漫反射近红外光谱技术快速识别马尾松松脂和湿地松松脂

不同种类的松脂因其化学组成的差异而对松脂下游产品的质量产生影响，因此确保松脂种类的稳定性是保证松脂下游产品质量的关键，针对在松脂原料采购过程中松脂种类识别困难的问题，提出了一种基于漫反射近红外光谱和偏最小二乘判别分析(PLSDA)相结合的分析技术，该技术能够快速识别马尾松松脂和湿地松松脂，为松脂原料采购提供可靠的种类信息。以在广西区内武鸣、防城、富川、梧州、百色、乐业共6个不同松脂产区采集所得的82个松脂样本进行建模，包括湿地松松脂51个，马尾松松脂31个，利用i-spec型近红外光谱仪采集松脂样本在900～1 700 nm范围内的近红外光谱。利用子窗口随机化分析法(SPA)进行变量选择，从510个波长点中优选出300个波长点组成的变量子集，再通过重复双重交叉检验技术(RDCV)确定偏最小二乘判别分析建模的潜变量数(n=7)。结果表明，所建立的分类模型能够准确识别两种不同种类的松脂，模型对于外部测试集中的松脂样本识别准确率为96.30%，能够满足松脂行业在原料采购过程中质量控制的需要。该方法具有分析速度快、操作简便、分析成本低、样本无损等优势，适用于松脂原料采购环节的质量控制。

Discrimination of Oleoresins from Pinus msssoniana and Pinus elliottii by Near Infrared Spectroscopy

It is crucial to ensure the quality stability of rosin material since the different chemical constituents of various kinds of rosins will obviously influence the quality of down-stream product. Herein, a method based on near infrared spectroscopy and partial least squares discriminant analysis was proposed to discriminate oleoresins from Pinus msssoniana and Pinus elliottii, which could be helpful to identify the species of oleoresins during the purchasing process. Eighty-two oleoresin samples from six different producing areas of Guangxi, i. e. Wuming, Fangcheng, Fuchuan, Wuzhou, Baise and Leye, were collected to develop classification model. These collected samples were consisted of 51 Pinus msssoniana and 31 Pinus elliottii. Diffuse reflection modes were applied to obtain near infrared spectrum range from 900～1 700 nm. Then, several chemometrics techniques such as sub-window permutation analysis and repeated double cross validation were used to select optimal variables and the number of principal component. Finally, 300 variables were extracted from the original variable pool and the optimal number of principal component was set to 8. Results showed that the obtained model can accurately discriminate oleoresins from Pinus msssoniana and Pinus elliottii, and the classification accuracy of external test is 96.30%, which can meet the need of quality control. The proposed method is less time-consuming, easy to operate and low-cost, and it is suitable for the quality control of purchasing process.