基于高光谱苹果品种叶片铁含量估测模型

Estimation Models for Foliar Fe Content of Malus domestica Borkh.cv. Using Hyperspectral Reflectance

【目的】在叶片水平上构建基于高光谱的苹果品种叶片铁素含量估测模型,为探寻实时、高效、无损的果树树体营养诊断提供技术途径。【方法】以苹果品种岩富10号为材料,测定岩富10号叶片光谱数据和铁素含量,采用光谱分析和相关分析法,筛选与叶片铁素含量相关性较强的光谱组合,利用偏最小二乘法构建苹果叶片铁素含量光谱估测模型。【结果】岩富10号苹果叶片一阶微分光谱与铁素含量的敏感波段为R'_(990)、R'_(1 113)、R'_(1 360)、R'_(1 408),相关系数最高为-0.698 9。对敏感波段两两进行加、减、乘、除运算,最优波段组合形式R'_(990)×R'_(1 048)与铁素含量相关系数为0.846 2。估测模型拟合度(R^2)最高为0.827 5。【结论】苹果叶片一阶微分光谱组合与铁素含量显著相关(P<0.05),光谱组合能够明显提高其相关性,偏最小二乘法与逐步回归建模相比估算模型的精度更佳,可以用于苹果叶片铁素含量的光谱估算。

【Objective】A hyperspectral estimation model of ferritin content in leaves of Yanfu 10（Malus domestica Borkh. Yanfu No. 10） was constructed at the leaf level in order to provide a technical approach for exploring real-time,efficient and lossless nutrient diagnosis of fruit trees.【Method】First,we determined the spectral reflectance and foliar Fe content of Malus domestica Borkh. cv. Yanfu No. 10,and then analyzed them by adopting the spectral methods and correlation analysis technology. Next,we selected the spectral reflectance combinations with strong correlation with foliar Fe content. Finally,the partial least squares（PLS）regression models were established based on sensitive wavebands. 【Result】The results showed that the four sensitive wavebands R＇_（990）、R＇_（1,113）、R＇_（1,360）、R＇_（1,408） were obtained,and the highest correlation coefficient between foliar Fe content and first-order differential spectrum was -0. 698,9. Through the approach of adding,subtracting,multiplying and dividing the sensitive wavebands with each other to find the optimal spectral reflectance variant,the highest correlation coefficient was 0. 846,2. The R^2 of training model reached to 0. 827,5.【Conclusion】The correlation between foliar first differential spectrum and Fe content was significant（P ﹤ 0.05）. Spectral combination could significantly improve the correlation. The accuracy of partial least square（PLS） method is better than that of stepwise regression modeling,so it can be used to estimate the foliar Fe content of Malus domestica Borkh. cv. Yanfu No. 10 using hyperspectrum.