Measurement of soluble solids content of three fruit species using universal near infrared spectroscopy models

Based on the similar physical and chemical properties of Fuji apples, Okubo peaches and Hosui pears, universal near infrared spectroscopy models were developed to measure the fruits' soluble solids content (SSC). The second derivative eliminated the baseline drift caused by different integration times. The effective wavebands of the three species were selected by moving window partial leastsquares (MWPLS) regression, and the effective wavebands were combined to develop a universal partial least-squares regression (PLS) model. Then, the effective wavelengths (876 nm, 890 nm and 900 nm) were identified by the successive projections algorithm (SPA) from the combined region (840-920 nm) to develop a multiple linear regression (MLR) model. The coefficient of determination of calibration (R-2) and validation (r(2)) were 0.96 and 0.97, and the root mean square error of calibration (RMSEC) and validation (RMSEP) were 0.47 degrees Brix and 0.45 degrees Brix, respectively, for the PLS model and 0.96 degrees Brix, 0.96 degrees Brix and 0.49 degrees Brix and 0.46 degrees Brix, respectively, for the MLR model. The universal models were greatly simplified by variable selection and proved to be practical for the prediction of SSC in the three fruit species.