Article(id=1149789684961862194, tenantId=1146029695717560320, journalId=1146123166801305609, issueId=1149768563956826506, articleNumber=null, orderNo=null, doi=10.12404/j.issn.1671-1815.2404359, pmid=null, cstr=null, oa=null, hot=null, price=null, onlineType=0, articleFormat=0, articleType=null, articleTypeStr=null, receivedDate=1717603200000, receivedDateStr=2024-06-06, revisedDate=1741017600000, revisedDateStr=2025-03-04, acceptedDate=null, acceptedDateStr=null, onlineDate=1752060822954, onlineDateStr=2025-07-09, pubDate=1749312000000, pubDateStr=2025-06-08, doiRegisterDate=null, doiRegisterDateStr=null, onlineIssueDate=1752060822954, onlineIssueDateStr=2025-07-09, onlineJustAcceptDate=null, onlineJustAcceptDateStr=null, onlineFirstDate=null, onlineFirstDateStr=null, sourceXml=null, magXml=null, createTime=1752060822954, creator=13701087609, updateTime=1752060822954, updator=13701087609, issue=Issue{id=1149768563956826506, tenantId=1146029695717560320, journalId=1146123166801305609, year='2025', volume='25', issue='16', pageStart='6587', pageEnd='7021', issueExtLink='null', onlineDate='null', pubDate='null', beforeIssueId=null, nextIssueId=null, price=null, status=1, issueComplete=1, articleOrder=1, issueType=-1, specialIssue=0, createTime=1752055787314, creator=13701087609, updateTime=1768456850262, updator=13701087609, preIssue=null, nextIssue=null, ext={EN=IssueExt(id=1218559607937618069, tenantId=1146029695717560320, journalId=1146123166801305609, issueId=1149768563956826506, language=EN, specialIssueTitle=, coverIllustrator=, specialIssueEditor=, specialIssueAbout=), CN=IssueExt(id=1218559607937618070, tenantId=1146029695717560320, journalId=1146123166801305609, issueId=1149768563956826506, language=CN, specialIssueTitle=, coverIllustrator=, specialIssueEditor=, specialIssueAbout=)}, issueFiles=null}, startPage=6804, endPage=6811, ext={EN=ArticleExt(id=1149789685167383091, articleId=1149789684961862194, tenantId=1146029695717560320, journalId=1146123166801305609, language=EN, title=A Rapid Detection Method for Dust Deposition Based on Fast-UNet, columnId=1156262729162810294, journalTitle=Science Technology and Engineering, columnName=Papers·Automation and Computational Technology, runingTitle=null, highlight=null, articleAbstract=

Dust deposition can affect the normal operation of equipment. To accurately and efficiently detect dust on equipment and formulate a scientific cleaning strategy, a lightweight dust deposition detection method based on Fast-UNet was proposed. By effectively pruning UNet and adopting max pooling and bilinear interpolation for down-sampling and up-sampling operations, the parameter redundancy was reduced, and a compact basic network was obtained. The lightweight Ghost Module was used to replace the ordinary convolution in the basic network, further reducing the complexity of the network. An convolutional block attention module(CBAM) that integrated channel and spatial attention was embedded in the encoding process, which made the network pay more attention to the target area while introducing minimal parameters. Experiments on a simulated dust deposition dataset show that, compared with the original model, Fast-UNet reduces the number of parameters by 99.6%, decreases computational complexity by 98.7%, achieves an inference speed of 94.18 frames per second, and maintains a recognition accuracy of 91.17%. Compared with five other mainstream segmentation models, Fast-UNet also demonstrates advantages in both accuracy and speed. This method meets the needs of dust detection for both accuracy and efficiency, providing a technical reference for dust quantitative analysis.

, correspAuthors=Zhang-yan ZHAO, authorNote=null, correspAuthorsNote=null, copyrightStatement=null, copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=null, magXml=null, pdfUrl=null, pdf=null, pdfFileSize=null, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=null, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=null, mapNumber=null, authorCompany=null, fund=null, authors=null, authorsList=Yu-hui YU, Yu WANG, Ting-hui GAO, Cheng-hua ZHANG, Zhang-yan ZHAO), CN=ArticleExt(id=1149789705010635574, articleId=1149789684961862194, tenantId=1146029695717560320, journalId=1146123166801305609, language=CN, title=基于Fast-UNet的灰尘沉积快速检测方法, columnId=1156262729783567290, journalTitle=科学技术与工程, columnName=论文·自动化技术、计算机技术, runingTitle=null, highlight=null, articleAbstract=

灰尘沉积会影响设备的正常运行。为准确且高效地对设备灰尘进行检测,制定科学的清洁策略,提出一种基于Fast-UNet的轻量化灰尘沉积检测方法。通过对UNet进行合理剪枝,采用最大池化与双线性插值进行下采样与上采样操作,减少了参数冗余,得到了结构紧凑的基础网络;使用轻量化的幽灵卷积模块(Ghost Module)替换基础网络中的普通卷积,进一步降低了网络的复杂度;在编码过程中嵌入融合了通道和空间注意力的CBAM(convolutional block attention module),在引入极少参数量的同时,使网络更加关注于目标区域。在模拟灰尘沉积的数据集上进行实验,结果表明:Fast-UNet与原模型相比,参数量减少了99.6%,计算量减少了98.7%,推理速度可达94.18帧/s,且识别精度高达91.17%。与其他5种主流分割模型相比,Fast-UNet在精度与速度上均具有优势。所提方法能够满足灰尘检测对检测精度与效率的需求,为灰尘定量分析提供技术参考。

, correspAuthors=赵章焰, authorNote=null, correspAuthorsNote=
* 赵章焰(1963―),男,汉族,湖北武汉人,博士,教授。研究方向:摄影测量、机器视觉、金属结构设计与仿真、攀爬机器人设计与实现。E-mail:
, copyrightStatement=null, copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=f7t8dNrUV+mjNnIuImIg0A==, magXml=miQ9TYBug6ki70w2IjLIAg==, pdfUrl=null, pdf=zWYd3vbFCIhOKMKj24Ivtw==, pdfFileSize=null, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=null, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=sD+LxCuLyHuJsCuYgW7WfA==, mapNumber=null, authorCompany=null, fund=null, authors=

尉宇晖(1998―),男,汉族,山西大同人,硕士研究生。研究方向:机器视觉、图像处理。E-mail:

, authorsList=尉宇晖, 王煜, 高庭辉, 张城华, 赵章焰)}, authors=[Author(id=1179140096839266702, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, orderNo=0, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=yyh15513778898@whut.edu.cn, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1179140096914764176, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, authorId=1179140096839266702, language=EN, stringName=Yu-hui YU, firstName=Yu-hui, middleName=null, lastName=YU, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=School of Transportation and Logistics Engineering, Wuhan University of Technology, Wuhan 430063, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1179140096998650257, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, authorId=1179140096839266702, language=CN, stringName=尉宇晖, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=武汉理工大学交通与物流工程学院, 武汉 430063, bio={"content":"

尉宇晖(1998―),男,汉族,山西大同人,硕士研究生。研究方向:机器视觉、图像处理。E-mail:

"}, bioImg=null, bioContent=

尉宇晖(1998―),男,汉族,山西大同人,硕士研究生。研究方向:机器视觉、图像处理。E-mail:

