[1]董　庚１,王焱清１,２,等.基于 ResNet１８网络的油茶果壳籽分选研究[J].湖北工业大学学报,2023,(5):29-33+81.
　DONG Gen,WANG Yanqing,SUN Jiwei,et al.Sorting Model of Camellia Fruit Shells and TeaSeeds Based on ResNet18 Network[J].,2023,(5):29-33+81.
点击复制

基于 ResNet１８网络的油茶果壳籽分选研究()
分享到：

《湖北工业大学学报》[ISSN:1003-4684/CN:42-1752/Z]

卷:
期数:
2023年第5期
页码:
29-33+81
栏目:
出版日期:
2023-10-30

文章信息/Info

Title:
Sorting Model of Camellia Fruit Shells and TeaSeeds Based on ResNet18 Network
文章编号:
1003-4684(2023)05-0029-05
作者:
董　庚１; 王焱清１; ２; 孙记委１; ２; 段宇飞１; ２
１ 湖北工业大学农机工程研究设计院,湖北 武汉 ４３００６８;２ 湖北省农机装备智能化工程技术研究中心,湖北 武汉 ４３００６８
Author(s):
DONG Gen1 ;  WANG Yanqing1; 2; SUN Jiwei1 ;  DUAN Yufei1; 2
1 Agricultural Machinery Engin. Research and Design Institute,Hubei Univ. of Tech., Wuhan 430068, China;2 Hubei Agricultural Machinery Equipment IntelligentHubei Agricultural Machinery and Equipment Intelligent Engineering, Wuhan 430068,China
关键词:
深度学习; 油茶果; 分选; ResNet１８模型
Keywords:
deep learning;  camellia fruit;  sorting;  ResNet18 model
分类号:
TP391.4
文献标志码:
A
摘要:
油茶果脱壳后果壳与茶籽混合在一起,油茶果的果壳与茶籽如何分选以及如何提升分选的准确率,是油茶果产业化所面临的一个难题.基于改进 ResNet１８模型的油茶果果壳与茶籽的分选模型,通过油茶果分选机进行图片的采集和处理获取实验样本图像,比较不同批量尺寸、学习率、激活函数以及优化器的影响,进一步优化模 型.实验表明,经过改进后的 ResNet１８模型,与未改进前相比,在损失函数上有了明显的降低,并且验证集准确率得到提升,验证集平均准确率由之前的９７．０３％,而最终测试集准确率由９７．２５％提高至９７．７５％.较高的准确率可以满足油茶果分选的需要.
Abstract:
This paper proposes a sorting model based on the improved ResNet18 model for the sorting of camellia fruit shells and tea seeds, which is a difficult problem for the industrialization of camellia fruit. The experimental sample images were acquired and processed by a camellia fruit sorting machine, and the model was further optimized by comparing the effects of different batch sizes, learning rates, activation functions, and optimizers. The experiments show that the improved ResNet18 model has a significant reduction in the loss function compared with the unimproved one, and the accuracy of the validation set is improved, with the average accuracy of the validation set increasing from 97.03% before to 97.21% currently, while the accuracy of the final test set increases from 97.25% to 97.75%, an improvement of 0.5 percentage points, and the higher accuracy can meet the needs of camellia fruit sorting.

参考文献/References:

