Detection Of Grape Berries Based On Visible/Near Infrared Spectroscopy And Development Of Portable Equipment

The grape berries should be destrueted when detected by traditional detective methods,which is time-consuming and labor-intensive.The non-destructive detection for grape quality during its maturity and storage is now catching much attention.In this work,supported by National Key Technology Research and Development Program of the Ministry of Science and Technology of China(2015BAD19B03),the visible/near infrared spectroscopy(Vis/NIR)in 400-1100 nm were applied to study the CIE L*a*b*,soluble solid contents(SSC),total phenolic compounds(TP)of 'Manicure Finger' and 'Ugni Blanc' grape berries during maturity and storage.Besides,a portable equipment for detection of grape berries was developed.The research contents and results were as follows:1.The quality detection of grape berries during its maturity process based on visible/near infrared spectroscopy.The 400-1100 nm Vis/NIR spectroscopy of 'Manicure Finger' and 'Ugni Blanc' during maturity process were acquired.Based on the full band spectra and wavelengths selected by successive projections algorithm,prediction models of partial least squares and least squares-support vector machine for CIE L*a*b*,soluble solid contents(SSC)and total phenolic compunds(TP)were built,and the discrimanitive models for five developmental stages(green,pre-veraison,veraison,post-veraison and ripe)were also established.The results showed that the RC2,RMSEC,Rp2,RMSEP,RPD of LS-SVM were in the range of 0.582-0.992,0.531-0.929,0.115-4.711,0.150-5.161,1.473-3.698,respectively.A total of 72 wavelengths were selected by successive projections algorithm(SPA)to build support vector machine discrimination analysis(SVM-DA)models with the total accuracy of 95.0%for 'Manicure Finger'.The fiill-band SVM-DA was built for 'Ugni Blanc' with total accuracy of 98.3%.2.The quality detection of grape berries during its storage process based on visible/near infrared spectroscopy.The 400-1100 nm Vis/NIR spectroscopy of 'Manicure Finger' and 'Ugni Blanc' under the storage condition of 4oC and 85-95%relative humidity for 25 days were acquired.Based on the full band spectra and wavelengths selected by successive projections algorithm,prediction models of partial least squares and least squares-support vector machine for CIE L*a*b*,soluble solid contents(SSC)and total phenolic compunds(TP)were built,and the discrimanitive models for different storage times(0,5,10,15,20,25 d)were also established.The results showed that the Rc2,RMSEC,Rp2,RMSEP,RPD of LS-SVM models were in the range of 0.741-0.898,0.069-5.893,0.712-0.874,0.073-6.190,0.727-2.883,respectively.In addition,the total accuracies of 100.0%were obtained for two varieties by the full-band spectra combined with SVM-DA.3.The design and development of a portable detection euqipment for grape quality.The holographic concave diffractive grating combined with charge-coupled component was chosen as core part to obtain spectra date of grape samples in this system,and halogen light source combined with optical fiber were chosen to provide stable Vis/NIR light source.In order to adapt diverse varieties of grapes with different sample sizes,an adjustable sample cuvette with internal and external threads was designed.The software which can easily perform model updating was written by C#based on windows operating system.A total of 110 samples of each variety were used as external samples to exam the performances of device and established models.The results showed that the RMSE of L*,a*,b*,SSC and TP for 'Manicure Finger' were 5.21,3.15,3.08,1.39°Brix and 0.24 g/kg,respectively,and 3.05,0.78,3.05,1.56°Brix,0.22 g/kg for 'Ugni Blanc',respectively.The total accuracies of developmental stages and storage time were 94.0%and 100.0%for'Manicure Finger',respectively,and 96.0%and 96.0%for 'Ugni Blanc',respectively.