Article(id=1217789890234859578, tenantId=1146029695717560320, journalId=1146123166801305609, issueId=1217789884081820362, articleNumber=null, orderNo=null, doi=10.12404/j.issn.1671-1815.2406407, pmid=null, cstr=null, oa=null, hot=null, price=null, onlineType=0, articleFormat=0, articleType=null, articleTypeStr=research-article, receivedDate=1724601600000, receivedDateStr=2024-08-26, revisedDate=1744732800000, revisedDateStr=2025-04-16, acceptedDate=null, acceptedDateStr=null, onlineDate=1768273335274, onlineDateStr=2026-01-13, pubDate=1753632000000, pubDateStr=2025-07-28, doiRegisterDate=null, doiRegisterDateStr=null, onlineIssueDate=1768273335274, onlineIssueDateStr=2026-01-13, onlineJustAcceptDate=null, onlineJustAcceptDateStr=null, onlineFirstDate=null, onlineFirstDateStr=null, sourceXml=null, magXml=null, createTime=1768273335274, creator=13701087609, updateTime=1768273335274, updator=13701087609, issue=Issue{id=1217789884081820362, tenantId=1146029695717560320, journalId=1146123166801305609, year='2025', volume='25', issue='21', pageStart='8761', pageEnd='9209', issueExtLink='null', onlineDate='null', pubDate='null', beforeIssueId=null, nextIssueId=null, price=null, status=1, issueComplete=1, articleOrder=1, issueType=-1, specialIssue=null, createTime=1768273333807, creator=13701087609, updateTime=1768273602927, updator=13701087609, preIssue=null, nextIssue=null, ext={EN=IssueExt(id=1217791012932604619, tenantId=1146029695717560320, journalId=1146123166801305609, issueId=1217789884081820362, language=EN, specialIssueTitle=, coverIllustrator=null, specialIssueEditor=, specialIssueAbout=), CN=IssueExt(id=1217791012932604620, tenantId=1146029695717560320, journalId=1146123166801305609, issueId=1217789884081820362, language=CN, specialIssueTitle=, coverIllustrator=null, specialIssueEditor=, specialIssueAbout=)}, issueFiles=null}, startPage=9010, endPage=9017, ext={EN=ArticleExt(id=1217789890788507760, articleId=1217789890234859578, tenantId=1146029695717560320, journalId=1146123166801305609, language=EN, title=Face Detection Method Based on Improved MTCNN, columnId=1156262729162810294, journalTitle=Science Technology and Engineering, columnName=Papers·Automation and Computational Technology, runingTitle=null, highlight=null, articleAbstract=

Although the multi-task convolutional neural networks (MTCNN) face detection algorithm has achieved good results in some face recognition tasks, the accuracy of face detection needs to be improved in the face of some complex small-scale and multi-person face detection tasks. An improved MTCNN algorithm was proposed. Firstly, the intersection over union (IoU) threshold parameter was fine-tuned when creating the data set to classify face samples more accurately. Secondly, replacing the max pooling layer of the network with convolutional layers can improve network performance. Finally, the squeeze-excitation(SE) attention mechanism was introduced into the O-Net network to improve the feature expression ability of the network. The test results show that compared with the original MTCNN algorithm, the detection accuracy of the P-Net network and R-Net network of the improved algorithm has increased by 1%, and the detection accuracy of the O-Net network has increased by 0.5%. Moreover, the improved algorithm performs better in the actual face detection task.

, correspAuthors=Han ZHONG, 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=Tao LI, Han ZHONG), CN=ArticleExt(id=1217789894278169081, articleId=1217789890234859578, tenantId=1146029695717560320, journalId=1146123166801305609, language=CN, title=基于改进多任务级联卷积神经网络的人脸检测算法, 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=

多任务级联卷积神经网络(multi-task convolutional neural networks,MTCNN)人脸检测算法虽然在一些人脸检测任务上取得了不错的成绩,但是面对一些复杂的小尺度、多人脸检测任务,人脸检测的准确率仍有待提高。提出一种改进的MTCNN算法,首先,在制造数据集时微调交并比(intersection over union,IoU)阈值参数,来更精确地对人脸样本进行分类;其次,用卷积层替换网络的最大池化层,来提升网络性能;最后,在O-Net网络中引入SE(squeeze-excitation)注意力机制,提高网络的特征表达能力。试验结果表明:改进后的算法与原始MTCNN算法相比较,P-Net网络和R-Net网络检测准确率提高了1%,O-Net网络检测准确率提高了0.5%,且改进后的算法在实际人脸检测任务中表现更好。

, correspAuthors=钟寒, authorNote=null, correspAuthorsNote=
* 钟寒(1984—),女,汉族,山东济宁人,博士,副教授。研究方向:数据分析。E-mail:
, copyrightStatement=null, copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=+1zaPptbrWbEqesuYHBeOQ==, magXml=ZKtoHppO5Dqzz3TKPZvfvA==, pdfUrl=null, pdf=MhnL9jngqB/tUe6MXP0Rnw==, pdfFileSize=13845414, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=ZZiHmTkPC2534Dqdg86P+g==, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=JB2EHta1UZEMksh5dODFXw==, mapNumber=null, authorCompany=null, fund=null, authors=

李涛(1990—),男,汉族,河南郑州人,硕士研究生。研究方向:人脸识别。E-mail:

, authorsList=李涛, 钟寒)}, authors=[Author(id=1217860106927129592, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, orderNo=0, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=2022212138@stu.ppsuc.edu.cn, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1217860107078123520, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, authorId=1217860106927129592, language=EN, stringName=Tao LI, firstName=Tao, middleName=null, lastName=LI, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=College of Information & Cyber Security, People's Public Security University of China, Beijing 100038, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1217860107182981128, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, authorId=1217860106927129592, language=CN, stringName=李涛, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=中国人民公安大学信息网络安全学院, 北京 100038, bio={"content":"

李涛(1990—),男,汉族,河南郑州人,硕士研究生。研究方向:人脸识别。E-mail:

"}, bioImg=null, bioContent=

李涛(1990—),男,汉族,河南郑州人,硕士研究生。研究方向:人脸识别。E-mail:

