Article(id=1236276112595144961, tenantId=1146029695717560320, journalId=1235980550691926019, issueId=1236276104999268557, articleNumber=null, orderNo=null, doi=10.3969/j.issn.0253-6099.2025.04.024, pmid=null, cstr=null, oa=null, hot=null, price=null, onlineType=0, articleFormat=0, articleType=null, articleTypeStr=null, receivedDate=1740412800000, receivedDateStr=2025-02-25, revisedDate=null, revisedDateStr=null, acceptedDate=null, acceptedDateStr=null, onlineDate=1772680794142, onlineDateStr=2026-03-05, pubDate=1753977600000, pubDateStr=2025-08-01, doiRegisterDate=null, doiRegisterDateStr=null, onlineIssueDate=1772680794142, onlineIssueDateStr=2026-03-05, onlineJustAcceptDate=null, onlineJustAcceptDateStr=null, onlineFirstDate=null, onlineFirstDateStr=null, sourceXml=null, magXml=null, createTime=1772680794142, creator=13701087609, updateTime=1772680794142, updator=13701087609, issue=Issue{id=1236276104999268557, tenantId=1146029695717560320, journalId=1235980550691926019, year='2025', volume='45', issue='4', pageStart='1', pageEnd='200', issueExtLink='null', onlineDate='null', pubDate='null', beforeIssueId=null, nextIssueId=null, price=null, status=1, issueComplete=1, articleOrder=1, issueType=1, specialIssue=null, createTime=1772680792331, creator=13701087609, updateTime=1772681498687, updator=13701087609, preIssue=null, nextIssue=null, ext={EN=IssueExt(id=1236279067746562719, tenantId=1146029695717560320, journalId=1235980550691926019, issueId=1236276104999268557, language=EN, specialIssueTitle=, coverIllustrator=null, specialIssueEditor=, specialIssueAbout=), CN=IssueExt(id=1236279067746562720, tenantId=1146029695717560320, journalId=1235980550691926019, issueId=1236276104999268557, language=CN, specialIssueTitle=, coverIllustrator=null, specialIssueEditor=, specialIssueAbout=)}, issueFiles=null}, startPage=130, endPage=136, ext={EN=ArticleExt(id=1236276114675519818, articleId=1236276112595144961, tenantId=1146029695717560320, journalId=1235980550691926019, language=EN, title=Visual Recognition of Float Flowmeter Readings in Sintering Process of Lithium-Ion Battery Cathode Materials Based on YOLOFFM, columnId=1236276106727321817, journalTitle=Mining and Metallurgical Engineering, columnName=METALLURGY, runingTitle=null, highlight=null, articleAbstract=

To address problems of accuracy and efficiency in automatically reading of float flowmeter at low flow rates under a complex environment in sintering kilns for new energy materials, an innovative YOLOFFM algorithm was proposed. The core improvements include reconstructing the backbone network, enhancing the neck structure, introducing an asymmetric compression decoupling head, and optimizing the loss function, which can significantly improve the efficiency and detection performance of the algorithm. The results show that this YOLOFFM algorithm has accuracy up to 99.15% and a recall rate of 98.69%, significantly reducing the number of model parameters and computational costs. Compared with various advanced algorithm models, YOLOFFM can improve accuracy while reducing the computational cost by more than 90%, fully demonstrating its high efficiency and reliability for new energy materials in a complex environment of sintering kilns.

, correspAuthors=null, 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=Yong HUANG, Yang LIU, Xing XIA, Shengwang XIAO, Ran LI), CN=ArticleExt(id=1236276119327003291, articleId=1236276112595144961, tenantId=1146029695717560320, journalId=1235980550691926019, language=CN, title=基于YOLOFFM的锂电正极材料烧结过程浮子流量视觉识别研究, columnId=1236276107322913000, journalTitle=矿冶工程杂志, columnName=冶金, runingTitle=null, highlight=null, articleAbstract=

针对使用烧结窑炉生产新能源材料过程中浮子流量计在低流量和复杂环境下自动读取的准确性与效率问题,提出了一种改进的YOLOv8n-Pose关键点浮子流量计自动识别(YOLOFFM)算法。该算法的核心改进包括重构主干网络、增强颈部结构、引入非对称压缩解耦头和优化损失函数,这些措施显著提升了算法的效率和检测性能。结果表明,YOLOFFM算法准确率达99.15%,召回率为98.69%,显著降低了模型的参数量和计算成本。与多种先进模型算法相比,YOLOFFM在提升精度的同时,计算成本降低90%以上,充分展示了其在烧结窑炉复杂环境中的高效性和可靠性。

, correspAuthors=null, authorNote=null, correspAuthorsNote=null, copyrightStatement=null, copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=gHOFn1lZGo/VVP/YgHW6aA==, magXml=KThGYY0ea6wYwRl0JH1eIg==, pdfUrl=null, pdf=IDXWZqYveNuSp9qcMfKu1w==, pdfFileSize=1688946, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=Yk0NPFF351KXpD2EL0wWSQ==, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=nv6Uj5clDUZ9Jwfl9fRDxg==, mapNumber=null, authorCompany=null, fund=null, authors=

黄勇(1990—),男,湖南益阳人,硕士,高级工程师,主要研究方向为智能计算及选冶过程自动化。E-mail:

, authorsList=黄勇, 刘洋, 夏星, 肖盛旺, 李然)}, authors=[Author(id=1236348222998901695, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, orderNo=0, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=lyrhy@foxmail.com, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1236348223116342219, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, authorId=1236348222998901695, language=EN, stringName=Yong HUANG, firstName=Yong, middleName=null, lastName=HUANG, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=Changsha Research Institute of Mining and Metallurgy Co, Ltd, Changsha 410012, Hunan, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1236348223225394131, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, authorId=1236348222998901695, language=CN, stringName=黄勇, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=长沙矿冶研究院有限责任公司,湖南 长沙 410012, bio={"content":"

黄勇(1990—),男,湖南益阳人,硕士,高级工程师,主要研究方向为智能计算及选冶过程自动化。E-mail:

"}, bioImg=null, bioContent=

黄勇(1990—),男,湖南益阳人,硕士,高级工程师,主要研究方向为智能计算及选冶过程自动化。E-mail:

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1236348222847906731, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, xref=null, ext=[AuthorCompanyExt(id=1236348222856295340, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, companyId=1236348222847906731, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=Changsha Research Institute of Mining and Metallurgy Co, Ltd, Changsha 410012, Hunan, China), AuthorCompanyExt(id=1236348222860489645, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, companyId=1236348222847906731, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=长沙矿冶研究院有限责任公司,湖南 长沙 410012)])]), Author(id=1236348223347028963, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, 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=1236348223460275179, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, authorId=1236348223347028963, language=EN, stringName=Yang LIU, firstName=Yang, middleName=null, lastName=LIU, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=Changsha Research Institute of Mining and Metallurgy Co, Ltd, Changsha 410012, Hunan, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1236348223565132785, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, authorId=1236348223347028963, language=CN, stringName=刘洋, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=长沙矿冶研究院有限责任公司,湖南 长沙 410012, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1236348222847906731, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, xref=null, ext=[AuthorCompanyExt(id=1236348222856295340, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, companyId=1236348222847906731, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=Changsha Research Institute of Mining and Metallurgy Co, Ltd, Changsha 410012, Hunan, China), AuthorCompanyExt(id=1236348222860489645, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, companyId=1236348222847906731, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=长沙矿冶研究院有限责任公司,湖南 长沙 410012)])]), Author(id=1236348223690961915, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, 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=1236348223816790017, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, authorId=1236348223690961915, language=EN, stringName=Xing XIA, firstName=Xing, middleName=null, lastName=XIA, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=Changsha Research Institute of Mining and Metallurgy Co, Ltd, Changsha 410012, Hunan, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1236348223946813452, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, authorId=1236348223690961915, language=CN, stringName=夏星, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=长沙矿冶研究院有限责任公司,湖南 长沙 410012, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1236348222847906731, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, xref=null, ext=[AuthorCompanyExt(id=1236348222856295340, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, companyId=1236348222847906731, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=Changsha Research Institute of Mining and Metallurgy Co, Ltd, Changsha 410012, Hunan, China), AuthorCompanyExt(id=1236348222860489645, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, companyId=1236348222847906731, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=长沙矿冶研究院有限责任公司,湖南 长沙 410012)])]), Author(id=1236348224055865369, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, 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=1236348224177500197, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, authorId=1236348224055865369, language=EN, stringName=Shengwang XIAO, firstName=Shengwang, middleName=null, lastName=XIAO, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=Changsha Research Institute of Mining and Metallurgy Co, Ltd, Changsha 410012, Hunan, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1236348224286552109, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, authorId=1236348224055865369, language=CN, stringName=肖盛旺, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=长沙矿冶研究院有限责任公司,湖南 长沙 410012, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1236348222847906731, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, xref=null, ext=[AuthorCompanyExt(id=1236348222856295340, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, companyId=1236348222847906731, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=Changsha Research Institute of Mining and Metallurgy Co, Ltd, Changsha 410012, Hunan, China), AuthorCompanyExt(id=1236348222860489645, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, companyId=1236348222847906731, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=长沙矿冶研究院有限责任公司,湖南 长沙 410012)])]), Author(id=1236348224383021115, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, orderNo=4, 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=1236348224504655941, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, authorId=1236348224383021115, language=EN, stringName=Ran LI, firstName=Ran, middleName=null, lastName=LI, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=Changsha Research Institute of Mining and Metallurgy Co, Ltd, Changsha 410012, Hunan, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1236348224630485069, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, authorId=1236348224383021115, language=CN, stringName=李然, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=长沙矿冶研究院有限责任公司,湖南 长沙 410012, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1236348222847906731, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, xref=null, ext=[AuthorCompanyExt(id=1236348222856295340, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, companyId=1236348222847906731, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=Changsha Research Institute of Mining and Metallurgy Co, Ltd, Changsha 410012, Hunan, China), AuthorCompanyExt(id=1236348222860489645, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, companyId=1236348222847906731, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=长沙矿冶研究院有限责任公司,湖南 长沙 410012)])])], keywords=[Keyword(id=1236348224894726238, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, language=EN, orderNo=1, keyword=YOLOFFM), Keyword(id=1236348224999583845, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, language=EN, orderNo=2, keyword=automatic reading), Keyword(id=1236348225146384498, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, language=EN, orderNo=3, keyword=MSBlock), Keyword(id=1236348225238659193, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, language=EN, orderNo=4, keyword=EffQA-FPN), Keyword(id=1236348225389654145, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, language=EN, orderNo=5, keyword=sintering kilns), Keyword(id=1236348225532260489, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, language=EN, orderNo=6, keyword=float flowmeter), Keyword(id=1236348225645506701, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, language=EN, orderNo=7, keyword=visual recognition), Keyword(id=1236348225741975697, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, language=EN, orderNo=8, keyword=algorithm), Keyword(id=1236348225918136474, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, language=CN, orderNo=1, keyword=YOLOFFM), Keyword(id=1236348226073325733, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, language=CN, orderNo=2, keyword=自动读数), Keyword(id=1236348226203349166, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, language=CN, orderNo=3, keyword=MSBlock), Keyword(id=1236348226366927028, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, language=CN, orderNo=4, keyword=EffQA-FPN), Keyword(id=1236348226484367549, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, language=CN, orderNo=5, keyword=烧结窑炉), Keyword(id=1236348226622779591, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, language=CN, orderNo=6, keyword=浮子流量), Keyword(id=1236348226752803025, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, language=CN, orderNo=7, keyword=视觉识别), Keyword(id=1236348226882826460, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, language=CN, orderNo=8, keyword=算法)], refs=[Reference(id=1236348230078886309, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, doi=null, pmid=null, pmcid=null, year=2017, volume=40, issue=7, pageStart=572, pageEnd=584, url=null, language=null, rfNumber=[1], rfOrder=0, authorNames=TSAI M C, KO P J, journalName=Journal of the Chinese Institute of Engineers, refType=null, unstructuredReference=TSAI M C, KO P J. On-line condition monitoring of servo motor drive systems by HHT in Industry 4.0[J]. Journal of the Chinese Institute of Engineers, 2017, 40(7): 572-584., articleTitle=On-line condition monitoring of servo motor drive systems by HHT in Industry 4.0, refAbstract=null), Reference(id=1236348230217298350, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, doi=null, pmid=null, pmcid=null, year=2017, volume=null, issue=null, pageStart=43, pageEnd=47, url=null, language=null, rfNumber=[2], rfOrder=1, authorNames=GAO J W, XIE H T, ZUO L, journalName=null, refType=null, unstructuredReference=GAO J W, XIE H T, ZUO L, et al. A robust pointer meter reading recognition method for substation inspection robot[C]//International Conference on Robotics and Automation Sciences (ICRAS), 2017: 43-47., articleTitle=A robust pointer meter reading recognition method for substation inspection robot, refAbstract=null), Reference(id=1236348230338933171, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, doi=null, pmid=null, pmcid=null, year=2017, volume=4, issue=4, pageStart=830, pageEnd=847, url=null, language=null, rfNumber=[3], rfOrder=2, authorNames=LU S Y, ZHANG Y, SU J J, journalName=IEEE/CAA Journal of Automatica Sinica, refType=null, unstructuredReference=LU S Y, ZHANG Y, SU J J. Mobile robot for power substation inspection: A survey[J]. IEEE/CAA Journal of Automatica Sinica, 2017, 4(4): 830-847., articleTitle=Mobile robot for power substation inspection: A survey, refAbstract=null), Reference(id=1236348230477345205, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, doi=null, pmid=null, pmcid=null, year=2024, volume=null, issue=5, pageStart=76, pageEnd=82, url=null, language=null, rfNumber=[4], rfOrder=3, authorNames=康朝海, 刘杨, 任伟建, journalName=工业仪表与自动化装置, refType=null, unstructuredReference=康朝海, 刘杨, 任伟建, 等. 基于MaskR-CNN的油田井场指针仪表识别方法研究[J]. 工业仪表与自动化装置, 2024(5): 76-82., articleTitle=基于MaskR-CNN的油田井场指针仪表识别方法研究, refAbstract=null), Reference(id=1236348230619951548, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, doi=null, pmid=null, pmcid=null, year=2024, volume=null, issue=5, pageStart=76, pageEnd=82, url=null, language=null, rfNumber=[4], rfOrder=4, authorNames=KANG Chaohai, LIU Yang, REN Weijian, journalName=Industrial Instrumentation & Automation, refType=null, unstructuredReference=KANG Chaohai, LIU Yang, REN Weijian, et al. Research on MaskR-CNN-based pointer instrument recognition method for oilfield wellsite[J]. Industrial Instrumentation & Automation, 2024(5): 76-82., articleTitle=Research on MaskR-CNN-based pointer instrument recognition method for oilfield wellsite, refAbstract=null), Reference(id=1236348230770946498, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, doi=null, pmid=null, pmcid=null, year=2020, volume=null, issue=null, pageStart=9806, pageEnd=9815, url=null, language=null, rfNumber=[5], rfOrder=5, authorNames=LIU Y L, CHEN H, SHEN C H, journalName=null, refType=null, unstructuredReference=LIU Y L, CHEN H, SHEN C H, et al. ABCNet: Real-time scene text spotting with Adaptive Bezier-Curve Network[C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2020: 9806-9815., articleTitle=ABCNet: Real-time scene text spotting with Adaptive Bezier-Curve Network, refAbstract=null), Reference(id=1236348230930330057, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, doi=null, pmid=null, pmcid=null, year=2019, volume=null, issue=null, pageStart=108, pageEnd=113, url=null, language=null, rfNumber=[6], rfOrder=6, authorNames=HE P, ZUO L, ZHANG C, journalName=null, refType=null, unstructuredReference=HE P, ZUO L, ZHANG C, et al. A value recognition algorithm for pointer meter based on improved Mask-RCNN[C]/2019 9th International Conference on Information Science and Technology (ICIST), 2019: 108-113., articleTitle=A value recognition algorithm for pointer meter based on improved Mask-RCNN, refAbstract=null), Reference(id=1236348231043576271, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, doi=null, pmid=null, pmcid=null, year=2020, volume=null, issue=null, pageStart=283, pageEnd=287, url=null, language=null, rfNumber=[7], rfOrder=7, authorNames=ZHANG X, DANG X, LYU Q, journalName=null, refType=null, unstructuredReference=ZHANG X, DANG X, LYU Q, et al. A Pointer meter recognition algorithm based on deep learning[C]/2020 3rd International Conference on Advanced Electronic Materials, Computers and Software Engineering (AEMCSE), 2020: 283-287., articleTitle=A Pointer meter recognition algorithm based on deep learning, refAbstract=null), Reference(id=1236348231152628179, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, doi=null, pmid=null, pmcid=null, year=2020, volume=32, issue=12, pageStart=1976, pageEnd=1984, url=null, language=null, rfNumber=[8], rfOrder=8, authorNames=ZHOU D K, YANG Y, ZHU J, journalName=Journal of Computer-Aided Design & Computer Graphics, refType=null, unstructuredReference=ZHOU D K, YANG Y, ZHU J, et al. Tilt correction method of pointer meter based on deep learning[J]. Journal of Computer-Aided Design & Computer Graphics, 2020, 32(12): 1976-1984., articleTitle=Tilt correction method of pointer meter based on deep learning, refAbstract=null), Reference(id=1236348231270068695, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, doi=null, pmid=null, pmcid=null, year=2022, volume=null, issue=null, pageStart=14, pageEnd=null, url=null, language=null, rfNumber=[9], rfOrder=9, authorNames=LIU H, HUANG Z L, TIAN L, journalName=null, refType=null, unstructuredReference=LIU H, HUANG Z L, TIAN L, et al. Method for detecting and recognizing pointer meter reading based on deep learning[C]/Third International Conference on Electronics and Communication; Network and Computer Technology (ECNCT 2021). 