[１]　柏云爱,宋大海,张富强,等．油茶籽油与橄榄油营养价 值的比较[J]．中国油脂,２００８,３３(０３):３９Ｇ４１． [２]　段宇飞,皇甫思思,王焱清,等．基于机器视觉的油茶果 果壳与茶籽分选方法研究[J]．中国农机化学报,２０２０, ４１(０６):１７１Ｇ１７８． [３]　徐克生,杜鹏东,汤晶宇,等．油茶果预处理装备及发展 趋势[J]．林业机械与木工设备,２０２１,４９(０４):８Ｇ１４． [４]　周敬东,李敏慧,周明刚,等．油茶果的模糊聚类色选算 法[J]．中国农机化学报,２０１５,３６(０４):９４Ｇ９９． [５]　周敬东,黄云朋,李敏慧,等．一种基于 CCD 相机的通 道式油茶果色选机系统[J]．中国农机化学报,２０１５,３６ (０５):１２８Ｇ１３３． [６]　赵志衡,宋欢,朱江波,等．基于卷积神经网络的花生籽 粒完整性识别算法及应用[J]．农业工程学报,２０１８,３４ (２１):１９５Ｇ２０１． [７]　吕梦棋,张芮祥,贾浩,等．基于改进 ResNet玉米种子 分类方法 研 究 [J]．中 国 农 机 化 学 报,２０２１,４２(０４): ９２Ｇ９８． [８]　林丽惠,魏毅,潘俊虹．基于卷积神经网络的武夷岩茶 叶片分类方法[J]．宁德师范学院学报(自然科学版), ２０２１,３３(０４):３６３Ｇ３６９． [９]　谢为俊,丁冶春,王凤贺,等．基于卷积神经网络的油茶 籽完整性识别方法[J]．农业机械学报,２０２０,５１(０７): １３Ｇ２１． [１０]赵洋,梁迎春,许军,等．改进 ResNet１８网络模型的花卉 识别[J]．计算机技术与发展,２０２２,３２(０７):１６７Ｇ１７２． [１１]张怡,赵珠蒙,王校常,等．基于 ResNet卷积神经网络的 绿茶种类识别模型构建[J]．茶叶科学,２０２１,４１(０２): ２６１Ｇ２７１． [１２]HEK,ZHANGX,RENS,etal．DeepresiduallearnＧ ingforimagerecognition[C]．２０１６IEEEConferenceon ComputerVisionand Pattern Recognition (CVPR)． IEEE,２０１６． [１３]SIMONYAN K,ZISSERMAN A．VerydeepconvoluＧ tionalnetworksforlargeＧscaleimagerecognition．[J] ComputerScience,２０１４． [１４]KRIZHEVSKY,ALEX,SUTSKEVER,ILYA,HINＧ TON,GEOFFREY E．ImageNetclassification with deepconvolutionalneuralnetworks[J]．CommunicaＧ tionsoftheACM,２０１７,６０(０６):８４Ｇ９０． [１５]CHAO NI．Automaticinspection machineformaize kernelsbasedondeepconvolutionalneuralnetworks [J]．BiosystemsEngineering,２０１９,１７８ :１３１Ｇ１４４． [１６]ANDREASKAMILARIS,FRANCESCX,PRENAFEＧ TAＧBOLDU′ ．Deeplearninginagriculture:Asurvey [J]．ComputersandElectronicsinAgriculture,２０１８, １４７ :７０Ｇ９０． [１７]IOFFES,SZEGEDYC．Batchnormalization:AccelerＧ atingdeepnetworktrainingbyreducinginternalcovariＧ ateshift[C]．Proceedingsofthe３２ndInternational ConferenceonMachineLearning．２０１５:４４８Ｇ４５６． [１８]朱莉,宋绪秋,邢鑫,等．基于改进 ResNet３４网络的树种 识别 研 究 [J]．国 外 电 子 测 量 技 术,２０２２,４１(０７): １１９Ｇ１２５． [１９]黄毅,段修生,孙世宇,等．基于改进sigmoid激活函数 的深度神经网络训练算法研究[J]．计算机测量与控 制,２０１７,２５(０２):１２６Ｇ１２９．

相似文献/References:

[1]周 浩,周先军,邱书畅.下采样迭代和超分辨率重建的图像风格迁移[J].湖北工业大学学报,2020,(1):25.
　ZHOU Hao,ZHOU Xianjun,QIU Shuchang.Image Style Transfer Algorithm Based on Downsampling Iterative and Super-Resolution Reconstruction[J].,2020,(5):25.
[2]张 宇,程 玉,陈建峡,等. 基于深度学习的在线推荐学习系统设计与开发[J].湖北工业大学学报,2021,(5):64.
　ZHANG Yu,CHENG Yu,CHEN Jianxia,et al. Design and Development of Online Recommendation Learning System Based on Deep Learning[J].,2021,(5):64.
[3]霍 俊,王焱清,孙记委,等.多通道高速刨削式油茶果破壳机设计与试验[J].湖北工业大学学报,2023,(1):1.
　HUO Jun,WANG Yanqing,SUN Jiwei,et al. Design and Experimental Research of Multi-channel Blade High speed Planer for Camellia Oleifera Shelling[J].,2023,(5):1.
[4]武明虎,雷常鼎,刘 聪. 改进的轻量级YOLOv4输电线路鸟巢检测方法[J].湖北工业大学学报,2023,(2):11.
　WU Minghu,LEI Changding,LIU Cong. Improved YOLOv4 Transmission Lines Bird’s Nest Detection Method[J].,2023,(5):11.
[5]卢周成１,王焱清１,２,等.油茶果园施肥一体机的设计与试验[J].湖北工业大学学报,2024,39(2):70.
　LU Zhoucheng,WANG Yanqing,ZHENG Tuo.Design and Experiment of Integrated Fertilizer Machine ]for Camellia Oleifera Orchard[J].,2024,39(5):70.

备注/Memo

备注/Memo:
收稿日期 ２０２２ -１２ -０９ [基金项目]湖北省重点研发计划项目(２０２０BBA０４２) [第一作者]董　庚(１９９４－),男,河南南阳人,湖北工业大学硕士研究生,研究方向为农产品智能化无损检测技术. [通信作者]王焱清(１９６３－),男,湖北鄂州人,工学博士,湖北工业大学教授,研究方向为农业机械化及自动化.
更新日期/Last Update: 2023-10-24