Article(id=1208051034803511323, tenantId=1146029695717560320, journalId=1146123166801305609, issueId=1208051024368083510, articleNumber=null, orderNo=null, doi=10.12404/j.issn.1671-1815.2404963, pmid=null, cstr=null, oa=null, hot=null, price=null, onlineType=0, articleFormat=0, articleType=null, articleTypeStr=research-article, receivedDate=1719936000000, receivedDateStr=2024-07-03, revisedDate=1742313600000, revisedDateStr=2025-03-19, acceptedDate=null, acceptedDateStr=null, onlineDate=1765951411200, onlineDateStr=2025-12-17, pubDate=1751040000000, pubDateStr=2025-06-28, doiRegisterDate=null, doiRegisterDateStr=null, onlineIssueDate=1765951411200, onlineIssueDateStr=2025-12-17, onlineJustAcceptDate=null, onlineJustAcceptDateStr=null, onlineFirstDate=null, onlineFirstDateStr=null, sourceXml=null, magXml=null, createTime=1765951411200, creator=13701087609, updateTime=1765951411200, updator=13701087609, issue=Issue{id=1208051024368083510, tenantId=1146029695717560320, journalId=1146123166801305609, year='2025', volume='25', issue='18', pageStart='7455', pageEnd='7883', issueExtLink='null', onlineDate='null', pubDate='null', beforeIssueId=null, nextIssueId=null, price=null, status=1, issueComplete=1, articleOrder=1, issueType=-1, specialIssue=null, createTime=1765951408712, creator=13701087609, updateTime=1765951896766, updator=13701087609, preIssue=null, nextIssue=null, ext={EN=IssueExt(id=1208053071507198943, tenantId=1146029695717560320, journalId=1146123166801305609, issueId=1208051024368083510, language=EN, specialIssueTitle=, coverIllustrator=null, specialIssueEditor=, specialIssueAbout=), CN=IssueExt(id=1208053071507198944, tenantId=1146029695717560320, journalId=1146123166801305609, issueId=1208051024368083510, language=CN, specialIssueTitle=, coverIllustrator=null, specialIssueEditor=, specialIssueAbout=)}, issueFiles=null}, startPage=7785, endPage=7792, ext={EN=ArticleExt(id=1208051035420074075, articleId=1208051034803511323, tenantId=1146029695717560320, journalId=1146123166801305609, language=EN, title=Intelligent Identification Algorithm for Typical Rail Damage Based on Improved YOLOv8, columnId=1156262728772735295, journalTitle=Science Technology and Engineering, columnName=Papers·Traffics and Transportations, runingTitle=null, highlight=null, articleAbstract=

The internal damage of the steel rail is serious, but the non-destructive testing B-display detection image has a lot of noise and noise, and the spatiotemporal distribution characteristics of different damages are not obvious, making it difficult to effectively identify. In response to this situation, a rail screw hole crack B-image recognition algorithm based on improved YOLOv8 was studied to improve the accuracy of intelligent identification of rail damage. Firstly, to reduce the missed detection of small damage targets, RepHGNetv2 network was used to optimize the YOLOv8 backbone network and improve the detection recall rate. Then, in order to improve the adaptability of the model to different types of damage detection, the detection head of YOLOv8 was replaced with Effientnet to improve the detection accuracy of the model. Finally, the LSKA attention mechanism was introduced into the SPPF module to enhance the model’s anti-interference ability against noise signals and improve its accuracy. The actual line detection results have verified that the detection accuracy of the above model reaches 95.1%, the recall rate reaches 93.8%, and the average accuracy reaches 97.6%, which is improved compared to other commonly used algorithms.

, correspAuthors=Gui-yang XU, 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=Hou-xue XIANG, Gui-yang XU, Yu-hua ZHANG, Xiao-yan HUANG), CN=ArticleExt(id=1208051039358525852, articleId=1208051034803511323, tenantId=1146029695717560320, journalId=1146123166801305609, language=CN, title=基于改进YOLOv8的钢轨典型伤损智能识别算法, columnId=1156262730664366426, journalTitle=科学技术与工程, columnName=论文·交通运输, runingTitle=null, highlight=null, articleAbstract=

钢轨内部损伤危害严重,但无损探伤B显检测图像杂波噪声多、不同损伤时空分布特征不明显,难以有效识别。针对这一情况,提出了一种基于改进YOLOv8的钢轨螺孔裂纹B显图识别算法,从提高钢轨损伤智能识别精确率。首先,为减少小伤损目标漏检,采用RepHGNetv2网络优化YOLOv8主干网络,提高检测召回率;然后,为提升模型对不同类型损伤检测的适应性,将YOLOv8的检测头替换为Effcientnet,提高模型检测精度;最后,在SPPF模块中引入LSKA注意力机制,提升模型对噪声信号的抗干扰能力,提高模型精确率。经实际线路检测结果验证,上述模型检测精确率达到95.1%,召回率达到93.8%,平均精度达到97.6%,对比其他常用算法均有所提高。

, correspAuthors=许贵阳, authorNote=null, correspAuthorsNote=
* 许贵阳(1973—),男,满族,辽宁丹东人,博士,教授。研究方向:车辆工程。E-mail:
, copyrightStatement=null, copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=TUcxcVIeCDCznOp4BbVlrg==, magXml=mspFy2Ifo4e6icppcwpkVA==, pdfUrl=null, pdf=yKpo57xVgqj8QnH2M9UOfw==, pdfFileSize=6013867, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=1CUauXYw+NV5yGjScHTItQ==, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=mIsO1YvMi1AfIBjyt3aLvA==, mapNumber=null, authorCompany=null, fund=null, authors=

向厚雪(2002—),女,汉族,河南信阳人,硕士研究生。研究方向:钢轨探伤。E-mail:

, authorsList=向厚雪, 许贵阳, 张玉华, 黄筱妍)}, authors=[Author(id=1208085586242543803, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, orderNo=0, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=1810314336@qq.com, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1208085586326429895, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, authorId=1208085586242543803, language=EN, stringName=Hou-xue XIANG, firstName=Hou-xue, middleName=null, lastName=XIANG, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1 School of Mechanical, Electrical and Vehicle Engineering, Beijing University of Civil Engineering and Architecture, Beijing 102612, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1208085586448064721, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, authorId=1208085586242543803, language=CN, stringName=向厚雪, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1 北京建筑大学机电与车辆工程学院, 北京 102612, bio={"content":"

向厚雪(2002—),女,汉族,河南信阳人,硕士研究生。研究方向:钢轨探伤。E-mail:

"}, bioImg=null, bioContent=

向厚雪(2002—),女,汉族,河南信阳人,硕士研究生。研究方向:钢轨探伤。E-mail:

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1208085584871006358, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, xref=1, ext=[AuthorCompanyExt(id=1208085584879394968, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, companyId=1208085584871006358, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1 School of Mechanical, Electrical and Vehicle Engineering, Beijing University of Civil Engineering and Architecture, Beijing 102612, China), AuthorCompanyExt(id=1208085584887783577, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, companyId=1208085584871006358, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1 北京建筑大学机电与车辆工程学院, 北京 102612)])]), Author(id=1208085586548728028, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, orderNo=1, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=xuguiyang@bucea.edu.cn, emailSecond=null, emailThird=null, correspondingAuthor=1, authorType=1, ext={EN=AuthorExt(id=1208085586661974252, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, authorId=1208085586548728028, language=EN, stringName=Gui-yang XU, firstName=Gui-yang, middleName=null, lastName=XU, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, *, address=1 School of Mechanical, Electrical and Vehicle Engineering, Beijing University of Civil Engineering and Architecture, Beijing 102612, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1208085586791997691, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, authorId=1208085586548728028, language=CN, stringName=许贵阳, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, *, address=1 北京建筑大学机电与车辆工程学院, 北京 102612, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1208085584871006358, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, xref=1, ext=[AuthorCompanyExt(id=1208085584879394968, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, companyId=1208085584871006358, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1 School of Mechanical, Electrical and Vehicle Engineering, Beijing University of Civil Engineering and Architecture, Beijing 102612, China), AuthorCompanyExt(id=1208085584887783577, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, companyId=1208085584871006358, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1 北京建筑大学机电与车辆工程学院, 北京 102612)])]), Author(id=1208085586938798344, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, 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=1208085587106570521, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, authorId=1208085586938798344, language=EN, stringName=Yu-hua ZHANG, firstName=Yu-hua, middleName=null, lastName=ZHANG, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, address=2 Infrastructure Testing Research Institute of China Academy of Railway Sciences,Beijing 100081, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1208085587219816745, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, authorId=1208085586938798344, language=CN, stringName=张玉华, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, address=2 中国铁道科学研究院基础设施检测研究所, 北京 100081, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1208085586091548842, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, xref=2, ext=[AuthorCompanyExt(id=1208085586104131756, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, companyId=1208085586091548842, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2 Infrastructure Testing Research Institute of China Academy of Railway Sciences,Beijing 100081, China), AuthorCompanyExt(id=1208085586108326060, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, companyId=1208085586091548842, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2 中国铁道科学研究院基础设施检测研究所, 北京 100081)])]), Author(id=1208085587345645880, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, 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=1208085587484057929, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, authorId=1208085587345645880, language=EN, stringName=Xiao-yan HUANG, firstName=Xiao-yan, middleName=null, lastName=HUANG, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, address=2 Infrastructure Testing Research Institute of China Academy of Railway Sciences,Beijing 100081, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1208085587601498457, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, authorId=1208085587345645880, language=CN, stringName=黄筱妍, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, address=2 中国铁道科学研究院基础设施检测研究所, 北京 100081, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1208085586091548842, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, xref=2, ext=[AuthorCompanyExt(id=1208085586104131756, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, companyId=1208085586091548842, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2 Infrastructure Testing Research Institute of China Academy of Railway Sciences,Beijing 100081, China), AuthorCompanyExt(id=1208085586108326060, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, companyId=1208085586091548842, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2 中国铁道科学研究院基础设施检测研究所, 北京 100081)])])], keywords=[Keyword(id=1208085587928654203, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, language=EN, orderNo=1, keyword=rail flaw detection), Keyword(id=1208085588054483337, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, language=EN, orderNo=2, keyword=B-imaging), Keyword(id=1208085588218061206, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, language=EN, orderNo=3, keyword=damage detection), Keyword(id=1208085588318724517, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, language=EN, orderNo=4, keyword=YOLOv8), Keyword(id=1208085588436165043, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, language=EN, orderNo=5, keyword=bolt hole crack), Keyword(id=1208085588541022656, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, language=CN, orderNo=1, keyword=钢轨探伤), Keyword(id=1208085588662657485, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, language=CN, orderNo=2, keyword=B显图), Keyword(id=1208085588855595485, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, language=CN, orderNo=3, keyword=损伤检测), Keyword(id=1208085589044339177, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, language=CN, orderNo=4, keyword=YOLOv8), Keyword(id=1208085589203722736, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, language=CN, orderNo=5, keyword=螺孔裂纹)], refs=[Reference(id=1208085596807996366, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, doi=null, pmid=null, pmcid=null, year=2016, volume=null, issue=9, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=[1], rfOrder=0, authorNames=栗英, journalName=现代制造技术与装备, refType=null, unstructuredReference=栗英. 钢轨养护及一种钢轨轮廓检测方法[J]. 现代制造技术与装备, 2016(9): 94, 96., articleTitle=钢轨养护及一种钢轨轮廓检测方法, refAbstract=null), Reference(id=1208085596866716628, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, doi=null, pmid=null, pmcid=null, year=2016, volume=null, issue=9, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=[1], rfOrder=1, authorNames=Li Ying, journalName=Modern Manufacturing Technology and Equipment, refType=null, unstructuredReference=Li Ying. Rail maintenance and a rail contour detection method[J]. Modern Manufacturing Technology and Equipment, 2016(9): 94, 96., articleTitle=Rail maintenance and a rail contour detection method, refAbstract=null), Reference(id=1208085596963185627, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, doi=null, pmid=null, pmcid=null, year=2022, volume=null, issue=1, pageStart=25, pageEnd=27, url=null, language=null, rfNumber=[2], rfOrder=2, authorNames=王炳辉, journalName=减速顶与调速技术, refType=null, unstructuredReference=王炳辉. 普速铁路钢轨探伤及伤损原因分析[J]. 减速顶与调速技术, 2022(1): 25-27., articleTitle=普速铁路钢轨探伤及伤损原因分析, refAbstract=null), Reference(id=1208085597059654627, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, doi=null, pmid=null, pmcid=null, year=2022, volume=null, issue=1, pageStart=25, pageEnd=27, url=null, language=null, rfNumber=[2], rfOrder=3, authorNames=Wang Binghui, journalName=Decelerator and Speed Control Technology, refType=null, unstructuredReference=Wang Binghui. Analysis of defect detection and damage causes of ordinary speed railway rails[J]. Decelerator and Speed Control Technology, 2022(1): 25-27., articleTitle=Analysis of defect detection and damage causes of ordinary speed railway rails, refAbstract=null), Reference(id=1208085597189678060, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, doi=null, pmid=null, pmcid=null, year=2024, volume=24, issue=6, pageStart=2453, pageEnd=2461, url=null, language=null, rfNumber=[3], rfOrder=4, authorNames=彭红星, 袁畅, 柯威曳, journalName=科学技术与工程, refType=null, unstructuredReference=彭红星, 袁畅, 柯威曳, 等. 基于改进YOLOv5的高速公路隧道车辆和人员检测[J]. 科学技术与工程, 2024, 24(6): 2453-2461., articleTitle=基于改进YOLOv5的高速公路隧道车辆和人员检测, refAbstract=null), Reference(id=1208085597349061627, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, doi=null, pmid=null, pmcid=null, year=2024, volume=24, issue=6, pageStart=2453, pageEnd=2461, url=null, language=null, rfNumber=[3], rfOrder=5, authorNames=Peng Hongxing, Yuan Chang, Ke Weiye, journalName=Science Technology and Engineering, refType=null, unstructuredReference=Peng Hongxing, Yuan Chang, Ke Weiye, et al. Vehicle and personnel detection in highway tunnels based on improved YOLOv5[J]. Science Technology and Engineering, 2024, 24(6): 2453-2461., articleTitle=Vehicle and personnel detection in highway tunnels based on improved YOLOv5, refAbstract=null), Reference(id=1208085597533609991, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, doi=null, pmid=null, pmcid=null, year=2024, volume=24, issue=17, pageStart=7205, pageEnd=7213, url=null, language=null, rfNumber=[4], rfOrder=6, authorNames=宣以国, 余成波, 蒋启超, journalName=科学技术与工程, refType=null, unstructuredReference=宣以国, 余成波, 蒋启超, 等. 基于改进YOLOv7的道路裂缝和坑洞检测算法[J]. 科学技术与工程, 2024, 24(17): 7205-7213., articleTitle=基于改进YOLOv7的道路裂缝和坑洞检测算法, refAbstract=null), Reference(id=1208085597663633422, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, doi=null, pmid=null, pmcid=null, year=2024, volume=24, issue=17, pageStart=7205, pageEnd=7213, url=null, language=null, rfNumber=[4], rfOrder=7, authorNames=Xuan Yiguo, Yu Chengbo, Jiang Qichao, journalName=Science Technology and Engineering, refType=null, unstructuredReference=Xuan Yiguo, Yu Chengbo, Jiang Qichao, et al. Improved YOLOv7 road crack and pothole detection algorithm[J]. Science Technology and Engineering, 2024, 24(17): 7205-7213., articleTitle=Improved YOLOv7 road crack and pothole detection algorithm, refAbstract=null), Reference(id=1208085597827211286, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, doi=null, pmid=null, pmcid=null, year=2024, volume=24, issue=32, pageStart=13944, pageEnd=13952, url=null, language=null, rfNumber=[5], rfOrder=8, authorNames=孔建国, 张向伟, 赵志伟, journalName=科学技术与工程, refType=null, unstructuredReference=孔建国, 张向伟, 赵志伟, 等. 基于改进YOLOv8的机场飞鸟实时目标检测方法[J]. 科学技术与工程, 2024, 24(32): 13944-13952., articleTitle=基于改进YOLOv8的机场飞鸟实时目标检测方法, refAbstract=null), Reference(id=1208085597953040414, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, doi=null, pmid=null, pmcid=null, year=2024, volume=24, issue=32, pageStart=13944, pageEnd=13952, url=null, language=null, rfNumber=[5], rfOrder=9, authorNames=Kong Jianguo, Zhang Xiangwei, Zhao Zhiwei, journalName=Science Technology and Engineering, refType=null, unstructuredReference=Kong Jianguo, Zhang Xiangwei, Zhao Zhiwei, et al. Real time object detection method for airport flying birds based on improved YOLOv8[J]. Science Technology and Engineering, 2024, 24(32): 13944-13952., articleTitle=Real time object detection method for airport flying birds based on improved YOLOv8, refAbstract=null), Reference(id=1208085598108229669, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, doi=null, pmid=null, pmcid=null, year=2022, volume=null, issue=null, pageStart=1, pageEnd=8, url=null, language=null, rfNumber=[6], rfOrder=10, authorNames=张永雄, 蔡勇, journalName=第三届中国铁路发展论坛学术论文集, refType=null, unstructuredReference=张永雄, 蔡勇. 多模态人工智能技术在钢轨伤损判定中的探索[C]// 第三届中国铁路发展论坛学术论文集. 