基于高光谱的枣树叶片氮素表征方法

Method for characterizing nitrogen in jujube leaves based on hyperspectral analysis

为提高枣树种植过程中施用氮肥的精准性,本研究以南疆重要经济作物骏枣(Ziziphus jujuba Mill.)为研究对象,通过对枣树叶片原始光谱和一阶微分光谱与全氮含量的相关性进行分析,利用光谱敏感变量构建植被指数作为衍生变量,再以衍生变量作为变量建立多种线性和非线性的氮素含量预测模型,并对氮素含量预测模型进行精度检验。结果显示:基于枣树原始光谱和一阶微分光谱的模型拟合决定系数均大于0.75,原始光谱变量的预测效果整体好于一阶微分光谱;预测效果最好的是基于原始光谱变量4的幂函数模型:Nit=1.097x^(0.735),R^(2)为0.821,RMSE为0.024 5。研究表明,建立的氮素含量预测模型能够实现基于高光谱反射率特征对枣树氮素的较好监测效果,能够作为枣树营养素诊断的重要理论依据。

Jujube as one important economic crop in Southern Xinjiang was used to analyze the relationship between raw spectra and first-order differential spectra of jujube leaves and the content of total nitrogen with hyperspectral techniques.A model for predicting the content of nitrogen was established to provide a theoretical basis for nitrogen monitoring and precise fertilization during jujube cultivation.Spectral sensitive variables were used to construct vegetation indices as derivative variables.Multiple linear and nonlinear models for predicting the content of nitrogen were established using derivative variables as variables.The accuracy of models for predicting the content of nitrogen was tested.Results showed that the fitted decision coefficients of models based on the original spectra and first-order differential spectra of jujube trees were greater than 0.75.The overall prediction performance of the original spectral variables was better than that of first-order differential spectra.The best prediction was based on the power function model of the original spectral variables 4:Nit =1.097x^(0.735),R^(2)=0.821,and RMSE=0.024 5.It is indicated that the model established for predicting the content of nitrogen can achieve good effect of monitoring nitrogen in jujube tree based on hyperspectral reflectance characteristics,and can serve as an important theoretical basis for the nutrient diagnosis of jujube tree.