In the last decade, visible and near-infrared spectroscopy (including three modes reflectance, interactance and transmission) has increasingly been utilized in food industry. In this research, transmission mode was used to acquire spectra of kiwifruits, apples and oranges in the range of 400– 1,000 nm, with an emphasis on the effect of fruit orientation on the spectra pattern. For soluble solids content and acidity (pH) to be predicted, various calibration models were developed, based on different combinations of preprocessing techniques. The best model for each characteristic was obtained by standard normal variate transformation in combination with median filter and first derivative. The correlation coefficients for soluble solids content and pH were 0.93 and 0.943, and the root mean square errors of prediction for them were 0.259°Brix and 0.076, respectively. The results indicate that the occurrence of peaks is uniquely the inherent of the fruit, independent of orientation. The transmission calibration model for the estimation of fruit constituents could be improved by using proper preprocessing techniques.

Keywords