北京: 中国铁道学会, 2022: 1-8., articleTitle=多模态人工智能技术在钢轨伤损判定中的探索, refAbstract=null), Reference(id=1208085598263418922, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, doi=null, pmid=null, pmcid=null, year=2022, volume=null, issue=null, pageStart=1, pageEnd=8, url=null, language=null, rfNumber=[6], rfOrder=11, authorNames=Zhang Yongxiong, Cai Yong, journalName=Academic Papers Collection of the 3rd China Railway Development Forum, refType=null, unstructuredReference=Zhang Yongxiong, Cai Yong. Exploration of multimodal artificial intelligence technology in rail damage assessment[C]// Academic Papers Collection of the 3rd China Railway Development Forum. Beijing: China Railway Society, 2022: 1-8., articleTitle=Exploration of multimodal artificial intelligence technology in rail damage assessment, refAbstract=null), Reference(id=1208085599513321529, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, doi=null, pmid=null, pmcid=null, year=2022, volume=36, issue=1, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=[7], rfOrder=12, authorNames=Zhang B, Huang S, Zhang L, journalName=Journal of Performance of Constructed Facilities, refType=null, unstructuredReference=Zhang B, Huang S, Zhang L, et al. Rail defect recognition based on waveform subtraction and rule base[J]. Journal of Performance of Constructed Facilities, 2022, 36(1). DOI:10.1061/(ASCE)CF.1943-5509.0001684., articleTitle=Rail defect recognition based on waveform subtraction and rule base, refAbstract=null), Reference(id=1208085599651733566, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, doi=null, pmid=null, pmcid=null, year=2021, volume=2675, issue=11, pageStart=888, pageEnd=901, url=null, language=null, rfNumber=[8], rfOrder=13, authorNames=Chen Z X, Wang Q H, Yang K, journalName=Transportation Research Record, refType=null, unstructuredReference=Chen Z X, Wang Q H, Yang K, et al. Deep learning for the detection and recognition of rail defects in ultrasound B-scan images[J]. Transportation Research Record, 2021, 2675(11): 888-901., articleTitle=Deep learning for the detection and recognition of rail defects in ultrasound B-scan images, refAbstract=null), Reference(id=1208085599836282951, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, doi=null, pmid=null, pmcid=null, year=2022, volume=44, issue=12, pageStart=82, pageEnd=88, url=null, language=null, rfNumber=[9], rfOrder=14, authorNames=何庆, 陈正兴, 王启航, journalName=铁道学报, refType=null, unstructuredReference=何庆, 陈正兴, 王启航, 等. 基于改进YOLO V3的钢轨伤损B显图像识别研究[J]. 铁道学报, 2022, 44(12): 82-88., articleTitle=基于改进YOLO V3的钢轨伤损B显图像识别研究, refAbstract=null), Reference(id=1208085599949529173, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, doi=null, pmid=null, pmcid=null, year=2022, volume=44, issue=12, pageStart=82, pageEnd=88, url=null, language=null, rfNumber=[9], rfOrder=15, authorNames=He Qing, Chen Zhengxing, Wang Qihang, journalName=Journal of Railway, refType=null, unstructuredReference=He Qing, Chen Zhengxing, Wang Qihang, et al. Research on B-image recognition of rail damage based on improved YOLO V3[J] Journal of Railway, 2022, 44 (12): 82-88., articleTitle=Research on B-image recognition of rail damage based on improved YOLO V3, refAbstract=null), Reference(id=1208085600104718425, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, doi=null, pmid=null, pmcid=null, year=2024, volume=24, issue=32, pageStart=13665, pageEnd=13677, url=null, language=null, rfNumber=[10], rfOrder=16, authorNames=任青阳, 王彦丁, 施俭, journalName=科学技术与工程, refType=null, unstructuredReference=任青阳, 王彦丁, 施俭. 卷积神经网络目标检测算法研究进展[J]. 科学技术与工程, 2024, 24(32): 13665-13677., articleTitle=卷积神经网络目标检测算法研究进展, refAbstract=null), Reference(id=1208085600247324772, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, doi=null, pmid=null, pmcid=null, year=2024, volume=24, issue=32, pageStart=13665, pageEnd=13677, url=null, language=null, rfNumber=[10], rfOrder=17, authorNames=Ren Qingyang, Wang Yanding, Shi Jian, journalName=Science Technology and Engineering, refType=null, unstructuredReference=Ren Qingyang, Wang Yanding, Shi Jian. Advances in target detection algorithms for convolutional neural networks[J]. Science Technology and Engineering, 2024, 24(32): 13665-13677., articleTitle=Advances in target detection algorithms for convolutional neural networks, refAbstract=null), Reference(id=1208085600436068466, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, doi=null, pmid=null, pmcid=null, year=2024, volume=24, issue=6, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=[11], rfOrder=18, authorNames=Zhang L, Zheng J C, Li C P, journalName=Sensors, refType=null, unstructuredReference=Zhang L, Zheng J C, Li C P, et al. CCDN-DETR: a detection transformer based on constrained contrast denoising for multi-class synthetic aperture radar object detection[J]. Sensors, 2024, 24(6). DOI:10.3390/s24061793., articleTitle=CCDN-DETR: a detection transformer based on constrained contrast denoising for multi-class synthetic aperture radar object detection, refAbstract=null), Reference(id=1208085600553508987, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, doi=null, pmid=null, pmcid=null, year=2019, volume=null, issue=null, pageStart=6105, pageEnd=6114, url=null, language=null, rfNumber=[12], rfOrder=19, authorNames=Tan M, Le Q, journalName=International Conference on Machine Learning, refType=null, unstructuredReference=Tan M, Le Q. Efficientnet: rethinking model scaling for convolutional neural networks[C]// International Conference on Machine Learning. Chambridge: PMLR, 2019: 6105-6114., articleTitle=Efficientnet: rethinking model scaling for convolutional neural networks, refAbstract=null), Reference(id=1208085600721281155, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, doi=null, pmid=null, pmcid=null, year=2018, volume=null, issue=null, pageStart=7132, pageEnd=7141, url=null, language=null, rfNumber=[13], rfOrder=20, authorNames=Hu J, Li S, Gang S, journalName=Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, refType=null, unstructuredReference=Hu J, Li S, Gang S. Squeeze-and-excitation networks[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2018: 7132-7141., articleTitle=Squeeze-and-excitation networks, refAbstract=null), Reference(id=1208085600830333066, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, doi=null, pmid=null, pmcid=null, year=2023, volume=13, issue=1, pageStart=69, pageEnd=72, url=null, language=null, rfNumber=[14], rfOrder=21, authorNames=刘洪宁, 李容露, journalName=科技创新与应用, refType=null, unstructuredReference=刘洪宁, 李容露. 基于YOLOv5的钢材数量识别方法[J]. 科技创新与应用, 2023, 13(1): 69-72., articleTitle=基于YOLOv5的钢材数量识别方法, refAbstract=null), Reference(id=1208085600947773587, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, doi=null, pmid=null, pmcid=null, year=2023, volume=13, issue=1, pageStart=69, pageEnd=72, url=null, language=null, rfNumber=[14], rfOrder=22, authorNames=Liu Hongning, Li Ronglu, journalName=Technological Innovation and Application, refType=null, unstructuredReference=Liu Hongning, Li Ronglu. A method for identifying steel quantity based on YOLOv5[J]. Technological Innovation and Application, 2023, 13(1): 69-72., articleTitle=A method for identifying steel quantity based on YOLOv5, refAbstract=null), Reference(id=1208085601052631193, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, doi=null, pmid=null, pmcid=null, year=2019, volume=null, issue=6, pageStart=111, pageEnd=115, url=null, language=null, rfNumber=[15], rfOrder=23, authorNames=张淑静, 苗开超, 张亚力, journalName=计算机与现代化, refType=null, unstructuredReference=张淑静, 苗开超, 张亚力, 等. 基于Bi-LSTM-CRF算法的气象预警信息质控系统的实现[J]. 计算机与现代化, 2019(6): 111-115., articleTitle=基于Bi-LSTM-CRF算法的气象预警信息质控系统的实现, refAbstract=null), Reference(id=1208085601178460322, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, doi=null, pmid=null, pmcid=null, year=2019, volume=null, issue=6, pageStart=111, pageEnd=115, url=null, language=null, rfNumber=[15], rfOrder=24, authorNames=Zhang Shujing, Miao Kaichao, Zhang Yali, journalName=Computer and Modernization, refType=null, unstructuredReference=Zhang Shujing, Miao Kaichao, Zhang Yali, et al. Implementation of meteorological early warning information quality control system based on Bi-LSTM-CRF algorithm[J]. Computer and Modernization, 2019(6): 111-115., articleTitle=Implementation of meteorological early warning information quality control system based on Bi-LSTM-CRF algorithm, refAbstract=null), Reference(id=1208085601312678059, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, doi=null, pmid=null, pmcid=null, year=2022, volume=12, issue=1, pageStart=143, pageEnd=145, url=null, language=null, rfNumber=[16], rfOrder=25, authorNames=亓淼, 郑凯东, journalName=智能计算机与应用, refType=null, unstructuredReference=亓淼, 郑凯东. 基于注意力机制的足球视频目标检测[J]. 智能计算机与应用, 2022, 12(1): 143-145, 154., articleTitle=基于注意力机制的足球视频目标检测, refAbstract=null), Reference(id=1208085601442701491, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, doi=null, pmid=null, pmcid=null, year=2022, volume=12, issue=1, pageStart=143, pageEnd=145, url=null, language=null, rfNumber=[16], rfOrder=26, authorNames=Qi Miao, Zheng Kaidong, journalName=Intelligent Computers and Applications, refType=null, unstructuredReference=Qi Miao, Zheng Kaidong. Football video object detection based on attention mechanism[J]. Intelligent Computers and Applications, 2022, 12(1): 143-145, 154., articleTitle=Football video object detection based on attention mechanism, refAbstract=null), Reference(id=1208085601560142015, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, doi=null, pmid=null, pmcid=null, year=2023, volume=46, issue=2, pageStart=175, pageEnd=183, url=null, language=null, rfNumber=[17], rfOrder=27, authorNames=郑玉珩, 黄德启, journalName=电子测量技术, refType=null, unstructuredReference=郑玉珩, 黄德启. 改进MobileViT与YOLOv4的轻量化车辆检测网络[J]. 电子测量技术, 2023, 46(2): 175-183., articleTitle=改进MobileViT与YOLOv4的轻量化车辆检测网络, refAbstract=null), Reference(id=1208085601698554057, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, doi=null, pmid=null, pmcid=null, year=2023, volume=46, issue=2, pageStart=175, pageEnd=183, url=null, language=null, rfNumber=[17], rfOrder=28, authorNames=Zheng Yuheng, Huang Deqi, journalName=Electronic Measurement Technology, refType=null, unstructuredReference=Zheng Yuheng, Huang Deqi. Improving the lightweight vehicle detection network of MobileViT and YOLOv4[J]. Electronic Measurement Technology, 2023, 46(2): 175-183., articleTitle=Improving the lightweight vehicle detection network of MobileViT and YOLOv4, refAbstract=null), Reference(id=1208085601874714836, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, doi=null, pmid=null, pmcid=null, year=2023, volume=null, issue=8, pageStart=84, pageEnd=88, url=null, language=null, rfNumber=[18], rfOrder=29, authorNames=宋小飞, 晁永生, journalName=组合机床与自动化加工技术, refType=null, unstructuredReference=宋小飞, 晁永生. 基于改进YOLOv5s网络的机械零件目标检测[J]. 组合机床与自动化加工技术, 2023(8): 84-88., articleTitle=基于改进YOLOv5s网络的机械零件目标检测, refAbstract=null), Reference(id=1208085601996349664, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, doi=null, pmid=null, pmcid=null, year=2023, volume=null, issue=8, pageStart=84, pageEnd=88, url=null, language=null, rfNumber=[18], rfOrder=30, authorNames=Song Xiaofei, Chao Yongsheng, journalName=Combined Machine Tool and Automatic Processing Technology, refType=null, unstructuredReference=Song Xiaofei, Chao Yongsheng. Object detection of mechanical parts based on improved YOLOv5s network[J]. Combined Machine Tool and Automatic Processing Technology, 2023(8): 84-88., articleTitle=Object detection of mechanical parts based on improved YOLOv5s network, refAbstract=null)], funds=[Fund(id=1208085596342428594, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, awardId=L221027, language=CN, fundingSource=北京市自然科学基金(L221027), fundOrder=null, country=null), Fund(id=1208085596522783680, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, awardId=2023YJ039, language=CN, fundingSource=中国铁道科学研究院集团有限公司科技研究开发计划(2023YJ039), fundOrder=null, country=null)], companyList=[AuthorCompany(id=1208085584871006358, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, xref=1, ext=[AuthorCompanyExt(id=1208085584879394968, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, companyId=1208085584871006358, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1 School of Mechanical, Electrical and Vehicle Engineering, Beijing University of Civil Engineering and Architecture, Beijing 102612, China), AuthorCompanyExt(id=1208085584887783577, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, companyId=1208085584871006358, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1 北京建筑大学机电与车辆工程学院, 北京 102612)]), AuthorCompany(id=1208085586091548842, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, xref=2, ext=[AuthorCompanyExt(id=1208085586104131756, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, companyId=1208085586091548842, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2 Infrastructure Testing Research Institute of China Academy of Railway Sciences,Beijing 100081, China), AuthorCompanyExt(id=1208085586108326060, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, companyId=1208085586091548842, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2 中国铁道科学研究院基础设施检测研究所, 北京 100081)])], figs=[ArticleFig(id=1208085590726255124, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, language=EN, label=Fig.1, caption=YOLOv8 structural diagram., figureFileSmall=VghxFk2L4NQcTG8Vv9E9MA==, figureFileBig=YDZAru+3qdOOWxGjftUE7g==, tableContent=null), ArticleFig(id=1208085590898221599, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, language=CN, label=图1, caption=YOLOv8结构图

