摘要： Near-infrared spectroscopy (NIRS) is a rapid analysis method that is widely used for quantitative determination of the major constituents in many food products. NIRS was applied in conjunction with a chemometric algorithm, namely the partial least squares regression (PLSR), to develop the optimum model for predicting the qualities of osmotically dehydrated papaya (ODP). Two hundred ODP samples were collected from commercial products and from different laboratory ODP processes with varying sucrose concentrations (35oBrix, 45oBirx, 55oBrix and 65oBrix) at 40°C for 6 hr and drying times at 60°C for 2 hr, 4 hr, 6 hr, 8 hr, 10 hr and 12 hr. All samples were divided into a calibration set (n = 140) and a validation set (n = 60) before quality determination and NIRS analysis. Samples were scanned over the NIR spectral range of 800–2400 nm in reflectance mode and their spectra were pretreated using the second derivative method. Suitable predictive models were developed by applying full wavelength PLSR and two wavelength interval selection methods, named the moving window partial least squares regression (MWPLSR) and the searching combination moving window partial least squares regression (SCMWPLSR). The results showed that SCMWPLSR provided better performance than PLSR and MWPLSR. The root mean square error of prediction values of water activity, moisture content, total soluble solids and the sucrose, glucose and fructose contents from SCMWPLSR were 0.014, 0.69% (dry basis), 0.58oBrix, 14.44 g/100 g of sample, 6.72 g/100 g of sample and 4.89 g/100 g of sample, respectively, with correlation coefficients in the range 0.981–0.994.