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1179140096755380618, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, xref=null, ext=[AuthorCompanyExt(id=1179140096763769227, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, companyId=1179140096755380618, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=School of Transportation and Logistics Engineering, Wuhan University of Technology, Wuhan 430063, China), AuthorCompanyExt(id=1179140096772157836, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, companyId=1179140096755380618, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=武汉理工大学交通与物流工程学院, 武汉 430063)])]), Author(id=1179140097065759123, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, orderNo=1, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1179140097153839509, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, authorId=1179140097065759123, language=EN, stringName=Yu WANG, firstName=Yu, middleName=null, lastName=WANG, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=School of Transportation and Logistics Engineering, Wuhan University of Technology, Wuhan 430063, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1179140097220948374, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, authorId=1179140097065759123, language=CN, stringName=王煜, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=武汉理工大学交通与物流工程学院, 武汉 430063, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1179140096755380618, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, xref=null, ext=[AuthorCompanyExt(id=1179140096763769227, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, companyId=1179140096755380618, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=School of Transportation and Logistics Engineering, Wuhan University of Technology, Wuhan 430063, China), AuthorCompanyExt(id=1179140096772157836, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, companyId=1179140096755380618, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=武汉理工大学交通与物流工程学院, 武汉 430063)])]), Author(id=1179140097283862936, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, orderNo=2, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1179140097371943322, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, authorId=1179140097283862936, language=EN, stringName=Ting-hui GAO, firstName=Ting-hui, middleName=null, lastName=GAO, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=School of Transportation and Logistics Engineering, Wuhan University of Technology, Wuhan 430063, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1179140097439052187, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, authorId=1179140097283862936, language=CN, stringName=高庭辉, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=武汉理工大学交通与物流工程学院, 武汉 430063, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1179140096755380618, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, xref=null, ext=[AuthorCompanyExt(id=1179140096763769227, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, companyId=1179140096755380618, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=School of Transportation and Logistics Engineering, Wuhan University of Technology, Wuhan 430063, China), AuthorCompanyExt(id=1179140096772157836, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, companyId=1179140096755380618, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=武汉理工大学交通与物流工程学院, 武汉 430063)])]), Author(id=1179140097510355357, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, orderNo=3, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1179140097577464223, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, authorId=1179140097510355357, language=EN, stringName=Cheng-hua ZHANG, firstName=Cheng-hua, middleName=null, lastName=ZHANG, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=School of Transportation and Logistics Engineering, Wuhan University of Technology, Wuhan 430063, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1179140097631990176, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, authorId=1179140097510355357, language=CN, stringName=张城华, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=武汉理工大学交通与物流工程学院, 武汉 430063, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1179140096755380618, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, xref=null, ext=[AuthorCompanyExt(id=1179140096763769227, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, companyId=1179140096755380618, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=School of Transportation and Logistics Engineering, Wuhan University of Technology, Wuhan 430063, China), AuthorCompanyExt(id=1179140096772157836, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, companyId=1179140096755380618, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=武汉理工大学交通与物流工程学院, 武汉 430063)])]), Author(id=1179140097694904738, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, orderNo=4, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=zzy63277@163.com, emailSecond=null, emailThird=null, correspondingAuthor=1, authorType=1, ext={EN=AuthorExt(id=1179140097774596516, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, authorId=1179140097694904738, language=EN, stringName=Zhang-yan ZHAO, firstName=Zhang-yan, middleName=null, lastName=ZHAO, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=*, address=School of Transportation and Logistics Engineering, Wuhan University of Technology, Wuhan 430063, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1179140097837511077, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, authorId=1179140097694904738, language=CN, stringName=赵章焰, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=*, address=武汉理工大学交通与物流工程学院, 武汉 430063, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1179140096755380618, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, xref=null, ext=[AuthorCompanyExt(id=1179140096763769227, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, companyId=1179140096755380618, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=School of Transportation and Logistics Engineering, Wuhan University of Technology, Wuhan 430063, China), AuthorCompanyExt(id=1179140096772157836, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, companyId=1179140096755380618, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=武汉理工大学交通与物流工程学院, 武汉 430063)])])], keywords=[Keyword(id=1179140097980117414, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, language=EN, orderNo=1, keyword=information processing), Keyword(id=1179140098055614887, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, language=EN, orderNo=2, keyword=dust deposition), Keyword(id=1179140098114335144, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, language=EN, orderNo=3, keyword=attention mechanism), Keyword(id=1179140098168861097, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, language=EN, orderNo=4, keyword=lightweight model), Keyword(id=1179140098231775658, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, language=EN, orderNo=5, keyword=UNet), Keyword(id=1179140098298884523, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, language=CN, orderNo=1, keyword=信息处理), Keyword(id=1179140098370187692, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, language=CN, orderNo=2, keyword=灰尘沉积), Keyword(id=1179140098437296557, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, language=CN, orderNo=3, keyword=注意力机制), Keyword(id=1179140098500211118, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, language=CN, orderNo=4, keyword=轻量化模型), Keyword(id=1179140098558931375, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, language=CN, orderNo=5, keyword=UNet)], refs=[Reference(id=1179140100467339720, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, doi=null, pmid=null, pmcid=null, year=2023, volume=54, issue=12, pageStart=4893, pageEnd=4900, url=null, language=null, rfNumber=[1], rfOrder=0, authorNames=汪峰, 王志豪, 张小松, journalName=中南大学学报(自然科学版), refType=null, unstructuredReference=汪峰, 王志豪, 张小松, 等. 换热器翅片表面灰尘沉积特性的实验研究[J]. 中南大学学报(自然科学版), 2023, 54(12): 4893-4900., articleTitle=换热器翅片表面灰尘沉积特性的实验研究, refAbstract=null), Reference(id=1179140100538642889, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, doi=null, pmid=null, pmcid=null, year=2023, volume=54, issue=12, pageStart=4893, pageEnd=4900, url=null, language=null, rfNumber=[1], rfOrder=1, authorNames=Wang Feng, Wang Zhihao, Zhang Xiaosong, journalName=Journal of Central South University(Science and Technology), refType=null, unstructuredReference=Wang Feng, Wang Zhihao, Zhang Xiaosong, et al. Experimental study on characteristics of dust deposition on fin surface of heat exchanger[J]. Journal of Central South University(Science and Technology), 2023, 54(12): 4893-4900., articleTitle=Experimental study on characteristics of dust deposition on fin surface of heat exchanger, refAbstract=null), Reference(id=1179140100609946058, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, doi=null, pmid=null, pmcid=null, year=2021, volume=50, issue=12, pageStart=246, pageEnd=252, url=null, language=null, rfNumber=[2], rfOrder=2, authorNames=牛龙飞, 蒋一岚, 苗心向, journalName=红外与激光工程, refType=null, unstructuredReference=牛龙飞, 蒋一岚, 苗心向, 等. 大口径反射镜表面颗粒物洁净度控制实验研究[J]. 红外与激光工程, 2021, 50(12): 246-252., articleTitle=大口径反射镜表面颗粒物洁净度控制实验研究, refAbstract=null), Reference(id=1179140100677054923, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, doi=null, pmid=null, pmcid=null, year=2021, volume=50, issue=12, pageStart=246, pageEnd=252, url=null, language=null, rfNumber=[2], rfOrder=3, authorNames=Niu Longfei, Jiang Yilan, Miao Xinxiang, journalName=Infrared and Laser Engineering, refType=null, unstructuredReference=Niu Longfei, Jiang Yilan, Miao Xinxiang, et al. Experimental study of particles cleanliness control on the surface of large-aperture reflector[J]. Infrared and Laser Engineering, 2021, 50(12): 246-252., articleTitle=Experimental study of particles cleanliness control on the surface of large-aperture reflector, refAbstract=null), Reference(id=1179140100781912524, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, doi=null, pmid=null, pmcid=null, year=2024, volume=45, issue=4, pageStart=460, pageEnd=467, url=null, language=null, rfNumber=[3], rfOrder=4, authorNames=姚磊, 付豪, 马科飞, journalName=太阳能学报, refType=null, unstructuredReference=姚磊, 付豪, 马科飞, 等. 高速公路边坡光伏组件表面灰尘粘附性能研究[J]. 太阳能学报, 2024, 45(4): 460-467., articleTitle=高速公路边坡光伏组件表面灰尘粘附性能研究, refAbstract=null), Reference(id=1179140100857409997, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, doi=null, pmid=null, pmcid=null, year=2024, volume=45, issue=4, pageStart=460, pageEnd=467, url=null, language=null, rfNumber=[3], rfOrder=5, authorNames=Yao Lei, Fu Hao, Ma Kefei, journalName=Acta Energiae Solaris Sinica, refType=null, unstructuredReference=Yao Lei, Fu Hao, Ma Kefei, et al. Performance study of dust adhesion of PV module surfaces on highway slope[J]. Acta Energiae Solaris Sinica, 2024, 45(4): 460-467., articleTitle=Performance study of dust adhesion of PV module surfaces on highway slope, refAbstract=null), Reference(id=1179140100928713166, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, doi=null, pmid=null, pmcid=null, year=2021, volume=42, issue=3, pageStart=479, pageEnd=489, url=null, language=null, rfNumber=[4], rfOrder=6, authorNames=雷炳海, 杨志刚, 金哲岩, journalName=力学季刊, refType=null, unstructuredReference=雷炳海, 杨志刚, 金哲岩. 风洞中灰尘在不同特性玻璃表面上粘附形貌和粘附程度的实验研究[J]. 力学季刊, 2021, 42(3): 479-489., articleTitle=风洞中灰尘在不同特性玻璃表面上粘附形貌和粘附程度的实验研究, refAbstract=null), Reference(id=1179140101004210639, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, doi=null, pmid=null, pmcid=null, year=2021, volume=42, issue=3, pageStart=479, pageEnd=489, url=null, language=null, rfNumber=[4], rfOrder=7, authorNames=Lei Binghai, Yang Zhigang, Jin Zheyan, journalName=Chinese Quarterly of Mechanics, refType=null, unstructuredReference=Lei Binghai, Yang Zhigang, Jin Zheyan. Experimental study of adhesion morphology features and the degree of adhesion of dusts on glass surfaces of different characteristics in a wind tunnel[J]. Chinese Quarterly of Mechanics, 2021, 42(3): 479-489., articleTitle=Experimental study of adhesion morphology features and the degree of adhesion of dusts on glass surfaces of different characteristics in a wind tunnel, refAbstract=null), Reference(id=1179140101075513808, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, doi=null, pmid=null, pmcid=null, year=2020, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=[5], rfOrder=8, authorNames=张萌, journalName=掩膜版表面灰尘杂质检测系统设计, refType=null, unstructuredReference=张萌. 掩膜版表面灰尘杂质检测系统设计[D]. 西安: 西安石油大学, 2020., articleTitle=null, refAbstract=null), Reference(id=1179140101138428369, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, doi=null, pmid=null, pmcid=null, year=2020, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=[5], rfOrder=9, authorNames=Zhang Meng, journalName=The design of a dust and impurity detection system for mask surface, refType=null, unstructuredReference=Zhang Meng. The design of a dust and impurity detection system for mask surface[D]. Xi'an: Xi'an Shiyou University, 2020., articleTitle=null, refAbstract=null), Reference(id=1179140101201342930, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, doi=null, pmid=null, pmcid=null, year=2023, volume=21, issue=2, pageStart=51, pageEnd=57, url=null, language=null, rfNumber=[6], rfOrder=10, authorNames=王化超, 高爱华, 秦文罡, journalName=光学与光电技术, refType=null, unstructuredReference=王化超, 高爱华, 秦文罡, 等. 光伏板积尘散射检测中的去噪方法研究[J]. 