Input为输入层;Conv为卷积层;Concat为特征融合;UP SAMPLE为上采样层;YOLO Head为检测头

, figureFileSmall=VghxFk2L4NQcTG8Vv9E9MA==, figureFileBig=YDZAru+3qdOOWxGjftUE7g==, tableContent=null), ArticleFig(id=1208085591137296947, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, language=EN, label=Fig.2, caption=HGNetv2 network structure diagram, figureFileSmall=yrWwwgfvHjbXuRDpxVTexQ==, figureFileBig=1l2cuZO0uC1oGd3iWcgWQQ==, tableContent=null), ArticleFig(id=1208085591275708992, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, language=CN, label=图2, caption=HGNetv2网络结构图

Input为输入层;Stem为初始预处理层;HGBlock为RepHGNetv2中的一个核心组件;DW Conv为卷积层;GAP为全局平均池化,FC为全连接层

, figureFileSmall=yrWwwgfvHjbXuRDpxVTexQ==, figureFileBig=1l2cuZO0uC1oGd3iWcgWQQ==, tableContent=null), ArticleFig(id=1208085591430898256, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, language=EN, label=Fig.3, caption=RepHGBlock structural diagram, figureFileSmall=w0Xr3zYfq4C7Qiqf/6BDWA==, figureFileBig=gdQ2DBdaw6cyTuHgb2hVSg==, tableContent=null), ArticleFig(id=1208085591560921696, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, language=CN, label=图3, caption=RepHGBlock结构图

Rep Conv为通过重参数化技术实现的卷积结构;ESC为退出当前卷积

, figureFileSmall=w0Xr3zYfq4C7Qiqf/6BDWA==, figureFileBig=gdQ2DBdaw6cyTuHgb2hVSg==, tableContent=null), ArticleFig(id=1208085591716110960, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, language=EN, label=Fig.4, caption=MBConv structural diagram, figureFileSmall=JXR8HfXvj3GtLl4LxIoDFQ==, figureFileBig=gTXn2otC52jzyDDkJDdWqA==, tableContent=null), ArticleFig(id=1208085591816774269, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, language=CN, label=图4, caption=MBConv结构图, figureFileSmall=JXR8HfXvj3GtLl4LxIoDFQ==, figureFileBig=gTXn2otC52jzyDDkJDdWqA==, tableContent=null), ArticleFig(id=1208085591942603401, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, language=EN, label=Fig.5, caption=SE structural diagram, figureFileSmall=o27O9sxe+rfxgGPU+ZOfIQ==, figureFileBig=1iPaB77uhFDcnvkfvpt+9w==, tableContent=null), ArticleFig(id=1208085592055849623, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, language=CN, label=图5, caption=SE模块结构图, figureFileSmall=o27O9sxe+rfxgGPU+ZOfIQ==, figureFileBig=1iPaB77uhFDcnvkfvpt+9w==, tableContent=null), ArticleFig(id=1208085592257176237, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, language=EN, label=Fig.6, caption=SPPF_LSKA structural diagram, figureFileSmall=NPgj0IUhER1ht9D+AlQWIw==, figureFileBig=gr7XiB3NDiTdKl2up9BvOg==, tableContent=null), ArticleFig(id=1208085592412365506, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, language=CN, label=图6, caption=SPPF_LSKA结构图, figureFileSmall=NPgj0IUhER1ht9D+AlQWIw==, figureFileBig=gr7XiB3NDiTdKl2up9BvOg==, tableContent=null), ArticleFig(id=1208085592538194638, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, language=EN, label=Fig.7, caption=Graph of loss function for article model training, figureFileSmall=hy+nRBK3jBaIIWII9l4TYw==, figureFileBig=+5io85S5NWZBQGvJRGoAtw==, tableContent=null), ArticleFig(id=1208085592664023772, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, language=CN, label=图7, caption=文章模型训练损失函数曲线图, figureFileSmall=hy+nRBK3jBaIIWII9l4TYw==, figureFileBig=+5io85S5NWZBQGvJRGoAtw==, tableContent=null), ArticleFig(id=1208085592777269988, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, language=EN, label=Fig.8, caption=Algorithm testing effect diagram in the article, figureFileSmall=qsCYJ1z6HVK6uGPWsGOO8Q==, figureFileBig=CPWijjmNK7HxNTmbj/x6bw==, tableContent=null), ArticleFig(id=1208085592861156078, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, language=CN, label=图8, caption=本文算法测试效果图, figureFileSmall=qsCYJ1z6HVK6uGPWsGOO8Q==, figureFileBig=CPWijjmNK7HxNTmbj/x6bw==, tableContent=null), ArticleFig(id=1208085592978596601, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, language=EN, label=Fig.9, caption=Original algorithm test rendering, figureFileSmall=Or4DEHZ0xefRFjbZMhMdcA==, figureFileBig=GCTYc482IHFBCiXvNnk96g==, tableContent=null), ArticleFig(id=1208085593112814341, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, language=CN, label=图9, caption=原始算法测试效果图, figureFileSmall=Or4DEHZ0xefRFjbZMhMdcA==, figureFileBig=GCTYc482IHFBCiXvNnk96g==, tableContent=null), ArticleFig(id=1208085593247032086, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, language=EN, label=Table 1, caption=