2022: 14., articleTitle=Method for detecting and recognizing pointer meter reading based on deep learning, refAbstract=null), Reference(id=1236348231395897822, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, doi=null, pmid=null, pmcid=null, year=2024, volume=null, issue=2, pageStart=36, pageEnd=43, url=null, language=null, rfNumber=[10], rfOrder=10, authorNames=刘煜博, 吐松江·卡日, 伊力哈木·亚尔买买提, journalName=仪表技术与传感器, refType=null, unstructuredReference=刘煜博, 吐松江·卡日, 伊力哈木·亚尔买买提, 等. 基于VCA-UNet的全自动指针式仪表读数方法[J]. 仪表技术与传感器, 2024(2): 36-43., articleTitle=基于VCA-UNet的全自动指针式仪表读数方法, refAbstract=null), Reference(id=1236348231492366819, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, doi=null, pmid=null, pmcid=null, year=2024, volume=null, issue=2, pageStart=36, pageEnd=43, url=null, language=null, rfNumber=[10], rfOrder=11, authorNames=LIU Yubo, TUSONGJIANG Kari, YILIHAMU Yaermaimaiti, journalName=Instrument Technique and Sensor, refType=null, unstructuredReference=LIU Yubo, TUSONGJIANG Kari, YILIHAMU Yaermaimaiti, et al. Automatic pointer instrument reading method based on VCA-UNet[J]. Instrument Technique and Sensor, 2024(2): 36-43., articleTitle=Automatic pointer instrument reading method based on VCA-UNet, refAbstract=null), Reference(id=1236348231605613032, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, doi=null, pmid=null, pmcid=null, year=2016, volume=23, issue=10, pageStart=1499, pageEnd=1503, url=null, language=null, rfNumber=[11], rfOrder=12, authorNames=ZHANG K P, ZHANG Z P, LI Z F, journalName=IEEE Signal Processing Letters, refType=null, unstructuredReference=ZHANG K P, ZHANG Z P, LI Z F, et al. Joint face detection and alignment using multitask cascaded convolutional networks[J]. IEEE Signal Processing Letters, 2016, 23(10): 1499-1503., articleTitle=Joint face detection and alignment using multitask cascaded convolutional networks, refAbstract=null), Reference(id=1236348231697887725, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, doi=null, pmid=null, pmcid=null, year=2024, volume=41, issue=8, pageStart=311, pageEnd=318, url=null, language=null, rfNumber=[12], rfOrder=13, authorNames=林鑫, 沈建新, 秦顺, journalName=计算机应用与软件, refType=null, unstructuredReference=林鑫, 沈建新, 秦顺, 等. 改进的RetinaFace复杂构件关键点检测算法[J]. 计算机应用与软件, 2024, 41(8): 311-318., articleTitle=改进的RetinaFace复杂构件关键点检测算法, refAbstract=null), Reference(id=1236348231785968111, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, doi=null, pmid=null, pmcid=null, year=2024, volume=41, issue=8, pageStart=311, pageEnd=318, url=null, language=null, rfNumber=[12], rfOrder=14, authorNames=LIN Xin, SHEN Jianxin, QIN Shun, journalName=Computer Applications and Software, refType=null, unstructuredReference=LIN Xin, SHEN Jianxin, QIN Shun, et al. A keypoints detection algorithm for complex components based on improved retinaface[J]. Computer Applications and Software, 2024, 41(8): 311-318., articleTitle=A keypoints detection algorithm for complex components based on improved retinaface, refAbstract=null), Reference(id=1236348231865659890, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, doi=null, pmid=null, pmcid=null, year=2025, volume=19, issue=3, pageStart=682, pageEnd=692, url=null, language=null, rfNumber=[13], rfOrder=15, authorNames=傅裕, 高树辉, journalName=计算机科学与探索, refType=null, unstructuredReference=傅裕, 高树辉. 改进YOLOv8s-Pose多人姿态估计轻量化模型研究[J]. 计算机科学与探索, 2025, 19(3): 682-692., articleTitle=改进YOLOv8s-Pose多人姿态估计轻量化模型研究, refAbstract=null), Reference(id=1236348231945351669, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, doi=null, pmid=null, pmcid=null, year=2025, volume=19, issue=3, pageStart=682, pageEnd=692, url=null, language=null, rfNumber=[13], rfOrder=16, authorNames=FU Yu, GAO Shuhui, journalName=Journal of Frontiers of Computer Science and Technology, refType=null, unstructuredReference=FU Yu, GAO Shuhui. Research on lightweight model of multi-person pose estimation based on improved YOLOv8s-pose[J]. Journal of Frontiers of Computer Science and Technology, 2025, 19(3): 682-692., articleTitle=Research on lightweight model of multi-person pose estimation based on improved YOLOv8s-pose, refAbstract=null), Reference(id=1236348232025043448, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, doi=null, pmid=null, pmcid=null, year=2022, volume=14, issue=19, pageStart=12274, pageEnd=null, url=null, language=null, rfNumber=[14], rfOrder=17, authorNames=ZHANG Y, GUO Z Y, WU J Q, journalName=Sustainability, refType=null, unstructuredReference=ZHANG Y, GUO Z Y, WU J Q, et al. Real-time vehicle detection based on improved YOLO v5[J]. Sustainability, 2022, 14(19): 12274., articleTitle=Real-time vehicle detection based on improved YOLO v5, refAbstract=null), Reference(id=1236348232087958011, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, doi=null, pmid=null, pmcid=null, year=2022, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=[15], rfOrder=18, authorNames=LI C Y, LI L L, JIANG H L, journalName=arXiv preprint arXiv, refType=null, unstructuredReference=LI C Y, LI L L, JIANG H L, et al. YOLOv6: A single-stage object detection framework for industrial applications[J]. arXiv preprint arXiv, 2022: 2209.02976., articleTitle=YOLOv6: A single-stage object detection framework for industrial applications, refAbstract=null), Reference(id=1236348232167649790, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, doi=null, pmid=null, pmcid=null, year=2023, volume=null, issue=null, pageStart=7464, pageEnd=7475, url=null, language=null, rfNumber=[16], rfOrder=19, authorNames=WANG C, BOCHKOVSKIY A, LIAO H, journalName=null, refType=null, unstructuredReference=WANG C, BOCHKOVSKIY A, LIAO H. YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors[C]//2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2023: 7464-7475., articleTitle=YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors, refAbstract=null), Reference(id=1236348232259924480, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, doi=null, pmid=null, pmcid=null, year=2023, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=[17], rfOrder=20, authorNames=WENG K H, CHU X X, XU X M, journalName=arXiv, refType=null, unstructuredReference=WENG K H, CHU X X, XU X M, et al. EfficientRep: An efficient Repvgg-style convnets with hardware-aware neural network design[J]. arXiv, 2023: 2302.00386., articleTitle=EfficientRep: An efficient Repvgg-style convnets with hardware-aware neural network design, refAbstract=null), Reference(id=1236348232373170691, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, doi=null, pmid=null, pmcid=null, year=2024, volume=60, issue=21, pageStart=18, pageEnd=37, url=null, language=null, rfNumber=[18], rfOrder=21, authorNames=苏妍妍, 邱志良, 李帼, journalName=计算机工程与应用, refType=null, unstructuredReference=苏妍妍, 邱志良, 李帼, 等. 基于深度学习的二维单人姿态估计综述[J]. 计算机工程与应用, 2024, 60(21): 18-37., articleTitle=基于深度学习的二维单人姿态估计综述, refAbstract=null), Reference(id=1236348232444473861, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, doi=null, pmid=null, pmcid=null, year=2024, volume=60, issue=21, pageStart=18, pageEnd=37, url=null, language=null, rfNumber=[18], rfOrder=22, authorNames=SU Yanyan, QIU Zhiliang, LI Guo, journalName=Computer Engineering and Applications, refType=null, unstructuredReference=SU Yanyan, QIU Zhiliang, LI Guo, et al. Review on deep learning-based 2D single-person pose estimation[J]. Computer Engineering and Applications, 2024, 60(21): 18-37., articleTitle=Review on deep learning-based 2D single-person pose estimation, refAbstract=null), Reference(id=1236348232549331462, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, doi=null, pmid=null, pmcid=null, year=2016, volume=null, issue=null, pageStart=483, pageEnd=499, url=null, language=null, rfNumber=[19], rfOrder=23, authorNames=NEWELL A, YANG K Y, DENG J, journalName=null, refType=null, unstructuredReference=NEWELL A, YANG K Y, DENG J. Stacked hourglass networks for human pose estimation[C]//Computer Vision: ECCV 2016, 2016: 483-499., articleTitle=Stacked hourglass networks for human pose estimation, refAbstract=null), Reference(id=1236348232637411849, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, doi=null, pmid=null, pmcid=null, year=2020, volume=null, issue=null, pageStart=5385, pageEnd=5394, url=null, language=null, rfNumber=[20], rfOrder=24, authorNames=CHENG B, XIAO B, WANG J, journalName=null, refType=null, unstructuredReference=CHENG B, XIAO B, WANG J, et al. HigherHRNet: Scale-Aware Representation Learning for Bottom-Up Human Pose Estimation[C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2020: 5385-5394., articleTitle=HigherHRNet: Scale-Aware Representation Learning for Bottom-Up Human Pose Estimation, refAbstract=null), Reference(id=1236348232717103627, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, doi=null, pmid=null, pmcid=null, year=2021, volume=18, issue=null, pageStart=1037, pageEnd=1049, url=null, language=null, rfNumber=[21], rfOrder=25, authorNames=NEFF C, SHETH A, FURGURSON S, journalName=J Real-Time Image Proc, refType=null, unstructuredReference=NEFF C, SHETH A, FURGURSON S, et al. EfficientHRNet[J]. J Real-Time Image Proc, 2021, 18: 1037-1049., articleTitle=EfficientHRNet, refAbstract=null)], funds=[Fund(id=1236348229831422356, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, awardId=2021YFC2902701, language=CN, fundingSource=国家重点研发计划(2021YFC2902701), fundOrder=null, country=null)], companyList=[AuthorCompany(id=1236348222847906731, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, xref=null, ext=[AuthorCompanyExt(id=1236348222856295340, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, companyId=1236348222847906731, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=Changsha Research Institute of Mining and Metallurgy Co, Ltd, Changsha 410012, Hunan, China), AuthorCompanyExt(id=1236348222860489645, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, companyId=1236348222847906731, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=长沙矿冶研究院有限责任公司,湖南 长沙 410012)])], figs=[ArticleFig(id=1236348227235148016, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, language=EN, label=Fig.1, caption=Diagram of YOLOFFM architecture, figureFileSmall=qfkWNwcWfvpMiU4IT183+g==, figureFileBig=Jnf+2Pm/mMpaP8c2DS1NRQ==, tableContent=null), ArticleFig(id=1236348227369365756, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, language=CN, label=图1, caption=YOLOFFM结构图, figureFileSmall=qfkWNwcWfvpMiU4IT183+g==, figureFileBig=Jnf+2Pm/mMpaP8c2DS1NRQ==, tableContent=null), ArticleFig(id=1236348227646189829, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, language=EN, label=Fig.2, caption=Diagram of DAConv architecture, figureFileSmall=Pwetx1f3hC6h56IVXjCDmQ==, figureFileBig=ae/wK22o69eEJhl/mD5fUA==, tableContent=null), ArticleFig(id=1236348227788796174, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, language=CN, label=图2, caption=DAConv结构图, figureFileSmall=Pwetx1f3hC6h56IVXjCDmQ==, figureFileBig=ae/wK22o69eEJhl/mD5fUA==, tableContent=null), ArticleFig(id=1236348227910430999, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, language=EN, label=Fig.3, caption=Diagram of EffQA-FPN architecture, figureFileSmall=fl7h6BNaAa0SX2bhHvgc9w==, figureFileBig=n9c3kMPZ5KDSO3cUqdN/QQ==, tableContent=null), ArticleFig(id=1236348228057231648, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, language=CN, label=图3, caption=EffQA-FPN结构图, figureFileSmall=fl7h6BNaAa0SX2bhHvgc9w==, figureFileBig=n9c3kMPZ5KDSO3cUqdN/QQ==, tableContent=null), ArticleFig(id=1236348228162089252, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, language=EN, label=Fig.4, caption=Diagram of EQRNeXt architecture, figureFileSmall=48k8uYFEiSrrm8s8Pfl11A==, figureFileBig=Uy3TzbiiBYO0Al6mwYN5Wg==, tableContent=null), ArticleFig(id=1236348228271141166, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, language=CN, label=图4, caption=EQRNeXt结构图, figureFileSmall=48k8uYFEiSrrm8s8Pfl11A==, figureFileBig=Uy3TzbiiBYO0Al6mwYN5Wg==, tableContent=null), ArticleFig(id=1236348228375998772, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, language=EN, label=Fig.5, caption=Diagram of QARepVGG architecture, figureFileSmall=bNS4IJj9hUrUsq5wt+F3KA==, figureFileBig=KtgqSOUBDSGIfBLdxq6g/w==, tableContent=null), ArticleFig(id=1236348228518605115, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, language=CN, label=图5, caption=QARepVGG结构图, figureFileSmall=bNS4IJj9hUrUsq5wt+F3KA==, figureFileBig=KtgqSOUBDSGIfBLdxq6g/w==, tableContent=null), ArticleFig(id=1236348228627657030, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, language=EN, label=Fig.6, caption=Curves of loss, precision, recall, mAP0.5, and mAP0.50~0.95 during training of YOLOv8n-Pose and YOLOFFM, figureFileSmall=IimQ3XYzBYx8hRHIAa+L8A==, figureFileBig=qcwvlFBsnu4YUHAYPKpiug==, tableContent=null), ArticleFig(id=1236348228740903247, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, language=CN, label=图6, caption=YOLOv8n-Pose和YOLOFFM在训练过程中的损失、精确率、召回率、mAP0.5和mAP0.50~0.95对比曲线

