Article(id=1156949463970435804, tenantId=1146029695717560320, journalId=1146123166801305609, issueId=1156949362480861758, articleNumber=null, orderNo=null, doi=10.12404/j.issn.1671-1815.2401484, pmid=null, cstr=null, oa=null, hot=null, price=null, onlineType=0, articleFormat=0, articleType=null, articleTypeStr=research-article, receivedDate=1709481600000, receivedDateStr=2024-03-04, revisedDate=1732032000000, revisedDateStr=2024-11-20, acceptedDate=null, acceptedDateStr=null, onlineDate=1753767847292, onlineDateStr=2025-07-29, pubDate=1738944000000, pubDateStr=2025-02-08, doiRegisterDate=null, doiRegisterDateStr=null, onlineIssueDate=1753767847292, onlineIssueDateStr=2025-07-29, onlineJustAcceptDate=null, onlineJustAcceptDateStr=null, onlineFirstDate=null, onlineFirstDateStr=null, sourceXml=null, magXml=null, createTime=1753767847292, creator=13701087609, updateTime=1753767847292, updator=13701087609, issue=Issue{id=1156949362480861758, tenantId=1146029695717560320, journalId=1146123166801305609, year='2025', volume='25', issue='4', pageStart='1312', pageEnd='1751', issueExtLink='null', onlineDate='null', pubDate='null', beforeIssueId=null, nextIssueId=null, price=null, status=1, issueComplete=1, articleOrder=1, issueType=-1, specialIssue=0, createTime=1753767823094, creator=13701087609, updateTime=1755171161273, updator=13701087609, preIssue=null, nextIssue=null, ext={EN=IssueExt(id=1162835389472424814, tenantId=1146029695717560320, journalId=1146123166801305609, issueId=1156949362480861758, language=EN, specialIssueTitle=, coverIllustrator=, specialIssueEditor=, specialIssueAbout=), CN=IssueExt(id=1162835389472424815, tenantId=1146029695717560320, journalId=1146123166801305609, issueId=1156949362480861758, language=CN, specialIssueTitle=, coverIllustrator=, specialIssueEditor=, specialIssueAbout=)}, issueFiles=null}, startPage=1540, endPage=1546, ext={EN=ArticleExt(id=1156949464658301676, articleId=1156949463970435804, tenantId=1146029695717560320, journalId=1146123166801305609, language=EN, title=Video Action Recognition Based on Sport Feature Enhancement in Two-stream Network, columnId=1156262729162810294, journalTitle=Science Technology and Engineering, columnName=Papers·Automation and Computational Technology, runingTitle=null, highlight=null, articleAbstract=

To solve the problem of insufficient extraction of sport features by dual stream networks in current action recognition, which leads to low recognition accuracy, a action recognition method based on sport feature enhancement two-stream networks was proposed to improve accuracy. The network was divided into spatial stream and temporal stream, with the same structure but different inputs. The input of the spatial stream network was a video frame sequence, while the input of the temporal stream network was a video frame difference sequence. The network structure used Resnet50 as the backbone network, replacing the 3×3 convolution with the proposed global sport feature module and local sport feature module, fully extracting video sport information, and finally combining spatial and temporal stream to output the results. The results show that the accuracy of the model on the UCF101 and HMDB51 datasets reaches 96.8% and 75.3%, which is superior to traditional algorithms.

, correspAuthors=Chen ZHAO, authorNote=null, correspAuthorsNote=null, copyrightStatement=null, copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=null, magXml=null, pdfUrl=null, pdf=null, pdfFileSize=null, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=null, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=null, mapNumber=null, authorCompany=null, fund=null, authors=null, authorsList=Ruo-chen CAO, Xiu-fang FENG, Chen ZHAO), CN=ArticleExt(id=1156949516684447886, articleId=1156949463970435804, tenantId=1146029695717560320, journalId=1146123166801305609, language=CN, title=基于运动特征增强双流网络的视频行为识别, 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=为解决目前行为识别中双流网络对运动特征提取的不充分导致识别准确度低的问题,通过提出基于运动特征增强双流网络的行为识别方法提高准确率。该网络分为空间流和时间流,空间流网络和时间流网络结构相同,输入不同。空间流网络输入为视频帧序列,而时间流网络输入为视频帧差序列。网络结构以Resnet50为骨干网络,将3×3卷积替换为所提出的全局运动特征模块和局部运动特征模块,充分提取视频运动信息,最终将空间流和时间流结合输出结果。结果表明:该模型在UCF101和HMDB51数据集上准确率达到96.8%和75.3%,与传统算法相比有一定优越性。, correspAuthors=赵晨, authorNote=null, correspAuthorsNote=
*赵晨(1998—),男,汉族,辽宁沈阳人,硕士研究生。研究方向:行为识别。E-mail:
, copyrightStatement=null, copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=LWDQyX5JDSjwwdVOXSF8DA==, magXml=K3C5nk8R28lNpB6G2qiNhA==, pdfUrl=null, pdf=7Dn+kDlCtruW3uPmT+Ebug==, pdfFileSize=3501193, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=lLspzedrVt7MNfizIDSu1g==, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=/JlHjHb5sMFQ4NjH8wX51A==, mapNumber=null, authorCompany=null, fund=null, authors=

曹若琛(1990—),男,汉族,山西太原人,博士,讲师。研究方向:虚拟现实、行为识别。E-mail:

, authorsList=曹若琛, 冯秀芳, 赵晨)}, authors=[Author(id=1225944423352152727, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, orderNo=0, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=caory004@163.com, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1225944424803381931, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, authorId=1225944423352152727, language=EN, stringName=Ruo-chen CAO, firstName=Ruo-chen, middleName=null, lastName=CAO, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=School of Software, Taiyuan University of Technology, Jinzhong 030600, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1225944424966959804, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, authorId=1225944423352152727, language=CN, stringName=曹若琛, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=太原理工大学软件学院, 晋中 030600, bio={"content":"

曹若琛(1990—),男,汉族,山西太原人,博士,讲师。研究方向:虚拟现实、行为识别。E-mail:

"}, bioImg=null, bioContent=

曹若琛(1990—),男,汉族,山西太原人,博士,讲师。研究方向:虚拟现实、行为识别。E-mail:

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1225944423117271673, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, xref=null, ext=[AuthorCompanyExt(id=1225944423129854587, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, companyId=1225944423117271673, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=School of Software, Taiyuan University of Technology, Jinzhong 030600, China), AuthorCompanyExt(id=1225944423138243196, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, companyId=1225944423117271673, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=太原理工大学软件学院, 晋中 030600)])]), Author(id=1225944425067623112, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, orderNo=1, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1225944425331864304, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, authorId=1225944425067623112, language=EN, stringName=Xiu-fang FENG, firstName=Xiu-fang, middleName=null, lastName=FENG, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=School of Software, Taiyuan University of Technology, Jinzhong 030600, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1225944425503830795, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, authorId=1225944425067623112, language=CN, stringName=冯秀芳, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=太原理工大学软件学院, 晋中 030600, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1225944423117271673, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, xref=null, ext=[AuthorCompanyExt(id=1225944423129854587, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, companyId=1225944423117271673, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=School of Software, Taiyuan University of Technology, Jinzhong 030600, China), AuthorCompanyExt(id=1225944423138243196, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, companyId=1225944423117271673, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=太原理工大学软件学院, 晋中 030600)])]), Author(id=1225944425633854236, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, orderNo=2, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=906070493@qq.com, emailSecond=null, emailThird=null, correspondingAuthor=1, authorType=1, ext={EN=AuthorExt(id=1225944425780654894, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, authorId=1225944425633854236, language=EN, stringName=Chen ZHAO, firstName=Chen, middleName=null, lastName=ZHAO, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=*, address=School of Software, Taiyuan University of Technology, Jinzhong 030600, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1225944425872929589, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, authorId=1225944425633854236, language=CN, stringName=赵晨, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=*, address=太原理工大学软件学院, 晋中 030600, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1225944423117271673, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, xref=null, ext=[AuthorCompanyExt(id=1225944423129854587, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, companyId=1225944423117271673, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=School of Software, Taiyuan University of Technology, Jinzhong 030600, China), AuthorCompanyExt(id=1225944423138243196, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, companyId=1225944423117271673, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=太原理工大学软件学院, 晋中 030600)])])], keywords=[Keyword(id=1225944426116199257, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, language=EN, orderNo=1, keyword=action recognition), Keyword(id=1225944426250416998, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, language=EN, orderNo=2, keyword=deep learning), Keyword(id=1225944426363663222, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, language=EN, orderNo=3, keyword=sport feature), Keyword(id=1225944426493686661, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, language=EN, orderNo=4, keyword=two-stream network), Keyword(id=1225944426602738579, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, language=CN, orderNo=1, keyword=行为识别), Keyword(id=1225944426774705056, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, language=CN, orderNo=2, keyword=深度学习), Keyword(id=1225944427152192458, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, language=CN, orderNo=3, keyword=运动特征), Keyword(id=1225944427257050074, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, language=CN, orderNo=4, keyword=双流网络)], refs=[Reference(id=1225944434500612461, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, doi=null, pmid=null, pmcid=null, year=2022, volume=39, issue=8, pageStart=1, pageEnd=10, url=null, language=null, rfNumber=[1], rfOrder=0, authorNames=袁首, 乔勇军, 苏航, journalName=微电子学与计算机, refType=null, unstructuredReference=袁首, 乔勇军, 苏航, 等. 基于深度学习的行为识别方法综述[J]. 微电子学与计算机, 2022, 39(8): 1-10., articleTitle=基于深度学习的行为识别方法综述, refAbstract=null), Reference(id=1225944434639024508, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, doi=null, pmid=null, pmcid=null, year=2022, volume=39, issue=8, pageStart=1, pageEnd=10, url=null, language=null, rfNumber=[1], rfOrder=1, authorNames=Yuan Shou, Qiao Yongjun, Su Hang, journalName=Microelectronics and Computer Science, refType=null, unstructuredReference=Yuan Shou, Qiao Yongjun, Su Hang, et al. A review of behavior recognition methods based on deep learning[J]. Microelectronics and Computer Science, 2022, 39(8): 1-10., articleTitle=A review of behavior recognition methods based on deep learning, refAbstract=null), Reference(id=1225944434777436552, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, doi=null, pmid=null, pmcid=null, year=2022, volume=130, issue=5, pageStart=1366, pageEnd=1401, url=null, language=null, rfNumber=[2], rfOrder=2, authorNames=Kong Y, Fu Y, journalName=International Journal of Computer Vision, refType=null, unstructuredReference=Kong Y, Fu Y. Human action recognition and prediction: a survey[J]. International Journal of Computer Vision, 2022, 130(5): 1366-1401., articleTitle=Human action recognition and prediction: a survey, refAbstract=null), Reference(id=1225944434899071376, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, doi=null, pmid=null, pmcid=null, year=2022, volume=39, issue=3, pageStart=651, pageEnd=660, url=null, language=null, rfNumber=[3], rfOrder=3, authorNames=梁绪, 李文新, 张航宁, journalName=计算机应用研究, refType=null, unstructuredReference=梁绪, 李文新, 张航宁. 人体行为识别方法研究综述[J]. 计算机应用研究, 2022, 39(3): 651-660., articleTitle=人体行为识别方法研究综述, refAbstract=null), Reference(id=1225944434995540380, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, doi=null, pmid=null, pmcid=null, year=2022, volume=39, issue=3, pageStart=651, pageEnd=660, url=null, language=null, rfNumber=[3], rfOrder=4, authorNames=Liang Xu, Li Wenxin, Zhang Hangning, journalName=Computer Application Research, refType=null, unstructuredReference=Liang Xu, Li Wenxin, Zhang Hangning. Summary of research on human behavior recognition methods[J]. Computer Application Research, 2022, 39(3): 651-660., articleTitle=Summary of research on human behavior recognition methods, refAbstract=null), Reference(id=1225944435083620771, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, doi=null, pmid=null, pmcid=null, year=2022, volume=39, issue=3, pageStart=154, pageEnd=158, url=null, language=null, rfNumber=[4], rfOrder=5, authorNames=丁雪琴, 朱轶昇, 朱浩华, journalName=计算机应用与软件, refType=null, unstructuredReference=丁雪琴, 朱轶昇, 朱浩华, 等. 基于时空异构双流卷积网络的行为识别[J]. 计算机应用与软件, 2022, 39(3): 154-158., articleTitle=基于时空异构双流卷积网络的行为识别, refAbstract=null), Reference(id=1225944435217838508, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, doi=null, pmid=null, pmcid=null, year=2022, volume=39, issue=3, pageStart=154, pageEnd=158, url=null, language=null, rfNumber=[4], rfOrder=6, authorNames=Ding Xueqin, Zhu Yisheng, Zhu Haohua, journalName=Computer Applications and Software, refType=null, unstructuredReference=Ding Xueqin, Zhu Yisheng, Zhu Haohua, et al. Behavior recognition based on spatio-temporal heterogeneous double-stream convolution network[J]. Computer Applications and Software, 2022, 39(3): 154-158., articleTitle=Behavior recognition based on spatio-temporal heterogeneous double-stream convolution network, refAbstract=null), Reference(id=1225944435360444851, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, doi=null, pmid=null, pmcid=null, year=2019, volume=null, issue=null, pageStart=7083, pageEnd=7093, url=null, language=null, rfNumber=[5], rfOrder=7, authorNames=Lin J, Gan C, Han S, journalName=Proceedings of the IEEE/CVF International Conference on Computer vision, refType=null, unstructuredReference=Lin J, Gan C, Han S. TSM: temporal shift module for efficient video understanding[C]// Proceedings of the IEEE/CVF International Conference on Computer vision. New York: IEEE, 2019: 7083-7093., articleTitle=TSM: temporal shift module for efficient video understanding, refAbstract=null), Reference(id=1225944435490468282, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, doi=null, pmid=null, pmcid=null, year=2023, volume=44, issue=1, pageStart=284, pageEnd=291, url=null, language=null, rfNumber=[6], rfOrder=8, authorNames=金博, 王景林, 刘泓铄, journalName=计算机工程与设计, refType=null, unstructuredReference=金博, 王景林, 刘泓铄, 等. 基于双注意力时间卷积网络的人体行为识别[J]. 计算机工程与设计, 2023, 44(1): 284-291., articleTitle=基于双注意力时间卷积网络的人体行为识别, refAbstract=null), Reference(id=1225944435595325890, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, doi=null, pmid=null, pmcid=null, year=2023, volume=44, issue=1, pageStart=284, pageEnd=291, url=null, language=null, rfNumber=[6], rfOrder=9, authorNames=Jin Bo, Wang Jinglin, Liu Hongyue, journalName=Computer Engineering and Design, refType=null, unstructuredReference=Jin Bo, Wang Jinglin, Liu Hongyue, et al. Human behavior recognition based on dual attention time convolution network[J]. Computer Engineering and Design, 2023, 44(1): 284-291., articleTitle=Human behavior recognition based on dual attention time convolution network, refAbstract=null), Reference(id=1225944435754709451, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, doi=null, pmid=null, pmcid=null, year=2020, volume=null, issue=null, pageStart=203, pageEnd=213, url=null, language=null, rfNumber=[7], rfOrder=10, authorNames=Feichtenhofer C, journalName=Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, refType=null, unstructuredReference=Feichtenhofer C. X3D: expanding architectures for efficient video recognition[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2020: 203-213., articleTitle=X3D: expanding architectures for efficient video recognition, refAbstract=null), Reference(id=1225944435880538576, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, doi=null, pmid=null, pmcid=null, year=2021, volume=21, issue=5, pageStart=1656, pageEnd=null, url=null, language=null, rfNumber=[8], rfOrder=11, authorNames=Dong M, Fang Z, Li Y, journalName=Sensors, refType=null, unstructuredReference=Dong M, Fang Z, Li Y, et al. AR3D: attention residual 3D network for human action recognition[J]. Sensors, 2021, 21(5): 1656., articleTitle=AR3D: attention residual 3D network for human action recognition, refAbstract=null), Reference(id=1225944436035727829, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, doi=null, pmid=null, pmcid=null, year=2023, volume=40, issue=2, pageStart=192, pageEnd=196, url=null, language=null, rfNumber=[9], rfOrder=12, authorNames=闫雨寒, 陈天, 刘忠育, journalName=计算机应用与软件, refType=null, unstructuredReference=闫雨寒, 陈天, 刘忠育, 等. 