Article(id=1203753463927775504, tenantId=1146029695717560320, journalId=1146123166801305609, issueId=1203753457208504777, articleNumber=null, orderNo=null, doi=10.12404/j.issn.1671-1815.2308931, pmid=null, cstr=null, oa=null, hot=null, price=null, onlineType=0, articleFormat=0, articleType=null, articleTypeStr=research-article, receivedDate=1699891200000, receivedDateStr=2023-11-14, revisedDate=1729440000000, revisedDateStr=2024-10-21, acceptedDate=null, acceptedDateStr=null, onlineDate=1764926790458, onlineDateStr=2025-12-05, pubDate=1737129600000, pubDateStr=2025-01-18, doiRegisterDate=null, doiRegisterDateStr=null, onlineIssueDate=1764926790458, onlineIssueDateStr=2025-12-05, onlineJustAcceptDate=null, onlineJustAcceptDateStr=null, onlineFirstDate=null, onlineFirstDateStr=null, sourceXml=null, magXml=null, createTime=1764926790458, creator=13701087609, updateTime=1764926790458, updator=13701087609, issue=Issue{id=1203753457208504777, tenantId=1146029695717560320, journalId=1146123166801305609, year='2025', volume='25', issue='2', pageStart='439', pageEnd='878', issueExtLink='null', onlineDate='null', pubDate='null', beforeIssueId=null, nextIssueId=null, price=null, status=1, issueComplete=1, articleOrder=1, issueType=-1, specialIssue=null, createTime=1764926788856, creator=13701087609, updateTime=1764928745558, updator=13701087609, preIssue=null, nextIssue=null, ext={EN=IssueExt(id=1203761664261858014, tenantId=1146029695717560320, journalId=1146123166801305609, issueId=1203753457208504777, language=EN, specialIssueTitle=, coverIllustrator=null, specialIssueEditor=, specialIssueAbout=), CN=IssueExt(id=1203761664261858015, tenantId=1146029695717560320, journalId=1146123166801305609, issueId=1203753457208504777, language=CN, specialIssueTitle=, coverIllustrator=null, specialIssueEditor=, specialIssueAbout=)}, issueFiles=null}, startPage=657, endPage=666, ext={EN=ArticleExt(id=1203753464825356635, articleId=1203753463927775504, tenantId=1146029695717560320, journalId=1146123166801305609, language=EN, title=Small Target Ship Remote Sensing Image Detection Based on Improved YOLOv5s, columnId=1156262729162810294, journalTitle=Science Technology and Engineering, columnName=Papers·Automation and Computational Technology, runingTitle=null, highlight=null, articleAbstract=

Ship targets in remote sensing images have multi-scale characteristics, changeable backgrounds, and complex meteorological characteristics, which lead to low accuracy, false detection, and missed detection of small target ships. In response to the above situation, an improved small-target ship detection model based on YOLOv5s was proposed. First, in order to solve the problems of scale changes and background variability in ship detection, the ASFF(adaptive spatial feature fusion) module was introduced. Secondly, in order to reduce the calculation amount and parameter amount of the detection network, the BoTNet attention mechanism was introduced, and then in order to improve the overall network to improve the detection accuracy, the EIoU border loss function was used, and finally the Slim-neck network was introduced to ensure the overall lightweight of the network. Experiments show that on the main data set LEVIR-Ship, compared with the benchmark YOLOv5s, mAP@0.5 increased by 7.1% to 81.3%, the number of parameters is reduced by 0.44 M, the calculation amount is reduced by 0.6GFLOPs, and the weight was reduced by 0.9 M. The proposed method performs better in various key indicators and achieves high-precision small target ship detection in complex environments. Comparative experiments are conducted on the verification data set McShips. The experiments show that the proposed method still performs better, verifying the universal applicability of the proposed method.

, correspAuthors=Xiao-ling XIAO, 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=Zhi-ang LI, Xiao-ling XIAO, Shao-fa ZHOU), CN=ArticleExt(id=1203753467954307782, articleId=1203753463927775504, tenantId=1146029695717560320, journalId=1146123166801305609, language=CN, title=基于改进的YOLOv5s小目标船舶遥感图像检测, 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=

遥感图像中船舶目标具有多尺度特性、背景多变及气象复杂等特点,导致小目标船舶检测存在精度低,出现误检,漏检等情况。针对上述情况,提出了一种基于YOLOv5s的小目标船舶检测改进模型。首先,为解决船舶检测中尺度变化和背景多变问题,引入了适应空间特征融合(adaptive structure feature fusion,ASFF)模块,其次,为减少检测网络的计算量和参数量引入了BoTNet注意力机制,然后为提升网络整体的检测精确度,使用了EIoU边框损失函数,最后为保证网络整体的轻量化引入了Slim-neck颈部网络。实验显示,在主要数据集LEVIR-Ship上,相较于基准YOLOv5s,mAP@0.5提升了7.1%达到了81.3%,参数量降低了0.44 M,计算量降低了0.6GFLOPs,权重降低了0.9 M。本文方法在各项关键指标中表现更为优秀,实现了复杂环境下高精度的小目标船舶检测。在验证数据集McShips上进行对比实验。实验表明,本文方法依然表现更为优秀,验证了所提方法具有普适性。

, correspAuthors=肖小玲, authorNote=null, correspAuthorsNote=
* 肖小玲(1973—),女,汉族,湖南邵阳人,博士,教授。研究方向:智能信息处理和网络安全。E-mail:
, copyrightStatement=null, copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=kJ4c125JKdaclTZeijqPSw==, magXml=f5RrftQO0pYZ4eEM+cSeBA==, pdfUrl=null, pdf=ZITwFWyIDH8GrJvaCPU8bw==, pdfFileSize=11638731, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=kgYIunOVBGbIIz+O65zJbw==, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=DNBkr/m85+/N+dCxwqSLJA==, mapNumber=null, authorCompany=null, fund=null, authors=

李志昂(2000—),男,汉族,湖北武汉人,硕士研究生。研究方向:深度学习。E-mail:

, authorsList=李志昂, 肖小玲, 周绍发)}, authors=[Author(id=1203787145262637734, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, orderNo=0, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=2022710688@yangtzeu.edu.cn, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1203787145375883951, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, authorId=1203787145262637734, language=EN, stringName=Zhi-ang LI, firstName=Zhi-ang, middleName=null, lastName=LI, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=School of Computer Science, Yangtze University, Jingzhou 434023, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1203787145468158645, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, authorId=1203787145262637734, language=CN, stringName=李志昂, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=长江大学计算机科学学院, 荆州 434023, bio={"content":"

李志昂(2000—),男,汉族,湖北武汉人,硕士研究生。研究方向:深度学习。E-mail:

"}, bioImg=null, bioContent=

李志昂(2000—),男,汉族,湖北武汉人,硕士研究生。研究方向:深度学习。E-mail:

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1203787145174557342, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, xref=null, ext=[AuthorCompanyExt(id=1203787145178751647, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, companyId=1203787145174557342, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=School of Computer Science, Yangtze University, Jingzhou 434023, China), AuthorCompanyExt(id=1203787145187140257, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, companyId=1203787145174557342, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=长江大学计算机科学学院, 荆州 434023)])]), Author(id=1203787145543656119, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, orderNo=1, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=xxl@yangtzeu.edu.cn, emailSecond=null, emailThird=null, correspondingAuthor=1, authorType=1, ext={EN=AuthorExt(id=1203787145661096635, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, authorId=1203787145543656119, language=EN, stringName=Xiao-ling XIAO, firstName=Xiao-ling, middleName=null, lastName=XIAO, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=*, address=School of Computer Science, Yangtze University, Jingzhou 434023, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1203787145761759933, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, authorId=1203787145543656119, language=CN, stringName=肖小玲, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=*, address=长江大学计算机科学学院, 荆州 434023, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1203787145174557342, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, xref=null, ext=[AuthorCompanyExt(id=1203787145178751647, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, companyId=1203787145174557342, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=School of Computer Science, Yangtze University, Jingzhou 434023, China), AuthorCompanyExt(id=1203787145187140257, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, companyId=1203787145174557342, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=长江大学计算机科学学院, 荆州 434023)])]), Author(id=1203787145841451715, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, 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=1203787145958892234, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, authorId=1203787145841451715, language=EN, stringName=Shao-fa ZHOU, firstName=Shao-fa, middleName=null, lastName=ZHOU, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=School of Computer Science, Yangtze University, Jingzhou 434023, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1203787146084721363, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, authorId=1203787145841451715, language=CN, stringName=周绍发, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=长江大学计算机科学学院, 荆州 434023, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1203787145174557342, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, xref=null, ext=[AuthorCompanyExt(id=1203787145178751647, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, companyId=1203787145174557342, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=School of Computer Science, Yangtze University, Jingzhou 434023, China), AuthorCompanyExt(id=1203787145187140257, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, companyId=1203787145174557342, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=长江大学计算机科学学院, 荆州 434023)])])], keywords=[Keyword(id=1203787146294436579, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, language=EN, orderNo=1, keyword=ship detection), Keyword(id=1203787146374128358, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, language=EN, orderNo=2, keyword=YOLOv5s), Keyword(id=1203787146453820139, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, language=EN, orderNo=3, keyword=small target detection), Keyword(id=1203787146537706223, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, language=EN, orderNo=4, keyword=BoTNet attention mechanism), Keyword(id=1203787146638369524, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, language=CN, orderNo=1, keyword=船舶检测), Keyword(id=1203787146747421433, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, language=CN, orderNo=2, keyword=YOLOv5s), Keyword(id=1203787146843890433, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, language=CN, orderNo=3, keyword=小目标检测), Keyword(id=1203787146931970821, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, language=CN, orderNo=4, keyword=BoTNet注意力机制)], refs=[Reference(id=1203787154234253572, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, doi=null, pmid=null, pmcid=null, year=2021, volume=46, issue=11, pageStart=1703, pageEnd=1715, url=null, language=null, rfNumber=[1], rfOrder=0, authorNames=Song Z H, Sui H G, Li Y C, journalName=Geomatics and Information Science of Wuhan University, refType=null, unstructuredReference=Song Z H, Sui H G, Li Y C. A survey on ship detection technology in high-resolution optical remote sensing images[J]. Geomatics and Information Science of Wuhan University, 2021, 46(11): 1703-1715., articleTitle=A survey on ship detection technology in high-resolution optical remote sensing images, refAbstract=null), Reference(id=1203787154334916883, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, doi=null, pmid=null, pmcid=null, year=2023, volume=27, issue=1, pageStart=63, pageEnd=84, url=null, language=null, rfNumber=[2], rfOrder=1, authorNames=Muhanmmad Y, Wan J H, Xu M M, journalName=Soft Computing, refType=null, unstructuredReference=Muhanmmad Y, Wan J H, Xu M M, et al. Ship detection based on deep learning using SAR imagery: a systematic literature review[J]. Soft Computing, 2023, 27(1): 63-84., articleTitle=Ship detection based on deep learning using SAR imagery: a systematic literature review, refAbstract=null), Reference(id=1203787154490106153, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, doi=null, pmid=null, pmcid=null, year=2023, volume=61, issue=null, pageStart=1, pageEnd=14, url=null, language=null, rfNumber=[3], rfOrder=2, authorNames=Jiang M, Gu L, Li X, journalName=IEEE Transactions on Geoscience and Remote Sensing, refType=null, unstructuredReference=Jiang M, Gu L, Li X, et al. Ship contourextraction from SAR images based on faster R-CNN and Chan-Vese model[J]. IEEE Transactions on Geoscience and Remote Sensing, 2023, 61: 1-14., articleTitle=Ship contourextraction from SAR images based on faster R-CNN and Chan-Vese model, refAbstract=null), Reference(id=1203787154641101112, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, doi=null, pmid=null, pmcid=null, year=2021, volume=null, issue=11, pageStart=59, pageEnd=64, url=null, language=null, rfNumber=[4], rfOrder=3, authorNames=赵振强, 何水原, 梁永志, journalName=测绘通报, refType=null, unstructuredReference=赵振强, 何水原, 梁永志. 基于Faster R-CNN的遥感影像船舶检测识别[J]. 测绘通报, 2021(11): 59-64., articleTitle=基于Faster R-CNN的遥感影像船舶检测识别, refAbstract=null), Reference(id=1203787154758541630, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, doi=null, pmid=null, pmcid=null, year=2021, volume=null, issue=11, pageStart=59, pageEnd=64, url=null, language=null, rfNumber=[4], rfOrder=4, authorNames=Zhao Zhenqiang, He Shuiyuan, Liang Yongzhi, journalName=Bulletin of Surveying and Mapping, refType=null, unstructuredReference=Zhao Zhenqiang, He Shuiyuan, Liang Yongzhi, et al. Research on remote sensing image ship detection and identification based on Faster R-CNN[J] Bulletin of Surveying and Mapping, 2021(11): 59-64., articleTitle=Research on remote sensing image ship detection and identification based on Faster R-CNN, refAbstract=null), Reference(id=1203787154888565068, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, doi=null, pmid=null, pmcid=null, year=2018, volume=null, issue=null, pageStart=336, pageEnd=344, url=null, language=null, rfNumber=[5], rfOrder=5, authorNames=Yuan Z, Geng J, Dai T, journalName=Geo-Spatial Knowledge and Intelligence:5th International Conference, refType=null, unstructuredReference=Yuan Z, Geng J, Dai T. Ship detection from remote sensing images based on deep learning[C]// Geo-Spatial Knowledge and Intelligence:5th International Conference. Chiang Mai: Springer, 2018: 336-344., articleTitle=Ship detection from remote sensing images based on deep learning, refAbstract=null), Reference(id=1203787155014394206, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, doi=null, pmid=null, pmcid=null, year=2021, volume=21, issue=8, pageStart=3192, pageEnd=3198, url=null, language=null, rfNumber=[6], rfOrder=6, authorNames=袁小平, 马绪起, 刘赛, journalName=科学技术与工程, refType=null, unstructuredReference=袁小平, 马绪起, 刘赛. 改进YOLOv3的行人车辆目标检测算法[J]. 科学技术与工程, 2021, 21(8): 3192-3198., articleTitle=改进YOLOv3的行人车辆目标检测算法, refAbstract=null), Reference(id=1203787155144417650, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, doi=null, pmid=null, pmcid=null, year=2021, volume=21, issue=8, pageStart=3192, pageEnd=3198, url=null, language=null, rfNumber=[6], rfOrder=7, authorNames=Yuan Xiaoping, Ma Xuqi, Liu Sai, journalName=Science Technology and Engineering, refType=null, unstructuredReference=Yuan Xiaoping, Ma Xuqi, Liu Sai, et al. An improved algorithm of pedestrian and vehicle detection based on YOLOv3[J]. Science Technology and Engineering, 2021, 21(8): 3192-3198., articleTitle=An improved algorithm of pedestrian and vehicle detection based on YOLOv3, refAbstract=null), Reference(id=1203787155236692351, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, doi=null, pmid=null, pmcid=null, year=2023, volume=54, issue=9, pageStart=264, pageEnd=271, url=null, language=null, rfNumber=[7], rfOrder=8, authorNames=许小华, 陈豹, 王海菁, journalName=人民长江, refType=null, unstructuredReference=许小华, 陈豹, 王海菁, 等. 基于改进YOLOv4-Tiny的河湖船舶目标检测算法[J]. 人民长江, 2023, 54(9): 264-271., articleTitle=基于改进YOLOv4-Tiny的河湖船舶目标检测算法, refAbstract=null), Reference(id=1203787155358327190, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, doi=null, pmid=null, pmcid=null, year=2023, volume=54, issue=9, pageStart=264, pageEnd=271, url=null, language=null, rfNumber=[7], rfOrder=9, authorNames=Xu Xiaohua, Chen Bao, Wang Haijing, journalName=Yangtze River, refType=null, unstructuredReference=Xu Xiaohua, Chen Bao, Wang Haijing, et al. Object detection algorithm for ships on rivers and lakes based on improved YOLOv4-Tiny[J]. Yangtze River, 2023, 54(9): 264-271., articleTitle=Object detection algorithm for ships on rivers and lakes based on improved YOLOv4-Tiny, refAbstract=null), Reference(id=1203787155454796195, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, doi=null, pmid=null, pmcid=null, year=2022, volume=14, issue=4, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=[8], rfOrder=10, authorNames=Xu X, Zhang X, Zhang T, journalName=Remote Sensing, refType=null, unstructuredReference=Xu X, Zhang X, Zhang T. Lite-yolov5: a lightweight deep learning detector for on-board ship detection in large-scene sentinel-1 sar images[J]. Remote Sensing, 2022, 14(4). DOI:10.3390/rs14041018., articleTitle=Lite-yolov5: a lightweight deep learning detector for on-board ship detection in large-scene sentinel-1 sar images, refAbstract=null), Reference(id=1203787155572236720, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, doi=null, pmid=null, pmcid=null, year=2022, volume=42, issue=7, pageStart=2248, pageEnd=2255, url=null, language=null, rfNumber=[9], rfOrder=11, authorNames=凡文俊, 赵曙光, 郭力争, journalName=计算机应用, refType=null, unstructuredReference=凡文俊, 赵曙光, 郭力争. 基于改进RetinaNet的船舶检测算法[J]. 计算机应用, 2022, 42(7): 2248-2255., articleTitle=基于改进RetinaNet的船舶检测算法, refAbstract=null), Reference(id=1203787155719037373, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, doi=null, pmid=null, pmcid=null, year=2022, volume=42, issue=7, pageStart=2248, pageEnd=2255, url=null, language=null, rfNumber=[9], rfOrder=12, authorNames=Fan Wenjun, Zhao Shuguang, Guo Lizheng, journalName=Journal of Computer Applications, refType=null, unstructuredReference=Fan Wenjun, Zhao Shuguang, Guo Lizheng. Ship detection algorithm based on improved RetinaNet[J]. Journal of Computer Applications, 2022, 42(7): 2248-2255., articleTitle=Ship detection algorithm based on improved RetinaNet, refAbstract=null), Reference(id=1203787155891003860, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, doi=null, pmid=null, pmcid=null, year=2023, volume=44, issue=2, pageStart=18, pageEnd=24, url=null, language=null, rfNumber=[10], rfOrder=13, authorNames=叶乐, 李朝锋, journalName=上海海事大学学报, refType=null, unstructuredReference=叶乐, 李朝锋. 