Quantitative Classification and Prediction of Starkrimson Pear Maturity by Near-Infrared Spectroscopy

Scientific evaluation of pear maturity is important for commercial reasons. Near-infrared spectroscopy is a non-destructive method that could be used for rapid assessment of pear maturity. The aim of this study was to develop a reasonable and effective method for the assessment of Starkrimson pear maturity using near-infrared technology. Partial least squares regression and five classification methods were used for analysis of the data. Among the indices used with the competitive adaptive reweighting-partial least squares regression method for quantitation, the visual ripeness index had the best modeling effect (Rp2: 0.87; root mean square error of prediction: 0.39). The classification model constructed with the visual ripeness index and post-ripeness score gave a cross-validation neural network model with the best classification effect and the highest accuracy (classification accuracy: 88.7%). The results showed that combination of quality indices with near-infrared spectroscopy was effective for rapidly evaluating the maturity of Starkrimson pears.