基于双重注意力和3DResNet-BiLSTM行为识别方法[J]. 计算机应用与软件, 2023, 40(2): 192-196, 205., articleTitle=基于双重注意力和3DResNet-BiLSTM行为识别方法, refAbstract=null), Reference(id=1225944436169945566, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, doi=null, pmid=null, pmcid=null, year=2023, volume=40, issue=2, pageStart=192, pageEnd=196, url=null, language=null, rfNumber=[9], rfOrder=13, authorNames=Yan yuhan, Chen Tian, Liu Zhongyu, journalName=Computer Applications and Software, refType=null, unstructuredReference=Yan yuhan, Chen Tian, Liu Zhongyu, et al. Behavior recognition method based on dual attention and 3DResNet-BiLSTM[J]. Computer Applications and Software, 2023, 40(2): 192-196, 205., articleTitle=Behavior recognition method based on dual attention and 3DResNet-BiLSTM, refAbstract=null), Reference(id=1225944436266414563, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=[10], rfOrder=14, authorNames=Bertasius G, Wang H, Torresani L, journalName=International Conference of Machine Learning, refType=null, unstructuredReference=Bertasius G, Wang H, Torresani L. Is space-time attention all you need for video understanding?[C]// International Conference of Machine Learning. New York: IEEE, 2021: DOI:10.48550/arXiv.2102.05095., articleTitle=Is space-time attention all you need for video understanding?, refAbstract=null), Reference(id=1225944436367077867, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, doi=null, pmid=null, pmcid=null, year=2022, volume=null, issue=null, pageStart=2874, pageEnd=2884, url=null, language=null, rfNumber=[11], rfOrder=15, authorNames=Ranasinghe K, Naseer M, Khan S, journalName=Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, refType=null, unstructuredReference=Ranasinghe K, Naseer M, Khan S, et al. Self-supervised video transformer[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2022: 2874-2884., articleTitle=Self-supervised video transformer, refAbstract=null), Reference(id=1225944436471935473, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, doi=null, pmid=null, pmcid=null, year=2022, volume=null, issue=null, pageStart=3202, pageEnd=3211, url=null, language=null, rfNumber=[12], rfOrder=16, authorNames=Liu Z, Ning J, Cao Y, journalName=Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, refType=null, unstructuredReference=Liu Z, Ning J, Cao Y, et al. Video swin transformer[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2022: 3202-3211., articleTitle=Video swin transformer, refAbstract=null), Reference(id=1225944436627124728, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, doi=null, pmid=null, pmcid=null, year=2023, volume=23, issue=8, pageStart=3370, pageEnd=3378, url=null, language=null, rfNumber=[13], rfOrder=17, authorNames=姬晓飞, 赵东阳, journalName=科学技术与工程, refType=null, unstructuredReference=姬晓飞, 赵东阳. 人体检测与异常行为识别联合算法[J]. 科学技术与工程, 2023, 23(8): 3370-3378., articleTitle=人体检测与异常行为识别联合算法, refAbstract=null), Reference(id=1225944436727788032, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, doi=null, pmid=null, pmcid=null, year=2023, volume=23, issue=8, pageStart=3370, pageEnd=3378, url=null, language=null, rfNumber=[13], rfOrder=18, authorNames=Ji Xiaofei, Zhao Dongyang, journalName=Science Technology and Engineering, refType=null, unstructuredReference=Ji Xiaofei, Zhao Dongyang. Joint algorithm for human detection and abnormal behavior recognition[J]. Science Technology and Engineering, 2023, 23(8): 3370-3378., articleTitle=Joint algorithm for human detection and abnormal behavior recognition, refAbstract=null), Reference(id=1225944438195794442, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, doi=null, pmid=null, pmcid=null, year=2014, volume=null, issue=null, pageStart=568, pageEnd=576, url=null, language=null, rfNumber=[14], rfOrder=19, authorNames=Simonyan K, Zisserman A, journalName=Advances in Neural Information Processing Systems, refType=null, unstructuredReference=Simonyan K, Zisserman A. Two-stream convolutional networks for action recognition in videos[C]// Advances in Neural Information Processing Systems. New York: IEEE, 2014: 568-576., articleTitle=Two-stream convolutional networks for action recognition in videos, refAbstract=null), Reference(id=1225944438275486225, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, doi=null, pmid=null, pmcid=null, year=2016, volume=null, issue=null, pageStart=20, pageEnd=36, url=null, language=null, rfNumber=[15], rfOrder=20, authorNames=Wang L, Xiong Y, Wang Z, journalName=European Conference on Computer Vision, refType=null, unstructuredReference=Wang L, Xiong Y, Wang Z, et al. Temporal segment networks: towards good practices for deep action recognition[C]// European Conference on Computer Vision. Cham: Springer, 2016: 20-36., articleTitle=Temporal segment networks: towards good practices for deep action recognition, refAbstract=null), Reference(id=1225944438367760919, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, 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=Soomro K, Zamir A R, Shah M, journalName=arXiv Preprint, refType=null, unstructuredReference=Soomro K, Zamir A R, Shah M. UCF101: a dataset of 101 human actions classes from videos in the wild[J]. arXiv Preprint, 2012: arXiv:1212.0402., articleTitle=UCF101: a dataset of 101 human actions classes from videos in the wild, refAbstract=null), Reference(id=1225944438506172959, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, doi=null, pmid=null, pmcid=null, year=2011, volume=null, issue=null, pageStart=2556, pageEnd=2563, url=null, language=null, rfNumber=[17], rfOrder=22, authorNames=Kuehne H, Jhuang H, Garrote E, journalName=2011 International Conference on Computer Vision, refType=null, unstructuredReference=Kuehne H, Jhuang H, Garrote E, et al. HMDB: a large video database for human motion recognition[C]// 2011 International Conference on Computer Vision. New York: IEEE, 2011: 2556-2563., articleTitle=HMDB: a large video database for human motion recognition, refAbstract=null), Reference(id=1225944438623613479, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, doi=null, pmid=null, pmcid=null, year=2021, volume=null, issue=null, pageStart=7387, pageEnd=7394, url=null, language=null, rfNumber=[18], rfOrder=23, authorNames=Demir U, Rawat Y S, Shah M, journalName=null, refType=null, unstructuredReference=Demir U, Rawat Y S, Shah M. Tinyvirat:low-resolution video action recognition[C]//25th International Conference on Pattern Recognition ( ICPR).New York: IEEE, 2021: 7387-7394., articleTitle=Tinyvirat:low-resolution video action recognition, refAbstract=null), Reference(id=1225944438736859694, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, doi=null, pmid=null, pmcid=null, year=2022, volume=39, issue=12, pageStart=284, pageEnd=289, url=null, language=null, rfNumber=[19], rfOrder=24, authorNames=王琼, 王旭, 刘云麟, journalName=计算机仿真, refType=null, unstructuredReference=王琼, 王旭, 刘云麟, 等. 基于改进密集轨迹算法的人体行为识别[J]. 计算机仿真, 2022, 39(12): 284-289, 356., articleTitle=基于改进密集轨迹算法的人体行为识别, refAbstract=null), Reference(id=1225944438845911605, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, doi=null, pmid=null, pmcid=null, year=2022, volume=39, issue=12, pageStart=284, pageEnd=289, url=null, language=null, rfNumber=[19], rfOrder=25, authorNames=Wang Qiong, Wang Xu, Liu Yunlin, journalName=Computer Simulation, refType=null, unstructuredReference=Wang Qiong, Wang Xu, Liu Yunlin, et al. Human behavior recognition based on improved dense trajectory algorithm[J]. Computer Simulation, 2022, 39(12): 284-289, 356., articleTitle=Human behavior recognition based on improved dense trajectory algorithm, refAbstract=null), Reference(id=1225944438946574902, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, doi=null, pmid=null, pmcid=null, year=2022, volume=null, issue=null, pageStart=179, pageEnd=182, url=null, language=null, rfNumber=[20], rfOrder=26, authorNames=Yang M, Guo Y, Zhou F, journalName=International Conference on Image Processing, Computer Vision and Machine Learning(ICICML), refType=null, unstructuredReference=Yang M, Guo Y, Zhou F, et al. TS-D3D: a novel two-stream model for action recognition[C]// International Conference on Image Processing, Computer Vision and Machine Learning(ICICML). New York: IEEE, 2022: 179-182., articleTitle=TS-D3D: a novel two-stream model for action recognition, refAbstract=null), Reference(id=1225944439156290110, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, doi=null, pmid=null, pmcid=null, year=2021, volume=null, issue=null, pageStart=6664, pageEnd=6671, url=null, language=null, rfNumber=[21], rfOrder=27, authorNames=Bai S, Wang Q, Li X, journalName=null, refType=null, unstructuredReference=Bai S, Wang Q, Li X. MFI:multi-range feature interchange for video action recognition[C]//25th International Conference on Pattern Recognition(ICPR). New York: IEEE, 2021: 6664-6671., articleTitle=MFI:multi-range feature interchange for video action recognition[C], refAbstract=null), Reference(id=1225944439303090755, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, doi=null, pmid=null, pmcid=null, year=2018, volume=null, issue=null, pageStart=695, pageEnd=712, url=null, language=null, rfNumber=[22], rfOrder=28, authorNames=Zolfaghari M, Singh K, Brox T, journalName=Proceedings of the European Conference on Computer Vision(ECCV), refType=null, unstructuredReference=Zolfaghari M, Singh K, Brox T. Eco: efficient convolutional network for online video understanding[C]// Proceedings of the European Conference on Computer Vision(ECCV). New York: IEEE, 2018: 695-712., articleTitle=Eco: efficient convolutional network for online video understanding, refAbstract=null)], funds=[Fund(id=1225944434165068119, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, awardId=202102020101007, language=CN, fundingSource=山西省重点研发计划(202102020101007), fundOrder=null, country=null)], companyList=[AuthorCompany(id=1225944423117271673, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, xref=null, ext=[AuthorCompanyExt(id=1225944423129854587, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, companyId=1225944423117271673, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=School of Software, Taiyuan University of Technology, Jinzhong 030600, China), AuthorCompanyExt(id=1225944423138243196, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, companyId=1225944423117271673, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=太原理工大学软件学院, 晋中 030600)])], figs=[ArticleFig(id=1225944427563233286, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, language=EN, label=Fig.1, caption=Overall framework of the network, figureFileSmall=j7zxIrkJvG42ncJ1EcY8Uw==, figureFileBig=Qk71uJJH02xgqTEIwL84kA==, tableContent=null), ArticleFig(id=1225944427672285201, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, language=CN, label=图1, caption=网络整体框架, figureFileSmall=j7zxIrkJvG42ncJ1EcY8Uw==, figureFileBig=Qk71uJJH02xgqTEIwL84kA==, tableContent=null), ArticleFig(id=1225944427835863079, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, language=EN, label=Fig.2, caption=Network structure, figureFileSmall=uAZYwuU4i5pWD6R0ori7HA==, figureFileBig=R67yc6C4FBhfajuIQFXC7Q==, tableContent=null), ArticleFig(id=1225944427982663736, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, language=CN, label=图2, caption=网络结构