基于频域通道注意力的YOLOv3网络的雾天海洋图像船舶检测[J]. 上海海事大学学报, 2023, 44(2): 18-24., articleTitle=基于频域通道注意力的YOLOv3网络的雾天海洋图像船舶检测, refAbstract=null), Reference(id=1203787156000055778, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, doi=null, pmid=null, pmcid=null, year=2023, volume=44, issue=2, pageStart=18, pageEnd=24, url=null, language=null, rfNumber=[10], rfOrder=14, authorNames=Ye Le, journalName=Journal of Shanghai Maritime University, refType=null, unstructuredReference=Ye Le, Li Chaofeng. Frequency channel attention-based YOLOv3 network for ship detection from foggy marine images[J]. Journal of Shanghai Maritime University, 2023, 44(2): 18-24., articleTitle=Frequency channel attention-based YOLOv3 network for ship detection from foggy marine images, refAbstract=null), Reference(id=1203787156155245039, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, doi=null, pmid=null, pmcid=null, year=2021, volume=14, issue=null, pageStart=6694, pageEnd=6706, url=null, language=null, rfNumber=[11], rfOrder=15, authorNames=Zhu M, Hu G, Zhou H, journalName=IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, refType=null, unstructuredReference=Zhu M, Hu G, Zhou H, et al. Arbitrary-oriented ship detection based on RetinaNet for remote sensing images[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2021, 14: 6694-6706., articleTitle=Arbitrary-oriented ship detection based on RetinaNet for remote sensing images, refAbstract=null), Reference(id=1203787156289462784, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, doi=null, pmid=null, pmcid=null, year=2021, volume=null, issue=null, pageStart=483, pageEnd=487, url=null, language=null, rfNumber=[12], rfOrder=16, authorNames=Liu T, Zhou B J, Zhao Y S, journalName=6th International Conference on Automation, Control and Robotics Engineering (CACRE), refType=null, unstructuredReference=Liu T, Zhou B J, Zhao Y S, et al. Ship detection algorithm based on improved YOLO V5[C]// 6th International Conference on Automation, Control and Robotics Engineering (CACRE). Dalian: IEEE, 2021: 483-487., articleTitle=Ship detection algorithm based on improved YOLO V5, refAbstract=null), Reference(id=1203787156461429266, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, doi=null, pmid=null, pmcid=null, year=2020, volume=43, issue=24, pageStart=87, pageEnd=92, url=null, language=null, rfNumber=[13], rfOrder=17, authorNames=化嫣然, 张卓, 龙赛, journalName=电子测量技术, refType=null, unstructuredReference=化嫣然, 张卓, 龙赛, 等. 基于改进YOLO算法的遥感图像目标检测[J]. 电子测量技术, 2020, 43(24): 87-92., articleTitle=基于改进YOLO算法的遥感图像目标检测, refAbstract=null), Reference(id=1203787156604035616, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, doi=null, pmid=null, pmcid=null, year=2020, volume=43, issue=24, pageStart=87, pageEnd=92, url=null, language=null, rfNumber=[13], rfOrder=18, authorNames=Hua Yanran, Zhang Zhuo, Long Sai, journalName=Electro-nic Measurement Technology, refType=null, unstructuredReference=Hua Yanran, Zhang Zhuo, Long Sai, et al. Remote sensing image target detection based on improved YOLO algorithm[J]. Electro-nic Measurement Technology, 2020, 43(24): 87-92., articleTitle=Remote sensing image target detection based on improved YOLO algorithm, refAbstract=null), Reference(id=1203787156759224876, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, doi=null, pmid=null, pmcid=null, year=2022, volume=8, issue=3, pageStart=331, pageEnd=368, url=null, language=null, rfNumber=[14], rfOrder=19, authorNames=Guo M H, Xu T X, Liu J J, journalName=Computational Visual Media, refType=null, unstructuredReference=Guo M H, Xu T X, Liu J J, et al. Attention mechanisms in computer vision: a survey[J]. Computational Visual Media, 2022, 8(3): 331-368., articleTitle=Attention mechanisms in computer vision: a survey, refAbstract=null), Reference(id=1203787156893442612, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, doi=null, pmid=null, pmcid=null, year=2021, volume=null, issue=null, pageStart=16519, pageEnd=16529, url=null, language=null, rfNumber=[15], rfOrder=20, authorNames=Srinivas A, Lin T Y, Parmar N, journalName=Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, refType=null, unstructuredReference=Srinivas A, Lin T Y, Parmar N, et al. Bottleneck transformers for visual recognition[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2021: 16519-16529., articleTitle=Bottleneck transformers for visual recognition, refAbstract=null), Reference(id=1203787157027660353, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=[16], rfOrder=21, authorNames=Li H, Li J, Wei H, journalName=arXiv preprint arXiv: 2206, refType=null, unstructuredReference=Li H, Li J, Wei H, et al. Slim-neck by GSConv: a better design paradigm of detector architectures for autonomous vehicles[J]. arXiv preprint arXiv: 2206.02424, 2022., articleTitle=Slim-neck by GSConv: a better design paradigm of detector architectures for autonomous vehicles, refAbstract=null), Reference(id=1203787157786829387, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, doi=null, pmid=null, pmcid=null, year=2022, volume=506, issue=null, pageStart=146, pageEnd=157, url=null, language=null, rfNumber=[17], rfOrder=22, authorNames=Zhang Y F, Ren W, Zhang Z, journalName=Neurocomputing, refType=null, unstructuredReference=Zhang Y F, Ren W, Zhang Z, et al. Focal and efficient IOU loss for accurate bounding box regression[J]. Neurocomputing, 2022, 506: 146-157., articleTitle=Focal and efficient IOU loss for accurate bounding box regression, refAbstract=null), Reference(id=1203787157904269914, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, doi=null, pmid=null, pmcid=null, year=2021, volume=52, issue=8, pageStart=8574, pageEnd=8586, url=null, language=null, rfNumber=[18], rfOrder=23, authorNames=Zheng Z, Wang P, Ren D, journalName=IEEE Transactions on Cybernetics, refType=null, unstructuredReference=Zheng Z, Wang P, Ren D, et al. Enhancing geometric factors in model learning and inference for object detection and instance segmentation[J]. IEEE Transactions on Cybernetics, 2021, 52(8): 8574-8586., articleTitle=Enhancing geometric factors in model learning and inference for object detection and instance segmentation, refAbstract=null), Reference(id=1203787158009127525, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=[19], rfOrder=24, authorNames=Liu S, Huang D, Wang Y, journalName=arXiv preprint arXiv: 1911.09516, 2019, refType=null, unstructuredReference=Liu S, Huang D, Wang Y. Learning spatial fusion for single-shot object detection[J]. arXiv preprint arXiv: 1911.09516, 2019., articleTitle=Learning spatial fusion for single-shot object detection, refAbstract=null), Reference(id=1203787158130762355, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, doi=null, pmid=null, pmcid=null, year=2022, volume=60, issue=null, pageStart=1, pageEnd=14, url=null, language=null, rfNumber=[20], rfOrder=25, authorNames=Chen J, Chen K, Chen H, journalName=IEEE Transactions on Geoscience and Remote Sensing, refType=null, unstructuredReference=Chen J, Chen K, Chen H, et al. A degraded reconstruction enhancement-based method for tiny ship detection in remote sensing images with a new large-scale dataset[J]. IEEE Transactions on Geoscience and Remote Sensing, 2022, 60: 1-14., articleTitle=A degraded reconstruction enhancement-based method for tiny ship detection in remote sensing images with a new large-scale dataset, refAbstract=null), Reference(id=1203787158256591486, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, doi=null, pmid=null, pmcid=null, year=2020, volume=null, issue=null, pageStart=1, pageEnd=6, url=null, language=null, rfNumber=[21], rfOrder=26, authorNames=Zheng Y, Zhang S, journalName=IEEE International Conference on Multimedia and Expo (ICME), refType=null, unstructuredReference=Zheng Y, Zhang S. Mcships: a large-scale ship dataset for detection and fine-grained categorization in the wild[C]// IEEE International Conference on Multimedia and Expo (ICME). New York: IEEE, 2020: 1-6., articleTitle=Mcships: a large-scale ship dataset for detection and fine-grained categorization in the wild, refAbstract=null), Reference(id=1203787158390809223, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, doi=null, pmid=null, pmcid=null, year=2018, volume=null, issue=null, pageStart=7132, pageEnd=7141, url=null, language=null, rfNumber=[22], rfOrder=27, authorNames=Hu J, Shen L, Sun G, journalName=Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, refType=null, unstructuredReference=Hu J, Shen L, Sun G. Squeeze-and-excitation networks[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. London: IEEE, 2018: 7132-7141., articleTitle=Squeeze-and-excitation networks, refAbstract=null), Reference(id=1203787158504055442, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, doi=null, pmid=null, pmcid=null, year=2018, volume=null, issue=null, pageStart=3, pageEnd=19, url=null, language=null, rfNumber=[23], rfOrder=28, authorNames=Woo S, Park J, Lee J Y, journalName=Proceedings of the European Conference on Computer Vision (ECCV), refType=null, unstructuredReference=Woo S, Park J, Lee J Y, et al. Cbam: convolutional block attention module[C]// Proceedings of the European Conference on Computer Vision (ECCV). Munich: Springer, 2018: 3-19., articleTitle=Cbam: convolutional block attention module, refAbstract=null), Reference(id=1203787158634078879, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, doi=null, pmid=null, pmcid=null, year=2021, volume=null, issue=null, pageStart=11863, pageEnd=11874, url=null, language=null, rfNumber=[24], rfOrder=29, authorNames=Yang L, Zhang R Y, Li L, journalName=International Conference on Machine Learning, refType=null, unstructuredReference=Yang L, Zhang R Y, Li L, et al. Simam: a simple, parameter-free attention module for convolutional neural networks[C]// International Conference on Machine Learning. New York: PMLR, 2021: 11863-11874., articleTitle=Simam: a simple, parameter-free attention module for convolutional neural networks, refAbstract=null), Reference(id=1203787158730547876, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, doi=null, pmid=null, pmcid=null, year=2021, volume=null, issue=null, pageStart=13713, pageEnd=13722, url=null, language=null, rfNumber=[25], rfOrder=30, authorNames=Hou Q, Zhou D, Feng J, journalName=Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, refType=null, unstructuredReference=Hou Q, Zhou D, Feng J. Coordinate attention for efficient mobile network design[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2021: 13713-13722., articleTitle=Coordinate attention for efficient mobile network design, refAbstract=null)], funds=[Fund(id=1203787153978400997, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, awardId=61771354, language=CN, fundingSource=国家自然科学基金(61771354), fundOrder=null, country=null)], companyList=[AuthorCompany(id=1203787145174557342, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, xref=null, ext=[AuthorCompanyExt(id=1203787145178751647, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, companyId=1203787145174557342, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=School of Computer Science, Yangtze University, Jingzhou 434023, China), AuthorCompanyExt(id=1203787145187140257, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, companyId=1203787145174557342, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=长江大学计算机科学学院, 荆州 434023)])], figs=[ArticleFig(id=1203787147187823384, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, language=EN, label=Fig.1, caption=YOLOv5s network main structure diagram, figureFileSmall=8r0uMuMeT8rLs7N41Na8gQ==, figureFileBig=LYMtnFWGZTuhgL7/AH3DEw==, tableContent=null), ArticleFig(id=1203787147313652514, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, language=CN, label=图1, caption=YOLOv5s的网络主体结构图

