基于近红外光谱技术的恩施玉露原产地茶鲜叶快速无损检测与判别

Enshi yulu tea is a famous product of steaming green tea and geographical indication producted from the young tea shoots picked from the protected areas tea plants. However, it is an urgent problem on how to determine the geographical origin of fresh tea leaves. Currently, the methods of inductively coupled plasma massspectrometry, organic ingredients fingerprinting technology and electronic nose technology are used to determine the origin of agricultural products, but it was time consuming, laborious and low discrimination stability. In order to maintain the landmark feature of Enshi yulu, a method of near infrared spectroscopy (NIRS) combined with different chemometrics was put up to identify the geographical origin of young tea shoots in the research scientifically and effectively. Firstly, massive young tea leaves samples were picked from different villages, different cultivation ages and varieties of tea trees, different altitudes，different slope aspect and different picking time in the protected areas, and then near infrared spectroscopy was obtained. Secondly, the signal-to-noise ratio was improved by different pretreatment methods. Finally, the chemometrics of interval partial least squares(iPLS), partial least squares(PLS), synergy interval partial least squares(siPLS), least squares support vector machine(LS-SVM), back propagation-artificial neural network(BP-ANN) and the combined methods were used to establish the NIRS models evaluated by discriminant rate, which were used to predict the geographical origin of young tea leaves. When it is more than 98% discriminant rate, the NIRS model is established successful. The implementation of the study will not only provide scientific and effective methods and theoretical support discrimination for Enshi Yulu geographical origin of fresh leaves, but also provide a usefull reference for other products of geographical indication.

恩施玉露是我国著名蒸青绿茶，作为国家地理标志产品，要求加工的鲜叶原料采自保护区，但如何判别鲜叶是否产自区域内是一个亟待解决的难题。当前农产品原产地判别常用电子耦合等离子体质谱法、有机成分指纹分析技术和电子鼻技术进行，但存在费时、费力和判别稳定性不高等问题。本项目拟运用近红外光谱技术结合化学计量学方法快速准确地判别鲜叶来源。研究时，采集区内不同村组、不同品种、不同树龄、不同海拔高度、不同坡向和不同采摘时间的海量鲜叶样品，扫描近红外光谱滤除光谱噪声，再运用间隔区间偏最小二乘法、偏最小二乘法、联合区间偏最小二乘法、最小二乘支持向量机方法等线性方法、人工神经网络等非线性方法以及上述组合方法分别建立光谱模型，并进行比较，确定最佳预测模型，最后通过模型稳定性检验，判别率达98%以上为终极目标。本研究预期为恩施玉露原产地鲜叶来源的判别提供科学有效的方法和理论支撑，对其它地标产品的产地判别也具有借鉴意义。

恩施玉露是我国著名蒸青绿茶，作为国家地理标志产品，要求加工的鲜叶原料采自保护区，但如何判别鲜叶是否产自区域内是一个亟待解决的难题。当前农产品原产地判别常用电子耦合等离子体质谱法、有机成分指纹分析技术和电子鼻技术进行，但存在费时、费力和判别稳定性不高等问题。本项目运用近红外光谱技术结合化学计量学方法快速准确地判别鲜叶来源。在芭蕉侗族乡、屯堡乡、舞阳坝街道办事处、白杨坪镇和盛家坝乡等恩施玉露产地，采集了同一地点的春季、夏季和秋季恩施玉露原产地茶鲜叶样品。为验证鲜叶产地近红外光谱判别模型的稳健性，又采集了与原产地接壤的宣恩县茶鲜叶样品和远离恩施玉露原产地区域的咸丰县茶鲜叶样品。样品共1080个，背景信息分别涵盖了不同栽培树龄（3年以下，5-10年，30年、50年和100以上的老茶树）、不同茶树品种（当地群体种、龙井43、春波绿、福鼎大白、恩苔早、浙农117、鄂茶一号、中茶108、碧春早、福云6号、楮叶齐和名山131等）、不同海拔高度（500米、600米、700米、800米、900米和1000米以上）、不同坡向（阴坡和阳坡）、不同价格（140元/Kg、110元/Kg、100元/Kg、80元/Kg、60元/Kg、50元/Kg及以下）和不同采摘时间（春季、夏季和秋季的上午、下午）的样品；采集标准分为芽、第一叶、第二叶、第三叶、一芽一叶、一芽二叶和一芽三叶，然后扫描获得近红外光谱。为有效的降低光谱的随机噪声，高频噪声对后期建立模型的影响，对鲜叶光谱进行了矢量归一化预处理、一阶导数+5点平滑预处理和二阶导数+9点平滑预处理。在此基础上，建立了三层BP神经网络预测模型（8（输入节点）-4（隐含层节点）-1（输出节点）），对验证集鲜叶样品产地进行了预测，产地判别识别率为100%，实现了恩施玉露鲜叶产地的快速预测。项目实施过程中，发表论文6篇，申报国家发明专利6项。本研究为恩施玉露原产地鲜叶来源的判别提供科学有效的方法和理论支撑，对其它地标产品的产地判别也具有借鉴意义。