FC为全连接层

, figureFileSmall=uAZYwuU4i5pWD6R0ori7HA==, figureFileBig=R67yc6C4FBhfajuIQFXC7Q==, tableContent=null), ArticleFig(id=1225944429425504331, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, language=EN, label=Fig.3, caption=Global sport feature module structure, figureFileSmall=Rl4D5HBSycpoTBvzG5wcDQ==, figureFileBig=375j9xBsJ1Zm5wBGgVna8Q==, tableContent=null), ArticleFig(id=1225944429693939809, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, language=CN, label=图3, caption=全局运动特征模块结构

K为特征图序号;RELU和Sigmoid为激活函数

, figureFileSmall=Rl4D5HBSycpoTBvzG5wcDQ==, figureFileBig=375j9xBsJ1Zm5wBGgVna8Q==, tableContent=null), ArticleFig(id=1225944429974958195, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, language=EN, label=Fig.4, caption=Local sport feature module structure, figureFileSmall=Ur3udOP9bDEBAlEDVPpCMQ==, figureFileBig=l7CM6+RLf57PZLT+xdgF0Q==, tableContent=null), ArticleFig(id=1225944430188867714, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, language=CN, label=图4, caption=局部运动特征模块结构, figureFileSmall=Ur3udOP9bDEBAlEDVPpCMQ==, figureFileBig=l7CM6+RLf57PZLT+xdgF0Q==, tableContent=null), ArticleFig(id=1225944430444720277, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, language=EN, label=Fig.5, caption=Temporal shift module, figureFileSmall=vXs1q4MEqvjq9uvugpOKjw==, figureFileBig=pOpmKKXgP1Ln0CK0X5XaLg==, tableContent=null), ArticleFig(id=1225944430646046887, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, language=CN, label=图5, caption=时间位移模块, figureFileSmall=vXs1q4MEqvjq9uvugpOKjw==, figureFileBig=pOpmKKXgP1Ln0CK0X5XaLg==, tableContent=null), ArticleFig(id=1225944430801236149, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, language=EN, label=Fig.6, caption=Partial visual inference graph, figureFileSmall=wIeVVgh5i80pT/ZJVc8/hQ==, figureFileBig=DzRpKjTrAfmK5nbVTUkvag==, tableContent=null), ArticleFig(id=1225944430927065284, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, language=CN, label=图6, caption=部分可视化推断图, figureFileSmall=wIeVVgh5i80pT/ZJVc8/hQ==, figureFileBig=DzRpKjTrAfmK5nbVTUkvag==, tableContent=null), ArticleFig(id=1225944431010951374, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, language=EN, label=Table 1, caption=

UCF101dataset information

, figureFileSmall=null, figureFileBig=null, tableContent=
信息 数值
运动 101
帧数 13 320
组数 25
每组帧数 4~7
视频时间/min 1 600
帧率/FPS 25
尺寸 320×240
), ArticleFig(id=1225944431115808982, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, language=CN, label=表1, caption=

UCF101数据集信息

, figureFileSmall=null, figureFileBig=null, tableContent=
信息 数值
运动 101
帧数 13 320
组数 25
每组帧数 4~7
视频时间/min 1 600
帧率/FPS 25
尺寸 320×240
), ArticleFig(id=1225944431258415325, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, language=EN, label=Table 2, caption=