光学与光电技术, 2023, 21(2): 51-57, 77., articleTitle=光伏板积尘散射检测中的去噪方法研究, refAbstract=null), Reference(id=1179140101268451795, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, doi=null, pmid=null, pmcid=null, year=2023, volume=21, issue=2, pageStart=51, pageEnd=57, url=null, language=null, rfNumber=[6], rfOrder=11, authorNames=Wang Huachao, Gao Aihua, Qin Wengang, journalName=Optics & Optoelectronic Technology, refType=null, unstructuredReference=Wang Huachao, Gao Aihua, Qin Wengang, et al. Denoising method in dust scattering detection of photovoltaic panels[J]. Optics & Optoelectronic Technology, 2023, 21(2): 51-57, 77., articleTitle=Denoising method in dust scattering detection of photovoltaic panels, refAbstract=null), Reference(id=1179140101331366356, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, doi=null, pmid=null, pmcid=null, year=2015, volume=null, issue=null, pageStart=234, pageEnd=241, url=null, language=null, rfNumber=[7], rfOrder=12, authorNames=Ronneberger O, Fischer P, Brox T, journalName=18th International Conference on Medical Image Computing and Computer-Assisted Intervention, refType=null, unstructuredReference=Ronneberger O, Fischer P, Brox T. U-net: convolutional networks for biomedical image segmentation[C]// 18th International Conference on Medical Image Computing and Computer-Assisted Intervention. Munich: Springer International Publishing, 2015: 234-241., articleTitle=U-net: convolutional networks for biomedical image segmentation, refAbstract=null), Reference(id=1179140101390086613, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, doi=null, pmid=null, pmcid=null, year=2017, volume=39, issue=12, pageStart=2481, pageEnd=2495, url=null, language=null, rfNumber=[8], rfOrder=13, authorNames=Badrinarayanan V, Kendall A, Cipolla R, journalName=IEEE Transactions on Pattern Analysis and Machine Intelligence, refType=null, unstructuredReference=Badrinarayanan V, Kendall A, Cipolla R. Segnet: A deep convolutional encoder-decoder architecture for image segmentation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(12): 2481-2495., articleTitle=Segnet: A deep convolutional encoder-decoder architecture for image segmentation, refAbstract=null), Reference(id=1179140101457195478, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, doi=null, pmid=null, pmcid=null, year=2016, volume=null, issue=null, pageStart=770, pageEnd=778, url=null, language=null, rfNumber=[9], rfOrder=14, authorNames=He K, Zhang X, Ren S, journalName=Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, refType=null, unstructuredReference=He K, Zhang X, Ren S, et al. Deep residual learning for image recognition[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2016: 770-778., articleTitle=Deep residual learning for image recognition, refAbstract=null), Reference(id=1179140101520110039, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, doi=null, pmid=null, pmcid=null, year=2022, volume=43, issue=8, pageStart=2252, pageEnd=2259, url=null, language=null, rfNumber=[10], rfOrder=15, authorNames=李颖, 李秀宇, 卢兆林, journalName=计算机工程与设计, refType=null, unstructuredReference=李颖, 李秀宇, 卢兆林, 等. 基于深度学习的煤粉颗粒CT图像分割方法[J]. 计算机工程与设计, 2022, 43(8): 2252-2259., articleTitle=基于深度学习的煤粉颗粒CT图像分割方法, refAbstract=null), Reference(id=1179140101583024600, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, doi=null, pmid=null, pmcid=null, year=2022, volume=43, issue=8, pageStart=2252, pageEnd=2259, url=null, language=null, rfNumber=[10], rfOrder=16, authorNames=Li Ying, Li Xiuyu, Lu Zhaolin, journalName=Computer Engineering and Design, refType=null, unstructuredReference=Li Ying, Li Xiuyu, Lu Zhaolin, et al. Coal particle CT image segmentation method based on deep learning[J]. Computer Engineering and Design, 2022, 43(8): 2252-2259., articleTitle=Coal particle CT image segmentation method based on deep learning, refAbstract=null), Reference(id=1179140101641744857, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, doi=null, pmid=null, pmcid=null, year=2024, volume=24, issue=24, pageStart=10362, pageEnd=10369, url=null, language=null, rfNumber=[11], rfOrder=17, authorNames=汪南洋, 沈疆海, journalName=科学技术与工程, refType=null, unstructuredReference=汪南洋, 沈疆海. 基于MSHAM-UNet的岩心孔洞图像分割方法[J]. 科学技术与工程, 2024, 24(24): 10362-10369., articleTitle=基于MSHAM-UNet的岩心孔洞图像分割方法, refAbstract=null), Reference(id=1179140101713048026, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, doi=null, pmid=null, pmcid=null, year=2024, volume=24, issue=24, pageStart=10362, pageEnd=10369, url=null, language=null, rfNumber=[11], rfOrder=18, authorNames=Wang Nanyang, Shen Jianghai, journalName=Science Technology and Engineering, refType=null, unstructuredReference=Wang Nanyang, Shen Jianghai. Image segmentation method of rock core hole based on MSHAM-UNet[J]. Science Technology and Engineering, 2024, 24(24): 10362-10369., articleTitle=Image segmentation method of rock core hole based on MSHAM-UNet, refAbstract=null), Reference(id=1179140101763379675, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, doi=null, pmid=null, pmcid=null, year=2022, volume=22, issue=29, pageStart=12927, pageEnd=12935, url=null, language=null, rfNumber=[12], rfOrder=19, authorNames=覃本学, 沈疆海, 马丙鹏, journalName=科学技术与工程, refType=null, unstructuredReference=覃本学, 沈疆海, 马丙鹏, 等. 基于Debseg-Net的岩屑图像语义分割[J]. 科学技术与工程, 2022, 22(29): 12927-12935., articleTitle=基于Debseg-Net的岩屑图像语义分割, refAbstract=null), Reference(id=1179140101822099932, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, doi=null, pmid=null, pmcid=null, year=2022, volume=22, issue=29, pageStart=12927, pageEnd=12935, url=null, language=null, rfNumber=[12], rfOrder=20, authorNames=Qin Benxue, Shen Jianghai, Ma Bingpeng, journalName=Science Technology and Engineering, refType=null, unstructuredReference=Qin Benxue, Shen Jianghai, Ma Bingpeng, et al. Semantic segmentation of rock debris image based on Debseg-Net[J]. Science Technology and Engineering, 2022, 22(29): 12927-12935., articleTitle=Semantic segmentation of rock debris image based on Debseg-Net, refAbstract=null), Reference(id=1179140101914374621, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, doi=null, pmid=null, pmcid=null, year=2024, volume=24, issue=2, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=[13], rfOrder=21, authorNames=De Silva A, Ranasinghe R, Sounthararajah A, journalName=Sensors, refType=null, unstructuredReference=De Silva A, Ranasinghe R, Sounthararajah A, et al. Beyond conventional monitoring: a semantic segmentation approach to quantifying traffic-induced dust on unsealed roads[J]. Sensors, 2024, 24(2): DOI: 10.3390/s24020510., articleTitle=Beyond conventional monitoring: a semantic segmentation approach to quantifying traffic-induced dust on unsealed roads, refAbstract=null), Reference(id=1179140101973094878, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, doi=null, pmid=null, pmcid=null, year=2023, volume=32, issue=8, pageStart=295, pageEnd=302, url=null, language=null, rfNumber=[14], rfOrder=22, authorNames=徐小平, 张勇, 刘广钧, journalName=计算机系统应用, refType=null, unstructuredReference=徐小平, 张勇, 刘广钧, 等. 基于改进ShuffleNetV2模型的光伏板灰尘识别[J]. 计算机系统应用, 2023, 32(8): 295-302., articleTitle=基于改进ShuffleNetV2模型的光伏板灰尘识别, refAbstract=null), Reference(id=1179140102031815135, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, doi=null, pmid=null, pmcid=null, year=2023, volume=32, issue=8, pageStart=295, pageEnd=302, url=null, language=null, rfNumber=[14], rfOrder=23, authorNames=Xu Xiaoping, Zhang Yong, Liu Guangjun, journalName=Computer Systems and Applications, refType=null, unstructuredReference=Xu Xiaoping, Zhang Yong, Liu Guangjun, et al. Identification of dust on photovoltaic panel based on improved ShuffleNetV2 model[J]. Computer Systems and Applications, 2023, 32(8): 295-302., articleTitle=Identification of dust on photovoltaic panel based on improved ShuffleNetV2 model, refAbstract=null), Reference(id=1179140102094729696, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, doi=null, pmid=null, pmcid=null, year=2020, volume=null, issue=null, pageStart=1580, pageEnd=1589, url=null, language=null, rfNumber=[15], rfOrder=24, authorNames=Han K, Wang Y, Tian Q, journalName=Proceedings of the IEEE/CVF conference on Computer Vision and Pattern Recognition, refType=null, unstructuredReference=Han K, Wang Y, Tian Q, et al. Ghostnet: more features from cheap operations[C]// Proceedings of the IEEE/CVF conference on Computer Vision and Pattern Recognition. New York: IEEE, 2020: 1580-1589., articleTitle=Ghostnet: more features from cheap operations, refAbstract=null), Reference(id=1179140102153449953, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, doi=null, pmid=null, pmcid=null, year=2018, volume=null, issue=null, pageStart=3, pageEnd=19, url=null, language=null, rfNumber=[16], rfOrder=25, authorNames=Woo S, Park J, Lee J Y, journalName=Proceedings of the European Conference on Computer Vision(ECCV), refType=null, unstructuredReference=Woo S, Park J, Lee J Y, et al. CBAM: convolutional block attention module[C]// Proceedings of the European Conference on Computer Vision(ECCV). New York: IEEE, 2018: 3-19., articleTitle=CBAM: convolutional block attention module, refAbstract=null), Reference(id=1179140102224753122, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, doi=null, pmid=null, pmcid=null, year=2022, volume=null, issue=null, pageStart=12454, pageEnd=12463, url=null, language=null, rfNumber=[17], rfOrder=26, authorNames=Elkerdawy S, Elhoushi M, Zhang H, journalName=Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, refType=null, unstructuredReference=Elkerdawy S, Elhoushi M, Zhang H, et al. Fire together wire together: a dynamic pruning approach with self-supervised mask prediction[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2022: 12454-12463., articleTitle=Fire together wire together: a dynamic pruning approach with self-supervised mask prediction, refAbstract=null), Reference(id=1179140102300250595, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=[18], rfOrder=27, authorNames=Li X, Sun X, Meng Y, journalName=arXiv Preprint, refType=null, unstructuredReference=Li X, Sun X, Meng Y, et al. Dice loss for data-imbalanced NLP tasks[J]. arXiv Preprint, 2019: arXiv: 1911.02855., articleTitle=Dice loss for data-imbalanced NLP tasks, refAbstract=null)], funds=null, companyList=[AuthorCompany(id=1179140096755380618, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, xref=null, ext=[AuthorCompanyExt(id=1179140096763769227, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, companyId=1179140096755380618, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=School of Transportation and Logistics Engineering, Wuhan University of Technology, Wuhan 430063, China), AuthorCompanyExt(id=1179140096772157836, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, companyId=1179140096755380618, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=武汉理工大学交通与物流工程学院, 武汉 430063)])], figs=[ArticleFig(id=1179140098730897840, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, language=EN, label=Fig.1, caption=Structure diagram of Fast-UNet, figureFileSmall=OjrxNcxOk7dOWjyRwa39gw==, figureFileBig=GnDQnpezPqHX7gKrB/dRWA==, tableContent=null), ArticleFig(id=1179140098798006705, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, language=CN, label=图1, caption=Fast-UNet结构图, figureFileSmall=OjrxNcxOk7dOWjyRwa39gw==, figureFileBig=GnDQnpezPqHX7gKrB/dRWA==, tableContent=null), ArticleFig(id=1179140098856726962, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, language=EN, label=Fig.2, caption=Structure Diagram of Ghost Module, figureFileSmall=GNYZC/5P/XfCLglRwZIEpg==, figureFileBig=cuDISfiUlLLnagAfo7YcfA==, tableContent=null), ArticleFig(id=1179140098923835827, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, language=CN, label=图2, caption=Ghost Module结构图, figureFileSmall=GNYZC/5P/XfCLglRwZIEpg==, figureFileBig=cuDISfiUlLLnagAfo7YcfA==, tableContent=null), ArticleFig(id=1179140098982556084, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, language=EN, label=Fig.3, caption=Structure diagram of CBAM, figureFileSmall=UjFBmkHtvcuJV8ZjFKgffA==, figureFileBig=F0OvaomrNJpLXtAxg/60Gg==, tableContent=null), ArticleFig(id=1179140099045470645, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, language=CN, label=图3, caption=CBAM结构图, figureFileSmall=UjFBmkHtvcuJV8ZjFKgffA==, figureFileBig=F0OvaomrNJpLXtAxg/60Gg==, tableContent=null), ArticleFig(id=1179140099137745334, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, language=EN, label=Fig.4, caption=Microscope and imaging system, figureFileSmall=gzuaEGQtHPPu4PxJ3iwA6A==, figureFileBig=h4jEJcSs9W5M3cMZ4MtZXw==, tableContent=null), ArticleFig(id=1179140099204854199, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, language=CN, label=图4, caption=显微镜与成像系统, figureFileSmall=gzuaEGQtHPPu4PxJ3iwA6A==, figureFileBig=h4jEJcSs9W5M3cMZ4MtZXw==, tableContent=null), ArticleFig(id=1179140099259380152, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, language=EN, label=Fig.5, caption=Ultrasonic dust removal equipment, figureFileSmall=3yAaNF5Rb6TH93PRSDmmFg==, figureFileBig=OcC2JGej72Ots2KKAIcylg==, tableContent=null), ArticleFig(id=1179140099334877625, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, language=CN, label=图5, caption=工业超声波除尘设备, figureFileSmall=3yAaNF5Rb6TH93PRSDmmFg==, figureFileBig=OcC2JGej72Ots2KKAIcylg==, tableContent=null), ArticleFig(id=1179140099397792186, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, language=EN, label=Fig.6, caption=Schematic diagram of dataset acquisition, figureFileSmall=vokZ7Vuh3EINfZ0JGNIFyQ==, figureFileBig=yuZj4xWYgg2BM2KBjxFGSw==, tableContent=null), ArticleFig(id=1179140099464901051, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, language=CN, label=图6, caption=数据集获取示意图, figureFileSmall=vokZ7Vuh3EINfZ0JGNIFyQ==, figureFileBig=yuZj4xWYgg2BM2KBjxFGSw==, tableContent=null), ArticleFig(id=1179140099532009916, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, language=EN, label=Fig.7, caption=Data augmentation effect diagram, figureFileSmall=J0q/SSGEMS4e9uhaHEZqvg==, figureFileBig=tE5TvTvQJ6Kj30xRCR1LKw==, tableContent=null), ArticleFig(id=1179140099594924477, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, language=CN, label=图7, caption=数据增强效果图, figureFileSmall=J0q/SSGEMS4e9uhaHEZqvg==, figureFileBig=tE5TvTvQJ6Kj30xRCR1LKw==, tableContent=null), ArticleFig(id=1179140099657839038, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, language=EN, label=Fig.8, caption=Dust recognition results of different networks, figureFileSmall=C26Mp3vHcSOnRU4crbD0mw==, figureFileBig=ExQxSyu2eBO2lZat81su9g==, tableContent=null), ArticleFig(id=1179140099729142207, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, language=CN, label=图8, caption=不同网络的灰尘识别结果, figureFileSmall=C26Mp3vHcSOnRU4crbD0mw==, figureFileBig=ExQxSyu2eBO2lZat81su9g==, tableContent=null), ArticleFig(id=1179140099842388416, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, language=EN, label=Table 1, caption=

