Article(id=1261267654854193768, tenantId=1146029695717560320, journalId=1146123166801305609, issueId=1261262687258985194, articleNumber=null, orderNo=null, doi=10.12404/j.issn.1671-1815.2406421, pmid=null, cstr=null, oa=null, hot=null, price=null, onlineType=0, articleFormat=0, articleType=null, articleTypeStr=research-article, receivedDate=1724688000000, receivedDateStr=2024-08-27, revisedDate=1745337600000, revisedDateStr=2025-04-23, acceptedDate=null, acceptedDateStr=null, onlineDate=1778639242137, onlineDateStr=2026-05-13, pubDate=1752768000000, pubDateStr=2025-07-18, doiRegisterDate=null, doiRegisterDateStr=null, onlineIssueDate=1778639242137, onlineIssueDateStr=2026-05-13, onlineJustAcceptDate=null, onlineJustAcceptDateStr=null, onlineFirstDate=null, onlineFirstDateStr=null, sourceXml=null, magXml=null, createTime=1778639242137, creator=13701087609, updateTime=1778639242137, updator=13701087609, issue=Issue{id=1261262687258985194, tenantId=1146029695717560320, journalId=1146123166801305609, year='2025', volume='25', issue='20', pageStart='8317', pageEnd='8759', issueExtLink='null', onlineDate='null', pubDate='null', beforeIssueId=null, nextIssueId=null, price=null, status=1, issueComplete=1, articleOrder=1, issueType=-1, specialIssue=null, createTime=1778638057769, creator=13701087609, updateTime=1778753106634, updator=13701087609, preIssue=null, nextIssue=null, ext={EN=IssueExt(id=1261745237240722095, tenantId=1146029695717560320, journalId=1146123166801305609, issueId=1261262687258985194, language=EN, specialIssueTitle=, coverIllustrator=null, specialIssueEditor=, specialIssueAbout=), CN=IssueExt(id=1261745237240722096, tenantId=1146029695717560320, journalId=1146123166801305609, issueId=1261262687258985194, language=CN, specialIssueTitle=, coverIllustrator=null, specialIssueEditor=, specialIssueAbout=)}, issueFiles=null}, startPage=8595, endPage=8603, ext={EN=ArticleExt(id=1261267655760163439, articleId=1261267654854193768, tenantId=1146029695717560320, journalId=1146123166801305609, language=EN, title=Detection of Low-quality Image Safety Helmet in Coal Mine Based on Improved YOLOv7, columnId=1156262729162810294, journalTitle=Science Technology and Engineering, columnName=Papers·Automation and Computational Technology, runingTitle=null, highlight=null, articleAbstract=

Due to the complex underground environment, low lighting conditions, and the small size of hard hats, the detection results are not ideal. To address low-quality images in complex environments, an improved YOLOv7 for hard hat detection in low-quality images from underground coal mines was proposed. Firstly, addressing the limitation that image features were susceptible to noise interference under low-light conditions, a multi-scale MELAN module was introduced. By constructing a multi-scale attention mechanism, broader contextual information was captured, thereby enhancing feature extraction and effectively suppressing noise interference. Secondly, the OD-SMP module was constructed using soft pooling and full-dimensional dynamic convolution in the backbone network, which reduced information diffusion in feature mappings, retained more contextual information, and enhanced the detection capability for small targets. Finally, to address the varying quality of detection samples caused by the complex backgrounds and environments with different lighting and distances in underground coal mines, Wise-IoU was used as the loss function. Experimental results show that the average precision of the improved model is 94.9%, which is 13.5% higher than the original YOLOv7 model, demonstrating better detection performance.

, correspAuthors=Yan-jun ZHANG, 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=Cheng-yang KANG, Yan-jun ZHANG, Rui ZHANG), CN=ArticleExt(id=1261267661212758694, articleId=1261267654854193768, tenantId=1146029695717560320, journalId=1146123166801305609, language=CN, title=基于改进YOLOv7的煤矿井下低质影像安全帽检测, columnId=1156262729783567290, journalTitle=科学技术与工程, columnName=论文·自动化技术、计算机技术, runingTitle=null, highlight=null, articleAbstract=

由于井下环境复杂,低光照以及安全帽的物体较小导致的检测效果不理想。针对复杂环境下的低质影像,提出一种改进YOLOv7的煤矿井下低质影像安全帽检测。首先,针对低光照条件下的图像特征易受噪声干扰的局限性,提出多尺度MELAN模块,通过构建多尺度注意力机制来捕捉更广泛的上下文信息,进而强化特征信息的提取,并有效抑制噪声的干扰;其次,在主干网络使用软池化和全维动态卷积构建OD-SMP模块,减少了特征映射中的信息弥散,保留了更多上下文信息,增强对小目标的检测能力;最后,对于煤矿井下不同光照和距离的复杂背景和环境,导致检测样本质量出现参差不齐的情况,使用Wise-IoU作为损失函数。实验结果表明:改进模型的平均精度为94.9%,比原来的YOLOv7模型高出13.5%,具有更好的检测效果。

, correspAuthors=张延军, authorNote=null, correspAuthorsNote=
* 张延军(1982—),男,满族,河北承德人,博士,教授。研究方向:机电液传动及智能控制。E-mail:
, copyrightStatement=null, copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=eiYTZmLVqDc/atfPzQyOCQ==, magXml=DXO/4e9+QJutenb/ve0JEw==, pdfUrl=null, pdf=kzDCjpz72ZLelBep5Q/9kg==, pdfFileSize=11253700, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=GtaDrscAKBmG17Vx/zqMCw==, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=EKN1l/Y5W5V3nhp+5RiSbw==, mapNumber=null, authorCompany=null, fund=null, authors=

康程洋(2000—),男,汉族,河南商丘人,硕士研究生。研究方向:机器视觉及智能控制。E-mail:

, authorsList=康程洋, 张延军, 张睿)}, authors=[Author(id=1261267662856925884, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, orderNo=0, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=1654032526@qq.com, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1261267663284744897, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, authorId=1261267662856925884, language=EN, stringName=Cheng-yang KANG, firstName=Cheng-yang, middleName=null, lastName=KANG, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1 School of Mechanical Engineering, Taiyuan University of Science and Technology, Taiyuan 030024, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1261267665818104519, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, authorId=1261267662856925884, language=CN, stringName=康程洋, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1 太原科技大学机械工程学院, 太原 030024, bio={"content":"

康程洋(2000—),男,汉族,河南商丘人,硕士研究生。研究方向:机器视觉及智能控制。E-mail:

"}, bioImg=null, bioContent=

康程洋(2000—),男,汉族,河南商丘人,硕士研究生。研究方向:机器视觉及智能控制。E-mail:

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1261267661909013167, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, xref=1, ext=[AuthorCompanyExt(id=1261267661942567600, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, companyId=1261267661909013167, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1 School of Mechanical Engineering, Taiyuan University of Science and Technology, Taiyuan 030024, China), AuthorCompanyExt(id=1261267662005482162, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, companyId=1261267661909013167, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1 太原科技大学机械工程学院, 太原 030024)])]), Author(id=1261267665973293771, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, orderNo=1, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=hebei0735@126.com, emailSecond=null, emailThird=null, correspondingAuthor=1, authorType=1, ext={EN=AuthorExt(id=1261267666375946962, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, authorId=1261267665973293771, language=EN, stringName=Yan-jun ZHANG, firstName=Yan-jun, middleName=null, lastName=ZHANG, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, *, address=1 School of Mechanical Engineering, Taiyuan University of Science and Technology, Taiyuan 030024, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1261267666707296981, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, authorId=1261267665973293771, language=CN, stringName=张延军, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, *, address=1 太原科技大学机械工程学院, 太原 030024, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1261267661909013167, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, xref=1, ext=[AuthorCompanyExt(id=1261267661942567600, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, companyId=1261267661909013167, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1 School of Mechanical Engineering, Taiyuan University of Science and Technology, Taiyuan 030024, China), AuthorCompanyExt(id=1261267662005482162, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, companyId=1261267661909013167, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1 太原科技大学机械工程学院, 太原 030024)])]), Author(id=1261267667210613465, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, 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=1261267667835564767, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, authorId=1261267667210613465, language=EN, stringName=Rui ZHANG, firstName=Rui, middleName=null, lastName=ZHANG, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, address=2 School of Computer Science and Technology, Taiyuan University of Science and Technology, Taiyuan 030024, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1261267668301132515, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, authorId=1261267667210613465, language=CN, stringName=张睿, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, address=2 太原科技大学计算机科学与技术学院, 太原 030024, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1261267662261334709, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, xref=2, ext=[AuthorCompanyExt(id=1261267662269723319, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, companyId=1261267662261334709, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2 School of Computer Science and Technology, Taiyuan University of Science and Technology, Taiyuan 030024, China), AuthorCompanyExt(id=1261267662278111928, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, companyId=1261267662261334709, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2 太原科技大学计算机科学与技术学院, 太原 030024)])])], keywords=[Keyword(id=1261267668762505960, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, language=EN, orderNo=1, keyword=low illumination), Keyword(id=1261267669186130666, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, language=EN, orderNo=2, keyword=multi scale), Keyword(id=1261267670394090221, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, language=EN, orderNo=3, keyword=full dimensional dynamic convolution), Keyword(id=1261267670696080112, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, language=EN, orderNo=4, keyword=attention mechanism), Keyword(id=1261267671081956083, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, language=EN, orderNo=5, keyword=loss function), Keyword(id=1261267671664964344, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, language=CN, orderNo=1, keyword=低光照), Keyword(id=1261267672487047931, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, language=CN, orderNo=2, keyword=多尺度), Keyword(id=1261267673317520128, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, language=CN, orderNo=3, keyword=全维动态卷积), Keyword(id=1261267673569178372, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, language=CN, orderNo=4, keyword=注意力机制), Keyword(id=1261267673900528392, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, language=CN, orderNo=5, keyword=损失函数)], refs=[Reference(id=1261267681211200342, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, doi=null, pmid=null, pmcid=null, year=2022, volume=31, issue=1, pageStart=013017, pageEnd=null, url=null, language=null, rfNumber=[1], rfOrder=0, authorNames=Li X, Wang S, Liu B, journalName=Journal of Electronic Imaging, refType=null, unstructuredReference=Li X, Wang S, Liu B, et al. Improved YOLOv4 network using infrared images for personnel detection in coal mines[J]. Journal of Electronic Imaging, 2022, 31(1): 013017., articleTitle=Improved YOLOv4 network using infrared images for personnel detection in coal mines, refAbstract=null), Reference(id=1261267681437692760, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, doi=null, pmid=null, pmcid=null, year=2021, volume=47, issue=10, pageStart=298, pageEnd=305, url=null, language=null, rfNumber=[2], rfOrder=1, authorNames=徐先峰, 赵万福, 邹浩泉, journalName=计算机工程, refType=null, unstructuredReference=徐先峰, 赵万福, 邹浩泉, . 基于MobileNet-SSD的安全帽佩戴检测算法[J]. 计算机工程, 2021, 47(10): 298-305, 313., articleTitle=基于MobileNet-SSD的安全帽佩戴检测算法, refAbstract=null), Reference(id=1261267681521578842, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, doi=null, pmid=null, pmcid=null, year=2021, volume=47, issue=10, pageStart=298, pageEnd=305, url=null, language=null, rfNumber=[2], rfOrder=2, authorNames=Xu Xianfeng, Zhao Wanfu, Zou Haoquan, journalName=Computer Engineering, refType=null, unstructuredReference=Xu Xianfeng, Zhao Wanfu, Zou Haoquan, et al. Safety helmet wearing detection algorithm based on MobileNet-SSD[J]. Computer Engineering, 2021, 47(10): 298-305, 313., articleTitle=Safety helmet wearing detection algorithm based on MobileNet-SSD, refAbstract=null), Reference(id=1261267681718711133, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, doi=null, pmid=null, pmcid=null, year=2021, volume=57, issue=8, pageStart=192, pageEnd=197, url=null, language=null, rfNumber=[3], rfOrder=3, authorNames=李明山, 韩清鹏, 张天宇, journalName=计算机工程与应用, refType=null, unstructuredReference=李明山, 韩清鹏, 张天宇, . 改进SSD的安全帽检测方法[J]. 计算机工程与应用, 2021, 57(8): 192-197., articleTitle=改进SSD的安全帽检测方法, refAbstract=null), Reference(id=1261267681966175071, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, doi=null, pmid=null, pmcid=null, year=2021, volume=57, issue=8, pageStart=192, pageEnd=197, url=null, language=null, rfNumber=[3], rfOrder=4, authorNames=Li Mingshan, Han Qingpeng, Zhang Tianyu, journalName=Computer Engineering and Applications, refType=null, unstructuredReference=Li Mingshan, Han Qingpeng, Zhang Tianyu, et al. Improving the safety helmet detection method for SSDs[J]. Computer Engineering and Applications, 2021, 57(8): 192-197., articleTitle=Improving the safety helmet detection method for SSDs, refAbstract=null), Reference(id=1261267682045866850, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, doi=null, pmid=null, pmcid=null, year=2024, volume=60, issue=6, pageStart=230, pageEnd=237, url=null, language=null, rfNumber=[4], rfOrder=5, authorNames=李嘉信, 胡杨, 黄协舟, journalName=计算机工程与应用, refType=null, unstructuredReference=李嘉信, 胡杨, 黄协舟, . 面向小目标的多空间层次安全帽检测[J]. 计算机工程与应用, 2024, 60(6): 230-237., articleTitle=面向小目标的多空间层次安全帽检测, refAbstract=null), Reference(id=1261267682188473189, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, doi=null, pmid=null, pmcid=null, year=2024, volume=60, issue=6, pageStart=230, pageEnd=237, url=null, language=null, rfNumber=[4], rfOrder=6, authorNames=Li Jiaxin, Hu Yang, Huang Xiezhou, journalName=Computer Engineering and Applications, refType=null, unstructuredReference=Li Jiaxin, Hu Yang, Huang Xiezhou, et al. Multi spatial hierarchical safety helmet detection for small targets[J]. Computer Engineering and Applications, 2024, 60(6): 230-237., articleTitle=Multi spatial hierarchical safety helmet detection for small targets, refAbstract=null), Reference(id=1261267682310108008, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, doi=null, pmid=null, pmcid=null, year=2022, volume=45, issue=17, pageStart=72, pageEnd=78, url=null, language=null, rfNumber=[5], rfOrder=7, authorNames=丁田, 陈向阳, 周强, journalName=电子测量技术, refType=null, unstructuredReference=丁田, 陈向阳, 周强, . 基于改进YOLOX的安全帽佩戴实时检测[J]. 电子测量技术, 2022, 45(17): 72-78., articleTitle=基于改进YOLOX的安全帽佩戴实时检测, refAbstract=null), Reference(id=1261267682381411179, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, doi=null, pmid=null, pmcid=null, year=2022, volume=45, issue=17, pageStart=72, pageEnd=78, url=null, language=null, rfNumber=[5], rfOrder=8, authorNames=Ding Tian, Chen Xiangyang, Zhou Qiang, journalName=Electronic Measurement Technology, refType=null, unstructuredReference=Ding Tian, Chen Xiangyang, Zhou Qiang, et al. Real time detection of helmet wearing based on improved YOLOX[J]. Electronic Measurement Technology, 2022, 45(17): 72-78., articleTitle=Real time detection of helmet wearing based on improved YOLOX, refAbstract=null), Reference(id=1261267682482074478, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, doi=null, pmid=null, pmcid=null, year=2023, volume=32, issue=3, pageStart=033013, pageEnd=null, url=null, language=null, rfNumber=[6], rfOrder=9, authorNames=Shao X, Liu S, Li X, journalName=Journal of Electronic Imaging, refType=null, unstructuredReference=Shao X, Liu S, Li X, et al. FM-YOLOv7: an improved detection method for mine personnel helmet[J]. Journal of Electronic Imaging, 2023, 32(3): 033013., articleTitle=FM-YOLOv7: an improved detection method for mine personnel helmet, refAbstract=null), Reference(id=1261267682586932081, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, doi=null, pmid=null, pmcid=null, year=2023, volume=43, issue=6, pageStart=183, pageEnd=190, url=null, language=null, rfNumber=[7], rfOrder=10, authorNames=高凯, 董立红, 邓凡, journalName=矿业研究与开发, refType=null, unstructuredReference=高凯, 董立红, 邓凡. 