(a)训练,矩形框损失;(b)训练,关键点损失;(c)训练,分类损失;(d)验证,精确率;(e)验证,召回率;(f)验证,mAP0.5;(g)验证,mAP0.50~0.95;(h)验证,关键点损失

, figureFileSmall=IimQ3XYzBYx8hRHIAa+L8A==, figureFileBig=qcwvlFBsnu4YUHAYPKpiug==, tableContent=null), ArticleFig(id=1236348228841566549, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, language=EN, label=Fig.7, caption=Practical recognition results of YOLOFFM, figureFileSmall=rHp3mW/5PlRzQnfWiifIqA==, figureFileBig=k1JDiSqWGUY0rvABhYxH6Q==, tableContent=null), ArticleFig(id=1236348228988367194, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, language=CN, label=图7, caption=YOLOFFM真实实例检测结果, figureFileSmall=rHp3mW/5PlRzQnfWiifIqA==, figureFileBig=k1JDiSqWGUY0rvABhYxH6Q==, tableContent=null), ArticleFig(id=1236348229122584929, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, language=EN, label=Table 1, caption=

Environment variables in experiment

, figureFileSmall=null, figureFileBig=null, tableContent=
类别版本/参数
操作系统Windows 11
CPU13th Gen Intel(R)Core(TM)i9-13900K
GPURTX 4090(24 GB)*1
RAM128.0 GB
开发环境Python 3.9.10
深度学习框架PyTorch 2.1.0
加速计算架构CUDA 12.1
), ArticleFig(id=1236348229202276710, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, language=CN, label=表1, caption=

实验环境参数

, figureFileSmall=null, figureFileBig=null, tableContent=
类别版本/参数
操作系统Windows 11
CPU13th Gen Intel(R)Core(TM)i9-13900K
GPURTX 4090(24 GB)*1
RAM128.0 GB
开发环境Python 3.9.10
深度学习框架PyTorch 2.1.0
加速计算架构CUDA 12.1
), ArticleFig(id=1236348229307134318, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, language=EN, label=Table 2, caption=

Ablation study results

, figureFileSmall=null, figureFileBig=null, tableContent=
模型名称参数量/106计算量/109P/%R/%mAP0.5/%mAP0.50~0.95/%
基线模型YOLOv8n2.928.2497.9794.1196.9491.15
YOLOv8n+MSBlock3.189.2397.7294.1597.2292.34
YOLOv8n+EffQA-FPN2.246.7197.6392.8396.6590.92
YOLOv8n+MSBlock-EffQA-FPN2.016.8599.1598.6999.3897.86
), ArticleFig(id=1236348229416186232, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, language=CN, label=表2, caption=

消融实验结果

, figureFileSmall=null, figureFileBig=null, tableContent=
模型名称参数量/106计算量/109P/%R/%mAP0.5/%mAP0.50~0.95/%
基线模型YOLOv8n2.928.2497.9794.1196.9491.15
YOLOv8n+MSBlock3.189.2397.7294.1597.2292.34
YOLOv8n+EffQA-FPN2.246.7197.6392.8396.6590.92
YOLOv8n+MSBlock-EffQA-FPN2.016.8599.1598.6999.3897.86
), ArticleFig(id=1236348229516849536, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, language=EN, label=Table 3, caption=