Mass participation coefficient of vibration mode

, figureFileSmall=null, figureFileBig=null, tableContent=
模型 剪枝参数
UNet 原模型
UNet_1 通道数减至 1 2
UNet_2 通道数减至 1 4
UNet_3 减少一层网络
UNet_4 减少一层网络、通道数减至 1 2
UNet_5 减少一层网络、通道数减至 1 4
), ArticleFig(id=1179140099901108673, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, language=CN, label=表1, caption=

剪枝实验设计

, figureFileSmall=null, figureFileBig=null, tableContent=
模型 剪枝参数
UNet 原模型
UNet_1 通道数减至 1 2
UNet_2 通道数减至 1 4
UNet_3 减少一层网络
UNet_4 减少一层网络、通道数减至 1 2
UNet_5 减少一层网络、通道数减至 1 4
), ArticleFig(id=1179140099984994754, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, language=EN, label=Table 2, caption=

Pruning experiment results

, figureFileSmall=null, figureFileBig=null, tableContent=
模型 Params/M IoU/% Precision/% Recall/%
UNet 37.66 83.05 91.22 90.03
UNet _1 8.64 83.36 91.29 90.51
UNet _2 2.16 83.18 91.11 90.44
UNet _3 8.56 83.44 91.07 90.81
UNet _4 2.14 83.40 91.35 90.77
UNet _5 0.54 82.89 90.58 90.71
), ArticleFig(id=1179140100056297923, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, language=CN, label=表2, caption=

剪枝实验结果

, figureFileSmall=null, figureFileBig=null, tableContent=
模型 Params/M IoU/% Precision/% Recall/%
UNet 37.66 83.05 91.22 90.03
UNet _1 8.64 83.36 91.29 90.51
UNet _2 2.16 83.18 91.11 90.44
UNet _3 8.56 83.44 91.07 90.81
UNet _4 2.14 83.40 91.35 90.77
UNet _5 0.54 82.89 90.58 90.71
), ArticleFig(id=1179140100119212484, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, language=EN, label=Table 3, caption=

Ablation experiment results

, figureFileSmall=null, figureFileBig=null, tableContent=
模型 Params/
M
Flops/
G
FPS/
s-1
IoU/
%
Precision/
%
Recall/
%
UNet 37.66 143.93 18.92 83.05 91.22 90.03
UNet_4 2.14 26.44 64.03 83.40 91.35 90.77
UNet_4+Ghost Module 0.13 1.83 97.67 81.86 89.49 89.62
Fast-UNet 0.14 1.86 94.18 83.23 91.17 90.65
), ArticleFig(id=1179140100182127045, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, language=CN, label=表3, caption=

消融实验结果

, figureFileSmall=null, figureFileBig=null, tableContent=
模型 Params/
M
Flops/
G
FPS/
s-1
IoU/
%
Precision/
%
Recall/
%
UNet 37.66 143.93 18.92 83.05 91.22 90.03
UNet_4 2.14 26.44 64.03 83.40 91.35 90.77
UNet_4+Ghost Module 0.13 1.83 97.67 81.86 89.49 89.62
Fast-UNet 0.14 1.86 94.18 83.23 91.17 90.65
), ArticleFig(id=1179140100245041606, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, language=EN, label=Table 4, caption=

