Infrared Spectra Of Liquor Feature Extraction And Classification Applied Research

Infrared Spectroscopy is a substance component analysis method in the field ofanalytical chemistry. Because of convenient sample collection, it is widely used inChemical, food science and many other fields now. Liquor is a complex liquid mixture.Analysing liquor by quantity as well as quality, which bases on infrared spectroscopy, isa significant method when identifying, anti-counterfeiting, trasing the quality of Liquorproducts.Liquor infrared spectroscopy can rapidly identify liquor alcohol content, flavor andother information, achieving the purpose of real-time detection of liquor quality. Thisthesis studies the infrared spectroscopy of the different brands and series liquor. Themain contents are as follows:1. Using multi-scale feature selection algorithm to select the feature wave band fromthe original liquor infrared spectroscopy. Build the feature database of the liquor.2. Using partial least squares regression method to train the alcohol regressionmodel. Then predict the alcohol quantitatively by the regression model.3. Using support vector machine method to train the classification model of thebrands, and classify liquor brands.Form the studies we can conclude that multi-scale feature selection algorithm canextract the valid information accurately. It’s useful for further research on the white spirit.Based on the machine learning methods and PLS regression, the classification accuracyrate reach99%, and the regression error is less than1°.