基于电子鼻信号处理的牛肉中掺假猪肉判别模型

Discrimination Model for Adulterated Pork in Beef Based on Electronic Nose Signal Processing

由于牛肉经济价值较高,一些商贩为了获取更多利益对其进行掺假,危害了消费者的健康。牛肉掺假的常规检测技术具有耗时、费力的特点,而电子鼻因其快速无损的优势在牛肉掺假检测方面具有很大的潜力。本研究对纯牛肉、猪肉及牛肉中掺假猪肉的金属氧化物半导体(Metal Oxide Semiconductor,MOS)型电子鼻传感器数据进行处理,采用主成分分析实现传感器数据的去相关和降维,以累计贡献率大于90%的少数前几个主成分得分作为输入,对比采用Fisher判别分析(Fisher Linear Discriminate Analysis,Fisher LDA)和极限学习机(Extreme Learning Machine,ELM)构建牛肉中掺假猪肉的判别模型。结果显示,ELM模型训练集和测试集的识别准确率分别为99.64%和94.29%,均高于训练集和测试集识别准确率(分别为96.43%和75.71%)的Fisher LDA模型。结果表明,所构建的ELM模型能够用于基于MOS型电子鼻的牛肉中掺假猪肉的快速鉴别,可望为保障肉制品质量安全贡献积极力量。

Because of the high economic value of beef,some traders adulterate it in order to obtain more profits,endangering the health of consumers.The conventional detection technology of beef adulteration is time-consuming and laborious,and the electronic nose has great potential in the detection of beef adulteration because of its rapid and nondestructive advantages.In this study,the data of metal oxide semiconductor(MOS)electronic nose sensor of pure beef,pork and beef adulterated pork were processed,and the sensor data were de-correlated and dimensionality reduced by principal component analysis.The first few principal component scores with cumulative contribution rate greater than 90%were taken as input,and Fisher linear discriminate analysis was used for comparison(Fisher LDA).Fisher LDA and extreme learning machine(ELM)were used to establish the identification model of adulterated pork in beef.The results show that the recognition accuracy of the training set and the test set of the ELM model are 99.64%and 94.29%,respectively,which are higher than the Fisher LDA model with the recognition accuracy of the training set and the test set(96.43%and 75.71%,respectively).The results show that the ELM model can be used for the rapid identification of adulterated pork in beef based on MOS electronic nose,which is expected to contribute to the quality and safety of meat products.