【作者】 贾文珅；

【导师】 吴文福； 王纪华； 兰玉彬；

【作者基本信息】 吉林大学 ， 农业电气化与自动化， 2014， 博士

【摘要】 本文以龙井茶为研究对象，利用近红外光谱技术、二维条码技术结合模式识别等方法围绕―西湖龙井‖茶的真伪鉴别、品质以及内部成分等方面开展了系统的研究，主要取得以下研究结果：1.本文首先比较了两类近红外光谱仪（傅里叶变换型近红外光谱仪和法布里干涉型近红外光谱仪）在真伪鉴别―西湖龙井‖茶方面的效果。在此基础上，针对不同类型的光谱仪采用CARS算法对光谱鉴别模型进行了优化，分别筛选出傅里叶变换型近红外光谱仪与法布里干涉型近红外光谱仪―西湖龙井‖茶真伪鉴别的关键变量（15个与7个）。表明近红外光谱分析技术能够用于―西湖龙井‖茶快速真伪鉴别应用。2.为开发―便携式近红外光谱西湖龙井茶快速真伪鉴别仪‖，本文截取了傅里叶变换型光谱仪近红外光谱（12800cm-1-4000cm-1）中，10526cm-1-6060cm-1谱区的近红外光谱，并采用PLS-DA进行了建模分析，结果表明，上述谱区可以较好的区分―西湖龙井‖茶与浙江龙井茶，可作为―西湖龙井‖茶真伪鉴别谱区。3.在上述谱区筛选的基础上，本文设计开发了一种用于―西湖龙井‖茶快速真伪鉴别仪，该仪器结合了近红外光谱技术与二维条码技术的优势，利用―物‖、―码‖一致的思路（物代表龙井茶，码代表二维条码），对―西湖龙井‖茶进行真伪判定。本仪器设计从商品标签的数据存储需求，近红外光谱模块的选择，人机交互需求三个方面进行对比论证，选定二维条码作为―西湖龙井‖茶的商品标签，并选用QR编码作为二维条码编码格式。二维条码技术具有成本低廉，读取迅速，人机交互性好，可用于存储商品基本信息及―西湖龙井‖茶真伪鉴别加密信息等多种优良特性。根据预评估―西湖龙井‖茶真伪鉴别所需光谱谱区的结论，选取了基于线性渐变滤光片设计的超微型近红外光谱仪作为该仪器近红外光谱获取模块，该模块仅用USB数据线供电，单体重量不超过60g。最后，为了具有良好的人机交互性和较高的开发效率使用微软公司.NET框架下的C#语言进行系统开发，设计并开发了一套―西湖龙井‖茶光谱信息二维条码加密存储方案，该设计方案可防止不法分子，通过逆向破解盗取―西湖龙井‖茶快速真伪鉴别仪内嵌模型算法。本文所研发的仪器，强化了―西湖龙井‖茶真伪鉴别的技术手段，可为其他珍贵农产品真伪鉴别提供依据和示范。4.为了做到无损包装，对―西湖龙井‖茶进行真伪鉴别，提出了一种透过单层自封袋薄膜龙井茶光谱的采集方法，并设计了一种专用的―西湖龙井‖茶包装袋。5.为了进一步揭示机理，本文对龙井茶品质主要指标进行湿化学测量，然后利用高精度傅里叶变换近红外光谱仪，对龙井茶样本在不打碎，不磨粉预处理条件下所采集光谱进行了近红外模型的构建与优化。通过对所获取的龙井茶近红外光谱进行散射校正的光谱预处理，并比较原始光谱、一阶导数光谱、二阶导数光谱在不同点数（3、5、7、9点）平滑处理条件下，偏最小二乘回归模型效果，探索最优预处理方案。在此基础上，对最优预处理方案的光谱，运用CARS算法分别找出龙井茶中茶多酚、咖啡碱、氨基酸以及表没食子儿茶素没食子酸酯含量的近红外分析建模关键变量（74个、101个、101个和87个）。所优化出的关键建模变量大大提高模型计算速度，可满足对龙井茶品质在线、快速检测的要求。更多还原

