| In order to realize the use of hyperspectral technology to determine whether the rapeseed oil combination type is moss picking,and to fill the gap of hyperspectral technology in rapeseed oleic acid content monitoring research.In this study,the HH2 Sync handheld spectrometer was used to collect the spectral data of rape canopy at seedling stage,and the field Spec 3 ground object spectrometer was used to collect the spectral data of rape leaves at moss,flowering and kerneling stages.The Agilent GC-MS 7980B gas chromatography was used to analyze the oleic acid content of rapeseed.Build based on rapeseed bolting stage,flowering period,pod stage original spectrum characteristic wavelengths of discriminant model,support vector machine(SVM)based on spectral parameters in the whole stages of rape rapeseed oleic acid content of the binomial model,based on the first-order spectrum characteristic wavelengths in the whole stages of rapeseed oleic acid content(MLSR)with multiple linear stepwise regression and partial least squares regression(PLSR)estimate model,and by using independent sample T test to verify this model precision.The main research results are as follows:(1)There were 36 samples in the training set of the SVM discrimination model for rape in the moss-picking period,and 26 of them were successful in prediction,with an overall accuracy of 77.2%.There were 18 samples in the validation set,and 13 of them were successful in prediction,with an overall accuracy of 77.2%.There were 36 samples in the training set of the SVM discrimination model for the flowering stage of rape,and 31 of them were predicted successfully,with an overall accuracy of 86.1%.There were 18 samples in the validation set,and 14 of them were predicted successfully,with an overall accuracy of77.8%.There were 36 samples in the training set of the SVM discrimination model for the seed fruit stage of rape,34 of which were predicted successfully,and the overall accuracy reached 94.4%.There were 18 samples in the validation set,and 13 of which were predicted successfully,and the overall accuracy reached 77.2%.(2)The PLSR model was the best model for estimating the oleic acid content of rapeseed at the seedling stage,and its training set R~2=0.470,RMSE=0.770,modeling set R~2=0.465,RMSE=1.222.(3)The original spectral reflectance of rapeseed at moss-picking stage,flowering stage and fruit’stage would change,and the correlation between the original spectral data and the oleic acid content of rapeseed in three growth stages would change.(4)In the estimation model of oleic acid content of all rapeseed in the licking stage of rapeseed,the training set R~2=0.696 and RMSE=0.583,and the test set R~2=0.649 and RMSE=1.056 of PLSR model were the best estimation model for oleic acid content of first-level branch rapeseed without licking.Independent sample t-test was conducted on the model.P=0.328>0.05,the model is reasonable.The training set R~2=0.686 and RMSE=0.903,the test set R~2=0.722 and RMSE=0.702 of PLSR model were the best model to estimate the oleic acid content of rapeseed in the first-level branch of rapeseed.Independent sample T test was conducted on the model,and P=0.540>0.05,indicating that the model was reasonable.(5)In the estimation models for oleic acid content of all rapeseed at flowering stage,the training set R~2=0.648 and RMSE=0.68 of PLSR model,and the test set R~2=0.661 and RMSE=0.852 were the best model to estimate the oleic acid content of rapeseed in the first-level branch of rapeseed.Independent sample T test was conducted on the model.P=0.839>0.05,the model is reasonable.The training set R~2=0.773 and RMSE=0.874,and the test set R~2=0.752 and RMSE=0.467 of MLSR model were the best model to estimate the oleic acid content of rapeseed in the first-level branch of rapeseed.The independent sample T test of the model showed that P=0.858>0.05,which indicated that the model was reasonable.(6)In the estimation model of oleic acid content of all rapeseed at the kerneling stage,the training set R~2=0.820 and RMSE=0.447 of PLSR model,and the test set R~2=0.645 and RMSE=0.994 were the best estimation model of oleic acid content of first-level branch rapeseed without plinsed coating.Independent sample T test was conducted on the model.P=0.238>0.05,the model is reasonable.The MLSR training set R~2=0.493 and RMSE=1.148,and the test set R~2=0.267 and RMSE=1.132 were the best model to estimate the oleic acid content of rapeseed in the first-level branch of rapeseed.Independent sample T test was conducted on the model,and P=0.231>0.05,indicating that the model was reasonable. |
PDF Full Text Request