【作者】 张龙；

【导师】 朱诚；

【作者基本信息】 浙江大学 ， 植物学， 2012， 博士

【摘要】 本研究利用近红外光谱、稳定同位素和矿质元素技术探讨其在植源性农产品(茶叶、杭白菊、花生、蜂蜜和水稻)产地溯源和鉴别中的应用,旨在初步建立地理标志产品保护及鉴别技术的有效方法。研究结果如下：1.采集安徽省、重庆市、福建省、广东省、广西壮族自治区、海南省、湖北省、河南省、四川省、山东省、云南省和浙江省茶叶,采用近红外光谱技术结合典则判别分析方法对非发酵茶、半发酵茶和发酵茶进行判别。结果表明,不同发酵程度茶叶在典则判别分析原始判别率和交叉验证判别率分别达到100%和94.4%。同时应用近红外光谱技术和稳定同位素技术分别对采自不同产地绿茶、红茶和乌龙茶等进行了鉴别,结果表明,茶叶水溶液和茶叶固体粉末的近红外光谱技术结合典则判别分析方法对茶叶产地的原始判别率和交叉验证判别率分别为100%和94.6%；而利用碳和氮稳定同位素技术仅能将安徽、浙江和南方部分省份(福建、广东、海南和广西壮族自治区)的茶叶分开。2.采集浙江省地理标志产品西湖龙井茶和浙江龙井茶,采用近红外光谱技术分别结合反向传递神经网络、径向基神经网络和最小二乘支持向量机判别西湖龙井和浙江龙井,结果表明最小二乘支持向量机模型性能最高,两种茶叶的正确判别率均达到100%。3.采集山东、湖北、河南、辽宁、广西和广东、四川产地的花生,采用近红外光谱技术进行鉴别,结果表明近红外光谱结合小波转换和k最近邻分析(WT-KNN)对不同省份花生的原始判别率和交叉验证判别率分别为100.0%和55.9%。4.利用矿质元素技术对杭白菊地理标志产品进行溯源分析,结果表明采自桐乡、孟州和毫州杭白菊其判别模型的矿质元素为Pb、Sr、Ba、Ga和V,原始判别率和交叉验证判别率分别为100%和97.%。桐乡市三个种植区杭白菊的判别模型元素为Cd、Pb和Rb,原始和交叉验证判别率分别为76.7%和70.0%。5.近红外光技术应用于重金属污染水稻叶片和转基因大米快速鉴别。结果表明,小波转换采用db2函数第3分解水平对光谱预处理结合径向基人工神经网络对重金属胁迫下水稻叶片识别效果最优。对Hg、Cd、Pb污染土壤上和正常条件下生长的水稻叶片的识别正确率分别为95.5%、81.8%、91.3%和100.0%。转基因水稻分别通过近红外光谱进行鉴别,结果表明,偏最小二乘回归模型对转tctp和miR166基因水稻的正确预测率达到100%。6.利用碳稳定同位素和近红外光谱技术对收集于黑龙江、辽宁、浙江、福建、成都、湖北、四川、广西、广东、河南、陕西、新疆、西藏自治区和重庆产地的蜂蜜进行了掺假鉴定,结果表明近红外光谱技术结合偏最小二乘回归模型其校正集和验证集的准确率分别达到100%和93.3%,近红外光谱技术可代替碳稳定同位素技术对蜂蜜的掺假进行快速鉴定。同时建立了绵白糖和果葡糖浆掺假预测模型。更多还原

【Abstract】 In present study, the geographical origin and the authenticity of plant agricultral products (tea, peanut, hangbaiju, rice and honey) were determined with various technonolgies. And then the problems of safety and geographical origin of food could be solved. And the results were as following:1Near infrared spectroscopy was used to identify different kinds of fermented tea which collected from Anhui, Chongqing, Fujian, Guangdong, Guangxi, Hainan, Hubei, Henan, Sichuan, Shandong, Yunnan and Zhenjiang. It was shown that the classification rates of origin and cross-validation test were100%and94.4%, respectively, in canonical discriminant analysis. In order to discriminate the geographical origin of tea, near infrared spectroscopy and stable isotope technologies were used. It was shown that with the combination of spectroscopy of tea water and tea powder the classification rates of geographical origin of tea were100%and94.6%in origin and cross-validation test respectively; while the stable isotope techonolgy could only discriminate the geographical origin of tea from provinces of Anhui, Zhejiang and the southern China (Fujian, Guangdong, Guangxi and Hainan).2Near infrared spectroscopy was used to identify Xihu Longjing and Zhejiang Longjing using partial least squared regression discriminate analysis (LSSVM), Back propagation neural network (BPNN) and radial basis function neural network (RBFNN). Xihu Longjing and Zhejiang Longjing were discriminated from each other perfectively in LSSVM.3The geographical origin of peanuts, collected from Shandong, Hubei, Henan, Liaoning, Guangdong and Guangxi, Sichuan, was identificated, and the classification rates of origin and cross-validation test were100%and55.9%respectively with the K near neibours methods combined with near infrared spectroscopy technology.4It was shown that the elements of discriminant model of place of origin (POO) were Pb, Sr, Ba, Ga and V, and the original and cross-validated classification rates were100%and97%respectively. The elements of discriminant model of different growth area in Tongxiang were Cd, Pb and Rb and the original and cross-validated classification rates were76.70%and70.0%respectively.5It was shown that tansgenic rice, TCTP and mi166, and Zhonghua11were all well discriminated from each other with NIR and PLS-DA. Heavy metal (mercury, cadmium and lead) polluted leaves of rice were identificated with near infrared spectroscopy. It was also shown that the classification rates of mercury, cadmium and lead polluted and control leaves of rice were95.5%,81.8%,91.3%and100.0%respectively in RBFNN model with the treatment of wavelet transform function db2at3level.6In the experiment of identification of commercial honey, collected form Heilongjiang, Liaoning, Zhejiang, Fujian, Chengdu, Hubei, Sichuan, Guangxi, Guangdong, Shanxi, Xinjiang, Tibet, Chongqing, carbon stable isotope ratio technology was first used to identify the adulteration of honey, and then the near infrared spectroscopy technology was used to predict the aluteration of honey. It was showed that the classification rates of calibration and validation test were100%and93.3%respectively. The standarded model of adulterated honey was constructed with the high fructose corn sugar (HFCS), soft sugar (SC) and honey.更多还原