基于递归门控卷积和上下文注意力的煤块检测算法[J]. 矿业研究与开发, 2023, 43(6): 183-190., articleTitle=基于递归门控卷积和上下文注意力的煤块检测算法, refAbstract=null), Reference(id=1261267682658235252, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, doi=null, pmid=null, pmcid=null, year=2023, volume=43, issue=6, pageStart=183, pageEnd=190, url=null, language=null, rfNumber=[7], rfOrder=11, authorNames=Gao Kai, Dong Lihong, Deng Fan, journalName=Mining Research and Development, refType=null, unstructuredReference=Gao Kai, Dong Lihong, Deng Fan. Coal block detection algorithm based on recursive gated convolution and contextual attention[J]. Mining Research and Development, 2023, 43(6): 183-190., articleTitle=Coal block detection algorithm based on recursive gated convolution and contextual attention, refAbstract=null), Reference(id=1261267682725344119, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, doi=null, pmid=null, pmcid=null, year=2023, volume=49, issue=12, pageStart=18, pageEnd=24, url=null, language=null, rfNumber=[8], rfOrder=12, authorNames=宋立业, 赵小萱, 崔昊, journalName=工矿自动化, refType=null, unstructuredReference=宋立业, 赵小萱, 崔昊. 基于改进YOLOv7的矿用电铲检测算法[J]. 工矿自动化, 2023, 49(12): 18-24, 32., articleTitle=基于改进YOLOv7的矿用电铲检测算法, refAbstract=null), Reference(id=1261267682842784633, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, doi=null, pmid=null, pmcid=null, year=2023, volume=49, issue=12, pageStart=18, pageEnd=24, url=null, language=null, rfNumber=[8], rfOrder=13, authorNames=Song Liye, Zhao Xiaoxuan, Cui Hao, journalName=Industrial and Mining Automation, refType=null, unstructuredReference=Song Liye, Zhao Xiaoxuan, Cui Hao. Mining electric shovel detection algorithm based on improved YOLOv7[J]. Industrial and Mining Automation, 2023, 49(12): 18-24, 32., articleTitle=Mining electric shovel detection algorithm based on improved YOLOv7, refAbstract=null), Reference(id=1261267682947642234, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, doi=null, pmid=null, pmcid=null, year=2023, volume=null, issue=null, pageStart=1, pageEnd=5, url=null, language=null, rfNumber=[9], rfOrder=14, authorNames=Ouyang D L, He S, Zhang G Z, journalName=IEEE International Conference on Acoustics, Speech and Signal Processing, refType=null, unstructuredReference=Ouyang D L, He S, Zhang G Z, et al. Efficient multi-scale attention module with cross-spatial learning[C]// IEEE International Conference on Acoustics, Speech and Signal Processing. New York: IEEE, 2023: 1-5., articleTitle=Efficient multi-scale attention module with cross-spatial learning, refAbstract=null), Reference(id=1261267683077665661, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=http://kns.cnki.net/kcms/detail/34.1055.td.20240516.1134.006.html, language=null, rfNumber=[10], rfOrder=15, authorNames=顾清华, 周宇静, 王丹, journalName=null, refType=null, unstructuredReference=顾清华, 周宇静, 王丹, . 基于多尺度特征提取的露天矿区道路坑洼检测模型[J/OL].(2025-04-24)[2024-06-18]. http://kns.cnki.net/kcms/detail/34.1055.td.20240516.1134.006.html., articleTitle=基于多尺度特征提取的露天矿区道路坑洼检测模型, refAbstract=null), Reference(id=1261267683216077695, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=http://kns.cnki.net/kcms/detail/34.1055.td.20240516.1134.006.html, language=null, rfNumber=[10], rfOrder=16, authorNames=Gu Qinghua, Zhou Yujing, Wang Dan, journalName=null, refType=null, unstructuredReference=Gu Qinghua, Zhou Yujing, Wang Dan, et al. A detection model for road potholes in open-pit mining areas based on multi-scale feature extraction[J/OL]. (2025-04-24)[2024-06-18]. http://kns.cnki.net/kcms/detail/34.1055.td.20240516.1134.006.html., articleTitle=A detection model for road potholes in open-pit mining areas based on multi-scale feature extraction, refAbstract=null), Reference(id=1261267683417404288, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=[11], rfOrder=17, authorNames=Chao L, Aojun Z, Anbang Y, journalName=arXiv Preprint, refType=null, unstructuredReference=Chao L, Aojun Z, Anbang Y. Omni-dimensional dynamic convolution[J]. arXiv Preprint, 2019: arXiv: 2209.07947., articleTitle=Omni-dimensional dynamic convolution, refAbstract=null), Reference(id=1261267683509678978, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, doi=null, pmid=null, pmcid=null, year=2024, volume=60, issue=20, pageStart=189, pageEnd=197, url=null, language=null, rfNumber=[12], rfOrder=18, authorNames=崔丽群, 曹华维, journalName=计算机工程与应用, refType=null, unstructuredReference=崔丽群, 曹华维. 改进YOLOv7的航拍图像目标检测[J]. 计算机工程与应用, 2024, 60(20): 189-197., articleTitle=改进YOLOv7的航拍图像目标检测, refAbstract=null), Reference(id=1261267683979441028, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, doi=null, pmid=null, pmcid=null, year=2024, volume=60, issue=20, pageStart=189, pageEnd=197, url=null, language=null, rfNumber=[12], rfOrder=19, authorNames=Cui Liqun, Cao Huawei, journalName=Computer Engineering and Applications, refType=null, unstructuredReference=Cui Liqun, Cao Huawei. Improving YOLOv7 for aerial image object detection[J]. Computer Engineering and Applications, 2024, 60(20): 189-197., articleTitle=Improving YOLOv7 for aerial image object detection, refAbstract=null), Reference(id=1261267684260459399, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, doi=null, pmid=null, pmcid=null, year=2024, volume=24, issue=18, pageStart=7833, pageEnd=7842, url=null, language=null, rfNumber=[13], rfOrder=20, authorNames=王俊岭, 王晨晨, 熊玉华, journalName=科学技术与工程, refType=null, unstructuredReference=王俊岭, 王晨晨, 熊玉华. 基于YOLOv5l-Im的排水管道缺陷检测方法及效果分析[J]. 科学技术与工程, 2024, 24(18): 7833-7842., articleTitle=基于YOLOv5l-Im的排水管道缺陷检测方法及效果分析, refAbstract=null), Reference(id=1261267684394677129, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, doi=null, pmid=null, pmcid=null, year=2024, volume=24, issue=18, pageStart=7833, pageEnd=7842, url=null, language=null, rfNumber=[13], rfOrder=21, authorNames=Wang Junling, Wang Chenchen, Xiong Yuhua, journalName=Science Technology and Engineering, refType=null, unstructuredReference=Wang Junling, Wang Chenchen, Xiong Yuhua. Drainage pipeline defect detection method and effect analysis based on YOLOv5l-Im[J]. Science Technology and Engineering, 2024, 24(18): 7833-7842., articleTitle=Drainage pipeline defect detection method and effect analysis based on YOLOv5l-Im, refAbstract=null), Reference(id=1261267684579226506, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, doi=null, pmid=null, pmcid=null, year=2022, volume=22, issue=14, pageStart=5659, pageEnd=5665, url=null, language=null, rfNumber=[14], rfOrder=22, authorNames=秦子豪, 雷鸣, 宋文广, journalName=科学技术与工程, refType=null, unstructuredReference=秦子豪, 雷鸣, 宋文广, . 基于轻量化深度学习模型的安全帽检测方法[J]. 科学技术与工程, 2022, 22(14): 5659-5665., articleTitle=基于轻量化深度学习模型的安全帽检测方法, refAbstract=null), Reference(id=1261267684717638540, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, doi=null, pmid=null, pmcid=null, year=2022, volume=22, issue=14, pageStart=5659, pageEnd=5665, url=null, language=null, rfNumber=[14], rfOrder=23, authorNames=Qin Zihao, Lei Ming, Song Wenguang, journalName=Science Technology and Engineering, refType=null, unstructuredReference=Qin Zihao, Lei Ming, Song Wenguang, et al. Safety helmet detection method based on lightweight deep learning model[J]. Science Technology and Engineering, 2022, 22(14): 5659-5665., articleTitle=Safety helmet detection method based on lightweight deep learning model, refAbstract=null), Reference(id=1261267684797330318, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, doi=null, pmid=null, pmcid=null, year=2024, volume=24, issue=18, pageStart=7726, pageEnd=7732, url=null, language=null, rfNumber=[15], rfOrder=24, authorNames=桑雨, 李立权, 李铁, journalName=科学技术与工程, refType=null, unstructuredReference=桑雨, 李立权, 李铁. 轻量化YOLOv7-tiny的遥感图像小目标检测[J]. 科学技术与工程, 2024, 24(18): 7726-7732., articleTitle=轻量化YOLOv7-tiny的遥感图像小目标检测, refAbstract=null), Reference(id=1261267684872827792, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, doi=null, pmid=null, pmcid=null, year=2024, volume=24, issue=18, pageStart=7726, pageEnd=7732, url=null, language=null, rfNumber=[15], rfOrder=25, authorNames=Sang Yu, Li Liquan, Li Tie, journalName=Science Technology and Engineering, refType=null, unstructuredReference=Sang Yu, Li Liquan, Li Tie. Small target detection in remote sensing images of lightweight YOLOv7-tiny[J]. Science Technology and Engineering, 2024, 24(18): 7726-7732., articleTitle=Small target detection in remote sensing images of lightweight YOLOv7-tiny, refAbstract=null)], funds=[Fund(id=1261267680867267409, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, awardId=202102010101010, language=CN, fundingSource=山西省重点研发计划(202102010101010), fundOrder=null, country=null)], companyList=[AuthorCompany(id=1261267661909013167, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, xref=1, ext=[AuthorCompanyExt(id=1261267661942567600, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, companyId=1261267661909013167, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1 School of Mechanical Engineering, Taiyuan University of Science and Technology, Taiyuan 030024, China), AuthorCompanyExt(id=1261267662005482162, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, companyId=1261267661909013167, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1 太原科技大学机械工程学院, 太原 030024)]), AuthorCompany(id=1261267662261334709, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, xref=2, ext=[AuthorCompanyExt(id=1261267662269723319, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, companyId=1261267662261334709, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2 School of Computer Science and Technology, Taiyuan University of Science and Technology, Taiyuan 030024, China), AuthorCompanyExt(id=1261267662278111928, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, companyId=1261267662261334709, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2 太原科技大学计算机科学与技术学院, 太原 030024)])], figs=[ArticleFig(id=1261267674378679054, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, language=EN, label=Fig.1, caption=YOLOv7 network structure, figureFileSmall=aMNm3B5mQzraA3e8Y+f8Kg==, figureFileBig=GtaDrscAKBmG17Vx/zqMCw==, tableContent=null), ArticleFig(id=1261267675058156305, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, language=CN, label=图1, caption=YOLOv7网络结构