Input 为输入图像;Backbone为网络主干;Neck为颈部网络;Head为检测头;CBS为由卷积层+批量归一化+SiLU激活函数组成的模块;CSP1_x为Cross Stage Partial(CSP)连接的第1个子模块中的第x个部分;SPPF为一种空间金字塔池化结构;Upsample为上采样;CSP2为cross stage partial(CSP)连接的第2个子模块;Concat为拼接模块;Conv为卷积层;Detect为不同尺度的检测头

, figureFileSmall=8r0uMuMeT8rLs7N41Na8gQ==, figureFileBig=LYMtnFWGZTuhgL7/AH3DEw==, tableContent=null), ArticleFig(id=1203787147431093039, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, language=EN, label=Fig.2, caption=Structure diagram of YOLOv5s submodule, figureFileSmall=Lss8yNVTWe5JmQrsOjIRLA==, figureFileBig=9ecpJ/I10K3Xu5isTnVdog==, tableContent=null), ArticleFig(id=1203787147590476602, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, language=CN, label=图2, caption=YOLOv5s子模块的结构图

MaxPool为最大池化层;Cancat为相加模块;Resunit为残差单元;Conv为卷积层;BN为批量归一化;SiLU为一种激活函数

, figureFileSmall=Lss8yNVTWe5JmQrsOjIRLA==, figureFileBig=9ecpJ/I10K3Xu5isTnVdog==, tableContent=null), ArticleFig(id=1203787147703722817, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, language=EN, label=Fig.3, caption=Diagram of the structural differences between BoTNet and ResNet, figureFileSmall=Qd7WLH+ehe7aqG3Y5I8CtA==, figureFileBig=eVO2b4J1jufX8/Igomm1dg==, tableContent=null), ArticleFig(id=1203787147808580428, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, language=CN, label=图3, caption=BoTNet与ResNet结构差异图, figureFileSmall=Qd7WLH+ehe7aqG3Y5I8CtA==, figureFileBig=eVO2b4J1jufX8/Igomm1dg==, tableContent=null), ArticleFig(id=1203787147917632340, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, language=EN, label=Fig.4, caption=Standard convolution and depthwise separable convolution, figureFileSmall=8Xt7Z+yekvEPE4/8dndztA==, figureFileBig=xHR98oNkk2j9f7HWXv0akA==, tableContent=null), ArticleFig(id=1203787148018295644, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, language=CN, label=图4, caption=标准卷积和深度可分离卷积

3 channel Input为彩色图像的3通道输入;Filters为滤波器;Maps为输出的特征图

, figureFileSmall=8Xt7Z+yekvEPE4/8dndztA==, figureFileBig=xHR98oNkk2j9f7HWXv0akA==, tableContent=null), ArticleFig(id=1203787148114764646, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, language=EN, label=Fig.5, caption=Slim-neck composition structure diagram, figureFileSmall=90xGH8/37hZmIiTgiTZrGg==, figureFileBig=cZZ+f+VFGg1d4I6gNQQMzg==, tableContent=null), ArticleFig(id=1203787149310141300, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, language=CN, label=图5, caption=Slim-neck的组成结构图

Conv为标准卷积; input为输入;c1 channels为有c1个通道;c2 channels为有c2个通道;output为输出

, figureFileSmall=90xGH8/37hZmIiTgiTZrGg==, figureFileBig=cZZ+f+VFGg1d4I6gNQQMzg==, tableContent=null), ArticleFig(id=1203787149435970429, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, language=EN, label=Fig.6, caption=ASFF structure diagram, figureFileSmall=HVekEXjN2rUvUIFMtaVi3w==, figureFileBig=lzq3jf8SnOfCI7T/YbnWjA==, tableContent=null), ArticleFig(id=1203787149561799562, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, language=CN, label=图6, caption=ASFF结构图

Origin img为原始图片;Level为特征图的级别;stride为步长;Predict为预测结果

, figureFileSmall=HVekEXjN2rUvUIFMtaVi3w==, figureFileBig=lzq3jf8SnOfCI7T/YbnWjA==, tableContent=null), ArticleFig(id=1203787149679240088, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, language=EN, label=Fig.7, caption=Improved YOLOv5s structure diagram, figureFileSmall=Zrjt+gRYEueizO23ZZIjmA==, figureFileBig=NofZfmQc8Yo2vX8MVF2qfQ==, tableContent=null), ArticleFig(id=1203787149826040741, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, language=CN, label=图7, caption=改进后的YOLOv5s结构图, figureFileSmall=Zrjt+gRYEueizO23ZZIjmA==, figureFileBig=NofZfmQc8Yo2vX8MVF2qfQ==, tableContent=null), ArticleFig(id=1203787149956064177, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, language=EN, label=Fig.8, caption=Attention mechanism average accuracy mean comparison chart, figureFileSmall=bFomL9gQlZiZD+KlAbeA9g==, figureFileBig=JN7MILWSBij6tpQVQ3gL7A==, tableContent=null), ArticleFig(id=1203787150060921787, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, language=CN, label=图8, caption=注意力机制平均精度均值对比图, figureFileSmall=bFomL9gQlZiZD+KlAbeA9g==, figureFileBig=JN7MILWSBij6tpQVQ3gL7A==, tableContent=null), ArticleFig(id=1203787150224499662, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, language=EN, label=Fig.9, caption=Average accuracy mean comparison chart, figureFileSmall=kgSVSV2/5fwA6gfjvjjIzg==, figureFileBig=kYj/sCoy4PftJhXD1Abtjg==, tableContent=null), ArticleFig(id=1203787150346134491, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, language=CN, label=图9, caption=平均精度均值对比图, figureFileSmall=kgSVSV2/5fwA6gfjvjjIzg==, figureFileBig=kYj/sCoy4PftJhXD1Abtjg==, tableContent=null), ArticleFig(id=1203787150442603495, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, language=EN, label=Fig.10, caption=Marine environment example test comparison chart at night, figureFileSmall=q/cx20IbY8fPsuO4NWZGiA==, figureFileBig=rUiQTVnyFRy66a5fCwQAOQ==, tableContent=null), ArticleFig(id=1203787150601987057, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, language=CN, label=图10, caption=黑夜下的海洋环境实例测试对比图, figureFileSmall=q/cx20IbY8fPsuO4NWZGiA==, figureFileBig=rUiQTVnyFRy66a5fCwQAOQ==, tableContent=null), ArticleFig(id=1203787150706844665, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, language=EN, label=Fig.11, caption=Comparison chart of test examples of marine environment under cloud cover, figureFileSmall=04ndJvyI5Q+tvFD9R+PYbA==, figureFileBig=Jcf2zn6kgITssH9qzwM5NQ==, tableContent=null), ArticleFig(id=1203787150803312643, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, language=CN, label=图11, caption=云层遮挡下的海洋环境实例测试对比图, figureFileSmall=04ndJvyI5Q+tvFD9R+PYbA==, figureFileBig=Jcf2zn6kgITssH9qzwM5NQ==, tableContent=null), ArticleFig(id=1203787150966890515, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, language=EN, label=Table 1, caption=