EfficientNet network structure

, figureFileSmall=null, figureFileBig=null, tableContent=
阶段 操作 分辨率 通道 层数
1 Conv3×3 224×224 32 1
2 MBConv1,k3×3 112×112 16 1
3 MBConv6,k3×3 112×112 24 2
4 MBConv6,k5×5 56×56 40 2
5 MBConv6,k3×3 28×28 80 3
6 MBConv6,k5×5 14×14 112 3
7 MBConv6,k5×5 14×14 192 4
8 MBConv6,k3×3 7×7 320 1
9 Conv1×1&Pooling&FC 7×7 1 280 1
), ArticleFig(id=1208085593444164385, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, language=CN, label=表1, caption=

EfficientNet网络结构

, figureFileSmall=null, figureFileBig=null, tableContent=
阶段 操作 分辨率 通道 层数
1 Conv3×3 224×224 32 1
2 MBConv1,k3×3 112×112 16 1
3 MBConv6,k3×3 112×112 24 2
4 MBConv6,k5×5 56×56 40 2
5 MBConv6,k3×3 28×28 80 3
6 MBConv6,k5×5 14×14 112 3
7 MBConv6,k5×5 14×14 192 4
8 MBConv6,k3×3 7×7 320 1
9 Conv1×1&Pooling&FC 7×7 1 280 1
), ArticleFig(id=1208085593590965041, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, language=EN, label=Table 2, caption=

Article experiment hardware environment

, figureFileSmall=null, figureFileBig=null, tableContent=
硬件名称 型号
CPU Intel(R)Core(TM)i7-14700HX
GPU NVIDIAGeForceRTX4070
), ArticleFig(id=1208085593767125824, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, language=CN, label=表2, caption=

文章实验硬件环境

, figureFileSmall=null, figureFileBig=null, tableContent=
硬件名称 型号
CPU Intel(R)Core(TM)i7-14700HX
GPU NVIDIAGeForceRTX4070
), ArticleFig(id=1208085593892954952, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, language=EN, label=Table 3, caption=

Article experiment software environment

, figureFileSmall=null, figureFileBig=null, tableContent=
参数名称 版本
Pytorch 2.2.0+cu121
Python Python3.8
Cuda 12.1
Cudnn 8 801
), ArticleFig(id=1208085595084137303, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, language=CN, label=表3, caption=

文章实验软件环境

, figureFileSmall=null, figureFileBig=null, tableContent=
参数名称 版本
Pytorch 2.2.0+cu121
Python Python3.8
Cuda 12.1
Cudnn 8 801
), ArticleFig(id=1208085595214160735, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, language=EN, label=Table 4, caption=

Comparison results of YOLO different network experiments

, figureFileSmall=null, figureFileBig=null, tableContent=
实验内容 P/% R/% mAP/% 模型参
数/MB
计算力/
GFLOPs
YOLOv3 92.5 91.2 95.5 12.1 282.2
YOLOv5 92.8 91.2 96.9 2.5 7.1
YOLOv6 92.7 92.0 96.6 4.23 11.8
YOLOv8 91.6 93.1 96.0 3.0 8.1
本文模型 95.1 93.8 97.6 3.4 7.1
), ArticleFig(id=1208085595394515823, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, language=CN, label=表4, caption=

YOLO不同网络实验对比结果

, figureFileSmall=null, figureFileBig=null, tableContent=
实验内容 P/% R/% mAP/% 模型参
数/MB
计算力/
GFLOPs
YOLOv3 92.5 91.2 95.5 12.1 282.2
YOLOv5 92.8 91.2 96.9 2.5 7.1
YOLOv6 92.7 92.0 96.6 4.23 11.8
YOLOv8 91.6 93.1 96.0 3.0 8.1
本文模型 95.1 93.8 97.6 3.4 7.1
), ArticleFig(id=1208085595562287997, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, language=EN, label=Table 5, caption=

Results of ablation experiment

, figureFileSmall=null, figureFileBig=null, tableContent=
RepHGNetv2 EfficientHead LSKA P/% R/% mAP/% 模型参数/MB 计算力/GFLOPs
× × × 91.6 93.1 96.0 3.0 8.1
× × 90.3 93.6 96.6 2.3 6.9
× × 92.1 93.3 96.9 3.8 8.1
× × 95.0 89.2 96.6 3.3 8.3
× 93.5 91.2 95.9 2.6 7.1
× 93.2 91.6 97.0 4.1 8.3
× 93.0 89.9 95.3 3.2 6.9
95.1 93.8 97.6 3.4 7.1
), ArticleFig(id=1208085595738448779, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, language=CN, label=表5, caption=

消融实验结果

, figureFileSmall=null, figureFileBig=null, tableContent=
RepHGNetv2 EfficientHead LSKA P/% R/% mAP/% 模型参数/MB 计算力/GFLOPs
× × × 91.6 93.1 96.0 3.0 8.1
× × 90.3 93.6 96.6 2.3 6.9
× × 92.1 93.3 96.9 3.8 8.1
× × 95.0 89.2 96.6 3.3 8.3
× 93.5 91.2 95.9 2.6 7.1
× 93.2 91.6 97.0 4.1 8.3
× 93.0 89.9 95.3 3.2 6.9
95.1 93.8 97.6 3.4 7.1
), ArticleFig(id=1208085595918803861, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, language=EN, label=Table 6, caption=

Results of various damage detection algorithms in the article

, figureFileSmall=null, figureFileBig=null, tableContent=
类别 P/% R/% mAP/%
所有类别 95.1 93.8 97.6
正常螺孔(zc) 91.7 93.9 96.8
螺孔上裂(sl) 93.8 94.1 98.0
螺孔下裂(xl) 1 93.9 99.0
螺孔水平裂纹(sp) 92.9 90.7 96.5
螺孔组合裂纹(zh) 96.9 96.5 97.7
), ArticleFig(id=1208085596078187425, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1208051034803511323, language=CN, label=表6, caption=

本文算法各类伤损检测结果

, figureFileSmall=null, figureFileBig=null, tableContent=
类别 P/% R/% mAP/%
所有类别 95.1 93.8 97.6
正常螺孔(zc) 91.7 93.9 96.8
螺孔上裂(sl) 93.8 94.1 98.0
螺孔下裂(xl) 1 93.9 99.0
螺孔水平裂纹(sp) 92.9 90.7 96.5
螺孔组合裂纹(zh) 96.9 96.5 97.7
)], 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.1671-1815.2404963, detailUrlEn=https://castjournals.cast.org.cn/joweb/kxjsygc/EN/10.12404/j.issn.1671-1815.2404963, pdfUrlCn=https://castjournals.cast.org.cn/joweb/kxjsygc/CN/PDF/10.12404/j.issn.1671-1815.2404963, pdfUrlEn=https://castjournals.cast.org.cn/joweb/kxjsygc/EN/PDF/10.12404/j.issn.1671-1815.2404963, aliStartDate=null, aliEndDate=null, collectionFlag=false, citedCount=null, citedUrl=null, reference=null)
收藏切换
基于改进YOLOv8的钢轨典型伤损智能识别算法
收藏切换
PDF下载
向厚雪 1 , 许贵阳 1, * , 张玉华 2 , 黄筱妍 2
科学技术与工程 | 论文·交通运输 2025,25(18): 7785-7792
收起
收藏切换
科学技术与工程 | 论文·交通运输 2025, 25(18): 7785-7792
基于改进YOLOv8的钢轨典型伤损智能识别算法
全屏
向厚雪1 , 许贵阳1, * , 张玉华2, 黄筱妍2
作者信息
  • 1 北京建筑大学机电与车辆工程学院, 北京 102612
  • 2 中国铁道科学研究院基础设施检测研究所, 北京 100081
  • 向厚雪(2002—),女,汉族,河南信阳人,硕士研究生。研究方向:钢轨探伤。E-mail:

通讯作者:

* 许贵阳(1973—),男,满族,辽宁丹东人,博士,教授。研究方向:车辆工程。E-mail:
Intelligent Identification Algorithm for Typical Rail Damage Based on Improved YOLOv8
Hou-xue XIANG1 , Gui-yang XU1, * , Yu-hua ZHANG2, Xiao-yan HUANG2
Affiliations
  • 1 School of Mechanical, Electrical and Vehicle Engineering, Beijing University of Civil Engineering and Architecture, Beijing 102612, China
  • 2 Infrastructure Testing Research Institute of China Academy of Railway Sciences,Beijing 100081, China
出版时间: 2025-06-28 doi: 10.12404/j.issn.1671-1815.2404963
文章导航
收藏切换

钢轨内部损伤危害严重,但无损探伤B显检测图像杂波噪声多、不同损伤时空分布特征不明显,难以有效识别。针对这一情况,提出了一种基于改进YOLOv8的钢轨螺孔裂纹B显图识别算法,从提高钢轨损伤智能识别精确率。首先,为减少小伤损目标漏检,采用RepHGNetv2网络优化YOLOv8主干网络,提高检测召回率;然后,为提升模型对不同类型损伤检测的适应性,将YOLOv8的检测头替换为Effcientnet,提高模型检测精度;最后,在SPPF模块中引入LSKA注意力机制,提升模型对噪声信号的抗干扰能力,提高模型精确率。经实际线路检测结果验证,上述模型检测精确率达到95.1%,召回率达到93.8%,平均精度达到97.6%,对比其他常用算法均有所提高。

钢轨探伤  /  B显图  /  损伤检测  /  YOLOv8  /  螺孔裂纹

The internal damage of the steel rail is serious, but the non-destructive testing B-display detection image has a lot of noise and noise, and the spatiotemporal distribution characteristics of different damages are not obvious, making it difficult to effectively identify. In response to this situation, a rail screw hole crack B-image recognition algorithm based on improved YOLOv8 was studied to improve the accuracy of intelligent identification of rail damage. Firstly, to reduce the missed detection of small damage targets, RepHGNetv2 network was used to optimize the YOLOv8 backbone network and improve the detection recall rate. Then, in order to improve the adaptability of the model to different types of damage detection, the detection head of YOLOv8 was replaced with Effientnet to improve the detection accuracy of the model. Finally, the LSKA attention mechanism was introduced into the SPPF module to enhance the model’s anti-interference ability against noise signals and improve its accuracy. The actual line detection results have verified that the detection accuracy of the above model reaches 95.1%, the recall rate reaches 93.8%, and the average accuracy reaches 97.6%, which is improved compared to other commonly used algorithms.