Input为输入;Backbone为主干;Head为头部;CBS为卷积-批归一化Silu模块;cat为拼接操作;UPSample为上采样量模块;ELAN为高级层级集合网络;MP为最大池化双路径模块;SPPCSPC为空间金字塔池化-跨阶段部分连接;ELAN-H为头部增强型ELAN;CBM为卷积-批归一化Mish模块

, figureFileSmall=aMNm3B5mQzraA3e8Y+f8Kg==, figureFileBig=GtaDrscAKBmG17Vx/zqMCw==, tableContent=null), ArticleFig(id=1261267676723295002, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, language=EN, label=Fig.2, caption=MELAN module for aggregating peripheral information, figureFileSmall=wQQgKoV4iPmE/jDsNR5H/w==, figureFileBig=vX+2sKtfzn/8/Iwo/0dafQ==, tableContent=null), ArticleFig(id=1261267676903650076, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, language=CN, label=图2, caption=聚合周边信息的MELAN模块

CBS为卷积操作;cat为拼接操作;MEMA为注意力机制

, figureFileSmall=wQQgKoV4iPmE/jDsNR5H/w==, figureFileBig=vX+2sKtfzn/8/Iwo/0dafQ==, tableContent=null), ArticleFig(id=1261267677176279841, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, language=EN, label=Fig.3, caption=EMA module structure, figureFileSmall=q0uwOiYXer8wr+VfCCnAdg==, figureFileBig=YmxLr0A7Oq7xp3L+WX9GNw==, tableContent=null), ArticleFig(id=1261267677436326691, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, language=CN, label=图3, caption=EMA模块结构

Ave Pool为平均池化;Softmax为归一化指数函数;Conv为卷积;Concat为拼接操作;Sigmoid为激活函数;Re-weight为分配权重;Matmul为矩阵相乘函数;Input为输入;Output为输出;Groups为分组;GroupNorm为组归一化;Cross-spatiallearning为跨空间学习

, figureFileSmall=q0uwOiYXer8wr+VfCCnAdg==, figureFileBig=YmxLr0A7Oq7xp3L+WX9GNw==, tableContent=null), ArticleFig(id=1261267677599904550, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, language=EN, label=Fig.4, caption=MEMA module structure, figureFileSmall=Hcs35m0UuKMjJPZqWW7d9g==, figureFileBig=IDyBvujE6HysmPd8KyFVuw==, tableContent=null), ArticleFig(id=1261267677801231145, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, language=CN, label=图4, caption=MEMA模块结构

Ave Pool为平均池化;Max Pool为平均池化;Softmax为归一化指数函数;Conv为卷积;Concat为拼接操作;Sigmoid为激活函数;Re-weight为分配权重;Matmul为矩阵相乘函数;Input为输入;Output为输出;Groups为分组;GroupNorm为组归一化;Cross-spatial learning为跨空间学习

, figureFileSmall=Hcs35m0UuKMjJPZqWW7d9g==, figureFileBig=IDyBvujE6HysmPd8KyFVuw==, tableContent=null), ArticleFig(id=1261267677977391915, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, language=EN, label=Fig.5, caption=OD-SMP module structure, figureFileSmall=YY7D1y5JawHaU5rDCDZmKw==, figureFileBig=Q9M4o7VGwIQWUSzLQ69SvA==, tableContent=null), ArticleFig(id=1261267678107415342, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, language=CN, label=图5, caption=OD-SMP模块结构图

红色模块为改进部分;SoftPool为软池化;ODConv为全维动态卷积;cat为拼接操作

, figureFileSmall=YY7D1y5JawHaU5rDCDZmKw==, figureFileBig=Q9M4o7VGwIQWUSzLQ69SvA==, tableContent=null), ArticleFig(id=1261267678338102065, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, language=EN, label=Fig.6, caption=Full dimensional dynamic convolution calculation flowchart, figureFileSmall=qcadv9UZmjRLHRrFmbYKnQ==, figureFileBig=49I0ciI9YYxk5sySq1vOog==, tableContent=null), ArticleFig(id=1261267678531040054, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, language=CN, label=图6, caption=全维动态卷积计算流程图, figureFileSmall=qcadv9UZmjRLHRrFmbYKnQ==, figureFileBig=49I0ciI9YYxk5sySq1vOog==, tableContent=null), ArticleFig(id=1261267678837224248, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, language=EN, label=Fig.7, caption=Detection effect diagram, figureFileSmall=Ss+r+SZ53WehmcMXOQwqxQ==, figureFileBig=eDRvo9jEH0hbDA6oNl8Pmw==, tableContent=null), ArticleFig(id=1261267679256654651, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, language=CN, label=图7, caption=检测效果图, figureFileSmall=Ss+r+SZ53WehmcMXOQwqxQ==, figureFileBig=eDRvo9jEH0hbDA6oNl8Pmw==, tableContent=null), ArticleFig(id=1261267679449592637, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, language=EN, label=Fig.8, caption=mAP curves before and after improvement, figureFileSmall=YCRul3ke77jaSRpg46exEA==, figureFileBig=3zqg0aWDt67CXW0t9goonA==, tableContent=null), ArticleFig(id=1261267679659307838, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, language=CN, label=图8, caption=改进前后mAP曲线, figureFileSmall=YCRul3ke77jaSRpg46exEA==, figureFileBig=3zqg0aWDt67CXW0t9goonA==, tableContent=null), ArticleFig(id=1261267679764165441, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, language=EN, label=Table 1, caption=

Experimental environment

, figureFileSmall=null, figureFileBig=null, tableContent=
配置名称 版本参数
操作系统 Ubuntu 18.04
GPU NVIDIA GeForce RTX 3060
CPU Intel Core i9-12900H
RAM/G 16
深度学习框架 PyTorch
CUDA版本 11.6
Python版本 3.8
), ArticleFig(id=1261267679877411652, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, language=CN, label=表1, caption=

实验环境

, figureFileSmall=null, figureFileBig=null, tableContent=
配置名称 版本参数
操作系统 Ubuntu 18.04
GPU NVIDIA GeForce RTX 3060
CPU Intel Core i9-12900H
RAM/G 16
深度学习框架 PyTorch
CUDA版本 11.6
Python版本 3.8
), ArticleFig(id=1261267680103904071, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, language=EN, label=Table 2, caption=