Module ablation experiment

, figureFileSmall=null, figureFileBig=null, tableContent=
网络结构 空间流 时间流 双流融合
原结构 92.0 91.3 94.6
GSFM 93.5 92.6 95.3
LSFM 94.6 93.8 96.1
GSFM+LSFM 95.3 94.6 96.8
), ArticleFig(id=1225944431417798888, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, language=CN, label=表2, caption=

模块消融实验

, figureFileSmall=null, figureFileBig=null, tableContent=
网络结构 空间流 时间流 双流融合
原结构 92.0 91.3 94.6
GSFM 93.5 92.6 95.3
LSFM 94.6 93.8 96.1
GSFM+LSFM 95.3 94.6 96.8
), ArticleFig(id=1225944431556210930, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, language=EN, label=Table 3, caption=

Temporal shift module experiment

, figureFileSmall=null, figureFileBig=null, tableContent=
网络结构 空间流 时间流 双流融合
原结构 92.0 91.3 94.6
GSFM+TSM 94.9 94.3 96.6
GSFM+LSFM 95.3 94.6 96.8
), ArticleFig(id=1225944431682040063, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, language=CN, label=表3, caption=

时间位移模块实验

, figureFileSmall=null, figureFileBig=null, tableContent=
网络结构 空间流 时间流 双流融合
原结构 92.0 91.3 94.6
GSFM+TSM 94.9 94.3 96.6
GSFM+LSFM 95.3 94.6 96.8
), ArticleFig(id=1225944431761731849, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, language=EN, label=Table 4, caption=

Module connection experiment

, figureFileSmall=null, figureFileBig=null, tableContent=
网络结构 空间流 时间流 双流融合
模块串联1 94.9 94.0 96.5
模块串联2 94.7 93.9 96.3
模块并联 95.3 94.6 96.8
), ArticleFig(id=1225944431879172369, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, language=CN, label=表4, caption=

模块连接方式实验

, figureFileSmall=null, figureFileBig=null, tableContent=
网络结构 空间流 时间流 双流融合
模块串联1 94.9 94.0 96.5
模块串联2 94.7 93.9 96.3
模块并联 95.3 94.6 96.8
), ArticleFig(id=1225944432038555932, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, language=EN, label=Table 5, caption=

Model fusion method experiment

, figureFileSmall=null, figureFileBig=null, tableContent=
融合方式 准确率/%
最大值融合 96.3
平均融合 96.6
加权融合 96.8
), ArticleFig(id=1225944432189550886, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, language=CN, label=表5, caption=

模型融合方式实验

, figureFileSmall=null, figureFileBig=null, tableContent=
融合方式 准确率/%
最大值融合 96.3
平均融合 96.6
加权融合 96.8
), ArticleFig(id=1225944432386683194, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, language=EN, label=Table 6, caption=

Performance comparison of different methods

, figureFileSmall=null, figureFileBig=null, tableContent=
方法 参数量/
106
浮点数/
GFLOPs
UCF101 HMDB51
DVSR[18] 82.9 56.0
IABC[19] 86.2 62.4
SHT-CNN[4] 92.3 62.5
TSN[15] 21.4 32.0 94.0 68.5
TSM[5] 48.6 66.0 95.9 73.5
TS-D3D[20] 17.5 29.4 96.0 65.6
Timesformer[10] 121.4 590.0 92.1
MFI[21] 24.6 67.2 95.6
SVT[11] 173.2 590.0 93.7 67.2
X3D[7] 150.2 143.5 90.3 68.2
AR3D[8] 139.5 125.0 94.8 69.3
Eco[22] 150.0 267.0 94.8 72.4
本文方法 54.8 69.0 96.8 75.3
), ArticleFig(id=1225944433791775046, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1156949463970435804, language=CN, label=表6, caption=

不同方法性能对比

, figureFileSmall=null, figureFileBig=null, tableContent=
方法 参数量/
106
浮点数/
GFLOPs
UCF101 HMDB51
DVSR[18] 82.9 56.0
IABC[19] 86.2 62.4
SHT-CNN[4] 92.3 62.5
TSN[15] 21.4 32.0 94.0 68.5
TSM[5] 48.6 66.0 95.9 73.5
TS-D3D[20] 17.5 29.4 96.0 65.6
Timesformer[10] 121.4 590.0 92.1
MFI[21] 24.6 67.2 95.6
SVT[11] 173.2 590.0 93.7 67.2
X3D[7] 150.2 143.5 90.3 68.2
AR3D[8] 139.5 125.0 94.8 69.3
Eco[22] 150.0 267.0 94.8 72.4
本文方法 54.8 69.0 96.8 75.3
)], 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.1671-1815.2401484, detailUrlEn=https://castjournals.cast.org.cn/joweb/kxjsygc/EN/10.12404/j.issn.1671-1815.2401484, pdfUrlCn=https://castjournals.cast.org.cn/joweb/kxjsygc/CN/PDF/10.12404/j.issn.1671-1815.2401484, pdfUrlEn=https://castjournals.cast.org.cn/joweb/kxjsygc/EN/PDF/10.12404/j.issn.1671-1815.2401484, aliStartDate=null, aliEndDate=null, collectionFlag=false, citedCount=null, citedUrl=null, reference=null)
收藏切换
基于运动特征增强双流网络的视频行为识别
收藏切换
PDF下载
曹若琛 , 冯秀芳 , 赵晨 *
科学技术与工程 | 论文·自动化技术、计算机技术 2025,25(4): 1540-1546
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(4): 1540-1546
基于运动特征增强双流网络的视频行为识别
全屏
曹若琛 , 冯秀芳, 赵晨*
作者信息
  • 太原理工大学软件学院, 晋中 030600
  • 曹若琛(1990—),男,汉族,山西太原人,博士,讲师。研究方向:虚拟现实、行为识别。E-mail:

通讯作者:

*赵晨(1998—),男,汉族,辽宁沈阳人,硕士研究生。研究方向:行为识别。E-mail:
Video Action Recognition Based on Sport Feature Enhancement in Two-stream Network
Ruo-chen CAO , Xiu-fang FENG, Chen ZHAO*
Affiliations
  • School of Software, Taiyuan University of Technology, Jinzhong 030600, China
出版时间: 2025-02-08 doi: 10.12404/j.issn.1671-1815.2401484
文章导航
收藏切换
为解决目前行为识别中双流网络对运动特征提取的不充分导致识别准确度低的问题,通过提出基于运动特征增强双流网络的行为识别方法提高准确率。该网络分为空间流和时间流,空间流网络和时间流网络结构相同,输入不同。空间流网络输入为视频帧序列,而时间流网络输入为视频帧差序列。网络结构以Resnet50为骨干网络,将3×3卷积替换为所提出的全局运动特征模块和局部运动特征模块,充分提取视频运动信息,最终将空间流和时间流结合输出结果。结果表明:该模型在UCF101和HMDB51数据集上准确率达到96.8%和75.3%,与传统算法相比有一定优越性。
行为识别  /  深度学习  /  运动特征  /  双流网络

To solve the problem of insufficient extraction of sport features by dual stream networks in current action recognition, which leads to low recognition accuracy, a action recognition method based on sport feature enhancement two-stream networks was proposed to improve accuracy. The network was divided into spatial stream and temporal stream, with the same structure but different inputs. The input of the spatial stream network was a video frame sequence, while the input of the temporal stream network was a video frame difference sequence. The network structure used Resnet50 as the backbone network, replacing the 3×3 convolution with the proposed global sport feature module and local sport feature module, fully extracting video sport information, and finally combining spatial and temporal stream to output the results. The results show that the accuracy of the model on the UCF101 and HMDB51 datasets reaches 96.8% and 75.3%, which is superior to traditional algorithms.

