茶叶品种鉴别在茶叶的生产和销售中起着十分重要的作用。 深入研究一种方法简单、 易于操作、 检测速度快的茶叶品种的鉴别方法, 对于茶叶产品品种的鉴别有着十分重要的意义。 利用红外光谱检测技术结合模糊聚类算法对茶叶品种进行快速鉴别是茶叶品种检测中最有效的和最实用的技术之一。 为实现茶叶品种的快速分类, 以快速广义噪声聚类(FGNC)为基础, 提出一种新的广义噪声聚类(NGNC)。 NGNC将FGNC目标函数中的欧式距离的平方扩展为欧式距离的p次方, 提高了FGNC的聚类准确率。 试验以优质乐山竹叶青、 劣质乐山竹叶青和峨眉山毛峰三种茶叶为研究对象, 采用FTIR-7600型傅里叶红外光谱仪检测茶叶样本的红外漫反射光谱。 首先用主成分分析(PCA)对茶叶的高维红外光谱进行降维处理, 然后用线性判别分析(LDA)进行茶叶光谱数据的品种类别信息的提取, 最后分别运行FGNC和NGNC两种聚类算法进行茶叶红外光谱的聚类分析。 实验结果表明, 同FGNC相比较, NGNC具有更高的聚类准确率, 更快的收敛速度和更逼近真实的聚类中心。 总体而言, 采用红外光谱技术检测茶叶样本, 同时结合PCA, LDA和NGNC可实现快速、 准确地聚类茶叶的红外光谱, 能有效地实现茶叶品种的鉴别分析, 为实现基于红外光谱和模糊聚类的茶叶品种鉴别分析提供了一种新方法和新思路。

The discrimination of tea varieties plays very important role in the production and sale of tea. It is of great significance for the study of a fast, easy and simple method for the identification of tea varieties. The combination of infrared spectroscopy detection technology and fuzzy clustering algorithm is one of the most effective and practical techniques in the detection of tea varieties. To realize the rapid discrimination of tea varieties, a novel generalized noise clustering (NGNC) was proposed based on fast generalized noise clustering (FGNC). Euclidean distance in the objective function of FGNC was replaced with the pth power of Euclidean distance and clustering accuracy was being improved. Emeishan Maofeng; high quality Leshan trimeresurus and low quality Leshan trimeresurus were prepared as the research object and the infrared reflectance (IR) spectra of tea samples were collected with FTIR-7600 infrared spectrometer. Firstly, the high-dimensional IR spectra of tea samples were reduced by principal component analysis (PCA). Secondly, linear discriminant analysis (LDA) was used to extract the discriminant information from the low-dimensional data. Finally, FGNC and NGNC were performed to identify tea varieties. The experimental results showed that in comparision with FGNC, NGNC has higher clustering accuracy, better cluster centers and faster convergence speed. Infrared spectroscopy coupled with NGNC, PCA and LDA could cluster IR spectra of tea samples quickly and correctly, which provided a new method and new idea for identifying tea varieties based on infrared spectroscopy and fuzzy clustering.