秋玉米SPAD值的光谱估算模型研究

研究不同秋玉米品种叶片SPAD值与一阶微分光谱的相关关系,筛选出8个品种的敏感波长和一阶微分光谱参数,建立不同品种SPAD值的一元线性、指数、多项式和基于一阶微分光谱参数的预测模型,计算建模集和验证集的均方根误差RMSE和相对误差RE。结果发现,不同秋玉米品种SPAD值与光谱一阶微分相关系数较高,均在0.8以上;不同秋玉米品种光谱一阶微分的敏感波段位于650~680nm;基于敏感波长建立的SPAD值预测模型均表现出较好效果,多项式模型效果最佳,模型估测精度在不同品种间存在差异;在8个秋玉米品种中,正大999的多项式预测模型表现最佳,其建模集RMSE和RE分别为2.762和3.643%,其验证集RMSEv和REv分别为3.322和4.518%。

Hyperspectral Estimation of SPAD Values in Different Varieties of Autumn Maize

By studying the correlation between SPAD value of different autumn maize leaves varieties and first derivative spectrum, the sensitive wavelengths and first derivative spectrum parameters of 8 autumn maize varieties were screened out, and linear, exponential, polynomial and first derivative spectrum parameters’ prediction models of SPAD value of different autumn maize varieties were established, and RMSE and RE of modeling set and verification set were calculated. The results showed that higher correlation between the SPAD value and the first derivative spectrum of different autumn maize varieties, which reached more than 0.8. The sensitive bands of different autumn maize varieties in first derivative spectrum are between 650-680nm. The SPAD value prediction model based on the sensitive wavelength performed well, especially by the polynomial model, but its estimation accuracy was different among the varieties. Among the 8 varieties of autumn maize, the prediction model of Zhengda 999 had the best performance, its modeling set RMSE and RE were 2.762 and 3.643%, respectively, and its verification set RMSEv and REv were 3.322 and 4.518%, respectively.