Key parameter table

, figureFileSmall=null, figureFileBig=null, tableContent=
参数名称 初始值 参数名称 初始值
图像大小 640×640 IoU阈值 0.2
初始学习率 0.01 色调 0.015
优化函数 Adam(1×10-2) 饱和度 0.7
学习率动量 0.937 亮度 0.4
权重衰减系数 0.000 5 Mosaic概率 1.0
), ArticleFig(id=1203787151138856988, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, language=CN, label=表1, caption=

关键参数表

, figureFileSmall=null, figureFileBig=null, tableContent=
参数名称 初始值 参数名称 初始值
图像大小 640×640 IoU阈值 0.2
初始学习率 0.01 色调 0.015
优化函数 Adam(1×10-2) 饱和度 0.7
学习率动量 0.937 亮度 0.4
权重衰减系数 0.000 5 Mosaic概率 1.0
), ArticleFig(id=1203787151323406385, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, language=EN, label=Table 2, caption=

Attention mechanism comparison table

, figureFileSmall=null, figureFileBig=null, tableContent=
名称 mAP/% 参数量/M 计算量/G 权重/M
SE 73.9 7.05 15.8 14.5
CBAM 77.5 7.01 15.8 14.5
SIMAM 76.0 7.01 15.8 14.5
CA 76.1 7.04 15.8 14.5
BoTNet 77.2 6.69 15.4 13.8
), ArticleFig(id=1203787151428264000, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, language=CN, label=表2, caption=

注意力机制对比表

, figureFileSmall=null, figureFileBig=null, tableContent=
名称 mAP/% 参数量/M 计算量/G 权重/M
SE 73.9 7.05 15.8 14.5
CBAM 77.5 7.01 15.8 14.5
SIMAM 76.0 7.01 15.8 14.5
CA 76.1 7.04 15.8 14.5
BoTNet 77.2 6.69 15.4 13.8
), ArticleFig(id=1203787151784779863, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, language=EN, label=Table 3, caption=

Comparison of evaluation indicators of different networks

, figureFileSmall=null, figureFileBig=null, tableContent=
网络 图片大小 mAP/
%
精准
率/%
召回
率/%
参数
量/M
计算
量/G
权重/
M
CenterNet 512×512 69.7 100 10.0 32.67 70.2 124
YOLOv3 416×416 75.9 81.4 70.9 61.50 154.5 123.4
YOLOv4 416×416 68.4 75.9 65.0 9.12 20.8 35.4
YOLOv5s 640×640 74.2 81.2 71.9 7.01 15.8 14.5
YOLOv8s 640×640 79.3 84.1 73.4 11.1 28.4 22.5
本文方法 640×640 81.3 81.2 76.8 6.57 15.2 13.6
), ArticleFig(id=1203787152011272305, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, language=CN, label=表3, caption=

不同网络的评价指标对比

, figureFileSmall=null, figureFileBig=null, tableContent=
网络 图片大小 mAP/
%
精准
率/%
召回
率/%
参数
量/M
计算
量/G
权重/
M
CenterNet 512×512 69.7 100 10.0 32.67 70.2 124
YOLOv3 416×416 75.9 81.4 70.9 61.50 154.5 123.4
YOLOv4 416×416 68.4 75.9 65.0 9.12 20.8 35.4
YOLOv5s 640×640 74.2 81.2 71.9 7.01 15.8 14.5
YOLOv8s 640×640 79.3 84.1 73.4 11.1 28.4 22.5
本文方法 640×640 81.3 81.2 76.8 6.57 15.2 13.6
), ArticleFig(id=1203787152137101439, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, language=EN, label=Table 4, caption=

Ablation experiment results table

, figureFileSmall=null, figureFileBig=null, tableContent=
改进名称 ASFF BotNet EIoU Slim-neck P/% R/% mAP@0.5/% GFLOPs 权重/M
YOLOv5s × × × × 81.2 71.9 74.2 15.8 14.5
改进1 × × × 78.2 74.8 76.5 24.2 25.4
改进2 × × 85.1 69.6 77.2 15.4 13.8
改进3 × 83.6 73.4 80.8 15.8 14.5
改进4 81.2 76.8 81.3 15.2 13.6
), ArticleFig(id=1203787152363593884, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, language=CN, label=表4, caption=

消融实验结果表

, figureFileSmall=null, figureFileBig=null, tableContent=
改进名称 ASFF BotNet EIoU Slim-neck P/% R/% mAP@0.5/% GFLOPs 权重/M
YOLOv5s × × × × 81.2 71.9 74.2 15.8 14.5
改进1 × × × 78.2 74.8 76.5 24.2 25.4
改进2 × × 85.1 69.6 77.2 15.4 13.8
改进3 × 83.6 73.4 80.8 15.8 14.5
改进4 81.2 76.8 81.3 15.2 13.6
), ArticleFig(id=1203787152510394541, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, language=EN, label=Table 5, caption=

Compare experimental result tables on McShips

, figureFileSmall=null, figureFileBig=null, tableContent=
网络 mAP/% 精准率/% 召回率/%
YOLOv5s 88.4 88.6 83.3
YOLOv7-tiny 76.0 81.3 70.6
YOLOv8s 91.8 92.4 84.7
本文方法 92.3 92.4 87.4
), ArticleFig(id=1203787153701576897, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1203753463927775504, language=CN, label=表5, caption=

在McShips上对比实验结果表

, figureFileSmall=null, figureFileBig=null, tableContent=
网络 mAP/% 精准率/% 召回率/%
YOLOv5s 88.4 88.6 83.3
YOLOv7-tiny 76.0 81.3 70.6
YOLOv8s 91.8 92.4 84.7
本文方法 92.3 92.4 87.4
)], 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.1671-1815.2308931, detailUrlEn=https://castjournals.cast.org.cn/joweb/kxjsygc/EN/10.12404/j.issn.1671-1815.2308931, pdfUrlCn=https://castjournals.cast.org.cn/joweb/kxjsygc/CN/PDF/10.12404/j.issn.1671-1815.2308931, pdfUrlEn=https://castjournals.cast.org.cn/joweb/kxjsygc/EN/PDF/10.12404/j.issn.1671-1815.2308931, aliStartDate=null, aliEndDate=null, collectionFlag=false, citedCount=null, citedUrl=null, reference=null)
收藏切换
基于改进的YOLOv5s小目标船舶遥感图像检测
收藏切换
PDF下载
李志昂 , 肖小玲 * , 周绍发
科学技术与工程 | 论文·自动化技术、计算机技术 2025,25(2): 657-666
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(2): 657-666
基于改进的YOLOv5s小目标船舶遥感图像检测
全屏
李志昂 , 肖小玲* , 周绍发
作者信息
  • 长江大学计算机科学学院, 荆州 434023
  • 李志昂(2000—),男,汉族,湖北武汉人,硕士研究生。研究方向:深度学习。E-mail:

通讯作者:

* 肖小玲(1973—),女,汉族,湖南邵阳人,博士,教授。研究方向:智能信息处理和网络安全。E-mail:
Small Target Ship Remote Sensing Image Detection Based on Improved YOLOv5s
Zhi-ang LI , Xiao-ling XIAO* , Shao-fa ZHOU
Affiliations
  • School of Computer Science, Yangtze University, Jingzhou 434023, China
出版时间: 2025-01-18 doi: 10.12404/j.issn.1671-1815.2308931
文章导航
收藏切换