action recognition  /  deep learning  /  sport feature  /  two-stream network
曹若琛, 冯秀芳, 赵晨. 基于运动特征增强双流网络的视频行为识别. 科学技术与工程, 2025 , 25 (4) : 1540 -1546 . DOI: 10.12404/j.issn.1671-1815.2401484
Ruo-chen CAO, Xiu-fang FENG, Chen ZHAO. Video Action Recognition Based on Sport Feature Enhancement in Two-stream Network[J]. Science Technology and Engineering, 2025 , 25 (4) : 1540 -1546 . DOI: 10.12404/j.issn.1671-1815.2401484
随着人工智能迅速发展,计算机视觉取得了一系列研究成果,具有时间维度的视频行为识别领域得到广泛关注[1-3]。当前,视频行为识别方法分为传统方法和深度学习方法。相比传统方法,深度学习方法准确率更高且泛化能力更强,但仍存在缺陷。深度学习方法主要有双流卷积网络[4-6],3D卷积网络[7-9]以及transformer网络[10-13]。双流卷积网络相比3D卷积网络和transformer网络复杂度更低,不易过拟合,因此主要围绕双流卷积进行。
双流卷积网络由Simonyan等[14]于2014年提出,该方法解决了早期二维卷积无法提取时间信息的问题,提出对空间信息和时间信息分别提取再进行融合的方法来提取视频特征。双流网络分为空间流网络和时间流网络,其中,空间流网络输入图像用来提取视频空间信息,时间流网络通过输入光流提取时间信息,最后对两个网络的结果进行融合。虽然这种早期双流网络通过融合时间流获取视频时间信息,但其不能进行长时序建模,准确率较低,因此目前视频行为识别主要应用TSN(temporal segment networks)[15]和TSM(temporal shift module)[5]这种支持长时序建模的双流网络。
TSN和TSM是经典的双流卷积模型。其中,TSN对长时序连续视频帧进行等分,接着对每个等分片段提取特征,最终对多个结果进行加权平均。TSM则是融合了时间位移模块,通过在时间维度移动像素获取时间特征。TSN和TSM都采用长视频序列中截取若干帧的方式解决了传统双流网络无法对长时序进行建模的问题。但目前的双流网络仍存在对运动特征信息提取不充分的问题。TSN网络分段采样策略提取特征,但网络并未高效提取运动特征。TSM在TSN基础上加入时间位移模块高效提取了局部运动特征,但缺乏全局运动特征建模。丁雪琴等[4]提出时空异构双流卷积网络去除冗余特征信息,但该方法仍未能高效提取运动特征。鉴于此,提出一种运动特征增强双流网络(sport feature enhancement two-stream network,SFETN),通过引入本文模块取得较高行为识别精度;设计一种全局运动特征模块(global sport feature module,GSFM),通过扩大特征图差异,高效提取视频全局运动特征;设计一种局部运动特征模块(local sport feature module,LSFM),通过改进时间位移操作,高效提取视频局部运动特征,在双流网络中实现对运动特征的充分提取以提高行为识别准确率,以期通过增强并融合视频全局和局部运动特征,解决双流网络模型特征提取不充分的问题。
基于双流卷积结构,提出运动特征增强双流网络,整体框架如图1所示。
为了能够实现对视频的长时序建模,本文网络借鉴TSN网络思想,对长视频序列进行间隔片段抽取,同时时间流输入使用帧差(视频帧相减)代替光流,避免数据预处理时光流抽取带来的高计算和长耗时。网络首先将一段长序列视频帧分为若干个视频片段,然后在每个视频片段中选择一帧RGB视频图像和相邻视频帧的帧差图像,RGB图像通过空间流网络进行空间特征提取,帧差图送入时间流网络进行时间特征提取,依次对每个视频片段都进行如上操作,这样就得到了若干个空间流网络分类结果和若干个时间流网络分类结果。对这些空间流网络分类结果做Segmental Consensus (一种共识) 得到空间流最终结果,对相应的时间流网络分类结果做Segmental Consensus得到时间流最终结果,最后将空间流分类结果和时间流分类结果作加权融合得到最终精度。本文网络建模方法如式(1)、式(2)所示。
$G=g\left[F\right({T}_{1};w),F({T}_{2};w),\dots,F({T}_{k};w\left)\right]$
$L(y,G)=-\sum _{i=1}^{c}{y}_{i}({G}_{i}-lg\sum _{j=1}^{c}exp{G}_{j})$
式中:$({T}_{1},{T}_{2},\dots,{T}_{k})$为视频片段;F(Tk;w)为参数为w的网络;g为融合函数,对同一类得分取平均值;G为段共识函数,结合交叉熵损失得到最终损失函数L;c为类别数;y为每类的值;yi为类别i的真实标签;Gi为类别i的分数;Gj为类别j的分数。
本文网络分为空间流网络和时间流网络,空间流网络和时间流网络结构相同,输入不同。空间流输入为视频帧序列,时间流输入为帧差序列。网络结构使用Resnet50为骨干网络,提出全局运动特征模块GSFM和局部运动特征模块LSFM替换掉原Resnet50网络模块中的3×3卷积,以更加充分地提取视频的运动特征,网络结构如图2所示。
空间流网络输入为视频帧序列,最初经由一个7×7的卷积和3×3的最大池化来进行特征提取并减少计算量去除冗余信息。随后通过若干层SFETN模块进行特征提取。由于本文网络使用Resnet50为骨干网络,因此SFETN模块的数量对应Resnet50网络中卷积模块的数量。每个SFETN模块的结构在图2中所示,特征图进入SFETN模块时,首先会经过1×1卷积减小维度,缩小计算量。然后分为两个分支,一个分支经过全局运动特征模块GSFM提取视频的全局运动特征,使得网络学习到视频的整体关联性;另一个分支经过局部运动特征模块LSFM提取视频局部运动特征,使得网络能够关注到视频的局部关键运动。将两个分支相加后,网络整体能够充分学习到视频的运动信息,从而提高行为识别的准确率。在经过若干SFETN模块提取特征后,最后经由FC层和Softmax函数得到空间流的分类结果。时间流输入为视频帧差序列,经过同样的网络结构提取相应的时间信息后,经由FC层和Softmax函数得到时间流的分类结果。最终将空间流分类结果和时间流分类结果加权融合得到最终结果,如式(3)所示。
$R=\sum _{i=1}^{n}\frac{x{A}_{i}+{y}_{time}{B}_{i}}{n}$
式(3)中:Ai为空间流每一类分数;Bi为时间流每一类分数;x为空间流权重;ytime为时间流权重;n为类别数;R为分类结果。
提出全局运动特征模块GSFM提取视频的全局运动特征,以解决目前网络对全局运动特征提取不充分的问题,其结构如图3所示。
全局运动特征可以通过特征图之间的整体差异进行实现。模块输入特征为X,特征尺寸为[N,T,C,H,W],其中,N为批处理视频帧数,T为时间维度数,C为通道数,HW分别为特征图的长度和宽度。特征图首先经过1×1卷积进行通道下采样,目的是减少后续计算量,可表示为
${X}^{m}=Conv\left(X\right), {X}^{m}\in {R}^{N\times T\times {C}_{1}\times H\times W}$
式(4)中:Conv为1×1卷积;Xm为通道下采样后的特征;C1为通道下采样后的通道数。
在通过通道下采样后,特征图会在时间维度上进行等时序分割,这里以KK+1示意时序分割后的两个特征图,进行特征帧作差处理。但是由于每个特征帧之间差异不够明显,为了增大相邻特征帧之间的差异,对相邻帧的后一帧进行通道增强后减去前一帧进行处理。具体为对于特征尺寸为C1×H×W的第K+1帧特征,首先使用最大池化,使用最大池化的目的是放大不同帧之间的全局运动特征差异,同时压缩空间特征,将特征尺寸变为C1×1×1,使得网络获得一个全局描述,可表示为
$z=\frac{1}{HW}=\sum _{i=1}^{H}\sum _{j=1}^{W}{Y}_{ijc}, c\in [1,{C}_{1}]$
式(5)中:YK+1帧特征图;z为输出结果。
特征图经过最大池化特征尺寸变为C1×1×1后,由于长和宽被压缩为1,会使网络更关注到通道级别的特征。之后使用一个FC层降维,以降低计算量并增强关键特征。之后使用RELU函数增强网络各层之间的非线性关系,提高模型表达能力。再使用FC层升维,还原特征信息,最后通过sigmoid函数特征归一化得到特征图D1对于特征尺寸为C1×H×W的第K帧特征,使用平均池化压缩空间特征,使得特征尺寸变为C1×1×1,得到特征图D2,这里使用平均池化而非最大池化的目的是让两帧相减后的全局运动特征差异达到最大,从而让网络能够更加明显地关注到运动差异信息。在得到D1D2后,将两特征图进行相减,得到全局运动特征D0,可表示为
D0=D1-D2
将所有时序分割后的特征图都进行如上操作,然后将得到的帧差特征图进行拼接,此时特征时间维度为T-1,为了保持时间维度和输入时相同,用零表示时间步结束的运动特征,再经过1×1卷积还原通道数为C,得到大小为[N,T,C,1,1]的特征D。为还原原特征图大小,将特征D与和最初的特征X进行相乘,得到最终的原特征图大小[N,T,C,H,W]的全局运动特征,最后使用3×3卷积对空间特征进行增强,进一步提升网络性能。
