基于BP神经网络的近红外光谱法鉴别芝麻油品牌的研究

研究了一种用近红外光谱分析技术快速鉴别芝麻油品牌的方法。首先对芝麻油样品的近红外光谱采用主成分分析法进行聚类分析,加结合人工神经网络技术进行芝麻油品牌的鉴别。通过主成分分析,得到前15个主成分的累计可信度达到99.72%,再将55个校正集样品的前15个主成分数据作为BP网络输入变量,建立一个3层BP人工神经网络的芝麻油...

Discrimination of different brand sesame oil using near infrared spectroscopy analysis based on BP artificial neural network

A new method for the discrimination of different brands of sesame oil by near infrared spectroscopy（NIRS） was developed. Firstly, using near infrared spectroscopy and principal component analysis to classify the sesame oils, then using principal components and BP artificial neural network method to identify the six different brands of sesame oil.The principal component analysis showed that the reliabilities of the first 15 principal components reached 99.72%,so the 15 principal component data of 55 calibrated samples were used as the inputs of back-propagation artificial neural network （ANN-BP）, and the three layers ANN-BP discrimination model was built,then 18 tested samples were inputed to the model,results showed that the identification rate was 83.33%. Compared with the relatively low identification rate of conventional identification methods,it is reliable and practicable to use BP artificial neural network method to identify the different brands of sesame oil.