Comparative experiment results

, figureFileSmall=null, figureFileBig=null, tableContent=
模型 Params/
M
Flops/
G
FPS/
s-1
IoU/
%
Precision/
%
Recall/
%
PSPNet 46.59 93.70 24.23 75.87 86.98 87.15
SegNet 29.44 90.46 26.97 78.93 88.50 87.02
DeepLabV3+ 5.81 21.03 61.22 80.61 89.24 89.12
UNet 37.66 143.93 18.92 83.05 91.22 90.03
MobileNetV2 5.50 26.75 86.59 73.19 82.86 86.39
ShuffleNetV2 6.03 6.32 96.37 71.91 83.86 83.42
Fast-UNet 0.14 1.86 94.18 83.23 91.17 90.65
), ArticleFig(id=1179140100316344775, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149789684961862194, language=CN, label=表4, caption=

对比实验结果

, figureFileSmall=null, figureFileBig=null, tableContent=
模型 Params/
M
Flops/
G
FPS/
s-1
IoU/
%
Precision/
%
Recall/
%
PSPNet 46.59 93.70 24.23 75.87 86.98 87.15
SegNet 29.44 90.46 26.97 78.93 88.50 87.02
DeepLabV3+ 5.81 21.03 61.22 80.61 89.24 89.12
UNet 37.66 143.93 18.92 83.05 91.22 90.03
MobileNetV2 5.50 26.75 86.59 73.19 82.86 86.39
ShuffleNetV2 6.03 6.32 96.37 71.91 83.86 83.42
Fast-UNet 0.14 1.86 94.18 83.23 91.17 90.65
)], attaches=null, journal=Journal(id=1146119176004939786, delFlag=0, nameCn=科学技术与工程, nameEn=Science Technology and Engineering, nameHistory1=null, nameHistory2=null, issn=1671-1815, eissn=, cn=11-4688/T, coden=null, periodic=4, language=CN, oaType=是, ccby=null, superviseOffice=null, ownerOffice=null, pubOffice=null, editorOffice=null, officeType=null, aims=null, clcCode=null, officeProv=null, officeCity=null, officeAddr=null, officeZip=null, officeEmail=null, officePhone=null, editDirector=null, officeDirector=null, officeDirectorPhone=null, officeStaffNum=null, officeEmpNum=null, coverPicUrl=UKU/O7GSka5polgCTkbIIw==, journalPrice=null, startedYear=null, abbrevIsoEn=Sci Technol Eng, journalRemark=null, publicationField=null, createdTime=null, updatedTime=1754445529766, createdBy=null, updatedBy=13701087609, firstLetterCn=S, firstLetterEn=S, subjectCode=Natural Sciences, subjectName=自然科学, subjectCodeEn=Natural Sciences, subjectNameEn=null, picCn=UKU/O7GSka5polgCTkbIIw==, picEn=5hwlULoNwcbj3xUmVi9MAQ==, jcr=null, cjcr=null, exts=[JournalExt(id=1159791870395564357, language=CN, name=科学技术与工程, nameHistory1=null, nameHistory2=null, managedBy=, sponsoredBy=, publishedBy=, editorOffice=, officeProv=null, officeCity=null, officeAddr=, officeZip=, editDirector=null, officeDirector=null, officePhone=null, coverPicUrl=null, journalRemark=, submitArticleUrl=null, websiteUrl=http://www.stae.com.cn/jsygc/home, createdTime=1754445529793, updatedTime=1754445529793, createdBy=13701087609, updatedBy=13701087609, submissionGuidelinesUrl=http://www.stae.com.cn/jsygc/site/menus/20090429150146001, submissionAuthorUrl=http://www.stae.com.cn/jsygc/author/login, submissionEditorUrl=http://www.stae.com.cn/jsygc/editor/login, submissionReviewUrl=http://www.stae.com.cn/jsygc/reviewer/login, submissionCeEditorUrl=, submissionAeEditorUrl=, option={"copyright":""}), JournalExt(id=1159791870441701702, language=EN, name=Science Technology and Engineering, nameHistory1=null, nameHistory2=null, managedBy=, sponsoredBy=, publishedBy=, editorOffice=, officeProv=null, officeCity=null, officeAddr=, officeZip=, editDirector=null, officeDirector=null, officePhone=null, coverPicUrl=null, journalRemark=, submitArticleUrl=null, websiteUrl=http://www.stae.com.cn/jsygc/home, createdTime=1754445529804, updatedTime=1754445529804, createdBy=13701087609, updatedBy=13701087609, submissionGuidelinesUrl=, submissionAuthorUrl=http://www.stae.com.cn/jsygc/author/login, submissionEditorUrl=http://www.stae.com.cn/jsygc/editor/login, submissionReviewUrl=http://www.stae.com.cn/jsygc/reviewer/login, submissionCeEditorUrl=, submissionAeEditorUrl=, option={"copyright":""})], databaseList=null, tenantJournalId=1146123166801305609, websiteList=[Website(id=1148243202391400884, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1146123166801305609, journalNameCn=null, journalNameEn=null, grayFlag=null, tenantId=1146029695717560320, platformId=null, journalGroupId=null, journalGroupNameCn=null, journalGroupNameEn=null, type=1, domain=https://castjournals.cast.org.cn/joweb/kxjsygc/CN, language=CN, createTime=1751692112777, createBy=18614031015, updateTime=1753520965431, updateBy=18614031015, name=科学技术与工程-中文站点, tplId=1146099689490845704, title=科学技术与工程, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1148622798802673703, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1148243202391400884, code=articleTextType, value=kx, createTime=1751782615614, updateTime=1751782615614, creator=18614031015, updator=18614031015), WebsiteProps(id=1148622798781702180, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1148243202391400884, code=banner, value=null, createTime=1751782615609, updateTime=1751782615609, creator=18614031015, updator=18614031015), WebsiteProps(id=1148622798769119267, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1148243202391400884, code=logo, value=https://castjournals.cast.org.cn/joweb/kjdb/CN/file/pic?fileId=j86gbwi+p0Idkyl5SzIlmQ==, createTime=1751782615606, updateTime=1751782615606, creator=18614031015, updator=18614031015), WebsiteProps(id=1148622798794285094, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1148243202391400884, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/kjdb/CN/file/pic, createTime=1751782615612, updateTime=1751782615612, creator=18614031015, updator=18614031015), WebsiteProps(id=1148622798790090789, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1148243202391400884, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_cn_619/, createTime=1751782615611, updateTime=1751782615611, creator=18614031015, updator=18614031015)]), Website(id=1155914124811976731, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1146123166801305609, journalNameCn=null, journalNameEn=null, grayFlag=null, tenantId=1146029695717560320, platformId=null, journalGroupId=null, journalGroupNameCn=null, journalGroupNameEn=null, type=1, domain=https://castjournals.cast.org.cn/joweb/kxjsygc/EN, language=EN, createTime=1753521003206, createBy=18614031015, updateTime=1753521003206, updateBy=18614031015, name=科学技术与工程-英文站点, tplId=1146101810881728533, title=Science Technology and Engineering, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1155914371227308235, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1155914124811976731, code=articleTextType, value=kx, createTime=1753521061952, updateTime=1753521061952, creator=18614031015, updator=18614031015), WebsiteProps(id=1155914371210531016, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1155914124811976731, code=banner, value=null, createTime=1753521061947, updateTime=1753521061947, creator=18614031015, updator=18614031015), WebsiteProps(id=1155914371202142407, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1155914124811976731, code=logo, value=https://castjournals.cast.org.cn/joweb/kjdb/CN/file/pic?fileId=j86gbwi+p0Idkyl5SzIlmQ==, createTime=1753521061945, updateTime=1753521061945, creator=18614031015, updator=18614031015), WebsiteProps(id=1155914371223113930, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1155914124811976731, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/kjdb/CN/file/pic, createTime=1753521061950, updateTime=1753521061950, creator=18614031015, updator=18614031015), WebsiteProps(id=1155914371218919625, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1155914124811976731, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_cn_619/, createTime=1753521061949, updateTime=1753521061949, creator=18614031015, updator=18614031015)])], journalTitle=科学技术与工程, weixinUrl=null, journalUrl=null, iacademicId=null, status=0, seqNo=null, journalTitleEn=Science Technology and Engineering, journalPhotoCn=UKU/O7GSka5polgCTkbIIw==, journalPhotoEn=5hwlULoNwcbj3xUmVi9MAQ==, journalFirstLetter=S, journalRecommend=null, journalNew=null, journalCollection=null, jcrJf=null, cjcrJf=null, jcrJfStr=null, cjcrJfStr=null, submissionFirstDecision=null, sciSubjectClassification=null, casSubjectClassification=null, citeScore=null, totalCitationFrequency=null, icpCode=null, psCode=null, advertisingLicenseCode=null, copyrightInformation=null, country=null, option=null, provinceCode=null, provinceName=null, collectFlag=false), detailUrlCn=https://castjournals.cast.org.cn/joweb/kxjsygc/CN/10.12404/j.issn.1671-1815.2404359, detailUrlEn=https://castjournals.cast.org.cn/joweb/kxjsygc/EN/10.12404/j.issn.1671-1815.2404359, pdfUrlCn=https://castjournals.cast.org.cn/joweb/kxjsygc/CN/PDF/10.12404/j.issn.1671-1815.2404359, pdfUrlEn=https://castjournals.cast.org.cn/joweb/kxjsygc/EN/PDF/10.12404/j.issn.1671-1815.2404359, aliStartDate=null, aliEndDate=null, collectionFlag=false, citedCount=null, citedUrl=null, reference=null)
收藏切换
基于Fast-UNet的灰尘沉积快速检测方法
收藏切换
PDF下载
尉宇晖 , 王煜 , 高庭辉 , 张城华 , 赵章焰 *
科学技术与工程 | 论文·自动化技术、计算机技术 2025,25(16): 6804-6811
收起
收藏切换
科学技术与工程 | 论文·自动化技术、计算机技术 2025, 25(16): 6804-6811
基于Fast-UNet的灰尘沉积快速检测方法
全屏
尉宇晖 , 王煜, 高庭辉, 张城华, 赵章焰*
作者信息
  • 武汉理工大学交通与物流工程学院, 武汉 430063
  • 尉宇晖(1998―),男,汉族,山西大同人,硕士研究生。研究方向:机器视觉、图像处理。E-mail:

