基于深度学习的香蕉成熟度自动分级  

Automatic classification of banana ripeness based on deep learning

作　　者：王灵敏 蒋瑜 WANG Ling-min;JIANG Yu(Guilin University of Technology at Nannning,Nanning,Guangxi 530001,China;Guangxi Agricultural Vocational and Technical University,Nanning,Guangxi 530007,China)

机构地区：[1]桂林理工大学南宁分校,广西南宁530001 [2]广西农业职业技术大学,广西南宁530007

出　　处：《食品与机械》2022年第11期149-154,共6页Food and Machinery

基　　金：广西高校中青年教师科研基础能力提升项目(编号:2020KY36006)。

摘　　要：目的:快速、准确分类香蕉成熟度。方法:采集不同成熟度的香蕉图像并建立图库,利用多种神经网络作为分类器提取香蕉特征,通过迁移学习对香蕉6个成熟度等级进行分类,并对最适合进行香蕉成熟度分类的网络模型进行改进,设计简易香蕉成熟度实时检测界面,最后验证模型的可行性和实用性。结果:AlexNet模型最适合用于香蕉成熟度分类,准确率最高,可达到95.56%;通过修改其全连接层结构改进AlexNet模型,模型准确率再提升1.11%。结论:AlexNet模型可快速准确识别并分类不同成熟度的香蕉。Objective:To classify banana ripeness quickly and accurately.Methods:Collect the bananas images of different maturity and establish gallery,using a variety of different neural networks as a classifier,banana feature extracting by migration study classifying banana six maturity level,access to the most suitable for banana maturity classification network model,network model,based on the improved and easily banana maturity real-time detection interface design,Finally,the feasibility and practicability of the model were verified.Results:AlexNet model was most suitable for banana maturity classification with the highest accuracy of 95.56%.AlexNet model was improved by modifying its full-connection layer structure,and the model accuracy was further improved by 1.11%.Conclusion:AlexNet model can quickly and accurately identify and classify bananas of different maturity.

关 键 词：香蕉 成熟度 自动分级 迁移学习 

分 类 号：TS255.7[轻工技术与工程—农产品加工及贮藏工程] TP18[轻工技术与工程—食品科学与工程] TP391.41[自动化与计算机技术—控制理论与控制工程]