烟叶化学成分协调性的近红外光谱检测模型研究.PDFVIP

中国农学通报 2015,31(26):90-95 Chinese Agricultural Science Bulletin 烟叶化学成分协调性的近红外光谱检测模型研究 方世诚，唐延林，方智文，魏晓楠，刘 芸 （贵州大学理学院，贵阳550025） 摘 要：对烟粉全部近红外光谱数据采取不同的预处理方法来探究烟叶化学成分协调性（施木克值、糖 碱比、氮碱比），基于有效波长光谱数据建立相应的近红外光谱的检测模型，并利用偏最小二乘法(PLS) 通过训练集的交叉验证建立回归模型。通过全部波长数据中值滤波平滑处理后建立施木克值的回归模 型，预测集r=0.9861、RMSE=0.0548 ；通过全部波长数据卷积平滑二阶求导处理后建立糖碱比的回归模 型，预测集r=0.9498、RMSE=0.9095 ；在对光谱数据处理较优的Norris 一阶导数数组基础上选取30 个有 效波长建立氮碱比的回归模型，预测集r=0.9202、RMSE=0.6947。结果表明：利用近红外光谱可以较好 地预测烤烟施木克值、糖碱比和氮碱比。 关键词：近红外光谱；糖碱比；氮碱比；施木克值 中图分类号：S127 文献标志码：A 论文编号：cas NIRSpectralDetectionModelontheCoordinationofTobaccoChemicalCompositions FangShicheng,TangYanlin,FangZhiwen,WeiXiaonan,LiuYun (CollegeofScience,GuizhouUniversity,Guiyang550025) Abstract:Inordertoexplorethecoordinationoftobaccochemicalcompositions (shmuckvalue,ratioofsugar andnicotine,ratioofnitrogenandnicotine),thedataofnearinfraredspectrum were pretreatedbydifferent methods. The NIR spectral detection models were established based on the effective wavelengths. The regressionmodelwascheckedbyPLStocrosstrainingset.Theregressionmodelofshmuckvaluewas establishedaftermedianfiltersmoothingoffullspectrawithcoefficientr=0.9861,RMSE=0.0548inprediction set.Theregressionmodeloftheratioofsugarandnicotinewasestablishedafterconvolutionsmoothingand 2nd derivative of full spectra with coefficient r=0.9498, RMSE=0.9095 in prediction set. 30 effective wavelengthdatainprocessingarraywereselectedtosetuptheregressionmodeloftheratioofnitrogenand nicotinebasedonthebetterNorrisfirstderivativearray,whichhadtheregressioncoefficientr=0.9202,RMSE= 0.6947.TheresultsindicatedthattheNIRspectracould better predicttheshmuckvalue,ratioofsugarand nicotineandratioofnitrogenandnicotineoftobacco. Keywords:NIRsp