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1217860105589146601, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, xref=null, ext=[AuthorCompanyExt(id=1217860105601729514, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, companyId=1217860105589146601, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=College of Information & Cyber Security, People's Public Security University of China, Beijing 100038, China), AuthorCompanyExt(id=1217860105605923820, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, companyId=1217860105589146601, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=中国人民公安大学信息网络安全学院, 北京 100038)])]), Author(id=1217860107333976083, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, orderNo=1, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=zhonghan@ppsuc.edu.cn, emailSecond=null, emailThird=null, correspondingAuthor=1, authorType=1, ext={EN=AuthorExt(id=1217860107472388125, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, authorId=1217860107333976083, language=EN, stringName=Han ZHONG, firstName=Han, middleName=null, lastName=ZHONG, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=*, address=College of Information & Cyber Security, People's Public Security University of China, Beijing 100038, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1217860107594022950, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, authorId=1217860107333976083, language=CN, stringName=钟寒, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=*, address=中国人民公安大学信息网络安全学院, 北京 100038, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1217860105589146601, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, xref=null, ext=[AuthorCompanyExt(id=1217860105601729514, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, companyId=1217860105589146601, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=College of Information & Cyber Security, People's Public Security University of China, Beijing 100038, China), AuthorCompanyExt(id=1217860105605923820, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, companyId=1217860105589146601, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=中国人民公安大学信息网络安全学院, 北京 100038)])])], keywords=[Keyword(id=1217860107799543862, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, language=EN, orderNo=1, keyword=MTCNN), Keyword(id=1217860107929567298, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, language=EN, orderNo=2, keyword=face detection), Keyword(id=1217860108072173646, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, language=EN, orderNo=3, keyword=SE attention mechanism), Keyword(id=1217860108256723027, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, language=CN, orderNo=1, keyword=MTCNN), Keyword(id=1217860108407717981, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, language=CN, orderNo=2, keyword=人脸检测), Keyword(id=1217860108516769892, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, language=CN, orderNo=3, keyword=SE注意力机制)], refs=[Reference(id=1217860113541546476, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, doi=null, pmid=null, pmcid=null, year=2024, volume=12, issue=null, pageStart=118406, pageEnd=118426, url=null, language=null, rfNumber=[1], rfOrder=0, authorNames=Ho H, Nguyen L, Le T, journalName=IEEE Access, refType=null, unstructuredReference=Ho H, Nguyen L, Le T, et al. Face detectionusing eigenfaces: a comprehensive review[J]. IEEE Access, 2024, 12: 118406-118426., articleTitle=Face detectionusing eigenfaces: a comprehensive review, refAbstract=null), Reference(id=1217860113658987003, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, doi=null, pmid=null, pmcid=null, year=2021, volume=21, issue=30, pageStart=12787, pageEnd=12795, url=null, language=null, rfNumber=[2], rfOrder=1, authorNames=赵立新, 邢润哲, 白银光, journalName=科学技术与工程, refType=null, unstructuredReference=赵立新, 邢润哲, 白银光, 等. 深度学习在目标检测的研究综述[J]. 科学技术与工程, 2021, 21(30): 12787-12795., articleTitle=深度学习在目标检测的研究综述, refAbstract=null), Reference(id=1217860113856119311, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, doi=null, pmid=null, pmcid=null, year=2021, volume=21, issue=30, pageStart=12787, pageEnd=12795, url=null, language=null, rfNumber=[2], rfOrder=2, authorNames=Zhao Lixin, Xing Runzhe, Bai Yinguang, journalName=Science Technology and Engineering, refType=null, unstructuredReference=Zhao Lixin, Xing Runzhe, Bai Yinguang, et al. Review on survey of deep learning in target detection[J]. Science Technology and Engineering, 2021, 21(30): 12787-12795., articleTitle=Review on survey of deep learning in target detection, refAbstract=null), Reference(id=1217860113977754143, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, doi=null, pmid=null, pmcid=null, year=2021, volume=null, issue=null, pageStart=6458, pageEnd=6467, url=null, language=null, rfNumber=[3], rfOrder=3, authorNames=Li J M, Xie H T, Li J H, journalName=Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition(CVPR), refType=null, unstructuredReference=Li J M, Xie H T, Li J H, et al. Frequency-aware discriminative feature learning supervised by single-center loss for face forgery detection[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition(CVPR). Los Alamitos: IEEE Computer Society Press, 2021: 6458-6467., articleTitle=Frequency-aware discriminative feature learning supervised by single-center loss for face forgery detection, refAbstract=null), Reference(id=1217860114170692154, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, doi=null, pmid=null, pmcid=null, year=2021, volume=null, issue=null, pageStart=7617, pageEnd=7627, url=null, language=null, rfNumber=[4], rfOrder=4, authorNames=Albiero V, Chen X Y, Yin X, journalName=Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition(CVPR), refType=null, unstructuredReference=Albiero V, Chen X Y, Yin X, et al. Img2pose: face alignment and detection via 6DoF, face pose estimation[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition(CVPR). Los Alamitos: IEEE Computer Society Press, 2021: 7617-7627., articleTitle=Img2pose: face alignment and detection via 6DoF, face pose estimation, refAbstract=null), Reference(id=1217860114351047252, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, doi=null, pmid=null, pmcid=null, year=2023, volume=59, issue=18, pageStart=1, pageEnd=13, url=null, language=null, rfNumber=[5], rfOrder=5, authorNames=张阳婷, 黄德启, 王东伟, journalName=计算机工程与应用, refType=null, unstructuredReference=张阳婷, 黄德启, 王东伟, 等. 基于深度学习的目标检测算法研究与应用综述[J]. 计算机工程与应用, 2023, 59(18): 1-13., articleTitle=基于深度学习的目标检测算法研究与应用综述, refAbstract=null), Reference(id=1217860114564956784, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, doi=null, pmid=null, pmcid=null, year=2023, volume=59, issue=18, pageStart=1, pageEnd=13, url=null, language=null, rfNumber=[5], rfOrder=6, authorNames=Zhang Yangting, Huang Deqi, Wang Dongwei, journalName=Computer Engineering and Applications, refType=null, unstructuredReference=Zhang Yangting, Huang Deqi, Wang Dongwei, et al. Review on research and application of deep learning-based target detection algorithms[J]. Computer Engineering and Applications, 2023, 59(18): 1-13., articleTitle=Review on research and application of deep learning-based target detection algorithms, refAbstract=null), Reference(id=1217860114858558084, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, doi=null, pmid=null, pmcid=null, year=2021, volume=null, issue=null, pageStart=1425, pageEnd=1434, url=null, language=null, rfNumber=[6], rfOrder=7, authorNames=Chaudhary B, Aghdaie P, Soleymani S, journalName=Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition(CVPR), refType=null, unstructuredReference=Chaudhary B, Aghdaie P, Soleymani S, et al. Differential morph face detection using discriminative wavelet sub-bands[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition(CVPR). Los Alamitos: IEEE Computer Society Press, 2021: 1425-1434., articleTitle=Differential morph face detection using discriminative wavelet sub-bands, refAbstract=null), Reference(id=1217860116003603086, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, doi=null, pmid=null, pmcid=null, year=2019, volume=null, issue=null, pageStart=168, pageEnd=177, url=null, language=null, rfNumber=[7], rfOrder=8, authorNames=Gao H X, Tao W, Wen D C, journalName=Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition(CVPR), refType=null, unstructuredReference=Gao H X, Tao W, Wen D C, et al. DupNet: towards very tiny quantized CNN with improved accuracy for face detection[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition(CVPR). Los Alamitos: IEEE Computer Society Press, 2019: 168-177., articleTitle=DupNet: towards very tiny quantized CNN with improved accuracy for face detection, refAbstract=null), Reference(id=1217860116137820826, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, doi=null, pmid=null, pmcid=null, year=2024, volume=38, issue=6, pageStart=194, pageEnd=202, url=null, language=null, rfNumber=[8], rfOrder=9, authorNames=徐铭, 李华, journalName=重庆理工大学学报(自然科学), refType=null, unstructuredReference=徐铭, 李华. 