Article(id=1245407859320668488, tenantId=1146029695717560320, journalId=1146123166801305609, issueId=1156262727438951343, articleNumber=null, orderNo=null, doi=10.12404/j.issn.1671-1815.2309833, pmid=null, cstr=null, oa=null, hot=null, price=null, onlineType=0, articleFormat=0, articleType=null, articleTypeStr=research-article, receivedDate=1702396800000, receivedDateStr=2023-12-13, revisedDate=1720454400000, revisedDateStr=2024-07-09, acceptedDate=null, acceptedDateStr=null, onlineDate=1774857972220, onlineDateStr=2026-03-30, pubDate=1741363200000, pubDateStr=2025-03-08, doiRegisterDate=null, doiRegisterDateStr=null, onlineIssueDate=1774857972220, onlineIssueDateStr=2026-03-30, onlineJustAcceptDate=null, onlineJustAcceptDateStr=null, onlineFirstDate=null, onlineFirstDateStr=null, sourceXml=null, magXml=null, createTime=1774857972220, creator=13701087609, updateTime=1774857972220, updator=13701087609, issue=Issue{id=1156262727438951343, tenantId=1146029695717560320, journalId=1146123166801305609, year='2025', volume='25', issue='7', pageStart='2193', pageEnd='3077', issueExtLink='null', onlineDate='null', pubDate='null', beforeIssueId=null, nextIssueId=null, price=null, status=1, issueComplete=1, articleOrder=1, issueType=-1, specialIssue=0, createTime=1753604116544, creator=13701087609, updateTime=1753771263994, updator=13701087609, preIssue=null, nextIssue=null, ext={EN=IssueExt(id=1156963794699248405, tenantId=1146029695717560320, journalId=1146123166801305609, issueId=1156262727438951343, language=EN, specialIssueTitle=, coverIllustrator=, specialIssueEditor=, specialIssueAbout=), CN=IssueExt(id=1156963794699248406, tenantId=1146029695717560320, journalId=1146123166801305609, issueId=1156262727438951343, language=CN, specialIssueTitle=, coverIllustrator=, specialIssueEditor=, specialIssueAbout=)}, issueFiles=null}, startPage=2691, endPage=2702, ext={EN=ArticleExt(id=1245407859836567894, articleId=1245407859320668488, tenantId=1146029695717560320, journalId=1146123166801305609, language=EN, title=Improved Automatic Crack Identification for Electrical Imaging Logging Using PSPNet, columnId=1156262729351549255, journalTitle=Science Technology and Engineering, columnName=Papers·Astronomy and Geosciences, runingTitle=null, highlight=null, articleAbstract=

An improved PSPNet(pyramid scene parseing network) network was proposed to automatically identify fractures in electrical imaging logging images, which was difficult to extract fracture features and led to low segmentation accuracy and large calculation of network parameters. Firstly, the backbone network in PSPNet was replaced with the optimized MobileNetV3 network, which could significantly reduce the number of network parameters and the amount of computation. Secondly, the asymptotic feature pyramid network(AFPN) was introduced to increase the interaction of multi-scale information and enhance the recognition ability of small cracks. Then, multi-depthwise Conv head transposed attention(MDTA) was introduced to extract global features and improve the extraction ability of key information. Finally, the combination of Focal Loss and Dice Loss were used as a loss function to solve the problem of unbalanced proportion of data sets. The experimental results show that the improved PSPNet network has a good segmentation effect on the fracture in the electrical imaging logging. Compared with the PSPNet network, mIoU(mean intersection over union) improved by 3.17% and mPA(mean pixel accuracy) improved by 6.38%. In addition, the number of parameters, calculation amount and weight of the proposed algorithm are reduced by 94.3%, 95.7% and 93.8% respectively compared with the original model. At the same time, the crack identification system based on CIFLog is developed, which can meet the practical needs of the electrical imaging logging.

