Calibration transfer of near infrared spectroscopy based on DS algorithm

Abstract-In order to share analysis model of near infrared spectroscopy between two Fourier transform near infrared spectrophotometers, navel orange was used as experimental materials. The samples were divided into calibrating samples set, standardization samples set and testing samples set by Kennard-stone algorithm, and direct standardization (DS) algorithm was applied to the calibration transfer between two machines, the model established by master instrument was used to predicted the content of vitamin C of the testing sample spectrums; this study also relates to the dependence of calibration transfer on the number of standardization samples. The results showed that the spectrums in the slave instrument were applicable to the model based on the master instrument just after calibration transfer. When the standardization samples reached 70, the results of calibration transfer was best, the correlation coefficient(R) was 0.887, while standard error of prediction (RMSEP) was 3.01mg/100g. The DS algorithm can be used to successfully transfer the calibration model for vitamin C of navel orange developed on a FT-NIR spectrophotometer to another.