Results of ablation experiment

, figureFileSmall=null, figureFileBig=null, tableContent=
YOLOv7 MELAN OD-SMP Wise-IoU mAP/
%
mAP@0.5/
%
GFLOPs
81.4 82.6 26.7
83.9 82.9 30.0
83.1 83.2 23.8
86.3 85.2 26.7
89.5 88.3 30.0
90.9 87.6 23.8
91.2 90.1 27.2
94.9 93.5 27.2
), ArticleFig(id=1261267680208761674, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, language=CN, label=表2, caption=

消融实验结果

, figureFileSmall=null, figureFileBig=null, tableContent=
YOLOv7 MELAN OD-SMP Wise-IoU mAP/
%
mAP@0.5/
%
GFLOPs
81.4 82.6 26.7
83.9 82.9 30.0
83.1 83.2 23.8
86.3 85.2 26.7
89.5 88.3 30.0
90.9 87.6 23.8
91.2 90.1 27.2
94.9 93.5 27.2
), ArticleFig(id=1261267680296842060, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, language=EN, label=Table 3, caption=

Comparison results of different algorithms

, figureFileSmall=null, figureFileBig=null, tableContent=
不同算法 mAP/% mAP@0.5/% GFLOPs
SSD 58.4 49.2 343.37
Fast-RCNN 66.5 51.9 193.78
YOLOv5 75.0 38.0 15.80
YOLOv7 81.4 82.6 26.70
YOLOv7-A 93.5 91.7 28.30
YOLOv8 88.8 86.8 8.20
本文算法 94.9 93.5 27.20
), ArticleFig(id=1261267680460419918, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1261267654854193768, language=CN, label=表3, caption=

不同算法对比效果结果

, figureFileSmall=null, figureFileBig=null, tableContent=
不同算法 mAP/% mAP@0.5/% GFLOPs
SSD 58.4 49.2 343.37
Fast-RCNN 66.5 51.9 193.78
YOLOv5 75.0 38.0 15.80
YOLOv7 81.4 82.6 26.70
YOLOv7-A 93.5 91.7 28.30
YOLOv8 88.8 86.8 8.20
本文算法 94.9 93.5 27.20
)], 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.1671-1815.2406421, detailUrlEn=https://castjournals.cast.org.cn/joweb/kxjsygc/EN/10.12404/j.issn.1671-1815.2406421, pdfUrlCn=https://castjournals.cast.org.cn/joweb/kxjsygc/CN/PDF/10.12404/j.issn.1671-1815.2406421, pdfUrlEn=https://castjournals.cast.org.cn/joweb/kxjsygc/EN/PDF/10.12404/j.issn.1671-1815.2406421, aliStartDate=null, aliEndDate=null, collectionFlag=false, citedCount=null, citedUrl=null, reference=null)
收藏切换
基于改进YOLOv7的煤矿井下低质影像安全帽检测
收藏切换
PDF下载
康程洋 1 , 张延军 1, * , 张睿 2
科学技术与工程 | 论文·自动化技术、计算机技术 2025,25(20): 8595-8603
收起
收藏切换
科学技术与工程 | 论文·自动化技术、计算机技术 2025, 25(20): 8595-8603
基于改进YOLOv7的煤矿井下低质影像安全帽检测
全屏
康程洋1 , 张延军1, * , 张睿2
作者信息
  • 1 太原科技大学机械工程学院, 太原 030024
  • 2 太原科技大学计算机科学与技术学院, 太原 030024
  • 康程洋(2000—),男,汉族,河南商丘人,硕士研究生。研究方向:机器视觉及智能控制。E-mail:

通讯作者:

* 张延军(1982—),男,满族,河北承德人,博士,教授。研究方向:机电液传动及智能控制。E-mail:
Detection of Low-quality Image Safety Helmet in Coal Mine Based on Improved YOLOv7
Cheng-yang KANG1 , Yan-jun ZHANG1, * , Rui ZHANG2
Affiliations
  • 1 School of Mechanical Engineering, Taiyuan University of Science and Technology, Taiyuan 030024, China
  • 2 School of Computer Science and Technology, Taiyuan University of Science and Technology, Taiyuan 030024, China
出版时间: 2025-07-18 doi: 10.12404/j.issn.1671-1815.2406421
文章导航
收藏切换

由于井下环境复杂,低光照以及安全帽的物体较小导致的检测效果不理想。针对复杂环境下的低质影像,提出一种改进YOLOv7的煤矿井下低质影像安全帽检测。首先,针对低光照条件下的图像特征易受噪声干扰的局限性,提出多尺度MELAN模块,通过构建多尺度注意力机制来捕捉更广泛的上下文信息,进而强化特征信息的提取,并有效抑制噪声的干扰;其次,在主干网络使用软池化和全维动态卷积构建OD-SMP模块,减少了特征映射中的信息弥散,保留了更多上下文信息,增强对小目标的检测能力;最后,对于煤矿井下不同光照和距离的复杂背景和环境,导致检测样本质量出现参差不齐的情况,使用Wise-IoU作为损失函数。实验结果表明:改进模型的平均精度为94.9%,比原来的YOLOv7模型高出13.5%,具有更好的检测效果。

低光照  /  多尺度  /  全维动态卷积  /  注意力机制  /  损失函数

Due to the complex underground environment, low lighting conditions, and the small size of hard hats, the detection results are not ideal. To address low-quality images in complex environments, an improved YOLOv7 for hard hat detection in low-quality images from underground coal mines was proposed. Firstly, addressing the limitation that image features were susceptible to noise interference under low-light conditions, a multi-scale MELAN module was introduced. By constructing a multi-scale attention mechanism, broader contextual information was captured, thereby enhancing feature extraction and effectively suppressing noise interference. Secondly, the OD-SMP module was constructed using soft pooling and full-dimensional dynamic convolution in the backbone network, which reduced information diffusion in feature mappings, retained more contextual information, and enhanced the detection capability for small targets. Finally, to address the varying quality of detection samples caused by the complex backgrounds and environments with different lighting and distances in underground coal mines, Wise-IoU was used as the loss function. Experimental results show that the average precision of the improved model is 94.9%, which is 13.5% higher than the original YOLOv7 model, demonstrating better detection performance.