基于改进YOLOv5s-face的Face5系列人脸检测算法[J]. 重庆理工大学学报(自然科学), 2024, 38(6): 194-202., articleTitle=基于改进YOLOv5s-face的Face5系列人脸检测算法, refAbstract=null), Reference(id=1217860116293010088, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, doi=null, pmid=null, pmcid=null, year=2024, volume=38, issue=6, pageStart=194, pageEnd=202, url=null, language=null, rfNumber=[8], rfOrder=10, authorNames=Xu Ming, Li Hua, journalName=Journal of Chongqing University of Technology(Natural Science), refType=null, unstructuredReference=Xu Ming, Li Hua. Face5 series face detection algorithm based on improved YOLOv5s-face[J]. Journal of Chongqing University of Technology(Natural Science), 2024, 38(6): 194-202., articleTitle=Face5 series face detection algorithm based on improved YOLOv5s-face, refAbstract=null), Reference(id=1217860116481753786, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, doi=null, pmid=null, pmcid=null, year=2021, volume=null, issue=null, pageStart=1674, pageEnd=1684, url=null, language=null, rfNumber=[9], rfOrder=11, authorNames=Vesdapunt N, Wang B Y, journalName=Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition(CVPR), refType=null, unstructuredReference=Vesdapunt N, Wang B Y. Crface: confidence ranker for model-agnostic face detection refinement[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition(CVPR). Los Alamitos: IEEE Computer Society Press, 2021: 1674-1684., articleTitle=Crface: confidence ranker for model-agnostic face detection refinement, refAbstract=null), Reference(id=1217860116636943043, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, doi=null, pmid=null, pmcid=null, year=2024, volume=41, issue=11, pageStart=228, pageEnd=233, url=null, language=null, rfNumber=[10], rfOrder=12, authorNames=戴惠丽, journalName=计算机应用与软件, refType=null, unstructuredReference=戴惠丽. 融合深度字典学习和特征重建的遮挡人脸检测研究[J]. 计算机应用与软件, 2024, 41(11): 228-233., articleTitle=融合深度字典学习和特征重建的遮挡人脸检测研究, refAbstract=null), Reference(id=1217860116754383571, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, doi=null, pmid=null, pmcid=null, year=2024, volume=41, issue=11, pageStart=228, pageEnd=233, url=null, language=null, rfNumber=[10], rfOrder=13, authorNames=Dai huili, journalName=Computer Applications and Software, refType=null, unstructuredReference=Dai huili. An occlusion face detection method combining depth dictionary learning and feature reconstruction[J]. Computer Applications and Software, 2024, 41(11): 228-233., articleTitle=An occlusion face detection method combining depth dictionary learning and feature reconstruction, refAbstract=null), Reference(id=1217860116871824097, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, doi=null, pmid=null, pmcid=null, year=2021, volume=null, issue=null, pageStart=16195, pageEnd=16204, url=null, language=null, rfNumber=[11], rfOrder=14, authorNames=Wang W J, Yang W H, Liu J Y, journalName=Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition(CVPR), refType=null, unstructuredReference=Wang W J, Yang W H, Liu J Y. HLA-Face: joint high-low adaptation for low light face detection[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition(CVPR). Los Alamitos: IEEE Computer Society Press, 2021: 16195-16204., articleTitle=HLA-Face: joint high-low adaptation for low light face detection, refAbstract=null), Reference(id=1217860117031207665, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, doi=null, pmid=null, pmcid=null, year=2024, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=[12], rfOrder=15, authorNames=丁雅眉, journalName=复杂场景下的人脸检测方法研究, refType=null, unstructuredReference=丁雅眉. 复杂场景下的人脸检测方法研究[D]. 西安: 西安邮电大学, 2024., articleTitle=null, refAbstract=null), Reference(id=1217860117161231096, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=[12], rfOrder=16, authorNames=Ding Yamei, journalName=Research on face detection methods in complex Scenes, refType=null, unstructuredReference=Ding Yamei. Research on face detection methods in complex Scenes[D]. Xi'an: Xi'an University of Posts and Telecommunications, 2024., articleTitle=null, refAbstract=null), Reference(id=1217860117299643139, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, doi=null, pmid=null, pmcid=null, year=2023, volume=45, issue=10, pageStart=1838, pageEnd=1846, url=null, language=null, rfNumber=[13], rfOrder=17, authorNames=董子平, 陈世国, 廖国清, journalName=计算机工程与科学, refType=null, unstructuredReference=董子平, 陈世国, 廖国清. 基于YOLOv5s的密集多人脸检测算法[J]. 计算机工程与科学, 2023, 45(10): 1838-1846., articleTitle=基于YOLOv5s的密集多人脸检测算法, refAbstract=null), Reference(id=1217860117412889358, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, doi=null, pmid=null, pmcid=null, year=2023, volume=45, issue=10, pageStart=1838, pageEnd=1846, url=null, language=null, rfNumber=[13], rfOrder=18, authorNames=Dong Ziping, Chen Shiguo, Liao Guoqing, journalName=Computer Engineering and Science, refType=null, unstructuredReference=Dong Ziping, Chen Shiguo, Liao Guoqing. A dense multi-face detection algorithm based on YOLOv5s[J]. Computer Engineering and Science, 2023, 45(10): 1838-1846., articleTitle=A dense multi-face detection algorithm based on YOLOv5s, refAbstract=null), Reference(id=1217860117555495707, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, doi=null, pmid=null, pmcid=null, year=2016, volume=23, issue=10, pageStart=1499, pageEnd=1503, url=null, language=null, rfNumber=[14], rfOrder=19, authorNames=Zhang K, Zhang Z, Li Z, journalName=IEEE Signal Process, refType=null, unstructuredReference=Zhang K, Zhang Z, Li Z, et al. Joint face detection and alignment using multitask cascaded convolutional networks[J]. IEEE Signal Process, 2016, 23(10): 1499-1503., articleTitle=Joint face detection and alignment using multitask cascaded convolutional networks, refAbstract=null), Reference(id=1217860117710684970, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, doi=null, pmid=null, pmcid=null, year=2021, volume=40, issue=6, pageStart=58, pageEnd=63, url=null, language=null, rfNumber=[15], rfOrder=20, authorNames=刘晓军, 类成玉, 张晓玲, journalName=兵工自动化, refType=null, unstructuredReference=刘晓军, 类成玉, 张晓玲, 等. 基于卷积神经网络的人脸检测综述[J]. 兵工自动化, 2021, 40(6): 58-63., articleTitle=基于卷积神经网络的人脸检测综述, refAbstract=null), Reference(id=1217860117844902709, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, doi=null, pmid=null, pmcid=null, year=2021, volume=40, issue=6, pageStart=58, pageEnd=63, url=null, language=null, rfNumber=[15], rfOrder=21, authorNames=Liu Xiaojun, Lei Chengyu, Zhang Xiaoling, journalName=Ordnance Industry Automation, refType=null, unstructuredReference=Liu Xiaojun, Lei Chengyu, Zhang Xiaoling, et al. Face detection survey based on convolutional neural network[J]. Ordnance Industry Automation, 2021, 40(6): 58-63., articleTitle=Face detection survey based on convolutional neural network, refAbstract=null), Reference(id=1217860117949760323, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, doi=null, pmid=null, pmcid=null, year=2020, volume=39, issue=6, pageStart=637, pageEnd=641, url=null, language=null, rfNumber=[16], rfOrder=22, authorNames=蓝雯飞, 张盛兰, 朱容波, journalName=中南民族大学学报(自然科学版), refType=null, unstructuredReference=蓝雯飞, 张盛兰, 朱容波, 等. 基于改进MTCNN的人脸检测算法[J]. 中南民族大学学报(自然科学版), 2020, 39(6): 637-641., articleTitle=基于改进MTCNN的人脸检测算法, refAbstract=null), Reference(id=1217860118138504013, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, doi=null, pmid=null, pmcid=null, year=2020, volume=39, issue=6, pageStart=637, pageEnd=641, url=null, language=null, rfNumber=[16], rfOrder=23, authorNames=Lan Wenfei, Zhang Shenglan, Zhu Rongbo, journalName=Journal of South-Central University for Nationalities(Natural Science Edition), refType=null, unstructuredReference=Lan Wenfei, Zhang Shenglan, Zhu Rongbo, et al. Face detection algorithm based on improved MTCNN[J]. Journal of South-Central University for Nationalities(Natural Science Edition), 2020, 39(6): 637-641., articleTitle=Face detection algorithm based on improved MTCNN, refAbstract=null), Reference(id=1217860118293693267, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, doi=null, pmid=null, pmcid=null, year=2021, volume=35, issue=6, pageStart=114, pageEnd=120, url=null, language=null, rfNumber=[17], rfOrder=24, authorNames=顾梅花, 冯婧, 杨娜, journalName=西安工程大学学报, refType=null, unstructuredReference=顾梅花, 冯婧, 杨娜. 