rail flaw detection  /  B-imaging  /  damage detection  /  YOLOv8  /  bolt hole crack
向厚雪, 许贵阳, 张玉华, 黄筱妍. 基于改进YOLOv8的钢轨典型伤损智能识别算法. 科学技术与工程, 2025 , 25 (18) : 7785 -7792 . DOI: 10.12404/j.issn.1671-1815.2404963
Hou-xue XIANG, Gui-yang XU, Yu-hua ZHANG, Xiao-yan HUANG. Intelligent Identification Algorithm for Typical Rail Damage Based on Improved YOLOv8[J]. Science Technology and Engineering, 2025 , 25 (18) : 7785 -7792 . DOI: 10.12404/j.issn.1671-1815.2404963
在中国铁路行业高速发展的同时,铁路行车的安全和稳定,影响到人民生命和财产的安全。特别是中国高速铁路的飞速发展带来了列车运行速度和行车密度的提高,以及日益繁忙的对外贸易中铁路货运列车货物载重量的增加,对于列车运行的安全性和可靠性提出了更高的要求,特别是工务段等部门负责的铁路养护工作尤其重要[1]。钢轨在长时间服役期间会由于列车的挤压,磨耗而受损,逐渐形成如轨面缺陷、轨头横孔、轨底横孔、螺孔裂纹等不同部位、不同类型的伤损[2]。若未能及时对这些伤损进行检修则会导致伤损积累并扩大,最终造成钢轨断裂,从而导致严重的铁路交通事故,极大地危害国民经济安全。
研究人员不断探索YOLO算法在不同场景下的性能优化和应用扩展,彭红星等[3]研究了一种基于改进YOLOv5的高速公路隧道车辆和人员检测算法,通过高斯混合聚类优化锚框、引入CARAFE上采样算子和坐标注意力机制,提高了模型在复杂环境下的检测精度和鲁棒性。宣以国等[4]提出了一种基于改进YOLOv7的道路裂缝和坑洞检测算法,通过引入可变形卷积网络和基于Wasserstein距离的小目标检测评估方法,提高了道路裂缝和坑洞的检测精度和速度。孔建国等[5]研究了一种基于改进YOLOv8的机场飞鸟实时目标检测方法,通过设计特征融合C3模块和CSPPF模块,并改进数据集和应用数据增强技术,提高了机场飞鸟检测的精度和速度。现有钢轨伤损检测系统大都采用人工判伤的模式进行伤损分类,人工判伤存在着判伤精确率低,误报率高的问题[6]。随着YOLO算法在多个领域的成功应用,也为钢轨B显图识别算法提供了新的思路和解决方案。Zhang等[7]提出了一种基于YOLOv4算法检测钢轨缺陷,通过分析钢轨结构波形和钢轨缺陷的关系,使用算法检测波形,基于B型图像成像规律筛选缺陷图像,改进后的算法可获得具有完整边缘信息的缺陷轮廓定位,但对复杂钢轨图像的检测准确率较低。Chen等[8]提出了一种基于改进YOLOv3模型的典型钢轨伤损识别方法,改进后的网络相较于原始网络检测效率和精度均有提升,平均检测精度可达87%,但是还有提升空间。何庆等[9]提出了一种基于改进YOLOv3的钢轨伤损B型图像识别算法,通过对YOLOv3网络结构改进,提高探伤设备的伤损识别率,尽管模型整体表现良好,但在小目标数据集检测精度方面还有一定的提升空间。
此外,由于超声波钢轨探伤车运行速度较高,导致探伤图像分辨率相对较低且有较多杂波干扰,特征不明显等现象,目前针对钢轨B显图的伤损识别方法仍然存在精确率低、误报率高的问题。为此,现提出一种基于YOLOv8的伤损识别算法,提高钢轨伤损分类识别的有效性。为了提高模型的召回率,使用RepHGNetv2网络代替YOLOv8的原主干网络。为使模型达到更好的检测效果,将YOLOv8头部替换为Efficient Head检测头,结合了尺度、空间和任务3种注意力函数。为减少背景对伤损目标检测的干扰,提高特征提取能力,将Large Separable Kernel Attention与YOLOv8中SPPF模块结合。
Ultralytics于2023年发布了YOLOv8 模型,其建立在各YOLO版本优点的基础上,引入了新功能,以进一步提高性能和灵活性。由于YOLOv8快速、准确且易于使用,在目标检测和跟踪、实例分割、图像分类和姿态估计领域得到了大量应用[10]。YOLOv8的改进有如下的几个方面:在骨干网络(Backbone)中,YOLOv8采用CSP的思想,使用了C2f模块替换了YOLOv5中的C3模块,每个阶段的模块个数从[3,6,9,3]修改为[3,6,6,3],实现了进一步的轻量化;在颈部网络(Neck)中,将YOLOv5中PAN-FPN网络上采样阶段的卷积删除,并将C3模块替换为C2f模块;在最后的预测头网络(Head),采用目前主流的解耦头结构,将分类和检测头分离,同时抛弃了Anchor-Based,使用了Anchor-Free的思想,模型结构如图1所示。由于钢轨探伤需要准确地识别钢轨内部缺陷,且不同钢轨有不同的缺陷类型,因此对算法模型要求精度高,泛化能力强,YOLOv8算法满足这些需求,所以本文研究在YOLOv8的基础上进行模型优化,以达到更好的识别效果。
YOLOv8在目标检测任务中有不错的性能,但是YOLOv8的主干网络使用了参数较多的网络结构和较高的分辨率来提高检测性能,这导致了较高的计算资源需求,限制了在资源有限的设备上的部署和应用。本文研究使用RepHGNetv2替换YOLOv8原来的主干网络,RepHGNetv2是对HGNetv2[11]的进一步优化,有效地减少了网络模型的参数和降低了网络的计算力,同时提升了网络模型的召回率。HGNetv2的网络结构如图2所示。
HGNet的整体结构由1个主干层和4个阶段组成。Stem层是HGNetv2网络的初始预处理层,一般包含卷积层,从输入的原始数据中提取特征。HGBlock是网络的核心组件,用层次化的方式处理数据,能够逐步捕获图像中的关键信息。GAP是全局平均池化,通过减少特征图的空间维度,将其转化为一个特征向量,从而增强模型的鲁棒性。最后HGNetv2通过卷积和全连接层来完成复杂的分类和检测任务。其中HGBlock的细节如图3所示,本文算法将HGBlock的普通卷积替换为RepConv。RepConv是一种模型重参化技术,它可以在推理阶段将多个计算模块合并为一个,从而显著提高模型的效率和性能。这一技术不仅减少了模型的计算量,还使得模型在保持高精度的同时,能够更好地适应资源有限的设备。在目标检测等复杂任务中,RepConv已经取得了很好的效果,也证明了其在实际应用中的有效性。
YOLOv8原有头部的单尺度预测结构缺乏动态学习能力,给目标检测尤其是多尺度目标检测带来一定限制。本文研究用EfficientNet替换YOLOv8原来的检测头,EfficientNet(网络结构如表1所示)使用了一种称为复合缩放方法,可以同时调整网络深度和宽度,在有限的计算资源下实现更好的性能[12]。通过在不同深度和宽度的模型之间进行平衡,EfficientNet能够获得更好的性能和泛化能力。EfficientNet网络结构如表1所示,它使用了一种称为MBConv(mobile inverted bottleneck convolution)的模块来进行特征提取,这种模块结合了轻量级的深度可分离卷积和传统的卷积操作,能够在保持模型效率的同时提取更丰富的特征。
MBConv模块如图4所示,主要由一个1×1的普通卷积(Conv),一个k×k的DepwiseConv卷积(深度可分离卷积)一个SE模块,一个1×1的普通卷积,一个Dropout层(随机丢弃一部分神经元,并将其输出置为零)构成。其中k为卷积核的维度,其值主要有3×3和5×5两种情况,s1和s2表示卷积操作的步长,BN Swish表示批量归一化激活函数,其中SE模块[13]图5所示,由一个全局平均池化(Avgpooling),两个全连接层(FC1和FC2)组成。第一个全连接层的节点个数是输入该MBConv特征矩阵channels的1/4,且使用Swish激活函数(Swish),第二个全连接层的节点个数等于DepthwiseConv层输出的特征矩阵channels,且使用Sigmoid激活函数(Sigmoid)。
在深度学习应用中,特别是在基于图像识别的自动探伤系统中,探伤车采集的B显图往往由于设备精度、环境干扰等多种因素,不可避免地会携带一些噪声。这些噪声不仅会对后续的网络处理产生干扰,还会严重影响网络对数据中伤损特征的准确提取和识别能力。为了解决这一问题,在YOLOv8骨干网络中的SPPF模块引入LSKA注意力机制。LSKA将深度卷积层的二维卷积核分解为级联的水平和垂直一维卷积核,旨在通过精细地调控网络对图像信息的关注程度,有效地减少噪声对网络的干扰,从而确保网络能够提取到更多真实且有效的伤损特征。对比不同注意力机制,LSKA具有较高的灵活性,可以根据任务需求和数据特征进行调整。同时,由于其考虑了位置信息,LSKA的注意力权重具有较强的可解释性,可以帮助理解模型的决策过程。与传统的全局自注意力机制相比,LSKA能够更高效地计算注意力权重,因为它只考虑了输入序列的局部信息,从而减少了计算量。改进后的SPPF_LSKA结构如图6所示,这一结构通过LSKA的引入,不仅有效地减少了噪声对网络的干扰,还大大提升了网络对伤损特征的提取能力。
本文算法采用GPU进行训练,硬件环境如表2所示。
使用Pycharm作为编译软件,软件环境如表3所示。采用YOLOv8n作为基准网络模型,设置初始学习率为0.01,Batch size为32,Epoch为300。
钢轨螺孔是铁路轨道固定的关键位置之一,承受着列车的重量和运行时的巨大动态荷载,所以钢轨螺孔处裂纹的扩展速度往往比其他位置的裂纹要快。一旦螺孔发生裂纹,可能会导致轨道固定不牢,严重影响铁路线路的安全性。基于以上原因,文章重点研究螺孔处的裂纹伤损,将伤损按照裂纹走势划分为正常螺孔(zc)、螺孔上裂(sl)、螺孔下裂(xl)、螺孔水平裂纹(sp)、螺孔组合裂纹(zh)5个类别,使用LabelImg软件对上述目标进行标注。
为进行训练和测试,本文研究进行了标定线路检测实验,利用某型智能探伤车采集并整理了包含1 440张B显图的钢轨损伤数据集,按照8∶1∶1划分为训练集、测试集和验证集。
文章使用精确率(precision,P)、召回率(recall,R)、平均精度均值(mean average precision,mAP)和浮点运算次数(giga floating-point operations per second,GFLOPs)[14]对模型进行评价。其中PR的计算公式为
$ P=\frac{\mathrm{TP}}{\mathrm{TP}+\mathrm{FP}}$
$ R=\frac{\mathrm{TP}}{\mathrm{TP}+\mathrm{FN}}$
式中:TP为被正确分类的正样本;FP为被错误分类的负样本;FN为被错误分类的正样本[15]
AP值代表模型对每一类别预测的精度高低,mAP是对不同类别的AP值取平均[16],评估整个模型的好坏。
为了验证本文模型的性能,将其与当前流行的目标检测算法YOLO的不同版本模型进行对比,由实验结果(表4)可以发现,本文模型在精确率、召回率以及平均精度均值这3个关键指标上均取得了显著的优势。具体来说,在模型大小上,YOLOv5的模型参数最小,为2.5 MB,计算力为7.1 GFLOPs,在精确率方面,YOLOv8原始模型相较于其他几个模型,稍有逊色,但是经过优化之后,本文模型相比精确率较高的YOLOv5提高了2.3%,达到95.1%。这意味着在相同的测试条件下,本文模型能够更准确地识别出伤损,减少误判的可能性。在召回率方面,YOLOv8原始模型达到93.1%,改进之后提升了0.7%,达到93.8%。这表明本文模型在识别目标时,能够更全面地覆盖钢轨伤损,减少漏检的情况。而在平均精度均值方面,本文模型相较于性能较好的YOLOv5提高了0.7%,比YOLOv8更是提高了1.6%,达到97.6%。实验数据充分证明了文章模型在多个类别的综合性能要优于其他对比模型。
为了更好地验证改进的各模块以及改进模块之间的组合对于原模型性能提升的影响[17],本文设计了消融实验,实验结果如表5所示。
通过表5实验结果可以看出,随着一系列改进模块的逐步加入,模型在多个关键性能指标上均实现了显著的提升,这不仅验证了每个模块设计的有效性,也展示了它们之间的良好协同性。将主干网络从原始架构替换为RepHGNetv2,模型参数变为2.3 MB,相比原模型3 MB减少了0.7×106,计算力变为6.9 GFLOPs,降低了1.2 GFLOPs,同时召回率也提高了0.5%,这表明RepHGNetv2不仅更轻量,而且在识别伤损目标方面也具有更好的性能。加入EfficientHead检测头,精确率、召回率和平均精度均值都有明显提高,分别达到92.1%、93.3%和96.9%。加入LSKA注意力机制,模型的精确率有大幅度提高,相比YOLOv8提高了3.4%。最后将各个模块组合起来,充分利用各个模块的优势[18],组合后的模型在精确率、召回率和平均精度上都达到了最高水平。