low illumination  /  multi scale  /  full dimensional dynamic convolution  /  attention mechanism  /  loss function
康程洋, 张延军, 张睿. 基于改进YOLOv7的煤矿井下低质影像安全帽检测. 科学技术与工程, 2025 , 25 (20) : 8595 -8603 . DOI: 10.12404/j.issn.1671-1815.2406421
Cheng-yang KANG, Yan-jun ZHANG, Rui ZHANG. Detection of Low-quality Image Safety Helmet in Coal Mine Based on Improved YOLOv7[J]. Science Technology and Engineering, 2025 , 25 (20) : 8595 -8603 . DOI: 10.12404/j.issn.1671-1815.2406421
煤矿工人必须佩戴安全帽,这是采矿现场安全管理的关键,安全帽可以有效减少和分配冲击力。根据最近的矿山安全事故报告,由于缺乏安全防范意识,偶尔会发生不戴头盔造成的伤亡。因此,为了保护煤矿工人的生命安全,煤矿单位必须监督他们正确佩戴安全帽。在早期的人工监测任务中,经常使用人工监管来确定煤矿工人是否佩戴头盔,由于监管人员的主观性和难以了解矿山情况,这容易产生安全隐患,并增加了高昂的劳动力成本。5G、互联网和大数据等技术的发展和成熟,促进了智能矿山的建设。实时监控人员信息和提供安全保障的主要方法是“视频监控”相结合[1]。智能检测煤矿工人是否佩戴安全帽,对消除人工监测造成的缺陷具有重要的研究价值,这引起了学者们的广泛关注。
在当前的研究背景下,深度学习技术已被中外学者广泛应用于安全帽检测领域。徐先峰等[2]构建MobileNet-SSD算法提高检测速度,解决安全帽数据集规模较小、网络难以充分拟合特征的问题。李明山等[3]在SSD(single shot multibox detector)模型中引入特征融合机制,通过构建分支网络的方式实现了特征信息的有效整合。针对默认框的配置进行了优化调整,这些改进措施显著提高了算法在实际应用中对安全帽检测的准确性。李嘉信等[4]提出一种面向小目标的多空间层次安全帽佩戴检测算法,减少了检测中漏检的现象。丁田等[5]在主干网络中引入注意力机制并使用CIoU损失函数来计算损失,减少检测过程中的干扰,提高对小目标的检测能力。
尽管基于计算机视觉的安全帽检测算法已应用于建筑、工业车间等,但在煤矿井下仍存在一些特定的挑战。对于不同光照和距离的复杂背景和环境,检测算法需要具有更强的目标识别能力和智能处理能力[6]。鉴于此,提出改进YOLOv7的煤矿井下低质影像安全帽检测算法。煤矿井下安全帽佩戴检测做以下改进:在特征提取网络ELAN模块加入设计的多尺度注意力机制MEMA模块,增强获得的特征信息,减弱井下低光照环境下噪声的影响[7];在主干网络使用软池化构建OD-SMP模块,减少了特征映射中的信息弥散,保留更多上下文信息,增强对小目标的检测能力;对于煤矿井下不同光照和距离的复杂背景和环境,导致检测样本质量出现参差不齐的情况,使用Wise-IoU损失函数,平衡样本质量。
YOLOv7[8]的基本架构包括:输入、主干和头部。输入图像通过Input进行预处理,并将其设置为640×640 RGB图像。主干包括CBS(卷积-批归一化Silu模块)、ELAN(高级层级集合网络)和MP(最大池化双路径模块)。输入图像通过主干进行特征提取。然后,在头部采用多尺度特征融合模式,可以保留更多的细节,有助于提高模型对目标的表示能力。YOLOv7有3个检测头,特征融合后的特征将传递到输出层,输出层将预测目标的位置和类别,并生成相应的边界框。检测头输出3个特征尺度:20×20、40×40和80×80,3个尺度分别以大、中、小目标对应的尺度检测目标。YOLOv7在预测结果中应用非极大值抑制(non-maximum suppression,NMS)算法来去除冗余的边界框,并确保每个目标只有一个与其对应的边界框。YOLOv7网络结构如图1所示。
为减少低光照噪声信息对检测效果的干扰,在特征提取部分的ELAN(efficient layer aggregation network)模块中添加自设计MEMA注意力模块,通过多尺度注意力模块对全局通道间信息建模的能力,抑制不重要信息的能力,它能够使空间语义特征在每个特征组中均匀分布,对于煤矿井下低照度环境,产生更清晰的目标特征,有效地增强神经元信息,使网络更加关注数据的重要特征,增强安全帽特征,减弱低照度产生的噪声信息对检测器的影响,多尺度MELAN模块如图2所示。一种新型的注意力机制EMA(efficient multi-scale attention)模块,结构图如图3所示,该模块无需进行降维操作即可完整保留每个通道的信息,同时显著降低计算成本[9]。能够将部分通道巧妙地重构为批量维度,并进一步将这些通道分组为多个子特征集。这一设计确保了每个特征组内部的空间语义特征分布均匀,从而提升了数据处理的效果和效率。EMA模块不仅通过编码全局信息来动态调整每个并行分支的通道权重,还利用跨维度交互的机制,有效整合了两个并行分支的输出特征,旨在扩展网络的全局感受野、保留精确的位置信息,同时降低计算需求。然而,EMA的平均池化计算区域内像素平均值,会丢失图像显著特征,而MEMA注意力机制采用全局平均池化和最大池化的双重策略代替EMA模块的全局平均池化,分别沿水平和垂直方向学习图像的一维特征,更好的保留安全帽的细节信息[10]
MEMA注意力机制结构如图4所示,针对输入特征XRC×H×W(R为邻域;C为特征图的通道数;H为输出特征图的高度;W为输出特征图的宽度),MEMA首先在通道维度将其划分成G个子特征X=[X0,Xi,…,XG-1],XiRC//G×H×W,取GC,每组特征通过学习到的注意力权重进行增强。针对每个子特征,MEMA架构巧妙地设计3条并行处理路径,旨在从不同维度提取分组特征图的注意力强度系数。其中,前两个并行路径均嵌入了1×1卷积层,随后分别应用了自适应全局平均池化和最大池化技术,这两种池化方法沿着X轴和Y轴方向独立地对特征图的通道信息进行压缩编码,然后这两条路径输出的编码特征被巧妙地连接起来,形成一个融合了多方向信息的特征表示,紧接着,这一综合特征通过一个共享的1×1卷积层进行进一步处理,输出两个向量后使用Sigmoid激活函数进行线性拟合;第3个分支包含3×3卷积,旨在捕获多尺度特征表示。在跨空间学习(cross-spatial learning)部分,MEMA对不同空间维度方向的跨空间信息进行聚合。首先,对引入张量为1×1分支所输出的全局信息进行归一化操作;其次,对1×1分支与3×3分支各自生成的全局空间信息进行独立编码,分别采用2D全局平均池化和最大池化技术,再利用Softmax函数进行拟合线性变换,将输出中贡献度最小的分支在信道维度上进行特征转换,调整其形状以匹配所需的维度结构,这一步骤巧妙地实现了跨空间信息的聚合,通过强化或抑制特定分支的贡献,提升了整体特征表示的鲁棒性和信息量;最后,通过跨空间交互模块聚合两个通道注意力权重值,捕捉像素级的成对关系,并通过Sigmoid激活函数输出特征映射,从而增强原始输入特征,得到最终的输出特征图。
煤矿井下环境的图像含有的特征细节信息不明显,且易与背景信息混淆,主干网络在提取安全帽和头部特征时,MP模块中的最大池化会丢失一部分本就不丰富的目标细节特征,造成对目标有效特征的提取能力不足。为了减少MP模块的信息丢失,增强模型对目标的特征提取能力,本研究利用SoftPool,增强检测器对受到噪声影响的不明显特征信息的辨识能力,更好地捕捉小目标的细节特征,对于小目标的检测尤其有帮助;引入全维动态卷积ODConv,替代原有卷积层,从多个维度捕捉关键的上下文信息,然后为每个分支提取的特征图分配可训练的权重值,协助模型关注对输出特征贡献度大的信息,从而构建了一种联合动态卷积的加权采样模块(OD-SMP),提升模型在煤矿井下尺度变化剧烈场景中的识别能力,其结构如图5所示。
SoftPool采用一种独特的加权池化方法,即利用Softmax函数进行操作,这使得它与传统的池化技术有所不同。SoftPool不仅能保持特征的表达性,而且其操作是可微的,确保在神经网络训练过程中,每一次反向传播的梯度都能得到有效更新。SoftPool可以让池化内核中的每一个激活因子被充分利用,这一特性仅需要极少的额外内存开销。同时,增大了相近特征信息的区分度,又保留了整体感受野的特征信息,从而显著提高了算法的准确率。
SoftPool以自然指数e作为底数,使得较大的激活值在输出中占据更大的影响力。更重要的是,SoftPool的可微性使得在反向传播过程中,局部邻域R内的每个激活都能至少获得一个最小的梯度分配,从而保证了梯度更新的全面性和有效性。在SoftPool的池化过程中,每一个激活ai都会被赋予一个权重wi,该权重为该激活的自然指数相对于邻域R内所有激活的自然指数之和的比值,可表示为
wi=$\frac{{\mathrm{e}}^{{a}_{i}}}{\sum _{j\in R}{\mathrm{e}}^{{a}_{j}}}$
通过对内核邻域内所有激活的加权求和得到SoftPool输出值为
$\stackrel{-}{\mathrm{a}}$=$\sum _{i\in R}$wiai
全维动态卷积[11]采用多维注意力并行策略,通过在卷积核的4个维度上并行地分配注意力权重,打破传统卷积固有感受野的束缚,通过捕捉和强化显著表型特征,帮助模型适应不同尺寸和形状的特征,更好地捕捉安全帽的细节特征。
图6所示,ODConv模块在处理输入特征X时,利用全局平均池化层(GAP)对其进行空间维度的压缩,首先将特征图转化为通道数为Cin的特征向量,其次采用一个全连接层(FC)将特征向量映射成低维空间,然后采用ReLU激活函数对特征向量中的负值进行剔除,以筛选出关键信息。构建了4个独立的头部分支,每个分支专注于生成特定类别的注意力标量,4种注意力分支可以帮助模型提取丰富的上下文信息[12]。全维动态卷积表达式为