一种融合上下文信息特征的改进MTCNN人脸检测算法[J]. 西安工程大学学报, 2021, 35(6): 114-120., articleTitle=一种融合上下文信息特征的改进MTCNN人脸检测算法, refAbstract=null), Reference(id=1217860118570517349, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, doi=null, pmid=null, pmcid=null, year=2021, volume=35, issue=6, pageStart=114, pageEnd=120, url=null, language=null, rfNumber=[17], rfOrder=25, authorNames=Gu Meihua, Feng Jing, Yang Na, journalName=Journal of Xi'an Polytechnic University, refType=null, unstructuredReference=Gu Meihua, Feng Jing, Yang Na. Improved MTCNN face detection algorithm fused with context features[J]. Journal of Xi'an Polytechnic University, 2021, 35(6): 114-120., articleTitle=Improved MTCNN face detection algorithm fused with context features, refAbstract=null), Reference(id=1217860118704735088, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, doi=null, pmid=null, pmcid=null, year=2021, volume=11, issue=3, pageStart=172, pageEnd=176, url=null, language=null, rfNumber=[18], rfOrder=26, authorNames=周航, 蔡茂国, 吴涛, journalName=智能计算机与应用, refType=null, unstructuredReference=周航, 蔡茂国, 吴涛, 等. 一种改进的多任务级联网络人脸检测算法研究[J]. 智能计算机与应用, 2021, 11(3): 172-176., articleTitle=一种改进的多任务级联网络人脸检测算法研究, refAbstract=null), Reference(id=1217860118817981306, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, doi=null, pmid=null, pmcid=null, year=2021, volume=11, issue=3, pageStart=172, pageEnd=176, url=null, language=null, rfNumber=[18], rfOrder=27, authorNames=Zhou Hang, Cai Maoguo, Wu Tao, journalName=Intelligent Computer and Applications, refType=null, unstructuredReference=Zhou Hang, Cai Maoguo, Wu Tao, et al. Research on an improved multi-task cascade network face detection algorithm[J]. Intelligent Computer and Applications, 2021, 11(3): 172-176., articleTitle=Research on an improved multi-task cascade network face detection algorithm, refAbstract=null), Reference(id=1217860118943810435, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, doi=null, pmid=null, pmcid=null, year=2021, volume=38, issue=6, pageStart=1907, pageEnd=1912, url=null, language=null, rfNumber=[19], rfOrder=28, authorNames=庞殊杨, 芦莎, journalName=计算机应用研究, refType=null, unstructuredReference=庞殊杨, 芦莎. 基于改进MTCNN的多尺度安全帽识别[J]. 计算机应用研究, 2021, 38(6): 1907-1912, 1916., articleTitle=基于改进MTCNN的多尺度安全帽识别, refAbstract=null), Reference(id=1217860119103193997, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, doi=null, pmid=null, pmcid=null, year=2021, volume=38, issue=6, pageStart=1907, pageEnd=1912, url=null, language=null, rfNumber=[19], rfOrder=29, authorNames=Pang Shuyang, Lu Sha, journalName=Application Research of Computers, refType=null, unstructuredReference=Pang Shuyang, Lu Sha. Multi-scale safety helmet detection based on improved MTCNN[J]. Application Research of Computers, 2021, 38(6): 1907-1912, 1916., articleTitle=Multi-scale safety helmet detection based on improved MTCNN, refAbstract=null), Reference(id=1217860120038523797, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, doi=null, pmid=null, pmcid=null, year=2021, volume=18, issue=6, pageStart=111, pageEnd=118, url=null, language=null, rfNumber=[20], rfOrder=30, authorNames=刘彩云, 李雅雯, 刘倩, journalName=长江大学学报(自然科学版), refType=null, unstructuredReference=刘彩云, 李雅雯, 刘倩. 一种改进的多任务级联卷积神经网络人脸检测算法[J]. 长江大学学报(自然科学版), 2021, 18(6): 111-118., articleTitle=一种改进的多任务级联卷积神经网络人脸检测算法, refAbstract=null), Reference(id=1217860120244044708, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, doi=null, pmid=null, pmcid=null, year=2021, volume=18, issue=6, pageStart=111, pageEnd=118, url=null, language=null, rfNumber=[20], rfOrder=31, authorNames=Liu Caiyun, Li Yawen, Liu Qian, journalName=Journal of Yangtze University(Natural Science Edition), refType=null, unstructuredReference=Liu Caiyun, Li Yawen, Liu Qian. An improved face detection algorithm based on multi-task cascaded convolutional neural network[J]. Journal of Yangtze University(Natural Science Edition), 2021, 18(6): 111-118., articleTitle=An improved face detection algorithm based on multi-task cascaded convolutional neural network, refAbstract=null), Reference(id=1217860120416011177, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, doi=null, pmid=null, pmcid=null, year=2022, volume=35, issue=4, pageStart=651, pageEnd=655, url=null, language=null, rfNumber=[21], rfOrder=32, authorNames=李艳灵, 王莎莎, 杨志鹏, journalName=信阳师范学院学报(自然科学版), refType=null, unstructuredReference=李艳灵, 王莎莎, 杨志鹏. 一种改进的多任务级联卷积神经网络人脸检测算法[J]. 信阳师范学院学报(自然科学版), 2022, 35(4): 651-655., articleTitle=一种改进的多任务级联卷积神经网络人脸检测算法, refAbstract=null), Reference(id=1217860120634115002, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, doi=null, pmid=null, pmcid=null, year=2022, volume=35, issue=4, pageStart=651, pageEnd=655, url=null, language=null, rfNumber=[21], rfOrder=33, authorNames=Li Yanling, Wang Shasha, Yang Zhipeng, journalName=Journal of Xinyang Normal University(Natural Science Edition), refType=null, unstructuredReference=Li Yanling, Wang Shasha, Yang Zhipeng. An improved multi-task cascaded convolutional neural network face detection algorithm[J]. Journal of Xinyang Normal University(Natural Science Edition), 2022, 35(4): 651-655., articleTitle=An improved multi-task cascaded convolutional neural network face detection algorithm, refAbstract=null), Reference(id=1217860120797692865, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, doi=null, pmid=null, pmcid=null, year=2023, volume=51, issue=3, pageStart=298, pageEnd=304, url=null, language=null, rfNumber=[22], rfOrder=34, authorNames=李丽平, 许营坤, 王嘉航, journalName=浙江工业大学学报, refType=null, unstructuredReference=李丽平, 许营坤, 王嘉航. 基于锚框稀疏图像金字塔的MTCNN人脸检测方法[J]. 浙江工业大学学报, 2023, 51(3): 298-304, 318., articleTitle=基于锚框稀疏图像金字塔的MTCNN人脸检测方法, refAbstract=null), Reference(id=1217860120965465037, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, doi=null, pmid=null, pmcid=null, year=2023, volume=51, issue=3, pageStart=298, pageEnd=304, url=null, language=null, rfNumber=[22], rfOrder=35, authorNames=Li Liping, Xu Yingkun, Wang Jiahang, journalName=Journal of Zhejiang University of Technology, refType=null, unstructuredReference=Li Liping, Xu Yingkun, Wang Jiahang. Research on MTCNN face detection method of sparse image pyramid based on anchor box[J]. Journal of Zhejiang University of Technology, 2023, 51(3): 298-304, 318., articleTitle=Research on MTCNN face detection method of sparse image pyramid based on anchor box, refAbstract=null), Reference(id=1217860121095488468, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, doi=null, pmid=null, pmcid=null, year=2019, volume=68, issue=4, pageStart=3224, pageEnd=3233, url=null, language=null, rfNumber=[23], rfOrder=36, authorNames=Ma Z, Chang D, Xie J, journalName=EEE Transactions on Vehicular Technology, refType=null, unstructuredReference=Ma Z, Chang D, Xie J, et al. Fine-grained vehicle classification with channel max pooling modified CNNs[J]. EEE Transactions on Vehicular Technology, 2019, 68(4): 3224-3233., articleTitle=Fine-grained vehicle classification with channel max pooling modified CNNs, refAbstract=null), Reference(id=1217860121191957468, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, doi=null, pmid=null, pmcid=null, year=2014, volume=null, issue=null, pageStart=1412, pageEnd=null, url=null, language=null, rfNumber=[24], rfOrder=37, authorNames=Springenberg T J, Dosovitskiy A, Brox T, journalName=ArXiv, refType=null, unstructuredReference=Springenberg T J, Dosovitskiy A, Brox T, et al. Striving for simplicity: the all convolutional net[J]. ArXiv, 2014: 1412.6806., articleTitle=Striving for simplicity: the all convolutional net, refAbstract=null), Reference(id=1217860121321980901, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, doi=null, pmid=null, pmcid=null, year=2018, volume=null, issue=null, pageStart=11211, pageEnd=null, url=null, language=null, rfNumber=[25], rfOrder=38, authorNames=Woo S, Park J, Lee J Y, journalName=European Conference on Computer Vision(ECCV), refType=null, unstructuredReference=Woo S, Park J, Lee J Y. CBAM: convolutional block attention module[C]// European Conference on Computer Vision(ECCV). Munich: Springer, 2018: 11211., articleTitle=CBAM: convolutional block attention module, refAbstract=null), Reference(id=1217860121544279023, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, doi=null, pmid=null, pmcid=null, year=2018, volume=null, issue=null, pageStart=7132, pageEnd=7141, url=null, language=null, rfNumber=[26], rfOrder=39, authorNames=Hu J, Li S, Albanie S, journalName=Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition(CVPR), refType=null, unstructuredReference=Hu J, Li S, Albanie S, et al. Sueeze-and-excitation networks[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition(CVPR). Los Alamitos: IEEE Computer Society Press, 2018: 7132-7141., articleTitle=Sueeze-and-excitation networks, refAbstract=null), Reference(id=1217860121686885372, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, doi=null, pmid=null, pmcid=null, year=2023, volume=35, issue=7, pageStart=158, pageEnd=169, url=null, language=null, rfNumber=[27], rfOrder=40, authorNames=汪建伟, 游疆, 万敏, journalName=强激光与粒子束, refType=null, unstructuredReference=汪建伟, 游疆, 万敏, 等. 