基于光谱分析的库尔勒香梨叶片全氮含量估测研究

Prediction of the total nitrogen content in leaves of Kuerle fragrant pear by spectral analysis

【目的】为了实现无损、实时、快速地监测香梨养分状况,【方法】利用SVC HR—768便携式光谱仪测定大田20 a树龄库尔勒香梨叶片的光谱反射率,并结合室内叶片氮素含量分析,采用逐步回归法对香梨叶片的全氮含量与原始光谱、一阶导数光谱、高光谱参数之间的关系进行分析。【结果】全氮含量与原始光谱在761 nm处构建的线性模型,调整决定系数R2值达到0.986;在1 564 nm处一阶微分光谱与全氮含量构建的线性模型,调整决定系数R2值为0.997;对于所选的高光谱特征变量,叶片全氮含量与红边位置变量(Dr)和红边面积变量(SDr)的相关关系极显著,并由此建立的线性模型的调整决定系数R2值均达到0.9以上,说明这些变量预测香梨叶片全氮含量具有可行性。【结论】对所构建的方程进行检验,确定了基于1 564 nm的一阶导数光谱模型为香梨叶片全氮含量的最佳预测模型。

[Objective]In order to monitor the nutrient status in leaves of Korla fragrant pear non-destruc- tively, timely and quickly, [Method]the spectral reflectance of leaves of 20-year-old Korla fragrant pear tree was measured in the field using a SVC HR-768 portable spectrometer. Combined with total nitrogen content analysis in the lab., stepwise regression method was used to analyze the relationships among the total nitrogen content of leaves, the original spectrum, the first derivative spectrum, and the high spectral parameters. [Result]One linear model was built at 761 nm between the total nitrogen content and the orig- inal spectrum, the adjusted determination coefficient value R2 was 0.986. Another linear model was built at 1 564 nm between the total nitrogen content and the first order derivative spectrum, and the adjusted determination coefficient value R2 was 0.997. To select hyperspectral feature variables, the relationships among the total nitrogen content of leaves, the red edge position （Dr）, and the red edge area （SDr） were extremely significant, and the adjusted determination coefficient values R2 were all above 0.900, indicat- ing that it was feasible to predict the total nitrogen content of Korla fragrant pear leaves by these variables. [Conclusion]The model based on the first derivative spectrum at 1 564 nm was the best for predicting the total nitrogen content in leaves of Korla fragrant pear.