conducted between the spectral index calculated by all wavelength combinations and chlorophyll content.The maximum value of the correlation coefficient was taken as the index to extract the 14 optimal wavelength combinations,and the corresponding spectral index value was calculated as the optimal spectral index.Finally,the optimal spectral indices were divided into three groups as model input variables combined with the three methods of Partial least squares regression(PLS),Least squares support vector machine regression(LSSVM),and LASSO regression to model,then compare and analyze the results.The coefficients of determination R_(c)^(2),R_(p)^(2)and the root mean square error RMSEC and RMSEP as model evaluation indicators,then soybean chlorophyll content inversion model with the highest accuracy,were finally selected.The results show that the 14 optimal spectral index wavelength combinations are RI(728,727),DI(735,732),NDVI(728,727),mSR(728,727),mNDI(728,727),SAVI(728,727),TVI(1007,708),FDRI(727,708),FDDI(727,

关 键 词：大豆 最优光谱指数 叶绿素含量 反演模型 

分 类 号：S565.1[农业科学—作物学]