利用红外光谱法对天津产地玫瑰香葡萄酒识别鉴定, 提出了用支持向量机(SVM）算法对干白葡萄酒红外光谱图进行分析, 详细介绍了干白葡萄酒谱图的预处理过程, 包括基线校正、 消除噪声、 标准归一化和去除异常样本点。 选取了不同产地的玫瑰香原酒样品511个、 贵人香原酒样品438个、 霞多丽原酒样品307个、 白玉霓原酒样品29个、 白羽原酒样品44个、 龙眼原酒样品31个、 泽香原酒样品79个, 针对不同分类将80%样品谱图用SVM建立分类模型, 并用剩余20%样品谱图对这些模型进行验证, 实验结果表明该方法行之有效, 不同葡萄品种干白葡萄酒模型正确率、 识别率和拒绝率达到97%以上, 不同产地玫瑰香干白葡萄酒正确率、 识别率和拒绝率达到98%以上。 该方法可以对天津产地玫瑰香葡萄酒进行定性识别鉴定。

The purpose was to achieve the identification of Muscat Hamburg wines produced in Tianjin region through scanning and analyzing dry white wine samples of different grape varieties and regions by infrared spectroscopy technology. A support vector machine (SVM) based method was introduced to analyze infrared spectra of dry white wines. The pretreatment processes of the IR spectra were also elaborated, including baseline adjustment, noise Elimination, standard normalization and eliminating the main component of abnormal sample points. The authors selected great quantity of dry white wine samples of different grape regions including 511 Muscat Hamburg wine samples, 438 Italian Riesling wine samples, 307 Chardonnay wine samples, 29 Ugni Blanc wine samples, 44 Rkatsiteli wine samples, 31 longan wine samples and 79 ZeHong wine samples. According to different classification problems, 80% of IR spectra of the wine samples were used to establish discrimination models with SVM-based method, and the remaining 20% of IR spectra were used for the validation of the discrimination models. Experimental results showed that the proposed method is effective, since high classification accuracy, identification rate and rejecting rate were achieved: over 97% for the white wine samples of different grape varieties, meanwhile over 98% for the Muscat Hamburg wine samples produced in different regions. So the method developed in this paper played a good role in the qualitative classification and discrimination of Muscat Hamburg wines produced in Tianjin region. This novel method has a considerable potential and a rosy application future due to the expeditiousness, stability and easy-operation of FTIR method, as well as the veracity and credibility of SVM method.