本文模型训练损失变化曲线如图7所示,在训练初期,损失值的下降速度较快,这是由于模型在初始阶段对参数进行了较大的调整,随着训练的进行,当模型逐渐接近最优解时,损失值的下降速度变缓,呈现出一种趋于平稳的趋势。在训练过程中,损失函数变化曲线会出现一些波动,这是由于训练数据中存在一些噪声。但整体来看,本文算法在训练过程中表现良好的收敛性和稳定性,模型能够逐步学习并优化其参数,为后续实验结果分析提供有力的支持。
使用训练好的模型,对验证集数据进行验证,各类别伤损结果如表6所示。
由实验结果可以看出,本文模型对各类伤损识别的精确率和召回率均可以达到90%以上,各类伤损的平均精度值均可以达到96.5%以上,这表明本文模型对于螺孔伤损的检测识别有较好的效果。
为了更直观地展示本文算法的优势,对测试集的数据进行测试,同时将原始YOLOv8算法在相同测试集上进行对比,本文算法的测试效果如图8所示,原始YOLOv8的测试效果如图9所示。
图8图9测试效果图可以看出,无论是组合裂纹还是正常裂纹,本文算法都能够以极高的置信度(1.0)进行准确识别,而YOLOv8算法虽然也能识别出裂纹,但其置信度却相对较低(0.9)。所以在面对一些复杂或模糊的伤损图像时,原始算法可能会产生误判或者漏判的情况,而本文算法则能够凭借更高的置信度,实现更加准确可靠的伤损识别。
提出了一种基于改进YOLOv8的钢轨典型伤损智能识别算法,针对钢轨B显图螺孔裂纹,解决了识别精确率低、误报率高的问题。通过采用HGNetv2作为主干网络,算法在特征提取和表示能力上得到了显著提升。HGNetv2的层次结构和多尺度特性使其能够捕获钢轨螺孔B显图中更为细微和复杂的特征,进而提高了识别召回率。在头部网络中,本文研究运用了EfficientNet作为检测头。EfficientNet的设计理念在于通过在不同深度和宽度的模型之间进行平衡,以实现最优的性能。这一设计使得本文算法在保持计算效率的同时,获得了更好的特征提取能力和泛化能力。最后在SPPF模块引入LSKA注意力机制,提高了算法对螺孔裂纹伤损特征的敏感度,使网络能够提取更多有效的特征信息,从而进一步提高识别精确率。
经过一系列的实验验证,相较于其他模型,本文算法在精确率、召回率以及平均精度值等方面均实现了显著的提高。特别是在钢轨螺孔裂纹B显图伤损识别任务中,本文算法展现出了更高的精确度和稳定性,为铁路安全检测提供了有力的技术支持。
  • 北京市自然科学基金(L221027)
  • 中国铁道科学研究院集团有限公司科技研究开发计划(2023YJ039)
参考文献 引证文献
排序方式:
[1]
栗英. 钢轨养护及一种钢轨轮廓检测方法[J]. 现代制造技术与装备, 2016(9): 94, 96.
Li Ying. Rail maintenance and a rail contour detection method[J]. Modern Manufacturing Technology and Equipment, 2016(9): 94, 96.
[2]
王炳辉. 普速铁路钢轨探伤及伤损原因分析[J]. 减速顶与调速技术, 2022(1): 25-27.
Wang Binghui. Analysis of defect detection and damage causes of ordinary speed railway rails[J]. Decelerator and Speed Control Technology, 2022(1): 25-27.
[3]
彭红星, 袁畅, 柯威曳, 等. 基于改进YOLOv5的高速公路隧道车辆和人员检测[J]. 科学技术与工程, 2024, 24(6): 2453-2461.
Peng Hongxing, Yuan Chang, Ke Weiye, et al. Vehicle and personnel detection in highway tunnels based on improved YOLOv5[J]. Science Technology and Engineering, 2024, 24(6): 2453-2461.
[4]
宣以国, 余成波, 蒋启超, 等. 基于改进YOLOv7的道路裂缝和坑洞检测算法[J]. 科学技术与工程, 2024, 24(17): 7205-7213.
Xuan Yiguo, Yu Chengbo, Jiang Qichao, et al. Improved YOLOv7 road crack and pothole detection algorithm[J]. Science Technology and Engineering, 2024, 24(17): 7205-7213.
[5]
孔建国, 张向伟, 赵志伟, 等. 基于改进YOLOv8的机场飞鸟实时目标检测方法[J]. 科学技术与工程, 2024, 24(32): 13944-13952.
Kong Jianguo, Zhang Xiangwei, Zhao Zhiwei, et al. Real time object detection method for airport flying birds based on improved YOLOv8[J]. Science Technology and Engineering, 2024, 24(32): 13944-13952.
[6]
张永雄, 蔡勇. 多模态人工智能技术在钢轨伤损判定中的探索[C]// 第三届中国铁路发展论坛学术论文集. 北京: 中国铁道学会, 2022: 1-8.
Zhang Yongxiong, Cai Yong. Exploration of multimodal artificial intelligence technology in rail damage assessment[C]// Academic Papers Collection of the 3rd China Railway Development Forum. Beijing: China Railway Society, 2022: 1-8.
[7]
Zhang B, Huang S, Zhang L, et al. Rail defect recognition based on waveform subtraction and rule base[J]. Journal of Performance of Constructed Facilities, 2022, 36(1). DOI:10.1061/(ASCE)CF.1943-5509.0001684.
[8]
Chen Z X, Wang Q H, Yang K, et al. Deep learning for the detection and recognition of rail defects in ultrasound B-scan images[J]. Transportation Research Record, 2021, 2675(11): 888-901.
[9]
何庆, 陈正兴, 王启航, 等. 基于改进YOLO V3的钢轨伤损B显图像识别研究[J]. 铁道学报, 2022, 44(12): 82-88.
He Qing, Chen Zhengxing, Wang Qihang, et al. Research on B-image recognition of rail damage based on improved YOLO V3[J] Journal of Railway, 2022, 44 (12): 82-88.
[10]
任青阳, 王彦丁, 施俭. 卷积神经网络目标检测算法研究进展[J]. 科学技术与工程, 2024, 24(32): 13665-13677.
Ren Qingyang, Wang Yanding, Shi Jian. Advances in target detection algorithms for convolutional neural networks[J]. Science Technology and Engineering, 2024, 24(32): 13665-13677.
[11]
Zhang L, Zheng J C, Li C P, et al. CCDN-DETR: a detection transformer based on constrained contrast denoising for multi-class synthetic aperture radar object detection[J]. Sensors, 2024, 24(6). DOI:10.3390/s24061793.
[12]
Tan M, Le Q. Efficientnet: rethinking model scaling for convolutional neural networks[C]// International Conference on Machine Learning. Chambridge: PMLR, 2019: 6105-6114.
[13]
Hu J, Li S, Gang S. Squeeze-and-excitation networks[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2018: 7132-7141.
[14]
刘洪宁, 李容露. 基于YOLOv5的钢材数量识别方法[J]. 科技创新与应用, 2023, 13(1): 69-72.
Liu Hongning, Li Ronglu. A method for identifying steel quantity based on YOLOv5[J]. Technological Innovation and Application, 2023, 13(1): 69-72.
[15]
张淑静, 苗开超, 张亚力, 等. 基于Bi-LSTM-CRF算法的气象预警信息质控系统的实现[J]. 计算机与现代化, 2019(6): 111-115.
Zhang Shujing, Miao Kaichao, Zhang Yali, et al. Implementation of meteorological early warning information quality control system based on Bi-LSTM-CRF algorithm[J]. Computer and Modernization, 2019(6): 111-115.
[16]
亓淼, 郑凯东. 基于注意力机制的足球视频目标检测[J]. 智能计算机与应用, 2022, 12(1): 143-145, 154.
Qi Miao, Zheng Kaidong. Football video object detection based on attention mechanism[J]. Intelligent Computers and Applications, 2022, 12(1): 143-145, 154.
[17]
郑玉珩, 黄德启. 改进MobileViT与YOLOv4的轻量化车辆检测网络[J]. 电子测量技术, 2023, 46(2): 175-183.
Zheng Yuheng, Huang Deqi. Improving the lightweight vehicle detection network of MobileViT and YOLOv4[J]. Electronic Measurement Technology, 2023, 46(2): 175-183.
[18]
宋小飞, 晁永生. 基于改进YOLOv5s网络的机械零件目标检测[J]. 组合机床与自动化加工技术, 2023(8): 84-88.
Song Xiaofei, Chao Yongsheng. Object detection of mechanical parts based on improved YOLOv5s network[J]. Combined Machine Tool and Automatic Processing Technology, 2023(8): 84-88.
2025年第25卷第18期
PDF下载
218
95
引用本文
BibTeX
文章信息
doi: 10.12404/j.issn.1671-1815.2404963
  • 接收时间:2024-07-03
  • 首发时间:2025-12-17
  • 出版时间:2025-06-28
补充材料
相关文章
文章信息
作者
出版历史
  • 收稿日期:2024-07-03
  • 修回日期:2025-03-19
基金
北京市自然科学基金(L221027)
中国铁道科学研究院集团有限公司科技研究开发计划(2023YJ039)
作者信息
    1 北京建筑大学机电与车辆工程学院, 北京 102612
    2 中国铁道科学研究院基础设施检测研究所, 北京 100081

通讯作者:

* 许贵阳(1973—),男,满族,辽宁丹东人,博士,教授。研究方向:车辆工程。E-mail:
参考文献
分享链接
https://castjournals.cast.org.cn/joweb/kxjsygc/CN/10.12404/j.issn.1671-1815.2404963
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