The Design And Realization Of Chinese Liquor Authentication And Traceability System Based On Infrared Spectroscopy

As a Chinese-specific wine, Chinese liquor has been existing for a long time with unique history. In the last few years, with the increasing demand of liquor, large amount of counterfeit liquor has been showing up in the market. Counterfeit liquor not only endangered the health and harmed the interests of consumers, but also lead to the serious winery economic losses, as well as disrupted the normal economic order. Therefore, how to identify the counterfeit liquor and make traceability of the counterfeit liquor quickly and effectively has become an important topic to the public security organizations and other law enforcement agencies.Infrared spectroscopy is a kind of spectroscopic technique rising in recent decades and has attracted a lot of scholars' attention. A lot of researches about it have been done. With the advantage of nondestructive analysis, fast measurement speed, information-rich, etc., it introduces a new way for liquor detecting.This paper selected the infrared spectroscopy of liquor as the studying object, focused on liquor infrared spectral feature extraction techniques, classification techniques and clustering techniques, aimed to design and develop a software system which has the functions of infrared spectral data storage, feature extraction, liquor authentication and traceability. Through this software system, the regulatory authorities can sanction liquor counterfeiting efficiently. The main contents and results are summarized as follows:(1) Several traditional infrared spectral feature extraction algorithms were introduced, both of the advantages and disadvantages of the algorithms were summarized, a wavelet-based multi-band feature extraction method was presented. The experimental results showed that the multi-band feature extraction method can effectively extract the characteristic points which play an important role in distinguishing in liquor infrared spectrum and improve forecast accuracy of the classification model.(2) The support vector machine and Mahalanobis distance discriminant method were combined to classify the collected liquor spectra, a hierarchical multi-classification scheme based on the attribute of liquor was proposed. The experimental results showed that the hierarchical classification scheme can effectively improve the classification accuracy. Experimental results show that classification model built by the method that combined the multi-band feature extraction method and the hierarchical classification scheme together get the highest classification accuracy of 91.14%.(3) The principal component analysis and DBSCAN algorithm were combined to cluster the counterfeit liquor infrared spectral data. The experiment results show that clustering can be used as an effective method of traceability.(4) Liquor IR feature extraction, classification and clustering methods proposed in this paper were integrated to design a comprehensive liquor authentication and traceability solution. Based on the solution, the liquor authentication and traceability system were designed and developed. According to the test, the system can achieve the goal of liquor authentication and traceability.