y=(αa1αb1αc1αd1W1+αa2αb2αc2αd2W2+…+αanαbnαcnαdnW1n)*X

式(3)中:X${R}^{H\times W\times {C}_{\mathrm{i}\mathrm{n}}}$y${R}^{H\times W\times {C}_{\mathrm{o}\mathrm{u}\mathrm{t}}}$分别为输入和输出特征,其中,HW分别为特征的长、宽,CinCout为通道数;Wi${R}^{k\times k\times {C}_{\mathrm{i}\mathrm{n}}}$为输出滤波器的第i个卷积核;αaiRK×K为沿着卷积核Wi核空间的注意标量;αbi${R}^{{C}_{\mathrm{i}\mathrm{n}}}$为沿着卷积核Wi输入通道维数的注意标量;αci${R}^{{C}_{\mathrm{o}\mathrm{u}\mathrm{t}}}$为沿着卷积核Wi输出通道维数的注意标量;αdiR为沿卷积核数的注意标量;☉为沿核空间不同维数的乘法运算;*为卷积运算。
全维动态卷积通过并行分支计算为卷积核沿核空间4个维度捕捉有效的上下文信息,4种注意力权重按照特定的顺序逐步累加到卷积核上,实现对不同大小的物体的动态调整[13]
在YOLOv7中,其损失函数CIoU是通过综合边界框回归损失、置信度损失以及分类概率损失来计算得出的。CIoU[14]的设计在特定情境下存在局限性,即在设定边框长宽比例时,若使长宽值相等但方向各异,会导致其兼顾的长宽比部分失效(v=0)。此外,CIoU在处理长宽变化时通常呈现负相关特性,即一边增加时另一边减少,这种趋势在边界框预测中易引发目标预测框与真实框的不匹配问题。上述两方面因素共同导致了模型预测框与真实目标框之间的不匹配,进而严重影响了CIoU的收敛性能,最终导致在安全帽佩戴检测任务中检测效果的不理想。引入损失函数Wise-IoU代替原损失函数进行模型优化,由于煤矿井下不同光照和距离的复杂背景和环境,导致检测样本质量出现参差不齐的情况,而Wise-IoU v3版损失函数能平衡这种情况[15]。Wise-IoU作为一种改进型的损失函数,在传统IoU基础上融入了类别权重的概念。其主要目标在于通过精细化的权重调整来缩小不同类别之间的性能差异,从而优化模型对测试结果的预测准确性。具体实现上,Wise-IoU为每个类别分配特定的权重值,并在计算IoU重叠度时应用这些权重,以实现不同类别间重叠区域的差异化评估,最终得到更为精确和可靠的模型性能评价结果。Wise-IoU共有3个版本,即Wise-IoU v1、Wise-IoU v2、Wise-IoU v3。
由于训练数据中不可避免地存在低质量样本,传统的几何度量(如距离、纵横比)在评估这些样本时可能会过度惩罚,进而造成模型的泛化性能下降。一个理想的损失函数设计应旨在当锚框与目标框达到良好匹配时,适度减少对几何度量误差的惩罚力度,避免对训练过程造成不必要的干扰,进一步提升模型的性能,强化了模型的泛化能力。基于这一改进,引入一种基于距离度量的注意力机制,即距离注意力,进而构建了具备双层注意力结构的Wise-IoU v1,以增强模型在煤矿井下复杂场景下的适应性。
RWIoU∈[1,e),RWIoU为记锚框和目标框中心点连接的归一化长度,这将显著放大普通质量锚框的LIoU
LIoU∈[0,1],LIoU为度量目标检测任务中预测框与真实框的重叠程度,这将大幅度减少高质量锚框RWIoU的冗余性,尤其是在锚框与目标框实现高度匹配时,更能有效减轻对中心点距离这一单一维度的过度依赖,进而优化模型的效能与精确度。

LWIoUv1=RWIoULWIoU

RWIoU=exp[$\frac{(x-{x}_{\mathrm{g}\mathrm{t}}{)}^{2}+(y-{y}_{\mathrm{g}\mathrm{t}}{)}^{2}}{({W}_{\mathrm{g}}^{2}+{H}_{\mathrm{g}}^{2}{)}^{\mathrm{*}}}$]
式中:LWIoU为Wise-IoU度量目标检测任务中预测框与真实框的重叠程度;LWIoUv1为 Wise-IoU v1度量目标检测任务中预测框与真实框的重叠程度;WgHg为最小的封闭框的宽度和长度;xyxgtygt分别为最小的封闭框内记锚框和目标框大小;为了防止放大普通质量锚框参数RWIoU对阻碍收敛的影响,将WgHg从计算中分离(上标*表示此操作)。
未引入新的度量指标的原因在于:它能够有效消除了阻碍收敛的因素。
在Wise-IoU v2中,受到Focal Loss设计思路的启发,对Wise-IoU v1进行优化,形成一个独特的单调聚焦系数${L}_{\mathrm{I}\mathrm{o}\mathrm{U}}^{\mathrm{*}}$。这一改进显著降低了简单样本在损失函数中的权重,从而引导模型更加专注于那些难以处理的复杂样本。这一策略显著增强了模型对困难示例的识别能力,进而实现了分类性能的显著提升,LWIoUv2为 Wise-IoU v2度量目标检测任务中预测框与真实框的重叠程度,可表示为
LWIoUv2=${L}_{\mathrm{I}\mathrm{o}\mathrm{U}}^{\mathrm{*}}$LWIoUv1
在模型训练过程中,随着LIoU的递减,${L}_{\mathrm{I}\mathrm{o}\mathrm{U}}^{\mathrm{*}}$也呈现下降趋势,这一现象导致训练后期收敛进程趋于迟缓。鉴于此,引入的均值LIoU作为归一化基准,可表示为
LWIoUv2=${\left(\frac{{L}_{\mathrm{I}\mathrm{o}\mathrm{U}}^{\mathrm{*}}}{{L}_{\mathrm{I}\mathrm{o}\mathrm{U}}}\right)}^{\gamma }$LWIoUv1
式(7)中:γ为调节因子,控制权重随IoU变化的敏感度。
在锚框的评估中,若其离群度偏低,这往往标志着其质量较高。鉴于此,为其设定一个较低的梯度增益权重,旨在确保边界框回归过程能够优先优化这些普通质量锚框的精确性,从而更有效地提升整体性能。而针对那些离群度显著较高的锚框,赋予其较小的梯度增益是一种合理的策略,这有助于防止这些低质量样本在训练过程中产生过大的、潜在的负面影响的梯度。
Wise-IoU v3利用离群度β构造一个非单调聚焦系数r,并将其应用于Wise-IoU v1,可表示为
r=$\frac{\beta }{\delta {\alpha }^{\beta -\delta }}$
式(8)中:αδ为超参数。
β=$\frac{{L}_{\mathrm{I}\mathrm{o}\mathrm{U}}^{\mathrm{*}}}{{\overline{L}}_{\mathrm{I}\mathrm{o}\mathrm{U}}}$∈[0,+∞)

