Detection Method Of Phenolic Compounds Content Of Wine Grape Using Hyperspectral Imaging Technology During Ripening

The main phenolic compounds of wine grape are anthocyanin and total phenols, which are the key factors of wine grape quality and impacts the quality of red wine significantly. Nowadays, the phenolic compounds content of wine grape is measured by chemical detection methods, which are destructive, expensive, time consuming, and contaminative to the environment. Therefore, the traditional chemical measurements can not meet the demands of modern wine industry. Faced with the above problems, the study is mainly focus on the Cabernet Sauvignon(CS) wine grapes at different ripening stages which are widely planted as the study object. Based on the hyperspectral imaging technology and multiple regression methods, the prediction models of anthocyanin and total phenols were built to accomplish the non-destructive and efficient detection of phenolics compounds content of wine grape during ripening in the band range of 900-1700 nm. It indicates a new detection method of measuring the phenolic compounds content of wine grape during its ripening stage in wine industry. The main content of the study are as follows:(1) The optimal spectral preprocessing method applied on the spectral data of wine grape in the study was determined. In the study of predicting the anthocyanin and total phenols content, the optimal performance of PLSR prediction model was obtained by S-G filtering method compared with other spectral preprocessing methods including MSC, SNV and derivation. The prediction coefficient of determination(P-R2) of PLSR prediction models of anthocyanin and total phenols content after S-G filtering was 0.8407 and 0.7551, the root mean square error of prediction(RMSEP) was 0.0129 and 0.0046, respectively.(2) The PLSR and SVR methods were applied to build anthocyanin content prediction models to accomplish the detection of anthocyanin content. By comparing the performance of the two above prediction models, it concluded that the performance of SVR prediction model could get a relatively good predictable performance with P-R2 of 0.9414 and RMSEP of 0.0046. The result indicated that it is capable of realizing the efficient and non-destructive detection of anthocyanin content of wine grape during ripening by using hyperspectral imaging technology.(3) The quantitative analysis model was built with spectral data of wine grape and the total phenols content by using hyperspectral imaging technology to accomplish the detection of total phenols content of wine grape during ripening. The PLSR and BPNN methods were adopted to build total phenols content prediction models. Compared with PLSR prediction model, BPNN prediction model could get better performance.. The P-R2 of BPNN prediction model was 0.7629 and the RMSEP was 0.0071.Given the above, the efficient and non-destructive detection of anthocyanin and total phenols content was accomplished by using hyperspectral imaging technology. A new determination method for measuring the content of phenolic compounds of wine grape was proposed, which indicates a promising application prospect.