复杂背景下的低空无人机检测与跟踪算法[J]. 强激光与粒子束, 2023, 35(7): 158-169., articleTitle=复杂背景下的低空无人机检测与跟踪算法, refAbstract=null), Reference(id=1217860121804324871, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, doi=null, pmid=null, pmcid=null, year=2023, volume=35, issue=7, pageStart=158, pageEnd=169, url=null, language=null, rfNumber=[27], rfOrder=41, authorNames=Wang Jianwei, You Jiang, Wan Min, journalName=High Power Laser and Particle Beams, refType=null, unstructuredReference=Wang Jianwei, You Jiang, Wan Min, et al. Low-altitude UAV detection and tracking algorithms incomplex backgrounds[J]. High Power Laser and Particle Beams, 2023, 35(7): 158-169., articleTitle=Low-altitude UAV detection and tracking algorithms incomplex backgrounds, refAbstract=null)], funds=[Fund(id=1217860113306665425, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, awardId=B20087, language=CN, fundingSource=高等学校学科创新引智基地(B20087), fundOrder=null, country=null)], companyList=[AuthorCompany(id=1217860105589146601, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, xref=null, ext=[AuthorCompanyExt(id=1217860105601729514, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, companyId=1217860105589146601, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=College of Information & Cyber Security, People's Public Security University of China, Beijing 100038, China), AuthorCompanyExt(id=1217860105605923820, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, companyId=1217860105589146601, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=中国人民公安大学信息网络安全学院, 北京 100038)])], figs=[ArticleFig(id=1217860108906840182, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, language=EN, label=Fig.1, caption=MTCNN network structure, figureFileSmall=JYFBi+AIUaPJ8dEAG4ZEFg==, figureFileBig=MLFyuXivSyrXJ0HpjSCXjw==, tableContent=null), ArticleFig(id=1217860109028475013, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, language=CN, label=图1, caption=MTCNN网络结构, figureFileSmall=JYFBi+AIUaPJ8dEAG4ZEFg==, figureFileBig=MLFyuXivSyrXJ0HpjSCXjw==, tableContent=null), ArticleFig(id=1217860109284327581, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, language=EN, label=Fig.2, caption=Intersection and merging of two regions, figureFileSmall=L9IQl5H93F4/3CLD8ZUEOQ==, figureFileBig=bI34YqYjF2CrAOQvktiOwA==, tableContent=null), ArticleFig(id=1217860109435322540, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, language=CN, label=图2, caption=两个区域的交与并, figureFileSmall=L9IQl5H93F4/3CLD8ZUEOQ==, figureFileBig=bI34YqYjF2CrAOQvktiOwA==, tableContent=null), ArticleFig(id=1217860109552763070, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, language=EN, label=Fig.3, caption=Example of maximum pooling, figureFileSmall=eHNKyMXQmpDBRnJS3tkIeA==, figureFileBig=KQnWbO7C0JsCHYheybtKvQ==, tableContent=null), ArticleFig(id=1217860109670203595, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, language=CN, label=图3, caption=最大池化运算示例, figureFileSmall=eHNKyMXQmpDBRnJS3tkIeA==, figureFileBig=KQnWbO7C0JsCHYheybtKvQ==, tableContent=null), ArticleFig(id=1217860109796032727, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, language=EN, label=Fig.4, caption=SE module, figureFileSmall=onxwK/kxnUo5lZMGpCMWDA==, figureFileBig=DzQuAuxJlQPjYcVYqJzoTw==, tableContent=null), ArticleFig(id=1217860109980582122, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, language=CN, label=图4, caption=SE模块, figureFileSmall=onxwK/kxnUo5lZMGpCMWDA==, figureFileBig=DzQuAuxJlQPjYcVYqJzoTw==, tableContent=null), ArticleFig(id=1217860110102216954, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, language=EN, label=Fig.5, caption=Optimized MTCNN network structure, figureFileSmall=PveQOa7ho0RdC8B6X5gNQg==, figureFileBig=c4GirF4AZrjreLBYGukHPQ==, tableContent=null), ArticleFig(id=1217860111473754377, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, language=CN, label=图5, caption=优化后的MTCNN网络结构, figureFileSmall=PveQOa7ho0RdC8B6X5gNQg==, figureFileBig=c4GirF4AZrjreLBYGukHPQ==, tableContent=null), ArticleFig(id=1217860111616360726, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, language=EN, label=Fig.6, caption=The accuracy of the P-Net, figureFileSmall=fZBgEQjTpql9NXuhChkJyA==, figureFileBig=Wioe2OmKV5JWq1+REchu0A==, tableContent=null), ArticleFig(id=1217860111834464549, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, language=CN, label=图6, caption=P-Net网络准确率, figureFileSmall=fZBgEQjTpql9NXuhChkJyA==, figureFileBig=Wioe2OmKV5JWq1+REchu0A==, tableContent=null), ArticleFig(id=1217860111977070904, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, language=EN, label=Fig.7, caption=The accuracy of the R-Net, figureFileSmall=NUhei2kZ62T7ISfZvp/sdg==, figureFileBig=nxaV6gofpZbXwAzbmv8Y4Q==, tableContent=null), ArticleFig(id=1217860112295838029, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, language=CN, label=图7, caption=R-Net网络准确率, figureFileSmall=NUhei2kZ62T7ISfZvp/sdg==, figureFileBig=nxaV6gofpZbXwAzbmv8Y4Q==, tableContent=null), ArticleFig(id=1217860112421667162, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, language=EN, label=Fig.8, caption=The accuracy of the O-Net, figureFileSmall=LoXRiG3tBo+093eGGrFuhg==, figureFileBig=LwPHQ3hUAFGvcUaSbrGNdA==, tableContent=null), ArticleFig(id=1217860112585245034, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, language=CN, label=图8, caption=O-Net网络准确率, figureFileSmall=LoXRiG3tBo+093eGGrFuhg==, figureFileBig=LwPHQ3hUAFGvcUaSbrGNdA==, tableContent=null), ArticleFig(id=1217860112715268473, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, language=EN, label=Fig.9, caption=Detection results of the original MTCNN algorithm, figureFileSmall=6ni/Z37wHqx8Uymlg6dK9A==, figureFileBig=asY6k8wxStEk6KKnH3lwBw==, tableContent=null), ArticleFig(id=1217860112845291914, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, language=CN, label=图9, caption=原始MTCNN算法模型检测结果, figureFileSmall=6ni/Z37wHqx8Uymlg6dK9A==, figureFileBig=asY6k8wxStEk6KKnH3lwBw==, tableContent=null), ArticleFig(id=1217860112954343835, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, language=EN, label=Fig.10, caption=Detection results of the model in this paper, figureFileSmall=W7Djh3smpVZt3nu/UJmDfw==, figureFileBig=M1jxlrZoedLY82PPrpEbSw==, tableContent=null), ArticleFig(id=1217860113075978668, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1217789890234859578, language=CN, label=图10, caption=本文模型检测结果, figureFileSmall=W7Djh3smpVZt3nu/UJmDfw==, figureFileBig=M1jxlrZoedLY82PPrpEbSw==, tableContent=null)], 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.1671-1815.2406407, detailUrlEn=https://castjournals.cast.org.cn/joweb/kxjsygc/EN/10.12404/j.issn.1671-1815.2406407, pdfUrlCn=https://castjournals.cast.org.cn/joweb/kxjsygc/CN/PDF/10.12404/j.issn.1671-1815.2406407, pdfUrlEn=https://castjournals.cast.org.cn/joweb/kxjsygc/EN/PDF/10.12404/j.issn.1671-1815.2406407, aliStartDate=null, aliEndDate=null, collectionFlag=false, citedCount=null, citedUrl=null, reference=null)
收藏切换
基于改进多任务级联卷积神经网络的人脸检测算法
收藏切换
PDF下载
李涛 , 钟寒 *
科学技术与工程 | 论文·自动化技术、计算机技术 2025,25(21): 9010-9017
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(21): 9010-9017
基于改进多任务级联卷积神经网络的人脸检测算法
全屏
李涛 , 钟寒*
作者信息
  • 中国人民公安大学信息网络安全学院, 北京 100038
  • 李涛(1990—),男,汉族,河南郑州人,硕士研究生。研究方向:人脸识别。E-mail:

通讯作者:

* 钟寒(1984—),女,汉族,山东济宁人,博士,副教授。研究方向:数据分析。E-mail:
Face Detection Method Based on Improved MTCNN
Tao LI , Han ZHONG*
Affiliations
  • College of Information & Cyber Security, People's Public Security University of China, Beijing 100038, China
出版时间: 2025-07-28 doi: 10.12404/j.issn.1671-1815.2406407
文章导航
收藏切换

多任务级联卷积神经网络(multi-task convolutional neural networks,MTCNN)人脸检测算法虽然在一些人脸检测任务上取得了不错的成绩,但是面对一些复杂的小尺度、多人脸检测任务,人脸检测的准确率仍有待提高。提出一种改进的MTCNN算法,首先,在制造数据集时微调交并比(intersection over union,IoU)阈值参数,来更精确地对人脸样本进行分类;其次,用卷积层替换网络的最大池化层,来提升网络性能;最后,在O-Net网络中引入SE(squeeze-excitation)注意力机制,提高网络的特征表达能力。试验结果表明:改进后的算法与原始MTCNN算法相比较,P-Net网络和R-Net网络检测准确率提高了1%,O-Net网络检测准确率提高了0.5%,且改进后的算法在实际人脸检测任务中表现更好。

MTCNN  /  人脸检测  /  SE注意力机制

Although the multi-task convolutional neural networks (MTCNN) face detection algorithm has achieved good results in some face recognition tasks, the accuracy of face detection needs to be improved in the face of some complex small-scale and multi-person face detection tasks. An improved MTCNN algorithm was proposed. Firstly, the intersection over union (IoU) threshold parameter was fine-tuned when creating the data set to classify face samples more accurately. Secondly, replacing the max pooling layer of the network with convolutional layers can improve network performance. Finally, the squeeze-excitation(SE) attention mechanism was introduced into the O-Net network to improve the feature expression ability of the network. The test results show that compared with the original MTCNN algorithm, the detection accuracy of the P-Net network and R-Net network of the improved algorithm has increased by 1%, and the detection accuracy of the O-Net network has increased by 0.5%. Moreover, the improved algorithm performs better in the actual face detection task.

