摘要
橡胶树叶片氮含量作为橡胶树营养状态的一个重要量化指标,实时、快速且准确地检测橡胶树叶片氮含量对于橡胶树营养诊断、实施变量施肥具有重要意义。本文使用美国ASD光谱仪采集橡胶树叶片光谱,通过随机选点、氮含量梯度选点以及主成分网格选点,以反射光谱、吸光度光谱、反射率导数光谱以及吸光度导数作为光谱信息,与其对应的橡胶树叶片氮素含量建立PLSR模型。结果表明,采用主成分网格选点建立的反射率导数以及吸光度导数光谱模型具有较好的相关性和预测能力,相关系数分别达到0.9599、0.9492,RMSEP仅0.0978、0.1061,所建立的PLSR模型可以达到无损、快速地预测橡胶树叶片氮含量的目的。
Real-time,fast and accurate detection of nitrogen status of rubber（Hevea brasiliensis）leaves is critical for prediction of nutrition diagnosis and prescription of nitrogen topdressing.In this paper,nitrogen content have been determined by kjeldahl nitrogen determination method,and spectrum（350-2500nm）have been obtained by ASD spectrometer to construct partial least square models（PLSR）for assessing nitrogen content of rubber leaves.Twelve models from different methods of selecting sample（Random,Content grads and PCA-grid）and various spectrum data（reflectance, absorbance,first derivative reflectance and first derivative Absorbance）have been contrasted by prediction ability.Results revealed that model using first derivative reflectance and first derivative absorbance data by PCA-grid method produces an acceptable model precision and accuracy.The model coefficient of correlation（r）reaches 0.9599,0.9492 and root mean square errors of prediction（RMSEP）are 0.0978,0.1061 for validation,respectively.It means that the model have nondestructive and rapid ability to predict nitrogen content of rubber leaves.
出处
《现代科学仪器》
2010年第2期126-129,共4页
Modern Scientific Instruments
基金
国家科技支撑计划项目(编号2009BADA1B03)
海南省自然科学基金(编号:30820)
公益性行业(农业)科研专项经费(编号:nyhyzx07-033-1)
中国热带农业科学院橡胶研究所基本科研业务费专项(编号:YWFZX09-01(N))
关键词
橡胶树叶片
氮素
光谱
模型优化
最小偏二乘回归
Leaves of Hevra brasiliensis
Nitrogen
Spectrum
Optimization model
Partial least squares regression