【Abstract】 With the research object of Longjing tea, this dissertation will apply the nearinfrared spectrum technology and two-dimensional bar code mode methods to studythe true and false identification, quality and interior ingredients for Xihu Longjing‘tea.The following are main results:1. Firstly, this dissertation compares effects of two NIR technologies (Fouriertransform and Fabry interference) on true and false identification for Xihu Longjing‘tea. On this basis, the CARS algorithm is used for different spectrometers to optimizethe spectrum identification module, and select key variables (15and7) for the two NIRspectrometers to identify true and false Longjing tea, which indicates the infraredspectroscopic analysis technology can be used to identify true and false Longjing teafast.2. In order to develop portable NIRS to identify true and false Longjing tea fast,this thesis selects the NIR on10526cm-1-6060cm-1band in Fourier transformspectrograph and applies PLS-DA to establish module and analyze, the result indicatesthe above band can distinguish Xihu Longjing‘tea and Zhejiang Longjing tea betterand be the band to identify true and false Xihu Longjing‘tea.3. On the basis of the above band selection, this dissertation designs anddevelops an identification device to identify true and false Xihu Longjing‘tea, whichcombines advantages of NIR technology and two-dimensional bar code and appliesthe concept of consistent object and code (object refers to Longjing tea and code refersto two-dimensional bar code) to identify true and false Xihu Longjing‘tea. Thisdesign development firstly starts with data storage demands of commodity label,selection of NIR module and man-machine interaction demands. Through comparisonand demonstration, due to low cost, fast reading and good man-machine interaction,the two-dimensional bar code technology can be used to store basic information ofcommodities and encryption information to identify Xihu Longjing‘tea, therefore,this dissertation selects such technology to be the commodity label of Xihu Longjing‘tea and QR coding as the format of two-dimensional bar code. However, according toresults of pre-evaluation for spectral range necessary to identify true and false XihuLongjing‘tea, the super-miniature near infrared spectrometer designed based on linearand tunable optical filter is selected to acquire module which is powered only withUSB data and do not exceed60g on single weight. Finally, in order to achieve better man-machine interaction and higher development efficiency, C#language under NETframework of Microsoft Corporation is used to develop system, design and develop aset of two-dimensional bar code storage plan written in spectral information of XihuLongjing‘tea, which can prevent the lawbreaker to steal the embedded modulealgorithm in identification device to fast identify true and false Xihu Longjing‘teathrough converse cracking. The instrument this dissertation researched and developedhas strengthened the identification technology and can provide basis anddemonstration for sources of quality safety of other agricultural products.4. In order to achieve lossless packing, and check the authenticity of XihuLongjing‘tea, an acquisition method of Longjing tea spectrum via single ziplock bagthin film is proposed and a packing bag is specially designed.5. In order further to reveal the mechanism, conducts wet chemical measurementon the main indicators of Longjing tea’s quality and applies Fourier transformspectrometer with high precision to predict the ingredient content of Longjing teasample under pre-handling conditions of no smash and grind. Firstly, the spectralpretreatment of scattering adjustment is implemented for achieved Longjing tea, andeffects of partial least squares regression model for original spectrum, first derivativespectrum and second derivative spectrum are compared on smooth treatmentconditions of different counts (3,5,7,9). On this basis, the CARS algorithm is appliedfor spectrum with optimal retreatment plan to respectively find out key modelingvariables related with each ingredient of Longjing tea in order to further improve theprecision of module. The last is to find out key variables for near infrared analysismodeling of Tea Polyphenols, caffeine, amino acid, and EGCG The number ofmodeling variables sharply reduces from4401to74,101,101and87and themodeling calculation speed is greatly improved, which can satisfy requirement ofonline and fast test for quality of Longjing tea.更多还原