基于OSC的土壤全氮近红外光谱测定  

Near Infrared Spectroscopy Determination of Soil Total Nitrogen Based on OSC

作　　者：杨超[1] 邢艳秋[1] 李俊明[1] 

机构地区：[1]东北林业大学森林作业与环境研究中心,哈尔滨150040

出　　处：《森林工程》2013年第3期25-28,共4页Forest Engineering

基　　金：国家自然科学基金资助项目(4087119)

摘　　要：采集农田、林地和盐碱地不同类型的土壤样本,采用偏最小二乘法结合OSC方法建立土壤有机质反演模型,运用交叉验证和外部验证相结合的评价方法进行比较分析。结果显示:采用平滑+MSC+OSC方法对光谱进行预处理,可以提高预测模型的精度。OSC因子个数和PLS主因子个数分别为6和4时,交叉验证决定系数R2为0.990 1,均方根误差为0.297 5,外部验证决定系数R2为0.926 1,均方根误差为0.283 6,模型达到最优。表明对光谱进行OSC预处理后建模是可行的,OSC降低与浓度阵无关的光谱信号,并且减少建立模型的主因子个数,进一步提高模型的精度和稳定性。Abstract： To explore the impact of OSC on the inversion model of soil total nitrogen in near-infrared spectra, this study used dif- ferent soil samples which were collected from farmland, woodland, and saline to establish the soil organic matters＇ inversion model by using partial least squares method combined with OSC. In order to evaluate the result we used cross-validation and external validation. It＇s concluded that ： the smooth ＋ MSC ＋ OSC method of spectral pre-processing can improve the prediction accuracy of the model. When the OSC number of factors and the number of PLS main factor were 6 and 4, the R2 and RMSE of the cross-validation coefficient were 0. 9901 and 0. 2975, respectively, and the R2 and RMSE of the external validation coefficient were 0. 9261 and 0. 2836, re- spectively. The model achieved the most superior. It means that the algorithm of OSC was proved to be the most effective and applica- ble, which had the ability of filtering spectroscopy signal, reduced the number of principal component for building model, improved the ability of prediction and enforced the robustness of calibration model.

关 键 词：近红外光谱 偏最小二乘法 土壤 全氮 正交信号校正 

分 类 号：S762[农业科学—森林保护学]