Comparison of mainstream algorithms

, figureFileSmall=null, figureFileBig=null, tableContent=
算法参数量/106计算量/109精确率/%召回率/%mAP0.5/%mAP0.50~0.95/%
Hourglass277.80583.3896.7792.4595.8988.46
HRNet28.5071.2596.9092.4295.1885.04
HigherHRNet63.80146.7496.8590.9294.5786.23
YOLOv5-Pose2.607.7097.5493.4296.0489.41
YOLOv6-Pose4.7011.4097.6392.8396.6590.92
YOLOv8n-Pose2.928.2497.9794.1196.9491.15
YOLOFFM2.016.8599.1598.6999.3897.86
), ArticleFig(id=1236348229592347013, tenantId=1146029695717560320, journalId=1235980550691926019, articleId=1236276112595144961, language=CN, label=表3, caption=

各主流算法结果对比

, figureFileSmall=null, figureFileBig=null, tableContent=
算法参数量/106计算量/109精确率/%召回率/%mAP0.5/%mAP0.50~0.95/%
Hourglass277.80583.3896.7792.4595.8988.46
HRNet28.5071.2596.9092.4295.1885.04
HigherHRNet63.80146.7496.8590.9294.5786.23
YOLOv5-Pose2.607.7097.5493.4296.0489.41
YOLOv6-Pose4.7011.4097.6392.8396.6590.92
YOLOv8n-Pose2.928.2497.9794.1196.9491.15
YOLOFFM2.016.8599.1598.6999.3897.86
)], attaches=null, journal=Journal(id=1235979682521329665, delFlag=0, nameCn=矿冶工程杂志, nameEn=Mining and Metallurgical Engineering, nameHistory1=null, nameHistory2=null, issn=0253-6099, eissn=null, cn=43-1104/TD, coden=null, periodic=1, language=CN, oaType=null, 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=sIAtluqp1t366oKMF09POA==, journalPrice=null, startedYear=null, abbrevIsoEn=Mining and Metallurgical Engineering, journalRemark=null, publicationField=null, createdTime=1772610119708, updatedTime=1772610494020, createdBy=18614031015, updatedBy=13701087609, firstLetterCn=M, firstLetterEn=M, subjectCode=Engineering, subjectName=null, subjectCodeEn=Engineering, subjectNameEn=null, picCn=sIAtluqp1t366oKMF09POA==, picEn=8lFm/e3VFuRPvwuEPHk4Qg==, jcr=null, cjcr=null, exts=[JournalExt(id=1235981252583551310, language=CN, name=矿冶工程杂志, nameHistory1=null, nameHistory2=null, managedBy=, sponsoredBy=, publishedBy=, editorOffice=, officeProv=null, officeCity=null, officeAddr=, officeZip=, editDirector=, officeDirector=null, officePhone=null, coverPicUrl=null, journalRemark=, submitArticleUrl=null, websiteUrl=, createdTime=1772610494037, updatedTime=1772610494037, createdBy=13701087609, updatedBy=13701087609, submissionGuidelinesUrl=, submissionAuthorUrl=http://www.kygczz.com/Journalx_kygc/authorLogOn.action, submissionEditorUrl=http://www.kygczz.com/Journalx_kygc/editorLogOn.action, submissionReviewUrl=http://www.kygczz.com/Journalx_kygc/expertLogOn.action, submissionCeEditorUrl=, submissionAeEditorUrl=, option={"copyright":""}), JournalExt(id=1235981252633882959, language=EN, name=Mining and Metallurgical Engineering, nameHistory1=null, nameHistory2=null, managedBy=, sponsoredBy=, publishedBy=, editorOffice=, officeProv=null, officeCity=null, officeAddr=, officeZip=, editDirector=, officeDirector=null, officePhone=null, coverPicUrl=null, journalRemark=, submitArticleUrl=null, websiteUrl=, createdTime=1772610494049, updatedTime=1772610494049, createdBy=13701087609, updatedBy=13701087609, submissionGuidelinesUrl=, submissionAuthorUrl=http://www.kygczz.com/Journalx_kygc/authorLogOn.action, submissionEditorUrl=http://www.kygczz.com/Journalx_kygc/editorLogOn.action, submissionReviewUrl=http://www.kygczz.com/Journalx_kygc/expertLogOn.action, submissionCeEditorUrl=, submissionAeEditorUrl=, option={"copyright":""})], databaseList=null, tenantJournalId=1235980550691926019, websiteList=[Website(id=1235982099019256290, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1235980550691926019, 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/kygczz/CN, language=CN, createTime=1772610695843, createBy=18614031015, updateTime=1772610913564, updateBy=18614031015, name=矿冶工程杂志-中文, tplId=1146099689490845704, title=矿冶工程杂志, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1235983631420805700, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1235982099019256290, code=articleTextType, value=kx, createTime=1772611061196, updateTime=1772611061196, creator=18614031015, updator=18614031015), WebsiteProps(id=1235983631378862657, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1235982099019256290, code=banner, value=null, createTime=1772611061186, updateTime=1772611061186, creator=18614031015, updator=18614031015), WebsiteProps(id=1235983631441777223, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1235982099019256290, code=grayFlag, value=0, createTime=1772611061201, updateTime=1772611061201, creator=18614031015, updator=18614031015), WebsiteProps(id=1235983631370474048, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1235982099019256290, code=logo, value=https://castjournals.cast.org.cn/joweb/kygczz/CN/file/pic?fileId=seRpqoNiVtKzuWGeM8/PzA==, createTime=1772611061184, updateTime=1772611061184, creator=18614031015, updator=18614031015), WebsiteProps(id=1235983631454360137, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1235982099019256290, code=minRunFlag, value=0, createTime=1772611061204, updateTime=1772611061204, creator=18614031015, updator=18614031015), WebsiteProps(id=1235983631408222787, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1235982099019256290, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/kygczz/CN/file/pic, createTime=1772611061193, updateTime=1772611061193, creator=18614031015, updator=18614031015), WebsiteProps(id=1235983631445971528, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1235982099019256290, code=silenceFlag, value=0, createTime=1772611061202, updateTime=1772611061202, creator=18614031015, updator=18614031015), WebsiteProps(id=1235983631387251266, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1235982099019256290, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_cn_619/, createTime=1772611061188, updateTime=1772611061188, creator=18614031015, updator=18614031015), WebsiteProps(id=1235983631429194309, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1235982099019256290, code=themeColor, value=null, createTime=1772611061198, updateTime=1772611061198, creator=18614031015, updator=18614031015), WebsiteProps(id=1235983631433388614, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1235982099019256290, code=themeStyle, value=null, createTime=1772611061199, updateTime=1772611061199, creator=18614031015, updator=18614031015)]), Website(id=1235982099262525932, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1235980550691926019, 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/kygczz/EN, language=EN, createTime=1772610695901, createBy=18614031015, updateTime=1772610917908, updateBy=18614031015, name=矿冶工程杂志-英文, tplId=1146101810881728533, title=Mining and Metallurgical Engineering, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1235983658880913998, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1235982099262525932, code=articleTextType, value=kx, createTime=1772611067743, updateTime=1772611067743, creator=18614031015, updator=18614031015), WebsiteProps(id=1235983658864136779, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1235982099262525932, code=banner, value=null, createTime=1772611067739, updateTime=1772611067739, creator=18614031015, updator=18614031015), WebsiteProps(id=1235983658897691217, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1235982099262525932, code=grayFlag, value=0, createTime=1772611067747, updateTime=1772611067747, creator=18614031015, updator=18614031015), WebsiteProps(id=1235983658855748170, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1235982099262525932, code=logo, value=https://castjournals.cast.org.cn/joweb/kygczz/EN/file/pic?fileId=seRpqoNiVtKzuWGeM8/PzA==, createTime=1772611067737, updateTime=1772611067737, creator=18614031015, updator=18614031015), WebsiteProps(id=1235983658910274131, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1235982099262525932, code=minRunFlag, value=0, createTime=1772611067750, updateTime=1772611067750, creator=18614031015, updator=18614031015), WebsiteProps(id=1235983658876719693, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1235982099262525932, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/kygczz/EN/file/pic, createTime=1772611067742, updateTime=1772611067742, creator=18614031015, updator=18614031015), WebsiteProps(id=1235983658901885522, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1235982099262525932, code=silenceFlag, value=0, createTime=1772611067748, updateTime=1772611067748, creator=18614031015, updator=18614031015), WebsiteProps(id=1235983658872525388, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1235982099262525932, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_en_623/, createTime=1772611067741, updateTime=1772611067741, creator=18614031015, updator=18614031015), WebsiteProps(id=1235983658889302607, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1235982099262525932, code=themeColor, value=null, createTime=1772611067745, updateTime=1772611067745, creator=18614031015, updator=18614031015), WebsiteProps(id=1235983658893496912, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1235982099262525932, code=themeStyle, value=null, createTime=1772611067746, updateTime=1772611067746, creator=18614031015, updator=18614031015)])], journalTitle=矿冶工程杂志, weixinUrl=null, journalUrl=http://www.kygczz.com/, iacademicId=null, status=1, seqNo=null, journalTitleEn=Mining and Metallurgical Engineering, journalPhotoCn=sIAtluqp1t366oKMF09POA==, journalPhotoEn=8lFm/e3VFuRPvwuEPHk4Qg==, journalFirstLetter=M, 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=, provinceCode=null, provinceName=null, collectFlag=false), detailUrlCn=https://castjournals.cast.org.cn/joweb/kygczz/CN/10.3969/j.issn.0253-6099.2025.04.024, detailUrlEn=https://castjournals.cast.org.cn/joweb/kygczz/EN/10.3969/j.issn.0253-6099.2025.04.024, pdfUrlCn=https://castjournals.cast.org.cn/joweb/kygczz/CN/PDF/10.3969/j.issn.0253-6099.2025.04.024, pdfUrlEn=https://castjournals.cast.org.cn/joweb/kygczz/EN/PDF/10.3969/j.issn.0253-6099.2025.04.024, aliStartDate=null, aliEndDate=null, collectionFlag=false, citedCount=null, citedUrl=null, reference=null)
收藏切换
基于YOLOFFM的锂电正极材料烧结过程浮子流量视觉识别研究
收藏切换
PDF下载
黄勇 , 刘洋 , 夏星 , 肖盛旺 , 李然
矿冶工程杂志 | 冶金 2025,45(4): 130-136
收起
收藏切换
矿冶工程杂志 | 冶金 2025, 45(4): 130-136
基于YOLOFFM的锂电正极材料烧结过程浮子流量视觉识别研究
全屏
黄勇 , 刘洋, 夏星, 肖盛旺, 李然
作者信息
  • 长沙矿冶研究院有限责任公司,湖南 长沙 410012
  • 黄勇(1990—),男,湖南益阳人,硕士,高级工程师,主要研究方向为智能计算及选冶过程自动化。E-mail:

Visual Recognition of Float Flowmeter Readings in Sintering Process of Lithium-Ion Battery Cathode Materials Based on YOLOFFM
Yong HUANG , Yang LIU, Xing XIA, Shengwang XIAO, Ran LI
Affiliations
  • Changsha Research Institute of Mining and Metallurgy Co, Ltd, Changsha 410012, Hunan, China
出版时间: 2025-08-01 doi: 10.3969/j.issn.0253-6099.2025.04.024
文章导航
收藏切换

针对使用烧结窑炉生产新能源材料过程中浮子流量计在低流量和复杂环境下自动读取的准确性与效率问题,提出了一种改进的YOLOv8n-Pose关键点浮子流量计自动识别(YOLOFFM)算法。该算法的核心改进包括重构主干网络、增强颈部结构、引入非对称压缩解耦头和优化损失函数,这些措施显著提升了算法的效率和检测性能。结果表明,YOLOFFM算法准确率达99.15%,召回率为98.69%,显著降低了模型的参数量和计算成本。与多种先进模型算法相比,YOLOFFM在提升精度的同时,计算成本降低90%以上,充分展示了其在烧结窑炉复杂环境中的高效性和可靠性。

YOLOFFM  /  自动读数  /  MSBlock  /  EffQA-FPN  /  烧结窑炉  /  浮子流量  /  视觉识别  /  算法

To address problems of accuracy and efficiency in automatically reading of float flowmeter at low flow rates under a complex environment in sintering kilns for new energy materials, an innovative YOLOFFM algorithm was proposed. The core improvements include reconstructing the backbone network, enhancing the neck structure, introducing an asymmetric compression decoupling head, and optimizing the loss function, which can significantly improve the efficiency and detection performance of the algorithm. The results show that this YOLOFFM algorithm has accuracy up to 99.15% and a recall rate of 98.69%, significantly reducing the number of model parameters and computational costs. Compared with various advanced algorithm models, YOLOFFM can improve accuracy while reducing the computational cost by more than 90%, fully demonstrating its high efficiency and reliability for new energy materials in a complex environment of sintering kilns.

