应用近红外光谱法估测小麦叶片糖氮比

糖氮比能够反映作物碳氮代谢的协调程度，及时、准确地监测糖氮比对于作物氮素营养诊断和调控具有重要意义.本研究以不同年份、品种、施氮水平的小麦大田试验为基础，获取鲜叶和粉末状干叶近红外(NIR)光谱及糖氮比信息，分别运用偏最小二乘法(partial least squares, PLS)、BP神经网络(back propagation neural network, BPNN)和小波神经网络(wavelet neural network, WNN)3种方法建立了小麦叶片糖氮比预测模型，并利用随机选择的样品集对所建模型进行测试和检验.结果表明： 小麦鲜叶光谱模型预测性能不佳；而干叶片预测模型表现了较好的准确性，在1655~2378 nm谱区范围内基于3种方法构建的干叶粉末糖氮比估算模型，其预测均方根误差均低于0.3%，决定系数均高于0.9.比较而言，WNN法表现最佳.总体显示，近红外光谱法可以准确预测小麦叶片糖氮比状况，为科学诊断糖氮比提供了理论基础和技术途径.

Estimation of sugar to nitrogen ratio in wheat leaves with near infrared spectrometry.

The soluble sugar to nitrogen ratio reflects the coordination degree of carbon (C) and nitrogen (N) metabolism. Precise and real time monitoring of soluble sugar to nitrogen ratio is of significant importance for nitrogen diagnosis and management regulation in wheat production. In this study, time course near infrared spectroscopy and soluble sugar to nitrogen ratio of fresh and dry leaves were obtained under different field experiments with varied years and cultivar and N rates. The methods of partial least squares (PLS), back propagation neural network (BPNN) and wavelet neural network (WNN) were used to develop the calibration models with the preprocessed spectra, respectively, and the dataset selected randomly was used to evaluate the constructed models. The results showed that the performance of the models for fresh leaves was not satisfied, but good for dry leaves with the root mean square errors of prediction (RMSEP) by PLS, BPNN and WNN models based on 1655-2378 nm less than 0.3% and with the coefficients of determination (R²) over than 0.9, respectively. In comparison, the model based on WNN was the best one. All these indicated that near infrared spectrometry could be applied to estimating the soluble sugar to nitrogen ratio in plant. The results provided the theoretical basis and technological approach for diagnosing crop C/N.