通讯作者:

* 赵章焰(1963―),男,汉族,湖北武汉人,博士,教授。研究方向:摄影测量、机器视觉、金属结构设计与仿真、攀爬机器人设计与实现。E-mail:
A Rapid Detection Method for Dust Deposition Based on Fast-UNet
Yu-hui YU , Yu WANG, Ting-hui GAO, Cheng-hua ZHANG, Zhang-yan ZHAO*
Affiliations
  • School of Transportation and Logistics Engineering, Wuhan University of Technology, Wuhan 430063, China
出版时间: 2025-06-08 doi: 10.12404/j.issn.1671-1815.2404359
文章导航
收藏切换

灰尘沉积会影响设备的正常运行。为准确且高效地对设备灰尘进行检测,制定科学的清洁策略,提出一种基于Fast-UNet的轻量化灰尘沉积检测方法。通过对UNet进行合理剪枝,采用最大池化与双线性插值进行下采样与上采样操作,减少了参数冗余,得到了结构紧凑的基础网络;使用轻量化的幽灵卷积模块(Ghost Module)替换基础网络中的普通卷积,进一步降低了网络的复杂度;在编码过程中嵌入融合了通道和空间注意力的CBAM(convolutional block attention module),在引入极少参数量的同时,使网络更加关注于目标区域。在模拟灰尘沉积的数据集上进行实验,结果表明:Fast-UNet与原模型相比,参数量减少了99.6%,计算量减少了98.7%,推理速度可达94.18帧/s,且识别精度高达91.17%。与其他5种主流分割模型相比,Fast-UNet在精度与速度上均具有优势。所提方法能够满足灰尘检测对检测精度与效率的需求,为灰尘定量分析提供技术参考。

信息处理  /  灰尘沉积  /  注意力机制  /  轻量化模型  /  UNet

Dust deposition can affect the normal operation of equipment. To accurately and efficiently detect dust on equipment and formulate a scientific cleaning strategy, a lightweight dust deposition detection method based on Fast-UNet was proposed. By effectively pruning UNet and adopting max pooling and bilinear interpolation for down-sampling and up-sampling operations, the parameter redundancy was reduced, and a compact basic network was obtained. The lightweight Ghost Module was used to replace the ordinary convolution in the basic network, further reducing the complexity of the network. An convolutional block attention module(CBAM) that integrated channel and spatial attention was embedded in the encoding process, which made the network pay more attention to the target area while introducing minimal parameters. Experiments on a simulated dust deposition dataset show that, compared with the original model, Fast-UNet reduces the number of parameters by 99.6%, decreases computational complexity by 98.7%, achieves an inference speed of 94.18 frames per second, and maintains a recognition accuracy of 91.17%. Compared with five other mainstream segmentation models, Fast-UNet also demonstrates advantages in both accuracy and speed. This method meets the needs of dust detection for both accuracy and efficiency, providing a technical reference for dust quantitative analysis.