遥感图像中船舶目标具有多尺度特性、背景多变及气象复杂等特点,导致小目标船舶检测存在精度低,出现误检,漏检等情况。针对上述情况,提出了一种基于YOLOv5s的小目标船舶检测改进模型。首先,为解决船舶检测中尺度变化和背景多变问题,引入了适应空间特征融合(adaptive structure feature fusion,ASFF)模块,其次,为减少检测网络的计算量和参数量引入了BoTNet注意力机制,然后为提升网络整体的检测精确度,使用了EIoU边框损失函数,最后为保证网络整体的轻量化引入了Slim-neck颈部网络。实验显示,在主要数据集LEVIR-Ship上,相较于基准YOLOv5s,mAP@0.5提升了7.1%达到了81.3%,参数量降低了0.44 M,计算量降低了0.6GFLOPs,权重降低了0.9 M。本文方法在各项关键指标中表现更为优秀,实现了复杂环境下高精度的小目标船舶检测。在验证数据集McShips上进行对比实验。实验表明,本文方法依然表现更为优秀,验证了所提方法具有普适性。

船舶检测  /  YOLOv5s  /  小目标检测  /  BoTNet注意力机制

Ship targets in remote sensing images have multi-scale characteristics, changeable backgrounds, and complex meteorological characteristics, which lead to low accuracy, false detection, and missed detection of small target ships. In response to the above situation, an improved small-target ship detection model based on YOLOv5s was proposed. First, in order to solve the problems of scale changes and background variability in ship detection, the ASFF(adaptive spatial feature fusion) module was introduced. Secondly, in order to reduce the calculation amount and parameter amount of the detection network, the BoTNet attention mechanism was introduced, and then in order to improve the overall network to improve the detection accuracy, the EIoU border loss function was used, and finally the Slim-neck network was introduced to ensure the overall lightweight of the network. Experiments show that on the main data set LEVIR-Ship, compared with the benchmark YOLOv5s, mAP@0.5 increased by 7.1% to 81.3%, the number of parameters is reduced by 0.44 M, the calculation amount is reduced by 0.6GFLOPs, and the weight was reduced by 0.9 M. The proposed method performs better in various key indicators and achieves high-precision small target ship detection in complex environments. Comparative experiments are conducted on the verification data set McShips. The experiments show that the proposed method still performs better, verifying the universal applicability of the proposed method.