YOLOFFM  /  automatic reading  /  MSBlock  /  EffQA-FPN  /  sintering kilns  /  float flowmeter  /  visual recognition  /  algorithm
黄勇, 刘洋, 夏星, 肖盛旺, 李然. 基于YOLOFFM的锂电正极材料烧结过程浮子流量视觉识别研究. 矿冶工程杂志, 2025 , 45 (4) : 130 -136 . DOI: 10.3969/j.issn.0253-6099.2025.04.024
Yong HUANG, Yang LIU, Xing XIA, Shengwang XIAO, Ran LI. Visual Recognition of Float Flowmeter Readings in Sintering Process of Lithium-Ion Battery Cathode Materials Based on YOLOFFM[J]. Mining and Metallurgical Engineering, 2025 , 45 (4) : 130 -136 . DOI: 10.3969/j.issn.0253-6099.2025.04.024
在使用烧结窑炉生产新能源材料过程中,流量的精确测量与控制至关重要,尤其是在处理低流速气体和液体时。每个烧结窑炉通常配备超过30个浮子流量计,而一个车间可能有数十台窑炉,这就意味着需要管理上千个流量计的读数,工作量相当庞大。浮子流量计用于确保原材料和气体的流动符合工艺要求。然而,烧结窑炉的工业生产常常处于高温、腐蚀性气体及其他恶劣环境中,人工监测流量不仅危险而且不切实际。因此,采用自动读数方法尤为必要[1]。利用检测机器人和计算机视觉技术进行自动读表不仅能提高数据采集的效率和准确性,还能减少人员在危险环境中的暴露时间[2-3]。这种自动化方法在新能源材料的生产应用中具有重大意义,有助于提升整体生产安全性和效率。
随着计算机硬件性能的提升和深度学习技术的发展,越来越多的深度学习算法被应用于仪表自动读数领域。然而,现有通过机器视觉实现浮子流量计自动读数的研究仍处于初期阶段。将先进的机器视觉技术应用于浮子流量计的自动监控系统,有望成为未来研究热点。
目前,数字和指针仪表识别领域已取得显著进展。数字仪表识别主要通过文字定位模型和光学字符识别(OCR)技术识别仪表上的文字[4-5]。指针仪表识别中,深度学习方法如Mask R-CNN、Faster R-CNN和EAST等算法在复杂自然场景下表现出卓越的鲁棒性[6-10]。尽管这些方法在指针仪表识别中取得了显著成效,但其高计算复杂度和资源消耗使得实时性和实际应用受到限制。
现有方法主要针对数字和指针式仪表,未能有效解决浮子流量计角点和浮子关键点的检测问题。当前领域中关于利用机器学习技术实现浮子流量计自动读数的研究仍处于初步阶段,相关工作较少或几乎空白。因此,要实现浮子流量计的自动化读数,需要首先完成图像校正,精准识别浮子流量计的4个角点及浮子位置,为后续处理提供可靠基础。关键点检测技术的进步为浮子流量计自动读数提供了可靠支持。这类技术最初主要应用于人脸对齐和人体姿态估计领域,近年来逐渐扩展至更多场景。当前广泛采用的算法包括MTCNN、RetinaFace和YOLOv8n-Pose[11-13]。特别是YOLOv8n-Pose,以其高精度、多目标检测能力及卓越的实时性能,成为浮子流量计关键点检测的优选方案。
YOLOv8n-Pose是目前YOLO系列最新的轻量级变体,在目标检测和姿态估计任务中表现出色,尤其适用于资源受限的实时监测场景。与前代版本YOLOv5-Pose、YOLOv6-Pose和YOLOv7-Pose[14-16]相比,YOLOv8n-Pose通过优化网络结构和改进训练策略,在速度和准确性之间实现了更佳平衡。
尽管如此,将YOLOv8n-Pose应用于浮子流量计检测仍存在挑战。在烧结窑炉的工业环境中,浮子流量计用于流量监测,但由于小目标特性和复杂环境,自动读数面临挑战。浮子尺寸小,易受背景干扰,光照变化、腐蚀气体和灰尘等因素影响图像质量,导致检测困难。针对这些问题,本文提出了一种改进的YOLOv8n-Pose关键点浮子流量计自动识别(YOLO float flow meter,YOLOFFM)算法,以实现浮子流量计关键点的高效、精准检测。
YOLOFFM结构如图1所示。YOLOFFM算法对C2f模块进行了部分替换,采用了新颖的混合尺度模块(MixScale Block,MSBlock),设计了EffQA-FPN网络结构以整合多尺度特征信息。
该YOLOFFM框架首先进行预处理,将输入图像调整尺寸并归一化至[0,1]的像素范围内。本研究中,预处理后的图像尺寸为(640,640,3)。这些图像随后被输入到YOLOFFM的主干网络中。最初,网络执行了2次标准卷积操作,生成了尺寸为(320,320,64)和(160,160,128)的特征图。接着,进行了3次连续的MSBlock特征提取操作,生成的特征图尺寸分别为(80,80,256)、(40,40,512)和(20,20,1 024)。随后,采用MSBlock+SPPF操作进一步优化这些特征,保持输出尺寸为(20,20,1 024)。在特征提取阶段,YOLOFFM识别了3个主要的特征层,分别位于不同的深度,即中层、中底层和底层,分别记为F1、F2和F3,对应的尺寸分别为(80,80,256)、(40,40,512)和(20,20,1024)。接下来,EffQA-FPN网络在架构的颈部被构建。通过ConvTranspose操作在F3层进行上采样并与F2层融合,生成尺寸为(40,40,512)的新特征图。随后,EQRNeXt(Enhanced QARepNeXt)模块对该特征图进行优化,生成F4层,尺寸同样为(40,40,512)。F4层再次进行上采样,并与F1层融合,EQRNeXt再次被使用,生成尺寸为(80,80,256)的out1特征层。该层通过3×3卷积进行下采样,并与F4层融合,生成尺寸为(40,40,768)的F4.5特征图。通过EQRNeXt模块的下一轮特征提取,生成尺寸为(40,40,512)的out2层。最后,通过3×3卷积对该层进行降维,并与F3层融合,之后由EQRNeXt提取最终的out3特征层,尺寸为(20,20,1 024)。EffQA-FPN网络最终生成了3个关键的特征图,尺寸分别为(20,20,1 024)、(40,40,512)和(80,80,256)。这些特征图随后被送入YOLOFFM的头部网络,生成最终的预测结果。
YOLOv8n的C2f模块在目标检测中性能优异,但在处理复杂场景和小目标检测时存在局限性,如高参数量和训练复杂度。工业图像中背景复杂、目标尺寸小,导致C2f模块难以有效提取特征。为解决这一问题,本文提出了MSBlock进一步优化特征提取,提高复杂场景和小目标检测性能,同时保持较高的计算效率。
扩张空洞卷积(dilated atrous convolution,DAConv)结构如图2所示。MSBlock首先对输入特征图进行1×1卷积操作,压缩通道数并实现初步特征融合。随后,特征图被划分为3个并行分支:Sub1、Sub2和Sub3,分别处理不同尺度的特征信息。Sub1通过1×1卷积进一步压缩通道数,并应用膨胀率为1的3×3膨胀卷积以扩展感受野,增强对小尺度局部特征的捕捉能力。之后,特征图通过1×1卷积进行特征融合,输出经过优化的小尺度特征。Sub2采用膨胀率为2的5×5膨胀卷积,有效捕捉中等尺度特征信息,增强对中尺度目标的检测能力。Sub3通过膨胀率为3的7×7膨胀卷积,专门提取大尺度特征和全局上下文信息,适用于大型目标的检测。
各分支完成特征提取后,所有特征图通过拼接操作合并为通道数为3C的特征图,实现多尺度信息的整合。最终,融合后的特征图通过1×1卷积压缩通道数,恢复到原始输入的通道维度,并进行批量归一化(BN)处理,生成最终输出特征图。该设计不仅实现了多尺度特征的有效融合,还通过卷积操作进一步精炼了特征表达,使模型在保持计算效率的同时,显著扩展感受野,提高对不同尺度目标的检测性能,特别在多尺度目标检测任务中表现优异。
YOLOv8n-Pose的颈部采用PANet(path aggregation network)整合不同尺度特征,但其简单归纳的方式在处理复杂背景下的小目标检测时效果有限,特别是在浮子流量计图像中。
为解决这一问题,文献[17]提出重复特征金字塔网络(Rep-PAN),通过重复结构和自适应特征聚合策略提升检测效率和精度。受此启发,本研究提出了EffQA-FPN,一种基于QARepVGG的特征金字塔网络,针对浮子流量计关键点检测任务进行优化,整合QARepNeXt和转置卷积等组件,在检测精度和计算效率间达到最佳平衡。该方法不仅提高了小目标检测能力,还增强了复杂背景下的鲁棒性。EffQA-FPN结构如图3所示。
基于EfficientRep网络,本研究提出EQRNeXt,设计了一个多路径特征处理模块。EQRNeXt结构如图4所示。主干路径通过QARepVGG模块处理输入特征,SE模块增强通道注意力;辅助路径采用DAConv捕获多尺度特征,并通过1×1卷积整合通道。两条路径的特征经过融合,通过1×1卷积生成输出特征。这种设计显著增强了网络特征提取能力和表达灵活性,特别适用于浮子流量计等关键点检测任务。
QARepVGG结构如图5所示。QARepVGG卷积结构由3×3卷积、1×1卷积、恒等映射和批量归一化(BN)层组成,推理时通过重参数化转换为单一3×3卷积。为避免协变量偏移,影响量化性能,QARepVGG增加了额外的BN操作,移除了1×1卷积和恒等层后的BN。基于此,本研究优化设计了EQRNeXt模块,增强了模型在量化场景下的适应性和性能。
为解决现有浮子流量计数据集在数量和质量方面的局限性,构建了一个大规模、高质量的数据集YOLOFFM。创建过程如下。
1)选取几种代表性浮子流量计,用专业相机从不同角度和浮子位置拍摄约100张高清图像,并用Labelme软件对这些图像进行精细标注,包括4个角点和5个浮子的精确位置。
2)采用创新的合成方法,随机选取1~5张标注的浮子流量计图像,将其合成到来自MINC数据集的各种背景纹理上。MINC数据集包含57 000张具有丰富纹理变化的图像,提供了广泛的背景选择。
3)对合成图像进行透视变换、旋转和光照变化等数据增强处理,以提高数据集的鲁棒性。最终生成的数据集包括55 000张训练图像和2 000张测评图像。
为全面评估YOLOFFM模型的性能,采用多维度的评估指标体系进行评价。该体系包括平均精度(mean average precision,mAP)、IoU阈值为0.5时的mAP(mAP0.5)、OKS阈值从0.50到0.95(步长为0.05)的多个阈值范围内的mAP值(mAP0.50~0.95)、召回率(R)、精确率(P)、参数大小(P1)以及每秒十亿次浮点运算数(GFLOPs)。
实验环境参数如表1所示。模型训练配置包括:①输入图像尺寸640像素×640像素,总批次数645,训练轮数100,每批次图像数16;②优化器选用Adam算法,学习率初始值0.02,动量系数0.937,权重衰减系数0.005。
在浮子流量计中,角点检测和浮子位置预测是关键任务。为了评估预测结果的准确性,采用对象关键点相似性(object keypoint similarity,OKS)作为主要指标[18],其表达式为:
式中:di为第i个预测关键点与实际角点之间的欧氏距离;s为浮子尺度;ki为关键点i的重要性常数;vi为关键点i的可见性标志;δvi>0)为指示函数。
此外,使用平均精度来评估浮子位置检测的精确度。在特定的OKS阈值i下,平均精度SAP的计算公式为:
式中:STP为真阳性数量;SFP为假阳性数量;SFN为假阴性数量。
召回率(R)用于评估模型在所有正样本中正确检测出的比例,反映了模型的检测能力,其计算公式为:
精确率(P)用于评估模型在所有检测出的样本中正样本的比例,反映了模型的准确性,其计算公式为:
为了评估各模块在浮子流量计关键点识别算法中的影响,以YOLOv8n作为基线模型,通过不同组合方式引入改进模块(MSBlock和EffQA-FPN)进行消融实验,结果如表2所示。从表2可以看出,相较于基线模型YOLOv8n,加入MSBlock模块后,模型的参数量增加了8.9%,计算量增加了12.0%,模型大小也相应增大,然而,精确率下降了0.25百分点,召回率仅略微提升了0.04百分点,mAP0.5和mAP0.50~0.95分别提升了0.28百分点和1.19百分点。尽管MSBlock模块增强了模型的特征提取能力,但其对性能的提升较为有限。引入EffQA-FPN模块后,参数量和计算量分别下降了23.3百分点和18.6百分点,召回率减少了1.28百分点,精确率、mAP0.5和mAP0.50~0.95略有下降。MSBlock和EffQA-FPN模块结合使用时,模型性能显著提升,参数量减少了31.2%,计算量下降了16.9百分点,召回率和精确率分别提升了4.58百分点和1.18百分点,mAP0.5和mAP0.50~0.95分别增加了2.44百分点和6.71百分点。这表明多模块组合显著提升了检测任务的整体性能。
对多个计算机视觉算法模型在YOLOFFM数据集上的表现进行了全面评估,重点分析了流量计4个角点和浮子位置的精确检测,结果如图6所示。随着训练轮次达到100次,模型收敛且无过拟合迹象。与YOLOv8n-Pose相比,YOLOFFM在训练和验证损失上始终表现出更低的数值,且下降趋势更明显,表明YOLOFFM具有更强的数据拟合能力。
