Abstract:The sugar content of kiwifruits is an important index to distinguish its maturity.A model of predicting the sugar content of different maturity of kiwifruits was established and optimized.The UV/Vis spectroscopy（200 nm-1 000 nm）acquisition system was used to collect reflectance spectra of different maturity of'Guichang'kiwifruits.The preprocessing effect of first-order derivative（D1st），multi-scatter calibration（MSC）and standard normal variation（SNV）on the original spectra was compared.The competitive adaptive reweighted sampling（CARS）was used to select characteristic spectra from preprocessed full spectra.The error back propagation（BP）network models were established based on full spectra and characteristic spectra to predict sugar content of kiwifruits，respectively.The results showed that the preprocessing effect of SNV was the best.29 characteristic wavelengths were extracted by CARS from 1 024 full wavelengths，and the working efficiency of prediction model was obviously improved.SNV-CARS-BP model had the best prediction ability（RP2=0.901，RMSEP=0.643%，RPD=3.217）.Therefore，it's possible to determine the sugar content of kiwifruits by UV/Vis spectroscopy combined with BP network and SNV-CARS-BP model was the best.