近红外全波段扫描技术建立数学模型鉴别地沟油方法研究

以7个品种的77份合格食用植物油、28份不合格植物油和118份地沟油作为研究对象，利用二极管阵列近红外光谱仪，以10 nm为步长对所有样品进行950~1 650 nm全波段扫描，通过对不同组别扫描数据的差异化分析，建立特征波长下不同组别的数学模型，鉴别地沟油与合格食用植物油及不合格植物油。结果表明：通过统计学分析手段建立的数学模型，对原始数据分类准确率为96.0%，交叉验证准确率为95.5%，该模型对未知样品的判定准确率达到95%以上。表明基于近红外全波段扫描技术鉴别地沟油的分析是可行的，且该方法具有操作简单、检测成本低、样品用量小等特点，可作为地沟油快速筛查方法使用。

Discerning of swill-cooked dirty oil by establishing mathematical model with near infrared full-band scanning technique

77 qualified edible vegetable oils of 7 varieties, 28 unqualified vegetable oils and 118 swill-cooked dirty oils were scanned in the full wavelength band at 950-1 650 nm with a step length of 10 nm by using diode array near infrared spectrometer. By analyzing the difference of the scanning data of different groups, the mathematical models of different groups under the characteristic wavelength were established to distinguish swill-cooked dirty oils from qualified edible vegetable oils and unqualified vegetable oils. The results showed that the accuracies of classification and cross-validation of the mathematical models established by statistical analysis tools for the original data were 96.0% and 95.5%, respectively,and the accuracy of the model for unknown samples reached over 95%. The results showed that it was feasible to identify swill-cooked dirty oils based on near infrared full-band scanning technique, and the method was simple in operation, low in cost and sample consumption,and could be used as a rapid discerning method for swill-cooked dirty oil.