LWIoUv3=rLWIoUv1

式(9)定义了离群度β作为衡量回归框品质的一个指标,通过引入动量为m的滑动平均值${\overline{L}}_{\mathrm{I}\mathrm{o}\mathrm{U}}$,该值被动态地维护与更新,旨在维持整体性能β于较高水平,从而有效缓解训练末期收敛速率减缓的问题。在此框架下,超参数αδ共同作用于决策过程:当某一回归框的离群度β达到预设阈值C时,该框将享有最大的梯度增益优势。值得注意的是,由于滑动平均值${\overline{L}}_{\mathrm{I}\mathrm{o}\mathrm{U}}$的动态特性,回归框品质的评判标准也随之灵活调整,确保了Wise-IoU v3能够实时制定最优的梯度增益分配策略,以适应训练过程中的各种情况。此外,Wise-IoU v3还特别关注于中等质量锚框的优化,进一步提升模型的定位精确度与性能。
使用从某煤矿上收集的图片,有两种图片类型,分别为安全帽和人,共2 000张。使用Labelimg对数据集中的图像进行了标注。随机以8∶2的比例分配为1 600张训练集和400张验证集,实验环境如表1所示。
在模型训练的过程中,输入图片尺寸设置为640×640,batch size设置为8,设定了动量因子为0.9,训练的学习率为0.001。通过这样的设置,进行300轮的模型迭代训练。
为了证实本研究所提出的MELAN、OD-SMP和WIoU的有效性,使用相同的数据集进行训练和验证,分别进行8组消融实验,结果如表2所示。
表2可知,原始网络对安全帽佩戴检测的平均准确率均值为81.4%;在不改变模型原始损失函数的情况下,将Backbone部分加入MELAN模块,检测模型的平均准确率均值提升2.5个百分点,参数量增加1.1%,原因在于:MEMA注意力模块增强有用的特征并抑制无用的特征;将Backbone部分加入OD-SMP模块后平均准确率均值提升1.7个百分点,表明SoftMax池化可以减少MP模块中的最大池化会丢失一部分本就不丰富的目标细节特征,造成对小目标有效特征的提取能力不足;使用Wise-IoU作为回归损失函数后,由于Wise-IoU能平衡检测样本质量出现参差不齐的情况,模型检测性能达到最佳,平均准确率均值提升至94.9%。检测效果如图7所示。
为了全面评估改进模型在检测任务中的性能,选取当前主流的几种模型进行对比分析。在确保实验条件与数据集保持一致的前提下,进行了全面的对比。实验结果如表3所示。结果表明,改进后的模型具有良好的检测效果。改进前后mAP曲线如图8所示。
(1)针对煤矿工人安全帽佩戴检测环境面临的低质影像存在低光照、小目标等痛点,提出一种基于改进YOLOv7的煤矿井下低质影像安全帽检测,提高了煤矿工人安全帽佩戴检测的精度,与YOLOv7相比具有较好的检测效果。本文算法加入MELAN模块,增强特征采样的感受野,提升浅层细节特征的利用率。改进原有MP模块的池化方法,全维动态卷积替代原有卷积层,进一步增强模型对小目标的特征提取能力。使用Wise-IoU损失函数替换原有的损失函数,应对井下恶劣多变的环境导致数据集质量参差不齐的情况。为验证方法的有效性,实验在自制数据集上完成。
(2)本文算法相较于原算法准确率提升13.5%。与SSD、Fast-RCNN、YOLOv5、YOLOv7-A和YOLOv8相比准确率分别提升36.5%、28.4%、19.9%、1.4%和6.1%;在模型参数计算量上,与SSD和Fast-RCNN均大幅度减少;与YOLOv5、YOLOv7和YOLOv8相比较,参数计算量略有增加,但mAP和mAP@0.5效果较好。因此,改进YOLOv7煤矿井下安全帽佩戴检测算法具有良好性能,能满足对煤矿井下安全帽佩戴的检测。
  • 山西省重点研发计划(202102010101010)
参考文献 引证文献
排序方式:
[1]
Li X, Wang S, Liu B, et al. Improved YOLOv4 network using infrared images for personnel detection in coal mines[J]. Journal of Electronic Imaging, 2022, 31(1): 013017.
[2]
徐先峰, 赵万福, 邹浩泉, . 基于MobileNet-SSD的安全帽佩戴检测算法[J]. 计算机工程, 2021, 47(10): 298-305, 313.
Xu Xianfeng, Zhao Wanfu, Zou Haoquan, et al. Safety helmet wearing detection algorithm based on MobileNet-SSD[J]. Computer Engineering, 2021, 47(10): 298-305, 313.
[3]
李明山, 韩清鹏, 张天宇, . 改进SSD的安全帽检测方法[J]. 计算机工程与应用, 2021, 57(8): 192-197.
Li Mingshan, Han Qingpeng, Zhang Tianyu, et al. Improving the safety helmet detection method for SSDs[J]. Computer Engineering and Applications, 2021, 57(8): 192-197.
[4]
李嘉信, 胡杨, 黄协舟, . 面向小目标的多空间层次安全帽检测[J]. 计算机工程与应用, 2024, 60(6): 230-237.
Li Jiaxin, Hu Yang, Huang Xiezhou, et al. Multi spatial hierarchical safety helmet detection for small targets[J]. Computer Engineering and Applications, 2024, 60(6): 230-237.
[5]
丁田, 陈向阳, 周强, . 基于改进YOLOX的安全帽佩戴实时检测[J]. 电子测量技术, 2022, 45(17): 72-78.
Ding Tian, Chen Xiangyang, Zhou Qiang, et al. Real time detection of helmet wearing based on improved YOLOX[J]. Electronic Measurement Technology, 2022, 45(17): 72-78.
[6]
Shao X, Liu S, Li X, et al. FM-YOLOv7: an improved detection method for mine personnel helmet[J]. Journal of Electronic Imaging, 2023, 32(3): 033013.
[7]
高凯, 董立红, 邓凡. 基于递归门控卷积和上下文注意力的煤块检测算法[J]. 矿业研究与开发, 2023, 43(6): 183-190.
Gao Kai, Dong Lihong, Deng Fan. Coal block detection algorithm based on recursive gated convolution and contextual attention[J]. Mining Research and Development, 2023, 43(6): 183-190.
[8]
宋立业, 赵小萱, 崔昊. 基于改进YOLOv7的矿用电铲检测算法[J]. 工矿自动化, 2023, 49(12): 18-24, 32.
Song Liye, Zhao Xiaoxuan, Cui Hao. Mining electric shovel detection algorithm based on improved YOLOv7[J]. Industrial and Mining Automation, 2023, 49(12): 18-24, 32.
[9]
Ouyang D L, He S, Zhang G Z, et al. Efficient multi-scale attention module with cross-spatial learning[C]// IEEE International Conference on Acoustics, Speech and Signal Processing. New York: IEEE, 2023: 1-5.
[10]
顾清华, 周宇静, 王丹, . 基于多尺度特征提取的露天矿区道路坑洼检测模型[J/OL].(2025-04-24)[2024-06-18]. http://kns.cnki.net/kcms/detail/34.1055.td.20240516.1134.006.html. http://kns.cnki.net/kcms/detail/34.1055.td.20240516.1134.006.html
Gu Qinghua, Zhou Yujing, Wang Dan, et al. A detection model for road potholes in open-pit mining areas based on multi-scale feature extraction[J/OL]. (2025-04-24)[2024-06-18]. http://kns.cnki.net/kcms/detail/34.1055.td.20240516.1134.006.html. http://kns.cnki.net/kcms/detail/34.1055.td.20240516.1134.006.html
[11]
Chao L, Aojun Z, Anbang Y. Omni-dimensional dynamic convolution[J]. arXiv Preprint, 2019: arXiv: 2209.07947.
[12]
崔丽群, 曹华维. 改进YOLOv7的航拍图像目标检测[J]. 计算机工程与应用, 2024, 60(20): 189-197.
Cui Liqun, Cao Huawei. Improving YOLOv7 for aerial image object detection[J]. Computer Engineering and Applications, 2024, 60(20): 189-197.
[13]
王俊岭, 王晨晨, 熊玉华. 基于YOLOv5l-Im的排水管道缺陷检测方法及效果分析[J]. 科学技术与工程, 2024, 24(18): 7833-7842.
Wang Junling, Wang Chenchen, Xiong Yuhua. Drainage pipeline defect detection method and effect analysis based on YOLOv5l-Im[J]. Science Technology and Engineering, 2024, 24(18): 7833-7842.
[14]
秦子豪, 雷鸣, 宋文广, . 基于轻量化深度学习模型的安全帽检测方法[J]. 科学技术与工程, 2022, 22(14): 5659-5665.
Qin Zihao, Lei Ming, Song Wenguang, et al. Safety helmet detection method based on lightweight deep learning model[J]. Science Technology and Engineering, 2022, 22(14): 5659-5665.
[15]
桑雨, 李立权, 李铁. 轻量化YOLOv7-tiny的遥感图像小目标检测[J]. 科学技术与工程, 2024, 24(18): 7726-7732.
Sang Yu, Li Liquan, Li Tie. Small target detection in remote sensing images of lightweight YOLOv7-tiny[J]. Science Technology and Engineering, 2024, 24(18): 7726-7732.
2025年第25卷第20期
PDF下载
67
30
引用本文
BibTeX
文章信息
doi: 10.12404/j.issn.1671-1815.2406421
  • 接收时间:2024-08-27
  • 首发时间:2026-05-13
  • 出版时间:2025-07-18
补充材料
相关文章
文章信息
作者
出版历史
  • 收稿日期:2024-08-27
  • 修回日期:2025-04-23
基金
山西省重点研发计划(202102010101010)
作者信息
    1 太原科技大学机械工程学院, 太原 030024
    2 太原科技大学计算机科学与技术学院, 太原 030024

通讯作者:

* 张延军(1982—),男,满族,河北承德人,博士,教授。研究方向:机电液传动及智能控制。E-mail:
参考文献
分享链接
https://castjournals.cast.org.cn/joweb/kxjsygc/CN/10.12404/j.issn.1671-1815.2406421
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