ship detection  /  YOLOv5s  /  small target detection  /  BoTNet attention mechanism
李志昂, 肖小玲, 周绍发. 基于改进的YOLOv5s小目标船舶遥感图像检测. 科学技术与工程, 2025 , 25 (2) : 657 -666 . DOI: 10.12404/j.issn.1671-1815.2308931
Zhi-ang LI, Xiao-ling XIAO, Shao-fa ZHOU. Small Target Ship Remote Sensing Image Detection Based on Improved YOLOv5s[J]. Science Technology and Engineering, 2025 , 25 (2) : 657 -666 . DOI: 10.12404/j.issn.1671-1815.2308931
中国是一个拥有广阔海域边境线的海洋大国,在海洋主权安全、海洋资源探索、海洋环境保护以及海域内情报探测等领域中,快速且准确地检测船舶是非常重要的。在数据来源上,一般有卫星合成孔径雷达(synthetic aperture radar,SAR)图像,高分可见光遥感图像。由于SAR可全天时、全天候,多维度成像,已经成为近年来的研究热点。但随着大量光学传感器的广泛使用,高空间分辨率的可见光图的数据量快速增加。由于其在数据获取方式上成本可控,方便快捷等优点。船舶检测使用遥感光学遥感图像已经成为一种新趋势[1]
检测方法上,大致可以分为传统方法检测,机器学习与深度学习3种方法,其中传统方法与机器学习方法需要手动的图像特征提取等复杂的数据处理过程,并且在环境发生变化时,不能快速适应,常常导致检测结果不理想[2]。深度学习方法凭借着其提取图像特征快速、检测精度高、泛化能力强等优点,已经在遥感图像检测领域广泛使用。以区域建议网络(region proposal network,RPN)为代表的二阶段检测方法(two stage)将检测任务拆分为分类和回归2个部分。在应用上,Jiang[3]使用Faster R-CNN网络将大场景SAR图像中的船舶切片。然后,应用FNLM滤波来降噪并增强目标切片的结构信息。提出了一种优化的Chan-Vese模型,在SSDD数据集上进行了一系列实验,结果表明,该方法不仅可以准确地提取观测船的轮廓,而且可以减少模型的计算时间。赵等[4]使用MASATI(mAritime sATellite imagery)海洋及湖泊船舶光学遥感数据集,实现了快速实时精准检测,其mAP达到了80%。在克服环境问题上,Yuan等[5]在数据预处理部分,将原始图像进行去雾处理后再放入Faster R-CNN网络,但是精确度的提升是以庞大的计算量作为代价的,并且还出现了图片颜色失真的情况。以上的二阶段检测方法多数在精确度上确实有提升,但是检测速度和模型大小都不太理想,对于后续部署实时检测并不友好。而一阶段检测算法在保证精确度的情况下速度更快,其代表算法有YOLOv3[6]、YOLOv4[7]、YOLOv5[8]和RetinaNet[9]等算法。叶等[10]使用改进的YOLOv3网络实现了雾天海洋图像船舶检测。Zhu等[11]通过改进RetinaNet的特征模块和损失函数解决了船舶检测中存在的任意方位、宽高比大、排列密集等问题。Liu等[12]将YOLOv5网络特征提取过程与GhostbottleNet算法相融合,大大提高了网络的精确性和实时性。化嫣然等[13]使用优化后的YOLOv3-Tiny实现了较为精确的遥感图像中大目标检测。以上多数检测方法要么是使用参数量多的大模型保证高精确度,要么牺牲精确度来实现模型的轻量。在数据集使用上,多数都是基于高分辨率下的遥感图像(通常为10 m及以下),很少有人使用中等分辨率的遥感图像检测(通常为10 m以上)。不同的分辨率在不同场景有着各自的优点。高分辨遥感图像凭借其图像质量优秀,且图像中的目标多为大目标,对于网络的检测精度具有质的提升,但是高分辨遥感图像覆盖的区域并不广阔,效率相对较低。中等分辨率遥感图像覆盖的区域比相同大小的高分辨图像要宽得多,在中等分辨率遥感图像中船舶都是小目标,在有限资源下,检测区域更广阔,更有利于快速检测船舶。
针对当前基于遥感图像船舶目标检测多数都是检测大型船舶,很少在复杂天气环境下检测小目标船舶。现提出一种基于YOLOv5s的改进网络。首先为保证后续研究的实时检测,采用YOLOv5s轻量化版本,为了让整个网络能够更好地学习到小目标船舶特征,在网络主干(backbone)部分加入Bottleneck Transformers注意力机制。为了让网络更轻量,在网络颈部(neck)采用轻量级结构:Slim-neck。针对小目标船舶尺度不一致的目标的特征提取问题,引入自适应空间特征融合(adaptive structure feature fusion,ASFF),同时在提升精确度上使用EIoU _loss边界框损失函数。通过以上改进实现在广阔海域上的小目标船舶检测。
YOLO系列网络的基本思想是将输入的特征图经过增强后,将特征图划分为S×S的格子(grid cells),每个格子负责对落入其中的目标进行检测,一次性预测所有各自所含目标的边界框、定位置信度、以及所有类别概率向量。相较于二阶段的目标检测方法,在保证精确度的情况下更加简捷快速。YOLOv5目前是YOLO系列第5个更新版本,且使用较为广泛的一个网络,根据不同使用场景及需求有YOLOv5m、YOLOv5l、YOLOv5s等版本。选择了网络宽度小、深度低、体积小、实时性较好的YOLOv5s版本。YOLOv5s网络结构大致可分为4个部分:输入端(input)、网络主干(backbone)、颈部网络(neck)、检测头(head)。其中输入端负责Mosaic数据增强,自适应锚框计算,自适应图片缩放等数据预处理功能。Backbone是由CBS、CSP1_X和快速金字塔池化(spatial pyramid pooling-fast,SPPF)组成,主要负责图像卷积后的特征提取、转换等工作。Neck由CSP2,上采样(Upsample)和拼接模块(Concat)构成,主要负责将上采样的图像特征进行拼接后传入head。Head主要负责图像预测。具体的整体结构及其子结构如图1图2所示。
注意力机制在目标检测领域已经应用非常广泛[14],其主要思想是借鉴了人类视觉在面对大量信息的处理决策过程。人类视网膜不同的部位具有不同程度的信息处理能力,且人类大脑对于事物关注程度取决于它的重要性,所以为了合理利用有限的视觉信息处理资源,需要将有限的处理资源用在重要信息部分。
为了使改进的YOLOv5s网络在参数量和计算量上没有大幅增加的情况下,有效地提升检测的精度。提出了在主干网络中添加BoTNet(bottleneck transformers)注意力机制,它是一种概念上简单但功能强大的主干架构[15],主要思想是将自注意力机制融入目标检测等多个计算机视觉任务。仅通过在ResNet的最后3个瓶颈块中用全局自注意力替换空间卷积,而不进行其他更改。从而达到增加网络整体精确度,减少或者不带来额外的计算量和参数量。具体结构如图3所示。
其中多头自注意力(multi-head self-attention,MHSA)的计算公式公式为
A t t e n t i o n ( Q , K , V ) = s o f t m a x Q K T d k V
h e a d i = A t t e n t i o n ( Q W i Q , K W i K , V W i V )
MultiHead(Q,K,V)=Concat(head1,head2,…,headh)WO
式中: A t t e n t i o n ( Q , K , V )为得到的注意力的值; Q K V分别为查询量(query)、键(key)和值(value); d k为key的维度; W i为参数矩阵; h e a d i为第 i个检测头的注意力得分; W O为参数矩阵;Concat为拼接操作; h为检测头个数; M u l t i H e a d ( Q , K , V )为多头自注意力机制的值。
船舶检测的最终目的是实时检测,那么网络的体量大小和结构就应该更轻量。在YOLOv5s的颈部网络使用了Slim-neck结构[16]。它主要由广义稀疏卷积技术(generalized-sparse convolution,GSConv)实现。其中GSConv是在标准卷积(standard convolution,SC)和深度可分离卷积(depthwise separable convolution,DSC)的基础上进行的改进。DSC相较于SC的优势是大幅度减少卷积的参数和计算量。但是缺点也很明显,输入图像的通道信息在计算过程中被分离。这一缺点将会导致DSC的特征提取和融合能力比SC低得多,最终网络会因学习了不完善的图像特征而导致精确度下降。而GSConv以较低的时间复杂度尽可能保留通道信息连接,保证了模型的准确性和运行速度之间的平衡。GS bottleneck、VoV-GSCSP是在GSConv基础上衍生的一系列轻量的颈部网络,具体结构变化如图4图5所示。
其中DSC由图4(b)逐通道卷积和图4(c)逐点卷积2个阶段组成。
边界框回归是决定目标定位性能的关键步骤,也是在目标检测领域中衡量一个网络性能重要指标之一。其主要的思想就是计算预测框与真实框的交并比值(intersection over union,IoU),IoU越大说明检测效果越好。但是IoU仅考虑预测框与真实框的重叠面积,没有考虑二者的位置与尺度信息,导致网络效果不好。并且无法处理特殊情况——当2个框不相交时,IoU=0,loss=0,导致梯度无法回传,网络无法训练。为此引入了EIoU[17]替换YOLOv5s使用的CIoU[18],它考虑了边界框回归的重叠面积、中心点距离、纵横比,一定程度上解决了此前的GIoU、DIoU的缺点。CIoU损失( L C I o U)与EIoU损失( L E I o U)的计算公式为
L C I o U = 1 - I o U + ρ 2 ( b , b g t ) c 2 + α v
L E I o U = L I o U + L d i s + L a s p = 1 - I o U + ρ 2 ( b , b g t ) c 2 + ρ 2 ( w , w g t ) c w 2 + ρ 2 ( h , h g t ) c h 2
式中:IoU为预测框与真实框的交并比值; ρ为两个中心点之间的欧式距离; α为权重函数; v为用来度量纵横比的相似性;边框损失函数包含3个部分:重叠损失( L I o U)、中心距离损失( L d i s)和宽高损失( L a s p); b g t w g t h g t b w h分别为实框和预测框的中心点、边宽、边长; c 2 c w 2 c h 2分别为真实框和预测框最小外接矩形的对角线、边宽、边长。
根据式(4)中 v反映的纵横比的差异,而不是宽高分别与其置信度的真实差异,所以有时会阻碍模型有效的优化相似性。而EIoU在CIoU的惩罚项基础上的第三部分将纵横比的影响因子拆开分别计算目标框和锚框的长和宽,有效地提高了收敛速度和定位精度。
在目标检测领域中,针对物体检测中尺度变化问题,现行常用的解决方法是使用金字塔特征方法。针对遥感图像中海上的小目标船舶,其大小和尺度变化多样,不同特征尺度之间的不一致性会导致金字塔特征方法提取的特征是有限制的。具体来说,当检测具有特征金字塔的对象时,采用启发式引导特征的选择:大实例通常与上特征图相关联,小实例与下特征图相联系。当某个对象在某一级别的特征图中被视为正相关时,其他级别的特征图相应区的域将被视为背景。因此,如果图像包含大小对象,则不同级别的特征之间的冲突往往占据特征金字塔的主要部分。这种不一致性干扰了训练期间的梯度计算,并降低了特征金字塔的有效性。自适应空间特征融合(adaptive structure feature fusion,ASFF)很大程度上解决了此类问题[19]
ASSF的具体解决方法-恒等缩放与自适应融合:对于某一级别的特征,首先将其他级别的特征集成并调整为相同的分辨率,然后进行训练以找到最佳融合。在每个空间位置,不同级别的特征被自适应地融合。并且引入的计算开销很小、实现简单。具体的网络结构如图6所示。
最后各个改进在网络中位置以及总体结构如图7所示。
本文使用的是开源数据集LEVIR-Ship,它是由Chen等[20]所提出。LEVIR-Ship中的图像由高分一号和高分六号卫星的多光谱相机拍摄,空间分辨率为16 m,包含1 973个正样本和1 923个负样本,覆盖了85个场景,如港口、海面、云层等一些复杂场景。超过3 000个小型船舶实例。所以LEVIR-Ship数据集的场景复杂性,实例数据量是非常符合复杂场景下的光感图像船舶检测。将3 896张图片按照8∶1∶1的比例划分为训练集、验证集、测试集。
为验证本文方法的普适性和有效性,选取了西北工业大学开源McShips数据集作为验证数据集[21]。该数据集包括14 709张带标注的图像,包含6类军舰和7类民用船舶,拍摄地点包含不同海域、不同天气条件和不同拍摄视角,即包含不同尺度的船舶。其中也包含一部分小目标船舶。实验中,该数据集被整体划分为民用船和军舰2个类别,各自包含5 945和5 382实体船舶。
实验平台是主要配置环境为 Windows 10专业版操作系统,CPU为6核12线程AMD R5 5600,GPU为NVIDIA RTX 3070,CUDA版本为11.6,基于PyTorch1.12.1框架下Python版本为3.8进行网络训练和测试。网络默认的关键参数如表1所示。
为验证本文算法的检测性能和速度,探究各个改进方法的有效性,通过对比基准YOLOv5s和改进方法的检测图像差异来评估模型性能。采用了目标检测领域常见的精准率(precision,P)、召回率(recall,R)、平均精度均值@0.5(mean average precision, mAP)和10亿次浮点运算量(GFLOPs)作为评价指标,具体计算见式为
P = T p T p + F p × 100 %
R = T p T p + F n × 100 %
A P = 0 1 P ( R ) d R
m A P = 1 n i = 1 m A P i
式中:Tp为预测正确的正样本数量;Fp为预测错误的正样本数量;Fn为预测错误的负样本数量;n为数据集中的类别数。
为证明所引入的BoTNet注意力机制的在所提的整个网络中优越性,将BoTNet与最近热门的注意力机制进行对比,如SE(squeeze-and-excitation networks)[22]、CBAM(convolutional block attention module)[23]、SIMAM(simple, parameter-free attention)[24]、CA(coordinate attention)[25]。实验条件都是使用基准YOLOv5s网格加上不同的注意力机制,对比的结果如图8表2所示。
为体现提出网络的优越性,将改进后YOLOv5s与基准YOLOv5s进行mAP@0.5数据对比,其结果如图9所示。同时将与CenterNet、YOLOv3、YOLOv4等热门网络进行对比,其中对比网络都采用默认参数设置。各项评价指标的实验结果如表3所示。
图9中可以看出改进后YOLOv5s在300个epoch迭代次数,从训练阶段的开始到结束mAP@0.5值全程是优于基准YOLOv5s的,证明了在船舶检测中,提出的方法更有效且对小目标检测表现更优秀。
