The Potato Internal Quality Testing Based On NIR Spec-troscopy Analysis Technology

Starch, reducing sugar, dry matter and protein content of potato flour are important factors in determining its nutritional value. As a new generation of optoelectronic detectioncc, near-infrared spectroscopy has become a hot topic with the advantage of fast, environmentally friendly, and accurate.In this paper, potato flour was chose as the research object. Near-infrared spectroscopy system (1200-2400nm) was investigated for non-destructive determination of starch, reducing sugar, dry matter and protein content in potato flour, combined with chemometric methods. These detection algorithms offer theoretical basis for the on-line nondestructive detection system. The main research contents are as follows:(1) Optimal wavelengths were selected by principal component regression based on multiplicative scatter correction spectrum. Characteristic wavelengths (Starch-1249nm、1316nm、 1437nm,1674nm、1781nm、1918nm、1989nm、2066nm、2215nm; reducing sugar-1320nm、1433nm、 1651nm、1788nm、1855nm、1909nm、1980nm、204nm、2143nm) were selected to build the principal component regression models. The correlation coefficient of calibration and validation models about starch and reducing sugar principal component regression models are 0.9582,0.9201 and 0.9693,0.9112, which is better than full wavelengths models and MLR models.(2) Six characteristic wavelengths (1429nm、1595nm、1901nm、1949nm、2036nm、2121nm) were selected by regression coefficient of partial least squares. PCR and PLSR models were build to predict the dry matter in potato flour. Compared to full wavelengths models and PCR models, PLSR models with optimal wavelengths has an excellent ability to predict the dry matter in potato flour. The correlation coefficient and root-mean-square error of calibration and validation models are 0.9778,0.2575 and 0.9804,0.3315, respectively.(3) Optimal wavelengths (PLSR-1459nm、1687nm、1787nm、1873nm、912nm、2000nm、 2074nm、2200nm；PCR-1259nm、1453nm、1683nm、1783nm、1874nm、1919nm、1999nm、2061nm、 2212nm) were selected by principal component regression and partial least squares regression. According to the absorption of near-infrared spectral region distribution of N-H groups in protein,5100 ～4360 cm-1 (1960～2293nm) was selected as the wavelengths interval.Compared to full wavelengths models, wavelengths interval models and PCR models, PLSR models with optimal wavelengths has an excellent ability to predict the protein content in potato flour. The correlation coefficient and root-mean-square error of calibration and validation models are0.9693,0.2937 and 0.9779,0.3304, respectively.