机器学习与计算机视觉技术在食品质量评价中的研究进展  

Research Progress on Machine Learning and Computer Vision Technology in Food Quality Evaluation

作　　者：黄晓琛 张凯利 刘元杰 陈洪[1] 黄凤洪[1] 魏芳[1,3] HUANG Xiaochen;ZHANG Kaili;LIU Yuanjie;CHEN Hong;HUANG Fenghong;WEI Fang(Hubei Key Laboratory of Lipid Chemistry and Nutrition,Oil Crops and Lipids Process Technology National&Local Joint Engineering Laboratory,Oil Crops Research Institute,Chinese Academy of Agricultural Sciences,Wuhan 430062,China;Key Laboratory of Agricultural Information Acquisition Technology,Ministry of Agriculture and Rural Affairs,College of Information and Electrical Engineering,China Agricultural University,Beijing 100083,China;Hubei Hongshan Laboratory,Wuhan 430070,China)

机构地区：[1]中国农业科学院油料作物研究所,油料脂质化学与营养湖北省重点实验室,油料油脂加工技术国家地方联合工程实验室,湖北武汉430062 [2]中国农业大学信息与电气工程学院,农业农村部农业信息获取技术重点实验室,北京100083 [3]湖北洪山实验室,湖北武汉430070

出　　处：《食品科学》2024年第12期1-10,共10页Food Science

基　　金：国家自然科学基金联合基金项目(U21A20274);“十四五”国家重点研发计划重点专项(2021YFD1600103);农业农村部油料作物生物学与遗传育种重点实验室开放课题项目(KF2023008);中国农业科学院创新工程项目(CAAS-ASTIP-2013-OCRI);湖北省自然科学基金创新群体项目(2023AFA042)。

摘　　要：近年来,随着社会对食品质量和安全的关注度不断提高,计算机视觉技术在食品质量评价领域逐渐受到重视并开始广泛应用。通过学习技术,如人工神经网络、卷积神经网络和支持向量机等,研究人员能够利用大量的食品图像和相关数据进行训练,从而实现对食品质量的自动评估和监测。特别是深度学习技术的发展,使得计算机能够更加准确地识别食品的外观、形状、颜色等特征,进而对其进行分类、预测和质量检测。除了在食品质量评价中的常规应用,学习技术还被用于更复杂的任务,如食品缺陷检测、异物检测、新鲜度评估等。这些技术不仅可以提高食品生产和加工的效率,还能够减少人为因素带来的误差,从而确保食品质量和安全。然而,尽管学习技术在食品质量评价中的应用取得了显著进展,但仍然存在一些挑战需要克服。例如,食品图像数据集的获取和标注成本较高,数据质量和数量的不足可能会影响模型的性能和泛化能力。此外,模型的可解释性和透明性也是一个重要问题,尤其是在需要对食品质量评价结果做出解释或决策的情况下。因此,未来的研究需要继续探索如何提高数据集的质量和规模、优化模型的鲁棒性和可解释性,以及开发更加高效和可持续的食品质量评价系统。In recent years,with rising concerns over food quality and safety,computer vision technology has gradually attracted attention and begun to be widely used in the field of food quality evaluation.Machine learning technologies such as artificial neural networks(ANN),convolutional neural networks(CNN),and support vector machines(SVM)allow automatic assessment and monitoring of food quality by training on large amounts of food images and related data.Particularly,with the development of deep learning,the computer is now able to more accurately recognize food features such as appearance,shape,and color,thereby allowing food classification,prediction and quality monitoring.In addition to its conventional application in food quality assessment,learning technologies also find application in more complex tasks such as defect detection,foreign object detection,and freshness assessment.These technologies not only improve the efficiency of food production and processing but also reduce errors caused by human factors,thereby ensuring food quality and safety.However,despite the significant progress in the application of learning technologies in food quality assessment,there are still challenges that need to be overcome.For instance,the high cost of acquiring and annotating food image datasets,as well as insufficient data quality and quantity,may affect the performance and generalization ability of models.Furthermore,the interpretability and transparency of models are important issues,especially when explaining or making decisions on food quality assessment results.Therefore,further research is needed to explore how to improve the quality and scale of datasets,optimize the robustness and interpretability of models,and develop more efficient and sustainable food quality assessment systems.

关 键 词：机器学习 食品质量 计算机视觉 食品检测 

分 类 号：TS207.7[轻工技术与工程—食品科学]