多谱图信息融合技术及其在白酒品牌鉴定中的应用

In this project, the application of comprehensive Feature Information in liquor brand is used as background and the temperature programmed thermal desorption ion mobility spectrometry(IMS) and time-of-flight mass spectrometry(TOFMS) is used as the analyzing tool, and in the view of the situation of limitations and some deficiencies of appraising the quality of complex mixtures depended on single spectra, therefore this project focuses on algorithms of information fusion of multiple-spectra and its application in order to improve the accuracy and robustness of quality inspection of liquor brand. This research has important academic and realistic meaning. There are four main aspects in the research and application of it. The first is preprocessing algorithms of spectra. The second is spectrum feature information extraction algorithm. The third is multispectral diagram IMS isomorphism information fusion algorithm and IMS/MS heterogeneous information fusion algorithm. The last is the sparse recognition algorithm of fusion spectrums. In this project, we organic combined the advanced theory of the pattern recognition with the advanced technology of the detection analysis, and SIMPLISMA is used to extract characteristics of the spectrogram information, and spectrum graph isomorphism and heterogeneous information fusion algorithm is also established. The research group in this project comes up with and sparse recognition algorithm of spectrum information feature extraction fast recognition, using the Yang He series liquor products as an example, and aiming at liquor brand identification and quality analysis, set up the characteristic spectral library and standard inspection process and method. With the spectrum of the gallery is unceasingly rich and perfect, we will provide technical support on the liquor quality identification and security crackdown on counterfeit goods.

本项目以综合特征信息在白酒品牌鉴定中的应用为基本背景，以程序升温热解析离子迁移谱和飞行时间质谱为分析工具，针对单一化学谱图来衡量复杂混合物构成的产品的质量中存在的不足和局限性，本项目拟开展多谱图信息融合算法与应用的研究，以提高白酒品牌鉴定与评价技术系统的准确性和鲁棒性。项目主要包括四方面的研究：1）谱图预处理算法；2）谱图特征信息提取算法；3）多谱图离子迁移谱同构信息融合算法和离子迁移谱/质谱异构信息融合算法；4）融合谱图的稀疏识别算法。本项目将模式识别中的先进理论与检测分析中的先进技术进行有机结合，应用快速识别的自模型算法进行特征提取，建立谱图同构和异构信息融合算法，提出融合谱图的稀疏识别算法，以洋河系列白酒产品为例，建立针对白酒品牌鉴别和品质分析的特征谱库和标准检验流程及方法，随着谱图库的不断丰富与完善，将对白酒品质鉴定和防伪打假提供技术保障。

本项目以综合特征信息在白酒品牌鉴定中的应用为基本背景，针对单一化学谱图来衡量复杂混合物构成的产品的质量中存在的不足和局限性，本项目开展了多谱图信息融合算法与应用的研究，以提高白酒品牌鉴定与评价技术系统的准确性和鲁棒性。项目主要包括三方面的研究：1）谱图预处理算法；2）谱图特征信息提取算法；3）多谱图信息融合鉴别算法。. 项目组联合运用了紫外可见光谱、拉曼光谱、离子迁移谱、荧光光谱等技术作为检测手段，对样品进行整体信息的检测，从而得到的反映物质整体特征的信息。使得采集的特征图谱能基本反映白酒的全貌，有专属性、模糊性和相对稳定性的特点，使其对白酒的质量控制指标由单一成分的测定上升为对整体品质的伪劣鉴别和质量控制，适用于基质成分较复杂的样品的质量评价模式，能够实现对白酒的整体、快速的检测，将对白酒品质鉴定和防伪打假提供技术保障。. 项目组在2014~2017年研究期间，共发表学术文章20篇，其中包括6篇英文论文。7篇SCI检索论文,1篇SSCI检索论文，1篇EI检索论文，12篇北大核心期刊检索论文。此外,项目组在研究期间获得专利3项，其中发明专利2项，实用新型专利1项。