| Currently the quality of fruit on tree and post-harvest is usually measured by traditional methods.These methods are destructive and time wasting.Meanwhile the near infrared analytical technology has the advantage of fast analysis,green,no damage and so on.This technology could realize evaluating the fruit in a fast and non-destructive way.(1)Soluble solids content(SSC)in grape was measured in a non-destructive way.The information of single and group grapes was collected in order to guide managing the orchard,setting the storage condition and to satisfy the different tastes of consumers.The near infrared spectroscopy(NIRS)of grapes were obtained by handheld NIRS MicroNIR 1700 between 950 and 1650 nm.PLS was used to build the SSC prediction model in grape.In order to decrease the redundant uninformative variables and increase the prediction accuracy and stability of model,UVE and RF coupled with PLS were used to select the important variables associated with the SSC in grape respectively.The result showed that the RF-PLS SSC prediction model is better than the full spectra PLS model and the UVE-PLS model.The RC,RCV and RP of RF-PLS model are 0.9801,0.9661 and 0.9646 respectively.The RMSEC,RMSECV and RMSEP of RF-PLS model are 0.6382 °Brix,0.8299 °Brix and 0.8688 °Brix respectively.The study show that it is completely feasible to predict the SSC in grape using handheld NIRS instrument after wavelength selection.The model has high prediction accuracy.(2)SSC,dry matter content(DMC),total phenol content(TPC)and total flavonoid content(TFC)are important indexes to evaluate the nutritional quality of mulberry fruit.Fast,real-time and non-destructive method was built to measure the nutritional content in mulberry with portable NIRS instrument.Firstly,the NIRS of mulberry was obtained by portable NIRS instrument MicroNIR 1700.After the spectra was preprocessed,PLS model was built to predict the SSC,DMC,TPC and TFC in mulberry.In order the increase the accuracy of PLS model,UVE,CARS and RF were used to select the better wavelength variables.The result show that CARS-PLS model has the least wavelength variables and better prediction accuracy.The RC of CARS-PLS model of SSC,DMC,TPC and TFC are 0.9807,0.9697,0.9491 and 0.9697 respectively.The RMSEC are 0.7035 °Brix,0.6266%,0.3823 mg/g(gallic acid)and 0.2666 mg/g(rutin)respectively.The RP are 0.9514,0.9430,0.8667 and 0.8771 respectively.The RMSEP are 1.2100 °Brix,0.9178%,0.6352 mg/g(gallic acid)and 0.6939 mg/g(rutin)respectively.The study show that handheld NIRS instrument MicroNIR 1700 combined with chemometric method could be applied in testing the SSC,DMC,TPC and TFC in mulberry on-site in a fast and non-destructive way.(3)The SSC prediction model in apple was transferred between the two same type instruments.K/S method was used to study the difference of RMSEP with different standard sample number.The best number was 18.The NIRS from slave instrument was corrected by direct standardization(DS)algorithm.25 wavelength variables were selected by CARS to build PLS model.Finally,the SSC prediction model in apple is successfully transferred from master instrument to slave instrument.The Rp and RMSEP are 0.8603,0.7914 °Brix respectively.DS algorithm combined with CARS could be effectively applied in the study of model transfer. |
