摘要
为验证高光谱成像技术在甜菜叶片氮素估测方面的可行性,以田间试验为基础,利用高光谱成像系统获取高光谱图像。通过分析不同预处理(Savitzky-golay平滑(S-G)、标准正态变换(SNV)、多元散射校正(MSC)以及一阶微分(FD))下光谱反射率与叶片ρN的相关关系,确定敏感波段,在此基础上,构建偏最小二乘(PLS)模型,探究不同预处理方法对模型的影响。为简化模型,利用连续投影法提取特征波长并构建光谱参数(RDI),分别以敏感波段范围内的全波长,特征波长和RDI为自变量,建立PLS和支持向量机(SVM)预测模型,研究不同输入变量下的模型预测精度。结果表明:甜菜叶片ρN的敏感波段范围为700～900 nm; SNV预处理后,模型的预测效果最好;利用连续投影算法(SPA)提取的特征波长为716、757、789、822和899 nm;基于RDI的SVM模型对甜菜叶片ρN的预测效果最佳,预测集的决定系数为0.78,均方根误差为3.08 g/kg,相对分析误差为1.47。研究为无损、快速、经济估测甜菜叶片氮素营养提供了理论参考。
Field experiments were applied to verify the feasibility of hyperspectral imaging techniques in the estimation of nitrogen in sugar beet leaves in this study.Savitzky-golay smoothing(SG),standard normal transform(SNV),multiple scatter correction(MSC)and the first-order differential(FD)were used to preprocess the original spectrum.By comparing the results of significant analysis of spectral reflectance and leaf nitrogen content under different pretreatments,it was found that the sensitivity band of nitrogen content in sugar beet leaf ranged from 700 to 900 nm.On this basis,partial least-square(PLS)prediction models were constructed to explore the effect of each pretreatment method on the model.The results showed that the prediction model using pretreatment method of SNV was the best.In order to simplify the model,a continuous projection algorithm(SPA)was used to extract the characteristic wavelengths,and the RDI spectral parameters are constructed by the extracted characteristic wavelengths.The multivariable linear regression(MLR),support vector machine(SVM)and PLS prediction models were established,using the full wavelength,characteristic wavelength,and RDI as the independent variables.The results showed that the SVM model based on RDI had the best prediction effect on nitrogen content in sugar beet leaves.The prediction set coefficient was 0.78,the root mean square error was 3.08 g/kg,and the relative analysis error was 1.47.This study provides a theoretical reference for non-destructive,rapid and economical estimation of nitrogen nutrition in sugar beet leaves.
作者
张晶
张珏
田海清
ZHANG Jing;ZHANG Jue;TIAN Haiqing(College of Mechanical and Electrical Engineering,Inner Mongolia Agricultural University, Hohhot 010018,China;College of Physics and Electronic Information,Inner Mongolia Normal University, Hohhot 010022,China)
出处
《中山大学学报（自然科学版）》
CAS
CSCD
北大核心
2018年第6期103-112,共10页
Acta Scientiarum Naturalium Universitatis Sunyatseni
基金
内蒙古自然科学基金(2016MS0346)
国家自然科学基金(41261084)
关键词
高光谱图像
甜菜
叶片ρN
特征波长
支持向量机
hyperspectral image
sugar beet
leaf nitrogen content
characteristic wavelength
support vector machines