提出能够高效对局部运动特征建模的局部运动特征模块LSFM,使用位移算子重排列思想,沿时间维度移动像素,使得视频帧拥有时间信息,同时使用卷积增强空间特征,提升性能。
局部运动特征模块由改进时间位移模块和3×3卷积构成,如图4所示。改进时间位移模块借鉴了时间位移模块TSM,并在其基础上进行改进。传统网络在时间维度上独立进行特征提取,因此没有时间建模效果(图5左侧)。时间位移模块简单来说就是令当前帧混合其相邻帧的像素信息,使其能够感知到时间信息。以卷积核大小为3的卷积过程为例,设输入X为一维向量,权值W=(w1,w2,w3),则进行时间位移的卷积操作可以分为两个过程,首先沿时间维度移动特征,可表示为
${X}_{i}^{-1}={X}_{i-1},{X}_{i}^{0}={X}_{i},{X}_{i}^{+1}={X}_{i+1}$
进行乘法累加,乘以权重W
$Y={w}_{1}{X}^{-1}+{w}_{2}{X}^{0}+{w}_{3}{X}^{+1}$
上述过程为一维向量时间位移卷积过程。可以看到,在没有增加任何参数量和计算量的情况下,就使得特征图拥有了局部运动特征。图5左侧表示含有通道维度和时间维度的张量,其中每一行代表不同时间特征,之后进行时间位移后,如图5右侧所示,将部分通道沿时间维度进行移动,这样对于每一行时间特征,就拥有其他通道的时间信息,即有表示一定运动特征的能力,但由于仅移动少部分通道,因此只能表示局部运动特征。
本文改进时间位移模块受到TSM的启发,使用变化通道代替固定的移位通道,假设输入维度为[N,T,C,H,W],对于TSM移位过程,移位通道数固定为C/4,其他通道固定不动,可表示为
$\left\{\begin{array}{l}[N,t,C,H,W]=[N,t+1,C,H,W],\\ 1\le t\le T-\mathrm{1,1}\le c\le \frac{C}{8},下移\\ [N,t,C,H,W]=[N,t-1,C,H,W],\\ 2\le t\le T,\frac{C}{8}\le c\le \frac{C}{4},上移\\ [N,t,C,H,W]=[N,t,C,H,W],\\ 1\le t\le T,\frac{C}{4}\le c\le C,不动\end{array}\right.$
由于TSM移动通道固定,因此对于局部特征的关注过于单一,缺乏特征多样性,因此为了改进这个问题,所提出的改进时间位移模块将固定通道改进为变化的通道,将移动通道数随机设为$\frac{C}{4}$$\frac{C}{2}$,并将移动通道的选取也进行随机化,通过以上操作,能够让网络提取的局部特征信息更加全面,再经过时间移位后,通过3×3卷积提取空间特征进一步提升网络性能,最后使用残差连接方式,将此时特征与模块输入时的特征进行相加得到最终结果。
2.1 数据集
在UCF101[16]和HMDB51[17]上进行消融和对比实验。鉴于HMDB51数据集规模较小且质量有限,将UCF101数据集作为主要数据集进行消融实验,而将HMDB51作为辅助数据集仅用在对比实验中。UCF101数据集中视频取自YouTube网站,共包含13 320个视频和101个类别。这些类别包含人与物交互,人和人互动,体育,乐器演奏等。每种动作分为25组,每组包含4~7个含有相似特征的视频。所有视频共6.5 GB,分辨率为320×240,帧率为25 FPS。UCF101的详细信息如表1所示。
HMDB51数据集包含个人动作,多人动作,人和物交互等51个类别,共有6 849个视频。视频分辨率为320×240。数据集收集自YouTube和谷歌等视频网站。使用官方划分的数据集结构进行实验。使用训练集训练网络并在测试集上验证网络性能。
本实验在64位Ubuntu系统下实现,计算机配置为Intel(R) Xeon(R) Platinum 8255C@CPU 2.00 GHz,Nvidia GeForce RTX 3090 24 G 显卡,基于Python 3.8+Pytorch 1.7.0 + CUDA 11.0 + CUDNN 7.6实现。对空间流网络和时间流网络分开训练,之后对两个结果进行最大值融合。空间流网络使用Kinetics400作为预训练数据集,时间流网络使用ImageNet作为预训练数据集。长时序视频分割段数设置为7,batch size设置为16,学习率设置为0.001,每10个epoch衰减为原来的1/10。图像尺寸设置为224×224。模型一共训练30个epoch。
为验证本文全局运动特征模块GSFM和局部运动特征模块LSFM的作用,以及探究在什么情况下模型性能达到最佳,在UCF101上进行消融实验。
首先验证全局运动特征模块GSFM和局部运动特征模块LSFM是否能够高效提取视频的全局运动特征和局部运动特征。模型采用Resnet50为骨干网络,因此对比基础骨干网络为Resnet50,然后分别使用GSFM和LSFM对Resnet50中的3×3卷积进行替换,具体为全局运动特征模块替换3×3卷积、局部运动特征模块替换3×3卷积、以及全局运动特征模块和局部运动特征模块并联后替换3×3卷积,分别在空间流网络和时间流网络上进行实验,结果如表2所示。
表2可知,无论是空间流还是时间流,所提的全局运动特征模块和局部运动特征模块都能使得模型准确率得到提升,并且在同时加入两个模块后准确率最佳。
为验证本文改进时间位移模块的有效性,对局部运动特征模块结构分别试用原时间位移模块和本文时间位移模块,实验结果如表3所示。可以看出,使用所提出的改进时间位移模块更能捕捉局部运动特征,准确率更高。
对全局运动特征模块GSFM和局部运动特征模块LSFM的连接方式进行实验,目的是探究两个模块如何能够最大程度提升准确率,实验分3种情况,分别为全局运动特征模块串联局部运动特征模块(模块串联1)、局部运动特征模块串联全局运动特征模块(模块串联2)和全局运动特征模块并联局部运动特征模块(模块并联),实验结果如表4所示,可以看出,当使用并联方式连接模块时,网络性能最佳。
为进一步提升网络识别准确率,对空间流和时间流结果融合方式进行实验,分别实验3种融合方式,分别为最大值融合,平均融合以及加权融合,结果如表5所示,可以看出,网络在加权融合方式下准确率最高。
本文模型分别在UCF101和HMDB51数据集上进行实验,对比视频动作识别领域的其他方法,其中包括传统方法以及深度学习方法,结果如表6所示。
表6可知,本文网络在UCF101和HMDB51数据集上准确率分别达到96.8%和75.3%,优于其他网络。相对于传统方法(如DVSR、IABC等),本文准确率明显更高。相对于双流网络(如TSM、SHT-CNN等),本文模型使用时间流使用帧差代替光流,避免了光流抽取,同时在参数量和浮点数相差不过大的情况下,准确率更高。相对于3D网络(如X3D、Eco等)和Transformer网络(如Timesformer、SVT等),本文网络参数量和浮点数远低于这些网络,同时准确率更高。
为直观展示本文模型的性能优势,从UCF101数据集中抽取部分视频片段在TSN和本文模型上进行可视化推断,结果如图6所示。可以看出,TSN错误地将冲浪运动识别为皮划艇运动,而本文模型全部识别正确,性能更高。
所提出运动特征增强双流网络行为识别方法。设计全局运动特征模块GSFM和局部运动特征模块LSFM解决双流网络提取运动特征不足的问题并提升识别准确率。同时,模型使用帧差代替光流避免数据预处理阶段额外的计算和耗时。本文算法在视频数据集UCF101和HMDB51上的识别Top-1精度分别达到96.8%和75.3%,在行为识别任务上取得了显著优势。未来的工作将致力于进一步提升行为识别准确率,并尝试探索更轻量化的结构来满足实际应用需求。
  • 山西省重点研发计划(202102020101007)
参考文献 引证文献
排序方式:
[1]
袁首, 乔勇军, 苏航, 等. 基于深度学习的行为识别方法综述[J]. 微电子学与计算机, 2022, 39(8): 1-10.
Yuan Shou, Qiao Yongjun, Su Hang, et al. A review of behavior recognition methods based on deep learning[J]. Microelectronics and Computer Science, 2022, 39(8): 1-10.
[2]
Kong Y, Fu Y. Human action recognition and prediction: a survey[J]. International Journal of Computer Vision, 2022, 130(5): 1366-1401.
[3]
梁绪, 李文新, 张航宁. 人体行为识别方法研究综述[J]. 计算机应用研究, 2022, 39(3): 651-660.
Liang Xu, Li Wenxin, Zhang Hangning. Summary of research on human behavior recognition methods[J]. Computer Application Research, 2022, 39(3): 651-660.
[4]
丁雪琴, 朱轶昇, 朱浩华, 等. 基于时空异构双流卷积网络的行为识别[J]. 计算机应用与软件, 2022, 39(3): 154-158.