information processing  /  dust deposition  /  attention mechanism  /  lightweight model  /  UNet
尉宇晖, 王煜, 高庭辉, 张城华, 赵章焰. 基于Fast-UNet的灰尘沉积快速检测方法. 科学技术与工程, 2025 , 25 (16) : 6804 -6811 . DOI: 10.12404/j.issn.1671-1815.2404359
Yu-hui YU, Yu WANG, Ting-hui GAO, Cheng-hua ZHANG, Zhang-yan ZHAO. A Rapid Detection Method for Dust Deposition Based on Fast-UNet[J]. Science Technology and Engineering, 2025 , 25 (16) : 6804 -6811 . DOI: 10.12404/j.issn.1671-1815.2404359
灰尘沉积是空气中微小颗粒逐渐沉降并附着在物体表面的过程,常见于工业设备、电子元器件和光学器件等领域,对设备的散热[1]、光学[2]和电气[3]性能产生不利影响。随着精密工程对设备洁净度的要求逐渐提高,愈发需要对灰尘沉积进行定期检测,因此设计一种高效且准确的灰尘沉积检测方法是有必要的。
早期研究主要聚焦于数字图像处理和光学方法。雷炳海等[4]采用区域自适应阈值分割方法,实现对玻璃表面附着灰尘的检测。张萌[5]通过分析灰度图及其直方图,判断掩膜版表面是否存在灰尘。王化超等[6]则利用分析激光照射后灰尘产生的反射光,检测光伏板上的积尘程度。这些方法在特定条件下能够取得较好的检测效果,但其较依赖于先验知识,且对外界环境噪声较为敏感,难以适应复杂多变的工程场景。
自卷积神经网络在计算机视觉领域取得巨大进展,相继出现UNet[7]、SegNet[8]、DeepLabV3+[9]等经典语义分割网络。深度学习方法因其无需依赖先验知识,且在精度和鲁棒性上表现优异,逐渐成为研究的热点并受到广泛青睐。李颖等[10]提出基于改进掩码区域卷积神经网络(mask region-based convolutional neural network,Mask R-CNN)的煤粉颗粒计算机断层扫描(computed tomography,CT)图像分割方法,以较大计算量为代价,得到了较为精确的分割结果。汪南洋等[11]设计了结合多尺度注意力机制的MSHAM-UNet,用于岩心孔洞的分割,有效实现了不同尺度特征的精准提取。覃本学等[12]提出了一种基于Debseg-Net的岩屑图像分割方法,其分割精度优于同类型网络。De Silva等[13]采用轻量化的MobileNetV3对道路交通扬尘进行分割,但其在边缘区域的分割效果仍有待提升。徐小平等[14]将坐标注意力与引入ShuffleNetV2,提出了一种高效的光伏板灰尘检测方法。这些研究表明深度学习方法可以用于颗粒状目标物的检测与分析。然而,如何在保障网络检测精度与抗噪性能的同时,降低其对计算和存储资源的占用仍需进一步研究。
为此,针对网络结构复杂、计算资源占用大的问题,在保证检测精度的前提下,提出一种基于UNet的轻量化灰尘检测算法Fast-UNet。首先,根据剪枝实验结果,对UNet进行了合理的网络层和通道剪枝,以减少冗余参数并简化网络结构;其次,引入轻量化的幽灵卷积(Ghost Module)[15]模块替代普通卷积,进一步降低网络复杂度;最后,在网络编码阶段引入融合通道与空间注意力的CBAM(convolutional block attention module)模块[16],以缓解参数量减少可能带来的检测精度损失,从而实现检测精度与效率之间的平衡。
UNet是一种经典的编码-解码结构的卷积神经网络。编码部分通过多次卷积和下采样提取不同尺度的特征,解码部分则通过多次卷积和上采样还原图像特征。编码与解码阶段通过跳跃连接融合各层次的信息。尽管UNet具备强大的特征学习能力,但其运行效率较低。Fast-UNet对网络结构进行了优化,通过网络层剪枝和通道剪枝有效减少参数冗余。下采样阶段采用高效的最大池化,上采样阶段采用双线性插值,进一步提升计算效率。此外,使用轻量化的幽灵卷积替代普通卷积,大幅降低了网络复杂度。同时,在编码阶段引入融合通道与空间注意力的CBAM模块,提升模型对关键特征区域的关注度,从而优化整体检测性能。网络结构如图1所示。
剪枝是网络轻量化的一种常用手段[17],其作用主要是在神经网络的基础上去除掉冗余参数,从而降低神经网络的大小和复杂度,减少其运算时间和储存占用,同时尽量不降低网络性能。
灰尘分布具有较强的随机性,其图像中的语义信息相对简单,但灰尘形态的多样性尤为显著,特别体现在复杂的轮廓结构上。在神经网络中,图像的语义信息主要集中在深层特征图中,而空间信息则更多体现在浅层特征图中。然而,深层特征提取过程中产生的冗余信息可能对网络的训练速度和识别效率产生不利影响,进而影响识别精度。此外,由于灰尘检测属于相对简单的二分类任务,复杂的网络结构可能导致过拟合和梯度不稳定等问题。为此,对UNet模型进行网络层和通道的剪枝优化,同时采用最大池化进行下采样,并通过双线性插值进行上采样,在保证性能的基础上得到了更为轻量化的基础网络。
训练不同剪枝参数的网络并进行测试,选择最优结果从而确定剪枝参数。通过分析后续的模型剪枝实验结果可知,最优的基础模型为只保留前4层的卷积网络结构,并将网络中的卷积通道数减1/2。剪枝实验设计如表1所示。
传统的特征提取方式是对输入特征图中的所有通道进行卷积操作,得到大量的输出特征图,这些特征图中存在较多的冗余信息,通过合理的策略减少运算,消除冗余信息可以大大简化网络,提高网络运算速度。
普通卷积操作需要同时生成所有特征图,计算成本较高,而生成的特征图中常包含大量相似的特征图,这些特征图实际上可以通过简单的线性变换或其他低成本操作生成,而不必进行完整的卷积计算[15]。基于此,引入幽灵卷积模块,替代网络中除输出层以外的普通卷积操作。幽灵卷积模块通过将一部分特征图的生成从高成本的卷积转移到低成本的线性变换,显著降低了参数规模和计算成本,同时保持了特征提取能力,其结构图如图2所示。
设输入特征图的高和宽分别为HW,通道数为C1、最终输出特征图通道数为C2、两次卷积使用的卷积核尺寸分别为K1K2,S为压缩系数。首先采用C2/S个大小为K1×K1×C1的卷积核对尺寸为H×W×C1的输入特征图进行卷积操作,得到尺寸为H×W×C2/S的原始特征图,这一步通过常规卷积提取关键特征,是整个操作中的高计算成本部分。然后使用S-1个大小为K2×K2的卷积核对上述特征图进行逐通道卷积Φk,共生成S-1组尺寸为H×W×C2/S的相似特征图,逐通道卷积仅在单一通道内进行操作,是整个操作中的低计算成本部分。将原始特征图通过恒等映射(Identity)直接保留,并将其与生成的相似特征图在通道维度上拼接,得到最终特征图。设普通卷积的尺寸为K3,其参数量为C2×K3×K3×C1,幽灵卷积模块的参数量为C2/S×K1×K1×C1+(S-1)×C2/S×K2×K2×1,普通卷积与幽灵卷积模块的参数比r可表示为
r= C 2 K 3 K 3 C 1 C 2 S K 1 K 1 C 1 + ( S - 1 ) C 2 S K 2 K 2 1
设定压缩系数S=2,卷积核大小K1=1、K2=3、K3=3,通过替换幽灵卷积模块,将网络的参数量压缩至原始基础网络的约 1 18,显著降低了网络的复杂度,同时有效提高了运行效率。
在灰尘识别过程中,灰尘附着的表面可能存在划痕、凹陷、阴影等干扰因素,导致大量无用的语义信息干扰识别精度。为增强网络对关键特征的聚焦能力并抑制干扰信息,同时避免显著增加网络的参数量与计算开销,在编码过程中引入融合通道与空间注意力的CBAM模块。该模块可分为两部分:通道注意力与空间注意力,通过顺序排列的方式融合来自通道维度与空间维度的注意力图,提高网络对重要信息的捕捉能力,其结构图如图3所示。
通道注意力的输入为原始特征图FRC×H×W,对特征图F进行全局最大池化与全局平均池化,得到两个大小为1×1×C的通道描述。将其送入参数共享的多层感知机(MLP),得到两个一维特征图。再将这两个特征图沿通道维度进行求和,并使用Sigmiod函数进行激活操作,得到通道权重系数Mc,将通道权重系数Mc与原始特征图F逐点相乘得到通道注意力的输出F',其计算过程为
Mc(F) = σ{MLP[AvgPool(F)] +MLP[MaxPool(F)]}
式(2)中:σ为Sigmiod激活函数;AvgPool、MaxPool分别为平均池化和最大池化操作。
空间注意力的输入为通道注意力的输出特征图F'RC×H×W。对特征图F'的全局最大池化与全局平均池化,得到两个大小为H×W×1的空间描述。将其沿通道维度进行拼接融合,经过一个7×7大小的卷积层后,使用Sigmiod函数进行激活操作,得到空间权重系数Ms,其计算公式为
$\boldsymbol{M}_{\mathrm{s}}(\boldsymbol{F})=\sigma\left\{f^{(7 \times 7)}\left[\operatorname{AvgPool}\left(\boldsymbol{F}^{\prime}\right) ; \operatorname{MaxPool}\left(\boldsymbol{F}^{\prime}\right)\right]\right\}$
式(3)中:f(7×7)为7×7大小的卷积核。
将空间权重系数Ms与特征图F'逐点相乘,得到空间注意力的输出F,即最终的输出特征图。
铜在工业上具有广泛用途,可以用作PCB板的基料和重要元件的制材,也是设备零件、外壳的主要原材料。综合考虑工程实际与实验可行性,选用铜箔作为被污染对象。
实验中,将铜箔按照指定尺寸均匀划分为多个观察区域,这些区域的污染状况能够较为全面地反映整个铜箔表面的污染情况。为模拟灰尘沉积的情形,先对铜箔进行碳粉污染并通过压实处理固定污染物,再利用显微镜和成像系统(图4)对污染后的铜箔表面进行多次定点观察和第一次图像采集。此外,为模拟少量顽固灰尘累积的情况,使用工业超声波除尘设备(图5)对污染后的铜箔进行除尘处理,随后再次对其表面进行多次定点观察与第二次图像采集。显微镜放大倍数为50倍,成像系统的分辨率为3 840×2 160,获得了两批反映铜箔表面污染情况的图像。
通过10×6的顺序裁剪与随机选取,得到300张大小为384×360的灰尘图像,使用labelme标注工具进行人工标注,最后随机选取其中200张作为训练集,100张作为验证集。数据集获取过程如图6所示。
灰尘图片边缘信息复杂,人工标注效率低,这使得数据集数量较少,无法满足训练需求,容易出现训练不稳定、过拟合等问题。考虑到灰尘图像的采集效果易受环境亮度、拍摄角度和镜头参数等因素的影响,对训练集进行多种数据增强综合操作。具体而言,通过水平翻转、垂直翻转和随机透视变换模拟不同拍摄角度下的灰尘图像分布;通过随机亮度和对比度,模拟不同环境光照条件及多种镜头参数设置对图像的影响。对数据集进行扩充后,共得到1 800张可用于训练的灰尘图片,显著提高了训练数据的多样性。数据增强后的效果如图7所示。
算法训练、测试所用计算机配置为:Intel 12400F处理器,NVIDIA GeForce GTX 1080ti显卡,显存11 GB,运行内存16 GB,Opencv版本为4.7.0.72,Python版本为3.9,Pytorch版本为1.10.0,CUDA版本为11.3。
采用Adam优化器进行损失函数的优化,初始学习率设置为0.000 5,学习率衰减倍数设置为0.99,模型训练的迭代批量大小设置为8,训练轮数设置为50。
灰尘数据集中灰尘像素点数量明显少于背景类像素,存在正负样本分布不平衡的问题。在综合考虑模型训练效率与分割结果精度后,采用二元交叉熵损失函数BCELoss与骰子损失函数DiceLoss[18]进行加权,组成BDLoss作为本文训练的损失函数。BCELoss使得模型关注目标区域与背景的差异性,DiceLoss使得模型关注目标区域的相似性,二者结合可以缓解样本分布不均匀的问题,其计算公式为
BDLoss=BCELoss+DiceLoss
式(4)中:
B C E L o s s = - 1 N i = 1 N [ y i l n ( p i ) + ( 1 - y i ) l n ( 1 - p i ) ] D i c e L o s s = 1 - 2 G T P r e d G T + P r e d
式(5)中:N为样本类别数;yi为真实标签;pi为模型预测值;GT为真实标签的集合;Pred为模型预测值的集合。
引入交并比(IoU)、召回率(Recall)、精确率(Precision)、模型参数量(Params)、浮点运算量(Flops)、每秒传输帧数(FPS)6个指标来评价模型的分割效果。IoU指预测区域和真实区域的交集区域与并集区域之比,Recall用于衡量模型寻找所有正例的能力,Precision用于衡量模型预测正例的准确性,Params和Flops用于衡量模型的复杂度,其值越小,说明模型对于内存和计算资源的占用越小,FPS指模型每秒推理图片的数量,该指标对于模型的实时检测能力具有重要意义。
IoU、Recall、Precision计算公式为
IoU= T P F N + F P + T P
Recall= T P T P + F N
Precision= T P T P + F P
式中:TP为将真正例预测为正例的像素个数;FP为将真反例预测为正例的像素个数;FN为将真正例预测为反例的像素个数;TN为将真反例预测为反例的像素个数。
为合理简化网络结构、消除冗余参数,设计网络剪枝实验,对不同剪枝参数的网络进行训练,并在测试集上进行评估,实验结果如表2所示。
表2可知,在对网络进行通道剪枝与神经层剪枝后,部分网络精度略微提高,这证明原网络存在一定的过拟合问题。剪枝后的网络未出现较大的精度下降,这表明原网络在通道与神经层的参数存在冗余,实验结果验证了剪枝实验的合理性和必要性。在综合考虑了网络精度、参数量以及后续改进带来的影响,选择UNet_4作为后续改进的基础网络。
通过剪枝实验,得到了结构紧凑的基础网络,对基础网络进行卷积替换,将网络中除输出端的普通卷积替换为幽灵卷积模块,进一步降低了网络的复杂度,提高网络的实时检测能力。在此基础上,将轻量化的注意力模块CBAM嵌入网络编码过程中,提高了网络对于目标特征的捕捉能力,抑制无关信息,从而有效缓解卷积替换带来的精度损失。为验证网络改进方案的有效性,设计消融实验,实验结果如表3所示。