, correspAuthors=Xiang ZHANG, authorNote=null, correspAuthorsNote=null, copyrightStatement=null, copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=null, magXml=null, pdfUrl=null, pdf=null, pdfFileSize=null, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=null, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=null, mapNumber=null, authorCompany=null, fund=null, authors=null, authorsList=Ke SHEN, Xiao-ling XIAO, Xiang ZHANG, Mao-shan LIN), CN=ArticleExt(id=1245407864005706347, articleId=1245407859320668488, tenantId=1146029695717560320, journalId=1146123166801305609, language=CN, title=改进PSPNet的电成像测井裂缝自动识别, columnId=1156262730077163858, journalTitle=科学技术与工程, columnName=论文·天文学、地球科学, runingTitle=null, highlight=null, articleAbstract=

针对裂缝特征提取困难导致裂缝分割精度低、网络参数量计算量大的问题,提出一种改进的PSPNet(pyramid scene parseing network)网络用于自动识别电成像测井图像中的裂缝。首先将PSPNet中的骨干网络替换为优化的MobileNetV3网络,减少网络参数量和计算量;其次,引入渐进特征金字塔(asymptotic feature pyramid network,AFPN),用于增加多尺度信息的交互,增强对细小裂缝的识别能力;接着,引入多深度卷积头转置注意力(multi-depthwise Conv head transposed attention,MDTA)进行全局特征的提取,提升关键信息的提取能力;最后,采用Focal Loss和Dice Loss组合相加作为损失函数,以解决数据集类别占比不平衡的问题。实验结果表明,改进的PSPNet网络对电成像测井裂缝具有较好的分割效果。与PSPNet网络相比,mIoU(mean intersection over union)提升了3.17%,mPA(mean pixel accuracy)提升了6.38%。此外,研究成果的参数量、计算量、权重分别比原模型减少94.3%、95.7%和93.8%。同时,开发了基于CIFLog的裂缝识别系统,该系统能够满足对电成像测井的实际需要。

, correspAuthors=张翔, authorNote=null, correspAuthorsNote=
* 张翔(1969—),男,汉族,湖北蕲春人,博士,教授。研究方向:成像测井与地球物理信息处理。E-mail:
, copyrightStatement=null, copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=w0hBdEZno74iJLXO4FWTjg==, magXml=8kkUwvUowAuv9rTsDymc5Q==, pdfUrl=null, pdf=L6TLqmSekiB3g3o1TB2Brg==, pdfFileSize=13037261, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=ewL3af8OgPjMoV4Qj1JvBw==, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=6/Rjn/D/pBBaCPrJ9t8f2Q==, mapNumber=null, authorCompany=null, fund=null, authors=

申科(1998—),男,汉族,湖北黄冈人,硕士研究生。研究方向:基于深度学习的测井数据处理。E-mail:

, authorsList=申科, 肖小玲, 张翔, 林茂山)}, authors=[Author(id=1245407864781652645, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, orderNo=0, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=2022710675@yangtzeu.edu.cn, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1245407864915870382, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, authorId=1245407864781652645, language=EN, stringName=Ke SHEN, firstName=Ke, middleName=null, lastName=SHEN, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1 School of Computer Science, Yangtze University, Jingzhou 434023, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1245407865184305862, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, authorId=1245407864781652645, language=CN, stringName=申科, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1 长江大学计算机科学学院, 荆州 434023, bio={"content":"

申科(1998—),男,汉族,湖北黄冈人,硕士研究生。研究方向:基于深度学习的测井数据处理。E-mail:

"}, bioImg=null, bioContent=

申科(1998—),男,汉族,湖北黄冈人,硕士研究生。研究方向:基于深度学习的测井数据处理。E-mail:

