基于电子鼻和机器视觉的鱼肉新鲜度检测研究  

Fish freshness detection based on an electronic nose and machine vision

作　　者：袁也 周博[1] 吴泽玮 YUAN Ye;ZHOU Bo;WU Zewei(Department of Mechanical Engineering,Yancheng Institute of Technology,Yancheng 224002,China)

机构地区：[1]盐城工学院机械工程学院,江苏盐城224002

出　　处：《食品与发酵工业》2024年第24期313-320,共8页Food and Fermentation Industries

基　　金：国家自然科学基金项目(22171239,31671583)。

摘　　要：为了提升鱼肉新鲜度检测的准确率,该研究采用了电子鼻、机器视觉和多数据融合技术快速地检测冷藏鱼肉的新鲜度。挥发性盐基氮含量与新鲜度密切相关且易于测量,因此被选定作为鱼肉新鲜度的指标;用机器视觉和电子鼻获取样品的图像和气味信息。应用反向传播神经网络、卷积神经网络(convolutional neural network,CNN)和卷积神经网络-门控循环单元-注意力(CNN-GRU-Attention)3种模型对鱼肉新鲜度进行3分类和7分类预测。结果表明,3分类和7分类实验中,3种模型利用电子鼻数据进行分类的效果均优于机器视觉方法。此外,对原始数据进行融合后,3个模型的分类准确率均有提升。特别是基于CNN-GRU-Attention模型的多感官数据融合方法在本次研究中效果最优,其在测试集上的准确率分别达97.61%和90.48%。研究结果表明,采用多感知检测技术结合CNN-GRU-Attention预测模型能够有效地提高鱼肉新鲜度检测的准确性。To improve the accuracy of fish freshness detection,electronic nose,machine vision,and multi-data fusion techniques were used to rapidly detect the freshness of refrigerated fish.Total volatile base nitrogen(TVB-N),which is closely related to freshness and is easy to measure,was selected as an indicator of fish freshness.Machine vision and electronic nose-acquired images as well as odor information were collected from samples.Three models,namely,the backpropagation neural network(BPNN),convolutional neural network(CNN),and convolutional neural network-gated recurrent unit-attention(CNN-GRU-Attention),were applied to fish freshness for 3-classification and 7-classification prediction.Results showed that the classification effect of the three models using the electronic nose data was better than that of the machine vision method,regardless of whether the application was 3-classification or 7-classification.In addition,the classification accuracy of the three models improved after the fusion of the original data.In particular,the multisensory data fusion method based on the CNN-GRU-Attention model performed the best in this study,with its accuracies on the test set reaching 97.61%and 90.48%,respectively.The results showed that multi-perception detection technology combined with the CNN-GRU-Attention prediction model could effectively improve the accuracy of fish freshness detection.

关 键 词：鱼肉新鲜度 电子鼻 机器视觉 数据融合 神经网络 

分 类 号：TS254.7[轻工技术与工程—水产品加工及贮藏工程] TP391.41[轻工技术与工程—食品科学与工程]