表3可以看出,CenterNet出现了精准率为100%但召回率仅为10%的情况,表明在预测结果时出现了没有误检但是有漏检的情况,除CenterNet的精准率外。在与较新的YOLOv8s对比中,改进后YOLOv5s在所有指标中依然是最优秀的,实现了在参数量,计算量和权重都下降的情况下实现了mAP的提升。
为验证所提出网络改进的有效性,对每个改进的模块进行消融实验,实验的结果如表4所示。
表4可以看出,每个部分的改进都有提升效果,自适应空间特征融合(ASFF)的引入对于网络的学习尺度不一致的目标时更准确,BotNet注意力机制使网络的计算量得到了有效降低,EIoU通过提升网络的定位性能提高最终的精确度,Slim-neck整体降低或减少了网络最终的计算量和权重大小。
为验证所提出的改进后YOLOv5s网络实际效果,选取了测试集的部分实例将提出的改进的YOLOv5s与对比的网络进行比较,同时为验证复杂环境下的检测效果,选取了黑夜下的海洋环境和云层遮挡的海洋环境,检测结果如图10图11所示。
图10可以看出,YOLOv5s出现误检的情况,YOLOv3凭借其网络的参数量巨大,得到了置信度最高的检测成绩,但是与改进的YOLOv5s相比没有太大优势。从图11可看出CenterNet出现了无法检测的情况,YOLOv3和YOLOv4都出现漏检的情况,只有改进的YOLOv5s没有出现误检,漏检的情况,总的来说,本文方法在实例检测中达到了更好的效果,实现了较为准确的小目标船舶检测。
在实验条件一致的情况下,将本文方法与基准YOLOv5s、YOLOv7-tiny、YOLOv8s进行对比实验,实验结果如表5所示。
表5可看出,本文方法在精准率、召回率、mAP评价指标上,全面由于对比网络。相较于基准YOLOv5s,mAP提升了3.9个百分点,在与YOLOv8s对比中,本文方法的mAP依旧提升了0.5个百分点。充分证明了本文方法的普适性。
针对遥感卫星需要在广大区域内快速检测船舶的需求,使用中等分辨率数据集——LEVIR-Ship,进行小目标船舶检测任务,在YOLOv5s基础网络结构上,进行了一系列优化改进,在进行实验与分析后,得到以下结论。
(1)优化了颈部网络,边框损失函数、引入了自适应空间特征融合(ASFF)以及BotNet注意力机制解决小目标船舶检测和网络体量问题,实验结果表明,改进后YOLOv5s方法实现了在降低计算量和权重大小的情况下,检测的mAP值得到了有效提升,相较于基准YOLOv5s和YOLOv8s,mAP分别提升了7.1个百分点和2个百分点,在云层和黑夜的复杂环境下实现了高精度小目标船舶检测。
(2)在船舶种类大而全的开源数据集McShips上,进行了对比实验,结果表明本文方法表现出了更优秀的性能,验证了所提出网络的普适性和有效性。
(3)后续研究将会收集海上更多气象环境下的船舶数据图像数据,增加数据集的多样性,并研究每个气象环境对于检测效果的影响,并思考解决不利气象环境下的模糊图片的船舶检测。
  • 国家自然科学基金(61771354)
参考文献 引证文献
排序方式:
[1]
Song Z H, Sui H G, Li Y C. A survey on ship detection technology in high-resolution optical remote sensing images[J]. Geomatics and Information Science of Wuhan University, 2021, 46(11): 1703-1715.
[2]
Muhanmmad Y, Wan J H, Xu M M, et al. Ship detection based on deep learning using SAR imagery: a systematic literature review[J]. Soft Computing, 2023, 27(1): 63-84.
[3]
Jiang M, Gu L, Li X, et al. Ship contourextraction from SAR images based on faster R-CNN and Chan-Vese model[J]. IEEE Transactions on Geoscience and Remote Sensing, 2023, 61: 1-14.
[4]
赵振强, 何水原, 梁永志. 基于Faster R-CNN的遥感影像船舶检测识别[J]. 测绘通报, 2021(11): 59-64.
Zhao Zhenqiang, He Shuiyuan, Liang Yongzhi, et al. Research on remote sensing image ship detection and identification based on Faster R-CNN[J] Bulletin of Surveying and Mapping, 2021(11): 59-64.
[5]
Yuan Z, Geng J, Dai T. Ship detection from remote sensing images based on deep learning[C]// Geo-Spatial Knowledge and Intelligence:5th International Conference. Chiang Mai: Springer, 2018: 336-344.
[6]
袁小平, 马绪起, 刘赛. 改进YOLOv3的行人车辆目标检测算法[J]. 科学技术与工程, 2021, 21(8): 3192-3198.
Yuan Xiaoping, Ma Xuqi, Liu Sai, et al. An improved algorithm of pedestrian and vehicle detection based on YOLOv3[J]. Science Technology and Engineering, 2021, 21(8): 3192-3198.
[7]
许小华, 陈豹, 王海菁, 等. 基于改进YOLOv4-Tiny的河湖船舶目标检测算法[J]. 人民长江, 2023, 54(9): 264-271.
Xu Xiaohua, Chen Bao, Wang Haijing, et al. Object detection algorithm for ships on rivers and lakes based on improved YOLOv4-Tiny[J]. Yangtze River, 2023, 54(9): 264-271.
[8]
Xu X, Zhang X, Zhang T. Lite-yolov5: a lightweight deep learning detector for on-board ship detection in large-scene sentinel-1 sar images[J]. Remote Sensing, 2022, 14(4). DOI:10.3390/rs14041018.
[9]
凡文俊, 赵曙光, 郭力争. 基于改进RetinaNet的船舶检测算法[J]. 计算机应用, 2022, 42(7): 2248-2255.
Fan Wenjun, Zhao Shuguang, Guo Lizheng. Ship detection algorithm based on improved RetinaNet[J]. Journal of Computer Applications, 2022, 42(7): 2248-2255.
[10]
叶乐, 李朝锋. 基于频域通道注意力的YOLOv3网络的雾天海洋图像船舶检测[J]. 上海海事大学学报, 2023, 44(2): 18-24.
Ye Le, Li Chaofeng. Frequency channel attention-based YOLOv3 network for ship detection from foggy marine images[J]. Journal of Shanghai Maritime University, 2023, 44(2): 18-24.
[11]
Zhu M, Hu G, Zhou H, et al. Arbitrary-oriented ship detection based on RetinaNet for remote sensing images[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2021, 14: 6694-6706.
[12]
Liu T, Zhou B J, Zhao Y S, et al. Ship detection algorithm based on improved YOLO V5[C]// 6th International Conference on Automation, Control and Robotics Engineering (CACRE). Dalian: IEEE, 2021: 483-487.
[13]
化嫣然, 张卓, 龙赛, 等. 基于改进YOLO算法的遥感图像目标检测[J]. 电子测量技术, 2020, 43(24): 87-92.
Hua Yanran, Zhang Zhuo, Long Sai, et al. Remote sensing image target detection based on improved YOLO algorithm[J]. Electro-nic Measurement Technology, 2020, 43(24): 87-92.
[14]
Guo M H, Xu T X, Liu J J, et al. Attention mechanisms in computer vision: a survey[J]. Computational Visual Media, 2022, 8(3): 331-368.
[15]
Srinivas A, Lin T Y, Parmar N, et al. Bottleneck transformers for visual recognition[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2021: 16519-16529.
[16]
Li H, Li J, Wei H, et al. Slim-neck by GSConv: a better design paradigm of detector architectures for autonomous vehicles[J]. arXiv preprint arXiv: 2206.02424, 2022.
[17]
Zhang Y F, Ren W, Zhang Z, et al. Focal and efficient IOU loss for accurate bounding box regression[J]. Neurocomputing, 2022, 506: 146-157.
[18]
Zheng Z, Wang P, Ren D, et al. Enhancing geometric factors in model learning and inference for object detection and instance segmentation[J]. IEEE Transactions on Cybernetics, 2021, 52(8): 8574-8586.
[19]
Liu S, Huang D, Wang Y. Learning spatial fusion for single-shot object detection[J]. arXiv preprint arXiv: 1911.09516, 2019.
[20]
Chen J, Chen K, Chen H, et al. A degraded reconstruction enhancement-based method for tiny ship detection in remote sensing images with a new large-scale dataset[J]. IEEE Transactions on Geoscience and Remote Sensing, 2022, 60: 1-14.
[21]
Zheng Y, Zhang S. Mcships: a large-scale ship dataset for detection and fine-grained categorization in the wild[C]// IEEE International Conference on Multimedia and Expo (ICME). New York: IEEE, 2020: 1-6.
[22]
Hu J, Shen L, Sun G. Squeeze-and-excitation networks[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. London: IEEE, 2018: 7132-7141.
[23]
Woo S, Park J, Lee J Y, et al. Cbam: convolutional block attention module[C]// Proceedings of the European Conference on Computer Vision (ECCV). Munich: Springer, 2018: 3-19.
[24]
Yang L, Zhang R Y, Li L, et al. Simam: a simple, parameter-free attention module for convolutional neural networks[C]// International Conference on Machine Learning. New York: PMLR, 2021: 11863-11874.
[25]
Hou Q, Zhou D, Feng J. Coordinate attention for efficient mobile network design[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2021: 13713-13722.
2025年第25卷第2期
PDF下载
254
106
引用本文
BibTeX
文章信息
doi: 10.12404/j.issn.1671-1815.2308931
  • 接收时间:2023-11-14
  • 首发时间:2025-12-05
  • 出版时间:2025-01-18
补充材料
相关文章
文章信息
作者
出版历史
  • 收稿日期:2023-11-14
  • 修回日期:2024-10-21
基金
国家自然科学基金(61771354)
作者信息
    长江大学计算机科学学院, 荆州 434023

通讯作者:

* 肖小玲(1973—),女,汉族,湖南邵阳人,博士,教授。研究方向:智能信息处理和网络安全。E-mail:
参考文献
分享链接
https://castjournals.cast.org.cn/joweb/kxjsygc/CN/10.12404/j.issn.1671-1815.2308931
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