表3可知,对基础网络UNet_4进行卷积替换后得到的新网络的参数量和计算量有大幅度的下降,推理速度提高了48%,但网络精度出现较为明显的下降。在编码器嵌入CBAM注意力模块后,以较小的计算成本提高了网络在空间和通道维度上对关键特征的捕捉能力,有效保证了网络精度。实验结果表明,改进后的Fast-UNet在保持轻量化的前提下,仍具有极佳的精度优势,符合灰尘检测场景对于检测精度与检测速度的要求。
为进一步验证本文方法的有效性与实用性,将Fast-UNet与UNet、PSPNet、SegNet、DeepLabV3+、MobileNetV2、ShufflenNetV2这6种语义分割网络进行对比实验。使用相同的实验平台和训练配置,分别对这6种语义分割网络进行训练,并在测试集上进行评估。各网络预测结果的可视化对比如图8所示,实验评估结果如表4所示。
可视化结果表明,与其他6种分割网络相比,Fast-UNet能够更有效地抑制背景噪声干扰,同时在灰尘边缘区域的分割表现更加精细。对比实验结果显示,Fast-UNet在精度和速度方面明显优于经典分割网络,如PspNet、SegNet和DeepLabv3+。与轻量化网络MobileNetV2和ShuffleNetV2相比,Fast-UNet在保持轻量化特性的同时,在精度上具有显著优势,这得益于CBAM注意力机制与UNet架构中的跳跃连接,能够更好地保留图像细节。
通过合理的剪枝策略和轻量化卷积模块,Fast-UNet显著提高了网络的运行效率,而引入CBAM注意力模块使其能够精准捕捉灰尘边缘的细节。相比于UNet,Fast-UNet在小颗粒灰尘的识别上更加灵敏,整体识别精度相当,但参数量减少了99.62%,计算量减少了98.71%,每秒处理图片的数量提升了近6倍,这对于灰尘图像的高效处理具有重要意义。
综上可知,Fast-UNet在保持网络高性能的同时,实现了结构轻量化,能够兼顾灰尘检测的精度需求和效率要求,具有较高的实用价值。
针对灰尘沉积检测问题,基于UNet设计轻量化方法Fast-UNet。为满足高效检测需求,对UNet进行了合理的网络层与通道剪枝,并引入轻量化的幽灵卷积模块,显著简化了网络结构。同时引入CBAM注意力机制,使网络能够更精准地聚焦于目标区域,提升分割性能。灰尘检测实验表明,Fast-UNet相比UNet参数量减少了99.62%,计算量减少了98.71%,但识别精度仍可达91.17%,且其推理速度达每秒94.18帧。相较于主流分割网络,Fast-UNet在检测速度与精度上具有显著优势,能够满足灰尘检测任务对于高效性与准确性的需求,为灰尘定量分析领域的发展提供了技术支撑。未来的研究将进一步探索该方法在其他场景下的应用潜力,并深入研究注意力机制与轻量化策略,以实现更高精度与实时检测效果。
参考文献 引证文献
排序方式:
[1]
汪峰, 王志豪, 张小松, 等. 换热器翅片表面灰尘沉积特性的实验研究[J]. 中南大学学报(自然科学版), 2023, 54(12): 4893-4900.
Wang Feng, Wang Zhihao, Zhang Xiaosong, et al. Experimental study on characteristics of dust deposition on fin surface of heat exchanger[J]. Journal of Central South University(Science and Technology), 2023, 54(12): 4893-4900.
[2]
牛龙飞, 蒋一岚, 苗心向, 等. 大口径反射镜表面颗粒物洁净度控制实验研究[J]. 红外与激光工程, 2021, 50(12): 246-252.
Niu Longfei, Jiang Yilan, Miao Xinxiang, et al. Experimental study of particles cleanliness control on the surface of large-aperture reflector[J]. Infrared and Laser Engineering, 2021, 50(12): 246-252.
[3]
姚磊, 付豪, 马科飞, 等. 高速公路边坡光伏组件表面灰尘粘附性能研究[J]. 太阳能学报, 2024, 45(4): 460-467.
Yao Lei, Fu Hao, Ma Kefei, et al. Performance study of dust adhesion of PV module surfaces on highway slope[J]. Acta Energiae Solaris Sinica, 2024, 45(4): 460-467.
[4]
雷炳海, 杨志刚, 金哲岩. 风洞中灰尘在不同特性玻璃表面上粘附形貌和粘附程度的实验研究[J]. 力学季刊, 2021, 42(3): 479-489.
Lei Binghai, Yang Zhigang, Jin Zheyan. Experimental study of adhesion morphology features and the degree of adhesion of dusts on glass surfaces of different characteristics in a wind tunnel[J]. Chinese Quarterly of Mechanics, 2021, 42(3): 479-489.
[5]
张萌. 掩膜版表面灰尘杂质检测系统设计[D]. 西安: 西安石油大学, 2020.
Zhang Meng. The design of a dust and impurity detection system for mask surface[D]. Xi'an: Xi'an Shiyou University, 2020.
[6]
王化超, 高爱华, 秦文罡, 等. 光伏板积尘散射检测中的去噪方法研究[J]. 光学与光电技术, 2023, 21(2): 51-57, 77.
Wang Huachao, Gao Aihua, Qin Wengang, et al. Denoising method in dust scattering detection of photovoltaic panels[J]. Optics & Optoelectronic Technology, 2023, 21(2): 51-57, 77.
[7]
Ronneberger O, Fischer P, Brox T. U-net: convolutional networks for biomedical image segmentation[C]// 18th International Conference on Medical Image Computing and Computer-Assisted Intervention. Munich: Springer International Publishing, 2015: 234-241.
[8]
Badrinarayanan V, Kendall A, Cipolla R. Segnet: A deep convolutional encoder-decoder architecture for image segmentation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(12): 2481-2495.
[9]
He K, Zhang X, Ren S, et al. Deep residual learning for image recognition[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2016: 770-778.
[10]
李颖, 李秀宇, 卢兆林, 等. 基于深度学习的煤粉颗粒CT图像分割方法[J]. 计算机工程与设计, 2022, 43(8): 2252-2259.
Li Ying, Li Xiuyu, Lu Zhaolin, et al. Coal particle CT image segmentation method based on deep learning[J]. Computer Engineering and Design, 2022, 43(8): 2252-2259.
[11]
汪南洋, 沈疆海. 基于MSHAM-UNet的岩心孔洞图像分割方法[J]. 科学技术与工程, 2024, 24(24): 10362-10369.
Wang Nanyang, Shen Jianghai. Image segmentation method of rock core hole based on MSHAM-UNet[J]. Science Technology and Engineering, 2024, 24(24): 10362-10369.
[12]
覃本学, 沈疆海, 马丙鹏, 等. 基于Debseg-Net的岩屑图像语义分割[J]. 科学技术与工程, 2022, 22(29): 12927-12935.
Qin Benxue, Shen Jianghai, Ma Bingpeng, et al. Semantic segmentation of rock debris image based on Debseg-Net[J]. Science Technology and Engineering, 2022, 22(29): 12927-12935.
[13]
De Silva A, Ranasinghe R, Sounthararajah A, et al. Beyond conventional monitoring: a semantic segmentation approach to quantifying traffic-induced dust on unsealed roads[J]. Sensors, 2024, 24(2): DOI: 10.3390/s24020510.
[14]
徐小平, 张勇, 刘广钧, 等. 基于改进ShuffleNetV2模型的光伏板灰尘识别[J]. 计算机系统应用, 2023, 32(8): 295-302.
Xu Xiaoping, Zhang Yong, Liu Guangjun, et al. Identification of dust on photovoltaic panel based on improved ShuffleNetV2 model[J]. Computer Systems and Applications, 2023, 32(8): 295-302.
[15]
Han K, Wang Y, Tian Q, et al. Ghostnet: more features from cheap operations[C]// Proceedings of the IEEE/CVF conference on Computer Vision and Pattern Recognition. New York: IEEE, 2020: 1580-1589.
[16]
Woo S, Park J, Lee J Y, et al. CBAM: convolutional block attention module[C]// Proceedings of the European Conference on Computer Vision(ECCV). New York: IEEE, 2018: 3-19.
[17]
Elkerdawy S, Elhoushi M, Zhang H, et al. Fire together wire together: a dynamic pruning approach with self-supervised mask prediction[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2022: 12454-12463.
[18]
Li X, Sun X, Meng Y, et al. Dice loss for data-imbalanced NLP tasks[J]. arXiv Preprint, 2019: arXiv: 1911.02855.
2025年第25卷第16期
PDF下载
375
143
引用本文
BibTeX
文章信息
doi: 10.12404/j.issn.1671-1815.2404359
  • 接收时间:2024-06-06
  • 首发时间:2025-07-09
  • 出版时间:2025-06-08
补充材料
相关文章
文章信息
作者
出版历史
  • 收稿日期:2024-06-06
  • 修回日期:2025-03-04
基金
作者信息
    武汉理工大学交通与物流工程学院, 武汉 430063

通讯作者:

* 赵章焰(1963―),男,汉族,湖北武汉人,博士,教授。研究方向:摄影测量、机器视觉、金属结构设计与仿真、攀爬机器人设计与实现。E-mail:
参考文献
分享链接
https://castjournals.cast.org.cn/joweb/kxjsygc/CN/10.12404/j.issn.1671-1815.2404359
分享至
全文二维码

扫描看全文

引用本文
BibTeX
本文的引用情况
2种不同金属材料的力学参数

Family
属数
Number of
genus
种数
Number of
species
占总种数比例
Percentage of
total species (%)

Genus
种数
Number of
species
占总种数比例
Percentage of total
species (%)
鹅膏菌科Amanitaceae 2 11 5.26 鹅膏菌属 Amanita 10 4.78
小菇科 Mycenaceae 2 12 5.74 丝盖伞属 Inocybe 5 2.39
多孔菌科 Polyporaceae 8 14 6.70 蜡蘑属 Laccaria 5 2.39
红菇科 Russulaceae 3 23 11.00 小皮伞属 Marasmius 6 2.87
小菇属 Mycena 11 5.26
光柄菇属 Pluteus 5 2.39
红菇属 Russula 17 8.13
栓菌属 Trametes 5 2.39
关闭全屏