MTCNN  /  face detection  /  SE attention mechanism
李涛, 钟寒. 基于改进多任务级联卷积神经网络的人脸检测算法. 科学技术与工程, 2025 , 25 (21) : 9010 -9017 . DOI: 10.12404/j.issn.1671-1815.2406407
Tao LI, Han ZHONG. Face Detection Method Based on Improved MTCNN[J]. Science Technology and Engineering, 2025 , 25 (21) : 9010 -9017 . DOI: 10.12404/j.issn.1671-1815.2406407
人脸检测是识别视频或图像中是否存在人脸的过程,如果有人脸存在,则进一步检测出人脸的位置、大小和人脸面部关键点的位置信息[1]。早期的人脸检测工作主要基于人工精心设计的局部描述进行特征提取[2-5],但随着深度学习的蓬勃发展,逐渐演变出许多基于不同深度学习架构的人脸检测方法,极大地改善了传统人脸检测在特征提取、精确度、可扩展性方面的诸多不足[6-9]。基于卷积神经网络的人脸检测最早可以追溯到1994年,随着研究的深入,基于卷积神经网络(convolutional neural networks,CNN)的人脸检测相对于其他人脸检测方法取得了更加优异的成绩,但在遮挡[10]、光线[11]、复杂背景[12-13]等情况下的人脸检测效果仍不理想。
Zhang等[14]在2016年ECCV会议上提出了一种多任务级联卷积神经网络(multi-task convolutional neural networks,MTCNN),MTCNN网络结构较小,效率高,在不受约束的环境下兼顾了人脸检测和人脸对齐功能,可应用于人脸支付、安防监控、社交平台等多个领域,成为工业界的人脸检测主流算法[15]。但研究表明,MTCNN网络在面对一些复杂的检测任务时,检测准确率仍有待提高,为解决这一问题,一些改进的MTCNN网络被相继提出。蓝雯飞等[16]通过对原始MTCNN算法进行迁移学习,使算法在自然环境中人脸检测的准确率有所提高。顾梅花等[17]提出集成上下文信息卷积模块,对多任务卷积神经网络MTCNN的R-Net和O-Net网络模块进行优化,有效提升了算法在课堂场景下的小人脸检测率。周航等[18]通过对MTCNN算法损失函数进行均方差函数求均值,使网络收敛更加平稳,检测准确率更高。庞殊杨等[19]通过将MobileNet模块融入MTCNN算法中,并适当调整网络卷积核和全连接层神经元个数使模型更适用于不同尺寸的安全帽识别。刘彩云等[20]通过设计一种新的集成图片信息卷积残差模块,并将其融入R-Net和O-Net网络层中来增大感受野,从而进一步提升了网络性能。李艳灵等[21]将多任务卷积神经网络MTCNN的R-Net和O-Net网络模块的NMS算法优化成Better-NMS算法,在不影响网络鲁棒性的同时,提升了人脸检测的准确率。李丽平等[22]采用锚框思路和深度可分离卷积思路,优化了多任务卷积神经网络MTCNN,进一步提升了模型性能。
基于上述研究,现从训练数据和MTCNN网络结构两个方面入手进行优化,进一步提升模型性能,使模型在面对复杂的小尺度、多人脸检测任务时,能够准确无误的识别出图像或视频中的全部人脸。首先,在制造数据集时微调交并比(intersection over union,IoU)阈值参数,来更精确地对人脸样本进行分类;其次,用卷积层替换网络的最大池化层来提升网络性能;最后,在O-Net网络中引入SE注意力机制,提高模型的特征表达能力。
MTCNN网络是一种高效的人脸检测算法,通过级联的3个CNN网络(P-Net、R-Net和O-Net)实现了人脸的快速、准确检测。MTCNN主要由3个框架组成,如图1所示。
P-Net(proposal network)12×12:主要负责获得人脸框的回归向量和人脸5个关键点位置,通过非极大值抑制(non-maximum suppression,NMS)来合并重叠度高的候选框。
R-Net(refine network)24×24:通过边界框回归和NMS进一步去掉非人脸区域,由于R-Net网络与P-Net相比较多了一个全连接层,因此能更好地抑制非人脸区域。
O-Net(output network)48×48:与R-Net相比较,O-Net多了一层卷积层,处理结果更加精准,同时对人脸区域进行更多监督,并输出一个人脸框和5个面部关键点坐标。
MTCNN的设计理念是通过不同网络层级之间的级联,利用低复杂度的模型快速生成候选框,再利用高复杂度的模型对候选框进行筛选,从而实现速度与精度的共同提升。MTCNN不仅能够进行人脸检测,还能够输出人脸关键点位置,可应用于人脸识别、人脸表情分析、人脸美化等多个领域,成为工业界的人脸检测主流算法。
采用CelebA数据集作为基准数据集生成各个子网络训练所需的样本。CelebA数据集是一个大型的人脸属性数据集,由香港中文大学多媒体实验室创建,包括10 177个身份、202 599张面部图像、5个地标位置以及每张图片40个二进制属性注释。
以CelebA数据集为基准生成样本后,需将样本分为正样本、部分样本、负样本对网络进行训练,使用NMS算法中交并比(IoU)对样本进行分类。IoU的本质是真实的人脸框与用于训练的人脸框之间的重叠值,示意图如图2所示。
IoU计算公式为
I o U = A B A B = A B A + B - A B
式(1)中:A为真实的人脸框的面积;B为用于训练的人脸框的面积;∪表示两者的并集;∩表示两者的交集。IoU数值小则重合度低,IoU数值大则重合度高,用于训练的样本的类别可根据IoU的大小去判断。
原文IoU阈值分别设置为0.65、0.4、0.3,当IoU>0.65时,代表样本为正样本;0.4<IoU<0.65时,代表样本为部分样本;IoU<0.3时,代表样本为负样本。而IoU阈值的设置则直接决定着样本的保留或删除,决定着样本标签的属性,若阈值设置不合理,则无法精准地划分样本类别,进而影响模型训练效果。经过反复实验证明:当IoU>0.7时,为正样本;当0.3<IoU<0.7时,为部分样本;当IoU<0.2为负样本。样本分类更准确,且将负样本、正样本、部分样本、坐标样本比例设置为7∶1∶1∶1时,模型训练效果更佳。
池化层(pooling layers)是卷积神经网络中的一种层结构,一般使用于卷积层与卷积层之间,主要作用是增大感受野、降低特征维度、防止过拟合、缩减模型大小、提高计算速度[23]。本文模型在连续卷积层中间使用最大池化来降低特征图大小,提取特征图中最重要的特征,最大池化运算示例如图3所示。可知,最大池化虽然降低了特征图大小,减少了后续层的参数量,但是原特征图将近8/9的信息丢失,严重影响了网络的性能。所以在目标检测任务中,使用最大池化层会导致大量信息丢失从而影响网络检测准确率[24],为解决这一问题,使用卷积层替换网络中的最大池化层,来提升网络的目标检测精度。
深度学习中的注意力机制可以模仿人类视觉和认知系统,它在处理输入数据时,可以模仿人类的神经系统,把注意力集中在重要信息上,通过引入注意力机制,神经网络能够像人类一样,拥有自主学习能力,并选择性地学习重要信息,提高网络的性能和泛化能力[25]。为了进一步提高网络检测精准度,在O-Net网络的第一次卷积运算后引入SE[26]注意力机制,提高网络的特征表达能力。
SE注意力机制通过自主学习的方式赋予有用特征通道高的权重,赋予无用特征通道低的权重,以此来获取更多有价值信息[27]。SE注意力机制主要采用压缩和激励两种操作来实现所述功能,模块结构如图4所示。
步骤1 Transformation(Ftr):给定一个输入的特征图X,通过Ftr操作,生成特征图U,公式如下所示。
$\boldsymbol{U}=\boldsymbol{v}_{\mathrm{c}} * \boldsymbol{X}=\sum_{s=1}^{C^{\prime}} \boldsymbol{v}_{\mathrm{c}}^{s} * \boldsymbol{X}^{s}$
式(2)中:X为输入特征图; ν c为二维空间卷积核;*表示卷积操作。
步骤2 Squeeze(Fsq):通过Fsq操作,将特征图U压缩为一个1×1×C的向量,公式如下所示。
$\boldsymbol{Z}_{\mathrm{c}}=F_{\mathrm{sq}}(\boldsymbol{U})=\frac{1}{H W} \sum_{i=1}^{H} \sum_{j=1}^{W} \boldsymbol{u}_{\mathrm{c}}(i, j)$
式(3)中:U为特征图;C为特征图通道数;HW为特征图空间维度; u c为单个通道的特征图。
步骤3 Exciation(Fex):对上一步得到的向量通过两个全连接层进行处理,得到通道权重值,赋予不同通道不同的权重,用不同的权重值表示不同通道的重要程度,公式如下所示。
$\begin{aligned} S & =F_{\mathrm{ex}}(\boldsymbol{Z}, \boldsymbol{W})=\sigma[g(\boldsymbol{Z}, \boldsymbol{W})] \\ & =\sigma\left[\boldsymbol{W}_{2} \delta\left(\boldsymbol{W}_{1}, \boldsymbol{Z}\right)\right] \end{aligned}$
式(4)中:Z为特征图;W1W2为两个全连接层的权重矩阵;δ为ReLU激活函数;σ为Sigmoid激活函数。
步骤4 Scale(Fscale):将上一步得到的权重向量对特征图U进行权重赋值,完成对原始特征的重标定,生成最终的特征图,公式如下所示。
$\boldsymbol{X}_{\mathrm{c}}=F_{\text {scale }}\left(\boldsymbol{u}_{\mathrm{c}}, \boldsymbol{s}_{\mathrm{c}}\right)=\boldsymbol{s}_{\mathrm{c}} \boldsymbol{u}_{\mathrm{c}}$
式(5)中: u c为单个通道特征图;sc为通道权重。
优化后的MTCNN网络结构如图5所示。
实验环境为Windows11,搭载Inter Core i7 12700H处理器,内存为16 G,采用Nvidia RTX 3060 laptop,显存为6 G,采用Python3.7进行代码编译。
由于MTCNN网络由3个子网络组成,因此其损失函数也由3部分组成。针对人脸识别问题,利用交叉熵损失函数来判断是否存在人脸,对于人脸框回归和人脸关键点位置定位,使用L2损失函数,最后将这3个损失函数的损失和自身的权重相乘然后累加起来,形成最后的总损失。在P-Net和R-Net训练的过程中,更加注重目标框的准确度,而较少关注关键点判定的损失,因此目标框损失所占权重较大,而关键点损失所占权重较小。对于O-Net的训练则恰恰相反,更加注重关键点的位置,因此关键点损失所占权重较大,目标框损失所占权重较小。
对于人脸识别的问题,将输入的样本xi,使用交叉熵代价函数,公式为
$L_{i}^{\mathrm{det}}=-\left[y_{i}^{\mathrm{det}} \ln p_{i}+\left(1-y_{i}^{\mathrm{det}}\right)\left(1-\ln p_{i}\right)\right]$
式(6)中: y i d e t为样本的真实标签;pi为样本为人脸的概率。
对于目标框的回归,采用欧氏距离,公式为
$L_{i}^{\text {box }}=\left\|\hat{\boldsymbol{y}}_{i}^{\text {box }}-\boldsymbol{y}_{i}^{\text {box }}\right\|_{2}^{2}$
式(7)中: y ^ i b o x为通过校正之后得到的边界框的坐标; y i b o x为目标的真实边界框。
对于关键点,也采用的是欧氏距离,公式为
$L_{i}^{\text {land }}=\left\|\hat{\boldsymbol{y}}_{i}^{\text {land }}-\boldsymbol{y}_{i}^{\text {land }}\right\|_{2}^{2}$
式(8)中: y ^ i l a n d为网络输出之后校正得到的边界框的坐标; y i l a n d为目标的真实边界框。
把上述3个损失函数按照不同的权重联合起来,得
$\min \sum_{i=1}^{N} \sum_{j \in\{\mathrm{det}, \text { box, land }\}} \alpha_{j} \beta_{i}^{j} L_{i}^{j}$
式(9)中:N为训练样本的总数; β i j∈{0,1}为样本类型指示符;αj为各个损失所占的权重。在训练P-Net和R-Net网络时,把αdetαboxαland 分别设置为1、0.5、0.5,在训练O-Net网络时,把αdetαboxαland 分别设置为1、0.5和1。
MTCNN网络采用级联的思想一步步筛选出符合标准的样本,将一个大的网络拆解成3个小型网络,即P-Net、R-Net、O-Net。在模型训练过程中,需要先用大小为12×12的数据训练P-Net网络,然后利用训练好的P-Net网络生成训练R-Net网络所需的大小为24×24的数据。生成的大小为24×24的数据将R-Net网络训练成功以后,利用训练好的R-Net网络生成训练O-Net网络所需的大小为48×48的数据,最后用生成的大小为48×48的数据训练O-Net网络。
在分别制作P-Net、R-Net、O-Net数据时,用 I o U阈值对数据进行分类,当 I o U>0.7时,为正样本;当0.3< I o U<0.7时,为部分样本;当 I o U<0.2时,为负样本。分类完成后,将负样本、正样本、部分样本、坐标样本比例设置为7∶1∶1∶1。
在训练P-Net网络时,先用微调后的 I o U阈值对数据进行分类,并将负样本、正样本、部分样本、坐标样本比例设置为7∶1∶1∶1,然后用普通卷积替代P-Net网络的最大池化层,用优化后的数据训练优化后的P-Net网络,实验结果如图6所示,可知,P-Net网络在原始MTCNN算法训练过程中准确率保持在92.5%~93%,而本文算法P-Net网络在训练过程中准确率保持94%,相比原始MTCNN算法准确率更高。
在训练R-Net网络时,同样先用微调后的 I o U阈值对数据进行分类,并将负样本、正样本、部分样本、坐标样本比例设置为7∶1∶1∶1,然后用普通卷积替代R-Net网络的最大池化层,用优化后的数据训练优化后的R-Net网络,实验结果如图7所示,可知,原始MTCNN算法R-Net网络在训练过程中准确率保持在94%~95%,而本文算法R-Net网络在训练过程中准确率保持在95%~96%,相比原始MTCNN算法准确率更高。
在训练O-Net网络时,同样先用微调后的 I o U阈值对数据进行分类,并将负样本、正样本、部分样本、坐标样本比例设置为7∶1∶1∶1,然后用普通卷积替代O-Net网络的最大池化层,并在第一层卷积后加入SE注意力机制,最后用优化后的数据训练优化后的O-Net网络。实验结果如图8所示。可知,原始MTCNN算法O-Net网络在训练过程中准确率保持在97%~97.5%,而本文算法O-Net网络在训练过程中准确率保持97.5%~98%,相比原始MTCNN算法的准确率更高。
把用原始MTCNN算法数据训练好的模型和本文模型训练好的模型进行多人脸检测,来对比优化前后模型的实际效果,原始MTCNN算法检测结果如图9所示,本文模型检测结果如图10所示。对比检测结果可知,原始MTCNN算法检测图片中有5个戴口罩人脸、一个严重遮挡人脸、两个既戴口罩又戴帽子人脸检测失败,本文模型检测图片中只有两个既戴口罩又戴帽子人脸检测失败。通过对比可知,本文模型的实际人脸检测效果要优于原始MTCNN算法模型。