Ding Xueqin, Zhu Yisheng, Zhu Haohua, et al. Behavior recognition based on spatio-temporal heterogeneous double-stream convolution network[J]. Computer Applications and Software, 2022, 39(3): 154-158.
[5]
Lin J, Gan C, Han S. TSM: temporal shift module for efficient video understanding[C]// Proceedings of the IEEE/CVF International Conference on Computer vision. New York: IEEE, 2019: 7083-7093.
[6]
金博, 王景林, 刘泓铄, 等. 基于双注意力时间卷积网络的人体行为识别[J]. 计算机工程与设计, 2023, 44(1): 284-291.
Jin Bo, Wang Jinglin, Liu Hongyue, et al. Human behavior recognition based on dual attention time convolution network[J]. Computer Engineering and Design, 2023, 44(1): 284-291.
[7]
Feichtenhofer C. X3D: expanding architectures for efficient video recognition[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2020: 203-213.
[8]
Dong M, Fang Z, Li Y, et al. AR3D: attention residual 3D network for human action recognition[J]. Sensors, 2021, 21(5): 1656.
[9]
闫雨寒, 陈天, 刘忠育, 等. 基于双重注意力和3DResNet-BiLSTM行为识别方法[J]. 计算机应用与软件, 2023, 40(2): 192-196, 205.
Yan yuhan, Chen Tian, Liu Zhongyu, et al. Behavior recognition method based on dual attention and 3DResNet-BiLSTM[J]. Computer Applications and Software, 2023, 40(2): 192-196, 205.
[10]
Bertasius G, Wang H, Torresani L. Is space-time attention all you need for video understanding?[C]// International Conference of Machine Learning. New York: IEEE, 2021: DOI:10.48550/arXiv.2102.05095.
[11]
Ranasinghe K, Naseer M, Khan S, et al. Self-supervised video transformer[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2022: 2874-2884.
[12]
Liu Z, Ning J, Cao Y, et al. Video swin transformer[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2022: 3202-3211.
[13]
姬晓飞, 赵东阳. 人体检测与异常行为识别联合算法[J]. 科学技术与工程, 2023, 23(8): 3370-3378.
Ji Xiaofei, Zhao Dongyang. Joint algorithm for human detection and abnormal behavior recognition[J]. Science Technology and Engineering, 2023, 23(8): 3370-3378.
[14]
Simonyan K, Zisserman A. Two-stream convolutional networks for action recognition in videos[C]// Advances in Neural Information Processing Systems. New York: IEEE, 2014: 568-576.
[15]
Wang L, Xiong Y, Wang Z, et al. Temporal segment networks: towards good practices for deep action recognition[C]// European Conference on Computer Vision. Cham: Springer, 2016: 20-36.
[16]
Soomro K, Zamir A R, Shah M. UCF101: a dataset of 101 human actions classes from videos in the wild[J]. arXiv Preprint, 2012: arXiv:1212.0402.
[17]
Kuehne H, Jhuang H, Garrote E, et al. HMDB: a large video database for human motion recognition[C]// 2011 International Conference on Computer Vision. New York: IEEE, 2011: 2556-2563.
[18]
Demir U, Rawat Y S, Shah M. Tinyvirat:low-resolution video action recognition[C]//25th International Conference on Pattern Recognition ( ICPR).New York: IEEE, 2021: 7387-7394.
[19]
王琼, 王旭, 刘云麟, 等. 基于改进密集轨迹算法的人体行为识别[J]. 计算机仿真, 2022, 39(12): 284-289, 356.
Wang Qiong, Wang Xu, Liu Yunlin, et al. Human behavior recognition based on improved dense trajectory algorithm[J]. Computer Simulation, 2022, 39(12): 284-289, 356.
[20]
Yang M, Guo Y, Zhou F, et al. TS-D3D: a novel two-stream model for action recognition[C]// International Conference on Image Processing, Computer Vision and Machine Learning(ICICML). New York: IEEE, 2022: 179-182.
[21]
Bai S, Wang Q, Li X. MFI:multi-range feature interchange for video action recognition[C]//25th International Conference on Pattern Recognition(ICPR). New York: IEEE, 2021: 6664-6671.
[22]
Zolfaghari M, Singh K, Brox T. Eco: efficient convolutional network for online video understanding[C]// Proceedings of the European Conference on Computer Vision(ECCV). New York: IEEE, 2018: 695-712.
2025年第25卷第4期
PDF下载
339
136
引用本文
BibTeX
文章信息
doi: 10.12404/j.issn.1671-1815.2401484
  • 接收时间:2024-03-04
  • 首发时间:2025-07-29
  • 出版时间:2025-02-08
补充材料
相关文章
文章信息
作者
出版历史
  • 收稿日期:2024-03-04
  • 修回日期:2024-11-20
基金
山西省重点研发计划(202102020101007)
作者信息
    太原理工大学软件学院, 晋中 030600

通讯作者:

*赵晨(1998—),男,汉族,辽宁沈阳人,硕士研究生。研究方向:行为识别。E-mail:
参考文献
分享链接
https://castjournals.cast.org.cn/joweb/kxjsygc/CN/10.12404/j.issn.1671-1815.2401484
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