The Study To Calibration Transfer Of Near-Infrared Spectroscopy Of Navel Orange

Modern near-infrared spectroscopy, as a rapid and non-destructive analytical technique, is increasingly conducted in many fields. Navel orange is a nutrient and popular fruit, many researchers had studied on orange internal quality quantitative analysis method using near infrared spectroscopy, however, in application the current near-infrared instrument is influenced by many factors, such as the optical path design in different instruments, the instrument performance, assembly errors, the environment of testing and so on, so the error exists between the measurement of the spectral data, thus the result can not be directly applied in another instrument. It is a high labour intensive and expensive operation to rebuild the model, therefore, it is necessary for us to solve the problem near-infrared model sharing using model transfer method.In this paper, navel orange was used as experimental materials. A rapid and nondestructive testing method to measure the sugar-acid ratio of orange was established based on near infrared spectroscopy, and the calibration transfer of near infrared spectroscopy on navel orange's internal quality based on the slope/bias method and the direct correction (DS) algorithm was conducted, the model based on total soluble sugar, total acid, vitamin C and soluble solids. The results are as follows:1. The near infrared spectroscopy model of navel orange sugar-acid ratio was established in VECTOR33N FT-NIR. The best pretreatment method was multiplication scattering correction method (MSC), the optimal spectral range was 7501.7～5449.8cm-1 cross validation correlation coefficient R was 0.809, root mean square error of cross validation (RMSECV) was 1.76, the correlation coefficient R of the prediction samples set was 0.846, root mean square error of prediction (RMSEP) is 2.13.2. Navel orange samples were divided into calibration samples set, prediction samples set and standardized samples set using Kennard-Stone algorithm. VECTOR22N type Fourier spectrometer as master instrument, VECTOR33N type Fourier spectrometer as slave instrument, the model transfer researched based on the different numbers of standardized samples. When the navel orange model of total soluble sugar was corrected by the slope/bias method, the RMSEP of prediction samples set for the slave instrument was reduced to 0.756%, the correlation coefficient R and RMSEP were 0.902 and 0.448% after model transfer by the DS algorithm; When the navel orange model of total acid was corrected by the slope/bias method, the RMSEP of prediction samples set was decreased to 0.0787%, the correlation coefficient R and RMSEP were 0.863 and 0.0451% after model transfer by the DS algorithm; when the navel orange model of Vitamin C corrected by the slope/bias method, the RMSEP of prediction samples set was reduced to 4.17mg/100g, the correlation coefficient R and RMSEP were 0.895 and 3.01mg/100g after model transfer by the DS algorithm; when the navel orange model of soluble solids corrected by thec method, the RMSEP of prediction samples set was reduced to 0.728%, the correlation coefficient R and RMSEP were 0.828 and 0.549% after model transfer by the DS algorithm.The results show that:under the experimental conditions, near infrared spectroscopy can be used to measure the sugar-acid ratio of navel orange rapidly. After model transferring, the number of the standardized samples was less impact on the slope/bias method, while the DS algorithm was great effect. Slope/bias method of calibration transfer had played an important role, but the capacity of the model calibration was limited. When the model was corrected by the DS algorithm, the results were obvious, so internal quality near-infrared analysis models of Navel orange could be shared between two Fourier spectrometers successfully.