基于改进EfficientNet的化纤色泽品质分类系统
罗彬豪，王杰*
（四川大学 机械工程学院，四川 成都 610065）
摘要：化纤长丝是常见的纺织品原料之一，为保证化纤丝筒产品质量，必须在出厂前对其进行全面检测。对此，设计了一套化纤丝筒外观品质分类系统，可以实现对化纤丝筒的快速检测。首先对实验平台进行建模并完成搭建，然后实地采集四类化纤丝筒照片，并利用数据增强制成化纤色泽品质分类数据集。之后，为更好地融合化纤图片的多尺度信息，引入PANet结构代替FPN层。此外，为让网络模型在多尺度融合时关注重点特征，在PANet中引入PSA注意力机制。实验结果表明，该方法在当前数据集上达到99.87%的准确率，与ResNet，VGG，DenseNet等网络相比在精确率和召回率上均有提升，能有效完成化纤色泽的分类判断。
关键词：化纤；检测；注意力机制；数据增强；EfficientNet
中图分类号：TH18；TS151.9 文献标志码：A doi：10.3969/j.issn.1006-0316.2024.02.007
文章编号：1006-0316 (2024) 02-0039-07
Color Quality Classification System for Chemical Fibers Based on Improved EfficientNet
LUO Binhao，WANG Jie
( School of Mechanical Engineering, Sichuan University, Chengdu 610065, China )
Abstract：Chemical filament long silk is one of the common textile raw materials. In order to ensure the quality of chemical filament spools, a thorough inspection must be carried out before leaving the factory. In this paper, a system for classifying the appearance quality of the chemical filament spools is designed, which enables rapid detection of the chemical filament spools. First, the experimental platform is modeled and constructed, and then four types of chemical filament spool photos are collected in the field and converted into a chemical color quality classification dataset using data augmentation. Furthermore, to better integrate the multiscale information of chemical images, the PANet structure is introduced to replace the FPN layer. Additionally, to ensure that the network model focused on the key features during the multiscale fusion, the PSA attention mechanism is introduced into the PANet. The experimental results show that this method achieved an accuracy of 99.87% on the current dataset with improvements in precision and recall compared with ResNet, VGG, and DenseNet, and can effectively classify the chemical color quality.
Key words：chemical fiber；testing；attention mechanism；data enhancement；EfficientNet
———————————————
收稿日期：2023-10-08
基金项目：四川省重点研发项目（2022YFG0058）
作者简介：罗彬豪（1999－），男，四川德阳人，硕士研究生，主要研究方向为故障诊断，E-mail：920356159@qq.com。*通讯作者：王杰（1964－），男，四川成都人，博士，教授，主要研究方向为机械制造和机器视觉等，E-mail：wangjie@scu.edu.cn。
（四川大学 机械工程学院，四川 成都 610065）
摘要：化纤长丝是常见的纺织品原料之一，为保证化纤丝筒产品质量，必须在出厂前对其进行全面检测。对此，设计了一套化纤丝筒外观品质分类系统，可以实现对化纤丝筒的快速检测。首先对实验平台进行建模并完成搭建，然后实地采集四类化纤丝筒照片，并利用数据增强制成化纤色泽品质分类数据集。之后，为更好地融合化纤图片的多尺度信息，引入PANet结构代替FPN层。此外，为让网络模型在多尺度融合时关注重点特征，在PANet中引入PSA注意力机制。实验结果表明，该方法在当前数据集上达到99.87%的准确率，与ResNet，VGG，DenseNet等网络相比在精确率和召回率上均有提升，能有效完成化纤色泽的分类判断。
关键词：化纤；检测；注意力机制；数据增强；EfficientNet
中图分类号：TH18；TS151.9 文献标志码：A doi：10.3969/j.issn.1006-0316.2024.02.007
文章编号：1006-0316 (2024) 02-0039-07
Color Quality Classification System for Chemical Fibers Based on Improved EfficientNet
LUO Binhao，WANG Jie
( School of Mechanical Engineering, Sichuan University, Chengdu 610065, China )
Abstract：Chemical filament long silk is one of the common textile raw materials. In order to ensure the quality of chemical filament spools, a thorough inspection must be carried out before leaving the factory. In this paper, a system for classifying the appearance quality of the chemical filament spools is designed, which enables rapid detection of the chemical filament spools. First, the experimental platform is modeled and constructed, and then four types of chemical filament spool photos are collected in the field and converted into a chemical color quality classification dataset using data augmentation. Furthermore, to better integrate the multiscale information of chemical images, the PANet structure is introduced to replace the FPN layer. Additionally, to ensure that the network model focused on the key features during the multiscale fusion, the PSA attention mechanism is introduced into the PANet. The experimental results show that this method achieved an accuracy of 99.87% on the current dataset with improvements in precision and recall compared with ResNet, VGG, and DenseNet, and can effectively classify the chemical color quality.
Key words：chemical fiber；testing；attention mechanism；data enhancement；EfficientNet
———————————————
收稿日期：2023-10-08
基金项目：四川省重点研发项目（2022YFG0058）
作者简介：罗彬豪（1999－），男，四川德阳人，硕士研究生，主要研究方向为故障诊断，E-mail：920356159@qq.com。*通讯作者：王杰（1964－），男，四川成都人，博士，教授，主要研究方向为机械制造和机器视觉等，E-mail：wangjie@scu.edu.cn。
- 上一篇：一种防风固沙植被培养机械的设计
- 下一篇：含膨胀元件的接触网与受电弓动力学特性分析
设为首页 | 加入收藏 | 免责条款
《机械》杂志版权所有 Copyright©2008-2012 Jixiezazhi.com All Rights Reserved
电话：028-85925070 传真：028-85925073 E-mail：jixie@vip.163.com
地址：四川省成都锦江工业开发区墨香路48号 邮编：610063
Powered by PageAdmin CMS