基于OSC–WT–PLS近红外光谱法测定烟草中芸香苷

采用正交信号校正(OSC)结合小波变换(WT)对烟草光谱进行光谱预处理,将预处理后的烟草光谱结合偏最小二乘法(PLS)建立了烟草光谱对芸香苷的预测模型。利用OSC滤除光谱中与芸香苷含量无关的光谱信息,确定OSC提取的最佳主成分数为7,再选择WT中的最佳小波基函数bior1.1对OSC预处理后的光谱进行压缩及进一步滤噪,然后进行PLS建模,OSC–WT–PLS所建模型决定系数r~2=0.874,校正标准偏差RMSEC=0.85,预测均方根误差RMSEP=0.743,交互验证系数Q_(ext)~2=0.887。结果表明,用OSC–WT–PLS可滤除光谱信息中与待测样品含量无关的信息、减少光谱数据量,降低建立模型的复杂度、提高建模速度及模型的预测能力、准确度。

Near Infrared Spectroscopy Determination of Rutin Content in Tobacco Based on Orthogonal Signal Correction Comnbined with Wavelet Transform and Partial Least Square Method

Tobacco spectrum was preprocessed by using orthogonal signal correction (OSC) with combinations wavelet transform (WT),then a model was established by this tobacco spectrum combined with partial least squares (PLS) to forecast rutin in tobacco. Spectral information unrelated to the content of rutin was filtered by OSC,the optimum number of component by OSC extracted determined was 7,the best wavelet basis function bior1.1 in WT was chosen to compress and further filter the noise in OSC preprocessed spectrum,then a model was established by using PLS. The OSC–WT –PLS model had a determination coefficient (r2) of 0.874,Root Mean Standard Error of Calibration(RMSEC) was 0.85, Root Mean Standard Error of prediction(RMSEP) was 0.743,and interaction coefficient of validation(Qext2) was 0.887. It shows by the result that OSC–WT–PLS can filter the information unrelated to sample content in spectral data to reducing spectral data and the complexity of models building, it can improve the speed of models building,theability and accuracy prediction.