训练过程中精确率、召回率、mAP0.5以及mAP0.50~0.95的对比曲线同样示于图6中。可见YOLOFFM在各指标上的表现均优于YOLOv8n-Pose。
图7展示了YOLOFFM对真实实例的识别结果。可以看出,YOLOFFM在光照不均、背景干扰等复杂工况下依然能够稳定识别关键目标,展现出良好的鲁棒性和适应性。这些表现进一步突显了YOLOFFM在流量计关键点检测任务中的优越性能,验证了其在关键点检测能力方面的显著提升。
为了充分展示改进后YOLOFFM网络的先进性和有效性,使用YOLOFFM数据集训练的结果与其他主流算法(Hourglass[19]、HRNet[20]、HigherHRNet[21]、YOLOv5-Pose[14]、YOLOv6-Pose[15]、YOLOv8n-Pose)进行比较,结果见表3。为了确保公平,所有网络在实验过程中都经过相同的参数进行训练。表3数据显示,相较于其他算法,YOLOFFM在多个方面均具有优越性,包括参数量、计算量、精确率和召回率等。相比于Hourglass,YOLOFFM参数量和计算量分别减少了99.3%和98.8%,精确率和召回率分别提高了2.38和6.24百分点,mAP0.5提升了3.49百分点。与HRNet相比,YOLOFFM参数量减少了92.9%,计算量减少了90.4%,精确率和召回率分别提高了2.25和6.27百分点,mAP0.5提高了4.20百分点。相较于HigherHRNet,YOLOFFM参数量减少了96.8%,计算量减少了95.3%,精确率和召回率分别提高了2.30和7.77百分点,mAP0.5提升了4.81百分点。与YOLOv5-Pose、YOLOv6-Pose和YOLOv8n-Pose的对比也表明YOLOFFM在性能上有显著提升。总体而言,YOLOFFM在大幅减少参数量和计算量的同时,保持了高水平的检测精度和召回率,特别是在mAP0.5指标上表现突出,证明了其在边缘设备上部署以实现浮子流量计关键点检测的实际优势和适用性。
针对烧结窑炉生产新能源材料过程中浮子流量计关键点检测所面临的诸多挑战,创新性提出了YOLOFFM算法。通过对YOLOv8n-Pose进行深度优化,YOLOFFM有效解决了现有方法在特征提取不足、网络复杂度高、资源受限、设备部署困难以及检测结果不准确等问题。其核心改进包括重构主干网络、增强颈部结构、引入非对称压缩解耦头和优化损失函数,这些措施显著提升了算法的效率和检测性能。实验结果表明,YOLOFFM在保持高精度的同时,展现出良好的硬件适应性和量化鲁棒性,为新能源材料烧结窑炉中浮子流量计的自动化读数提供了可靠解决方案。尽管在处理复杂场景时仍有改进空间,YOLOFFM的整体表现标志着浮子流量计关键点检测技术的重要进展。未来研究将致力于进一步提高算法在复杂环境下的鲁棒性,以及探索其在更广泛领域的工业应用。
  • 国家重点研发计划(2021YFC2902701)
参考文献 引证文献
排序方式:
[1]
TSAI M C, KO P J. On-line condition monitoring of servo motor drive systems by HHT in Industry 4.0[J]. Journal of the Chinese Institute of Engineers, 2017, 40(7): 572-584.
[2]
GAO J W, XIE H T, ZUO L, et al. A robust pointer meter reading recognition method for substation inspection robot[C]//International Conference on Robotics and Automation Sciences (ICRAS), 2017: 43-47.
[3]
LU S Y, ZHANG Y, SU J J. Mobile robot for power substation inspection: A survey[J]. IEEE/CAA Journal of Automatica Sinica, 2017, 4(4): 830-847.
[4]
康朝海, 刘杨, 任伟建, 等. 基于MaskR-CNN的油田井场指针仪表识别方法研究[J]. 工业仪表与自动化装置, 2024(5): 76-82.
KANG Chaohai, LIU Yang, REN Weijian, et al. Research on MaskR-CNN-based pointer instrument recognition method for oilfield wellsite[J]. Industrial Instrumentation & Automation, 2024(5): 76-82.
[5]
LIU Y L, CHEN H, SHEN C H, et al. ABCNet: Real-time scene text spotting with Adaptive Bezier-Curve Network[C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2020: 9806-9815.
[6]
HE P, ZUO L, ZHANG C, et al. A value recognition algorithm for pointer meter based on improved Mask-RCNN[C]/2019 9th International Conference on Information Science and Technology (ICIST), 2019: 108-113.
[7]
ZHANG X, DANG X, LYU Q, et al. A Pointer meter recognition algorithm based on deep learning[C]/2020 3rd International Conference on Advanced Electronic Materials, Computers and Software Engineering (AEMCSE), 2020: 283-287.
[8]
ZHOU D K, YANG Y, ZHU J, et al. Tilt correction method of pointer meter based on deep learning[J]. Journal of Computer-Aided Design & Computer Graphics, 2020, 32(12): 1976-1984.
[9]
LIU H, HUANG Z L, TIAN L, et al. Method for detecting and recognizing pointer meter reading based on deep learning[C]/Third International Conference on Electronics and Communication; Network and Computer Technology (ECNCT 2021). 2022: 14.
[10]
刘煜博, 吐松江·卡日, 伊力哈木·亚尔买买提, 等. 基于VCA-UNet的全自动指针式仪表读数方法[J]. 仪表技术与传感器, 2024(2): 36-43.
LIU Yubo, TUSONGJIANG Kari, YILIHAMU Yaermaimaiti, et al. Automatic pointer instrument reading method based on VCA-UNet[J]. Instrument Technique and Sensor, 2024(2): 36-43.
[11]
ZHANG K P, ZHANG Z P, LI Z F, et al. Joint face detection and alignment using multitask cascaded convolutional networks[J]. IEEE Signal Processing Letters, 2016, 23(10): 1499-1503.
[12]
林鑫, 沈建新, 秦顺, 等. 改进的RetinaFace复杂构件关键点检测算法[J]. 计算机应用与软件, 2024, 41(8): 311-318.
LIN Xin, SHEN Jianxin, QIN Shun, et al. A keypoints detection algorithm for complex components based on improved retinaface[J]. Computer Applications and Software, 2024, 41(8): 311-318.
[13]
傅裕, 高树辉. 改进YOLOv8s-Pose多人姿态估计轻量化模型研究[J]. 计算机科学与探索, 2025, 19(3): 682-692.
FU Yu, GAO Shuhui. Research on lightweight model of multi-person pose estimation based on improved YOLOv8s-pose[J]. Journal of Frontiers of Computer Science and Technology, 2025, 19(3): 682-692.
[14]
ZHANG Y, GUO Z Y, WU J Q, et al. Real-time vehicle detection based on improved YOLO v5[J]. Sustainability, 2022, 14(19): 12274.
[15]
LI C Y, LI L L, JIANG H L, et al. YOLOv6: A single-stage object detection framework for industrial applications[J]. arXiv preprint arXiv, 2022: 2209.02976.
[16]
WANG C, BOCHKOVSKIY A, LIAO H. YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors[C]//2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2023: 7464-7475.
[17]
WENG K H, CHU X X, XU X M, et al. EfficientRep: An efficient Repvgg-style convnets with hardware-aware neural network design[J]. arXiv, 2023: 2302.00386.
[18]
苏妍妍, 邱志良, 李帼, 等. 基于深度学习的二维单人姿态估计综述[J]. 计算机工程与应用, 2024, 60(21): 18-37.
SU Yanyan, QIU Zhiliang, LI Guo, et al. Review on deep learning-based 2D single-person pose estimation[J]. Computer Engineering and Applications, 2024, 60(21): 18-37.
[19]
NEWELL A, YANG K Y, DENG J. Stacked hourglass networks for human pose estimation[C]//Computer Vision: ECCV 2016, 2016: 483-499.
[20]
CHENG B, XIAO B, WANG J, et al. HigherHRNet: Scale-Aware Representation Learning for Bottom-Up Human Pose Estimation[C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2020: 5385-5394.
[21]
NEFF C, SHETH A, FURGURSON S, et al. EfficientHRNet[J]. J Real-Time Image Proc, 2021, 18: 1037-1049.
2025年第45卷第4期
PDF下载
44
17
引用本文
BibTeX
文章信息
doi: 10.3969/j.issn.0253-6099.2025.04.024
  • 接收时间:2025-02-25
  • 首发时间:2026-03-05
  • 出版时间:2025-08-01
补充材料
相关文章
文章信息
作者
出版历史
  • 收稿日期:2025-02-25
基金
国家重点研发计划(2021YFC2902701)
作者信息
    长沙矿冶研究院有限责任公司,湖南 长沙 410012
参考文献
分享链接
https://castjournals.cast.org.cn/joweb/kygczz/CN/10.3969/j.issn.0253-6099.2025.04.024
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