Rapid Detection Of Quality Index In Cabernet Sauvignon Red Wine Based On Near-infrared Spectroscopy

The study aimed to establish a method for rapid nondestructive determination by near-infrared spectroscopy for the content of quality indexes in Cabernet Sauvignon dry red wine.The spectroscopic date of diffused reflection was collected with near-infrared spectrograph,the chemical determination was performed on total sugar,reducing sugar,total acid,volatile acid,pH,alcohol,total extract substances,total SO₂,free SO₂,protein,total anthocyanin,tannins,total flavonoid,the total phenol delphinidin-3-gliucoside and cyanidin-3-glucoside of samples by standard convention-al method.The spectroscopic date and date from conventional determination were established model through regression calculation using partial least squares method.The main contents and results were as folloes:?1?Screening of near infrared spectral pretreatment method.The near infrared spectrums were processed using the moving average smoothing?MAS?,savitzky golag smoothing?S-G-S?,normalize,savitzky-golay first dervative?S-G-D1st?,savitzky-golay second dervative?S-G-D2nd?,standard normal variate?SNV?,savitzky-golay first dervative with standard normal variate?S-G-D1st+SNV?,savitzky-golay second dervative with standard normal variate S-G-D2nd+SNV.Determine the optimal method by comparing the eight pretreatment methods.The result showed that S-G-D1st+SNV was the best pretreatment for total sugar,total SO₂,free SO₂,protein,total anthocyanin,tannins,total flavonoid content.The best pretreatment result of total acid,reducing sugar,pH,total extract substances,total phenol were S-G-D2nd+SNV.The pretreatment of MAS was the best method for volatile acid and alcohol content.The pretreatment of S-G-D2nd and S-G-S were the best mothod for delphinidin-3-gliucoside and cyanidin-3-glucoside respectively.?2?Identified the selected wavenumber ranges in near infrared spectrum of key chemical composition.Comparing the wavenumber ranges of 950 nm-1300 nm,1300 nm-1460 nm,1460 nm-1650 nm and 950 nm-1650 nm on calibrating the near infrared models quality.Then identify the wavenumber range of the chemical composition by comparing the NIR spectrum of the samples.We finally identify the wavenumber range of chemical composition,which are reducing sugar,total extract substances and protein 950nm-1300 nm,alcohol acid 1300nm-1460 nm,total acid 1460 nm-1650 nm,total sugar,volatile,pH,total SO₂,free SO₂,total anthocyanin,tannins,total flavonoid,total phenol,delphinidin-3-gliucoside and cyanidin-3-glucoside 950 nm-1650 nm,.?3?Established optimal NIR calibration and validation models of the quality index of the sample.The models had a excellent prediction effect of volatile acid,pHvalue,alcohol,total extract,total phenol?PRD?10?,the coefficient of determination of calibration set?R_c²?and root mean square errors of calibration?RMSEC?were 99.83%,99.92%,99.99%,99.91%,99.44%and 0.0019,0.0096,0.0042,0.1548,0.0206,the coefficient of determination of validation set?R_v²?and root mean square errors of prediction?RMSEP?were 99.24%,99.87%,99.73%,99.90%,99.15%,0.0021,0.0105,0.0055,0.1594,0.0428,respectively;The models had a good prediction effect of total sugar,reducing sugar and total acid?5?PRD?10?,R_c² and RMSEC were 97.85%,98.31%,99.33%,0.0447,0.0386,0.0461,R_v² and RMSEP were97.20%,97.78%,97.81%,0.0521,0.0415,0.0499,respectively;The models of total SO₂,free SO_2,protein,total anthocyanin,tannin and total flavonoid were generally effective?2.5?PRD?5?,R_c² and RMSEC were 94.24%?86.99%,95.12%,88.19%,95.08%,96.71%,1.2850,0.8565,1.1790,0.4280,0.2222,2.8110,R_v² and RMSEP were 94.24%,86.99%,95.12%,87.20%,94.25%,95.23%,1.2893,0.9472,1.2011,0.4450,0.2415,2.9950,respectively;The models of anthocyanin-3-glucoside and anthocyanin-3-glucoside were poor?PRD?2.5?,R_c² and RMSEC were 80.55%,74.48%,0.2718,0.0317,R_v² and RMSEP were70.23%,69.98%,0.3021,0.0425,respectively.