针对在一些复杂的小尺度、多人脸检测任务中人脸检测的准确率低的问题,提出一种改进的MTCNN算法。首先,在制造数据集时微调IoU阈值参数,来获取更多、更精确的人脸样本;其次,用卷积层替换网络的最大池化层,来提升网络性能;最后,在O-Net网络中引入SE注意力机制,提高模型的特征表达能力。试验结果表明:改进后的算法与原始MTCNN算法相比较,P-Net网络和R-Net网络检测准确率提高了1%,O-Net网络检测准确率提高了0.5%,且改进后的算法在实际人脸检测任务中表现更好。在未来的研究中,将致力于研究遮挡情况下人脸检测数据集和遮挡情况下如何提升MTCNN算法人脸检测准确率,提升MTCNN算法在复杂环境下的性能。
  • 高等学校学科创新引智基地(B20087)
参考文献 引证文献
排序方式:
[1]
Ho H, Nguyen L, Le T, et al. Face detectionusing eigenfaces: a comprehensive review[J]. IEEE Access, 2024, 12: 118406-118426.
[2]
赵立新, 邢润哲, 白银光, 等. 深度学习在目标检测的研究综述[J]. 科学技术与工程, 2021, 21(30): 12787-12795.
Zhao Lixin, Xing Runzhe, Bai Yinguang, et al. Review on survey of deep learning in target detection[J]. Science Technology and Engineering, 2021, 21(30): 12787-12795.
[3]
Li J M, Xie H T, Li J H, et al. Frequency-aware discriminative feature learning supervised by single-center loss for face forgery detection[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition(CVPR). Los Alamitos: IEEE Computer Society Press, 2021: 6458-6467.
[4]
Albiero V, Chen X Y, Yin X, et al. Img2pose: face alignment and detection via 6DoF, face pose estimation[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition(CVPR). Los Alamitos: IEEE Computer Society Press, 2021: 7617-7627.
[5]
张阳婷, 黄德启, 王东伟, 等. 基于深度学习的目标检测算法研究与应用综述[J]. 计算机工程与应用, 2023, 59(18): 1-13.
Zhang Yangting, Huang Deqi, Wang Dongwei, et al. Review on research and application of deep learning-based target detection algorithms[J]. Computer Engineering and Applications, 2023, 59(18): 1-13.
[6]
Chaudhary B, Aghdaie P, Soleymani S, et al. Differential morph face detection using discriminative wavelet sub-bands[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition(CVPR). Los Alamitos: IEEE Computer Society Press, 2021: 1425-1434.
[7]
Gao H X, Tao W, Wen D C, et al. DupNet: towards very tiny quantized CNN with improved accuracy for face detection[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition(CVPR). Los Alamitos: IEEE Computer Society Press, 2019: 168-177.
[8]
徐铭, 李华. 基于改进YOLOv5s-face的Face5系列人脸检测算法[J]. 重庆理工大学学报(自然科学), 2024, 38(6): 194-202.
Xu Ming, Li Hua. Face5 series face detection algorithm based on improved YOLOv5s-face[J]. Journal of Chongqing University of Technology(Natural Science), 2024, 38(6): 194-202.
[9]
Vesdapunt N, Wang B Y. Crface: confidence ranker for model-agnostic face detection refinement[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition(CVPR). Los Alamitos: IEEE Computer Society Press, 2021: 1674-1684.
[10]
戴惠丽. 融合深度字典学习和特征重建的遮挡人脸检测研究[J]. 计算机应用与软件, 2024, 41(11): 228-233.
Dai huili. An occlusion face detection method combining depth dictionary learning and feature reconstruction[J]. Computer Applications and Software, 2024, 41(11): 228-233.
[11]
Wang W J, Yang W H, Liu J Y. HLA-Face: joint high-low adaptation for low light face detection[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition(CVPR). Los Alamitos: IEEE Computer Society Press, 2021: 16195-16204.
[12]
丁雅眉. 复杂场景下的人脸检测方法研究[D]. 西安: 西安邮电大学, 2024.
Ding Yamei. Research on face detection methods in complex Scenes[D]. Xi'an: Xi'an University of Posts and Telecommunications, 2024.
[13]
董子平, 陈世国, 廖国清. 基于YOLOv5s的密集多人脸检测算法[J]. 计算机工程与科学, 2023, 45(10): 1838-1846.
Dong Ziping, Chen Shiguo, Liao Guoqing. A dense multi-face detection algorithm based on YOLOv5s[J]. Computer Engineering and Science, 2023, 45(10): 1838-1846.
[14]
Zhang K, Zhang Z, Li Z, et al. Joint face detection and alignment using multitask cascaded convolutional networks[J]. IEEE Signal Process, 2016, 23(10): 1499-1503.
[15]
刘晓军, 类成玉, 张晓玲, 等. 基于卷积神经网络的人脸检测综述[J]. 兵工自动化, 2021, 40(6): 58-63.
Liu Xiaojun, Lei Chengyu, Zhang Xiaoling, et al. Face detection survey based on convolutional neural network[J]. Ordnance Industry Automation, 2021, 40(6): 58-63.
[16]
蓝雯飞, 张盛兰, 朱容波, 等. 基于改进MTCNN的人脸检测算法[J]. 中南民族大学学报(自然科学版), 2020, 39(6): 637-641.
Lan Wenfei, Zhang Shenglan, Zhu Rongbo, et al. Face detection algorithm based on improved MTCNN[J]. Journal of South-Central University for Nationalities(Natural Science Edition), 2020, 39(6): 637-641.
[17]
顾梅花, 冯婧, 杨娜. 一种融合上下文信息特征的改进MTCNN人脸检测算法[J]. 西安工程大学学报, 2021, 35(6): 114-120.
Gu Meihua, Feng Jing, Yang Na. Improved MTCNN face detection algorithm fused with context features[J]. Journal of Xi'an Polytechnic University, 2021, 35(6): 114-120.
[18]
周航, 蔡茂国, 吴涛, 等. 一种改进的多任务级联网络人脸检测算法研究[J]. 智能计算机与应用, 2021, 11(3): 172-176.
Zhou Hang, Cai Maoguo, Wu Tao, et al. Research on an improved multi-task cascade network face detection algorithm[J]. Intelligent Computer and Applications, 2021, 11(3): 172-176.
[19]
庞殊杨, 芦莎. 基于改进MTCNN的多尺度安全帽识别[J]. 计算机应用研究, 2021, 38(6): 1907-1912, 1916.
Pang Shuyang, Lu Sha. Multi-scale safety helmet detection based on improved MTCNN[J]. Application Research of Computers, 2021, 38(6): 1907-1912, 1916.
[20]
刘彩云, 李雅雯, 刘倩. 一种改进的多任务级联卷积神经网络人脸检测算法[J]. 长江大学学报(自然科学版), 2021, 18(6): 111-118.
Liu Caiyun, Li Yawen, Liu Qian. An improved face detection algorithm based on multi-task cascaded convolutional neural network[J]. Journal of Yangtze University(Natural Science Edition), 2021, 18(6): 111-118.
[21]
李艳灵, 王莎莎, 杨志鹏. 一种改进的多任务级联卷积神经网络人脸检测算法[J]. 信阳师范学院学报(自然科学版), 2022, 35(4): 651-655.
Li Yanling, Wang Shasha, Yang Zhipeng. An improved multi-task cascaded convolutional neural network face detection algorithm[J]. Journal of Xinyang Normal University(Natural Science Edition), 2022, 35(4): 651-655.
[22]
李丽平, 许营坤, 王嘉航. 基于锚框稀疏图像金字塔的MTCNN人脸检测方法[J]. 浙江工业大学学报, 2023, 51(3): 298-304, 318.
Li Liping, Xu Yingkun, Wang Jiahang. Research on MTCNN face detection method of sparse image pyramid based on anchor box[J]. Journal of Zhejiang University of Technology, 2023, 51(3): 298-304, 318.
[23]
Ma Z, Chang D, Xie J, et al. Fine-grained vehicle classification with channel max pooling modified CNNs[J]. EEE Transactions on Vehicular Technology, 2019, 68(4): 3224-3233.
[24]
Springenberg T J, Dosovitskiy A, Brox T, et al. Striving for simplicity: the all convolutional net[J]. ArXiv, 2014: 1412.6806.
[25]
Woo S, Park J, Lee J Y. CBAM: convolutional block attention module[C]// European Conference on Computer Vision(ECCV). Munich: Springer, 2018: 11211.
[26]
Hu J, Li S, Albanie S, et al. Sueeze-and-excitation networks[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition(CVPR). Los Alamitos: IEEE Computer Society Press, 2018: 7132-7141.
[27]
汪建伟, 游疆, 万敏, 等. 复杂背景下的低空无人机检测与跟踪算法[J]. 强激光与粒子束, 2023, 35(7): 158-169.
Wang Jianwei, You Jiang, Wan Min, et al. Low-altitude UAV detection and tracking algorithms incomplex backgrounds[J]. High Power Laser and Particle Beams, 2023, 35(7): 158-169.
2025年第25卷第21期
PDF下载
118
47
引用本文
BibTeX
文章信息
doi: 10.12404/j.issn.1671-1815.2406407
  • 接收时间:2024-08-26
  • 首发时间:2026-01-13
  • 出版时间:2025-07-28
补充材料
相关文章
文章信息
作者
出版历史
  • 收稿日期:2024-08-26
  • 修回日期:2025-04-16
基金
高等学校学科创新引智基地(B20087)
作者信息
    中国人民公安大学信息网络安全学院, 北京 100038

通讯作者:

* 钟寒(1984—),女,汉族,山东济宁人,博士,副教授。研究方向:数据分析。E-mail:
参考文献
分享链接
https://castjournals.cast.org.cn/joweb/kxjsygc/CN/10.12404/j.issn.1671-1815.2406407
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