论文部分内容阅读
研究利用近红外光谱(near-infrared,NIR)和化学计量学方法估测小麦(Triticum aestivum)新鲜叶片和粉末状干叶中全氮含量的可行性,并建立小麦叶片氮含量估测模型,以期为小麦氮素营养的精确管理提供理论依据。以3个小麦田间试验观测资料为基础,分别运用偏最小二乘法(partial least squares,PLS)、反向传播神经网络(back-propagation neural network,BPNN)和小波神经网络(wavelet neural network,WNN),建立小麦叶片氮含量的鲜叶和粉末状干叶近红外光谱估测模型,用随机选择的样品集对所建模型进行测试和检验。结果显示,利用PLS、BPNN和WNN3种方法构建的近红外光谱模型均能准确地估测小麦叶片氮含量,其中基于BPNN和WNN的模型优于基于PLS的模型,且以基于WNN的模型表现最好。对模型进行检验的结果显示,粉末状干叶模型的预测均方根误差(RMSEP)分别为0.147、0.101和0.094,鲜叶模型的RMSEP分别为0.216、0.175和0.169,模型的相关系数均在0.84以上。因此,利用近红外光谱估算小麦叶片氮素营养精确可行,对其他作物的氮素营养估测提供了借鉴和参考。
The feasibility of using near-infrared (NIR) and chemometric methods to estimate total nitrogen in fresh leaves and powdery dry leaves of wheat (Triticum aestivum) was studied. An estimation model of nitrogen content in wheat leaves was established, The precise management of nitrogen nutrition provides a theoretical basis. Based on the experimental data of three wheat fields, partial least squares (PLS), back-propagation neural network (BPNN) and wavelet neural network (WNN) , The leaf nitrogen content of wheat leaves and powder dry leaf near infrared spectroscopy estimation model was established, and the model was tested and tested by randomly selected samples. The results showed that NIR spectra of wheat leaves could be accurately estimated by using near-infrared spectral models constructed by PLS, BPNN and WNN methods. BPNN and WNN models were superior to PLS-based models and WNN-based models were the most effective it is good. The results of the model test showed that the root mean square error of prediction (RMSEP) of powdery dry leaf model was 0.147, 0.101 and 0.094 respectively, RMSEP of fresh leaf model was 0.216, 0.175 and 0.169 respectively, and the correlation coefficient of the model was 0.84 the above. Therefore, it is feasible and feasible to estimate nitrogen nutrition of wheat leaves by using near-infrared spectroscopy, and provide reference for estimating nitrogen nutrition of other crops.