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1245407864328667776, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, xref=1, ext=[AuthorCompanyExt(id=1245407864341250690, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, companyId=1245407864328667776, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1 School of Computer Science, Yangtze University, Jingzhou 434023, China), AuthorCompanyExt(id=1245407864353833603, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, companyId=1245407864328667776, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1 长江大学计算机科学学院, 荆州 434023)])]), Author(id=1245407865289163475, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, 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=1245407865448547043, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, authorId=1245407865289163475, language=EN, stringName=Xiao-ling XIAO, firstName=Xiao-ling, middleName=null, lastName=XIAO, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, address=1 School of Computer Science, Yangtze University, Jingzhou 434023, China
2 Oil and Gas Resources and Exploration Technology, Ministry of Education, Yangtze University, Wuhan 430100, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1245407865561793264, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, authorId=1245407865289163475, language=CN, stringName=肖小玲, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, address=1 长江大学计算机科学学院, 荆州 434023
2 油气资源与勘探技术教育部重点实验室(长江大学), 武汉 430100, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1245407864328667776, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, xref=1, ext=[AuthorCompanyExt(id=1245407864341250690, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, companyId=1245407864328667776, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1 School of Computer Science, Yangtze University, Jingzhou 434023, China), AuthorCompanyExt(id=1245407864353833603, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, companyId=1245407864328667776, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1 长江大学计算机科学学院, 荆州 434023)]), AuthorCompany(id=1245407864475468427, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, xref=2, ext=[AuthorCompanyExt(id=1245407864496439950, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, companyId=1245407864475468427, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2 Oil and Gas Resources and Exploration Technology, Ministry of Education, Yangtze University, Wuhan 430100, China), AuthorCompanyExt(id=1245407864513217168, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, companyId=1245407864475468427, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2 油气资源与勘探技术教育部重点实验室(长江大学), 武汉 430100)])]), Author(id=1245407865687622395, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, orderNo=2, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=zx_jr_xl@163.com, emailSecond=null, emailThird=null, correspondingAuthor=1, authorType=1, ext={EN=AuthorExt(id=1245407865800868615, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, authorId=1245407865687622395, language=EN, stringName=Xiang ZHANG, firstName=Xiang, middleName=null, lastName=ZHANG, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, *, address=2 Oil and Gas Resources and Exploration Technology, Ministry of Education, Yangtze University, Wuhan 430100, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1245407865918309137, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, authorId=1245407865687622395, language=CN, stringName=张翔, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, *, address=2 油气资源与勘探技术教育部重点实验室(长江大学), 武汉 430100, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1245407864475468427, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, xref=2, ext=[AuthorCompanyExt(id=1245407864496439950, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, companyId=1245407864475468427, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2 Oil and Gas Resources and Exploration Technology, Ministry of Education, Yangtze University, Wuhan 430100, China), AuthorCompanyExt(id=1245407864513217168, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, companyId=1245407864475468427, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2 油气资源与勘探技术教育部重点实验室(长江大学), 武汉 430100)])]), Author(id=1245407866027361059, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, orderNo=3, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1245407866132218674, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, authorId=1245407866027361059, language=EN, stringName=Mao-shan LIN, firstName=Mao-shan, middleName=null, lastName=LIN, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=3, address=3 Tuha Branch, China National Logging Corporation, Hami 839000, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1245407866258047807, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, authorId=1245407866027361059, language=CN, stringName=林茂山, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=3, address=3 中国石油集团测井有限公司吐哈分公司, 哈密 839000, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1245407864639046298, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, xref=3, ext=[AuthorCompanyExt(id=1245407864655823516, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, companyId=1245407864639046298, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3 Tuha Branch, China National Logging Corporation, Hami 839000, China), AuthorCompanyExt(id=1245407864672600734, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, companyId=1245407864639046298, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3 中国石油集团测井有限公司吐哈分公司, 哈密 839000)])])], keywords=[Keyword(id=1245407866404848462, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, language=EN, orderNo=1, keyword=PSPNet), Keyword(id=1245407866492928859, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, language=EN, orderNo=2, keyword=fracture identification), Keyword(id=1245407866585203559, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, language=EN, orderNo=3, keyword=electrical imaging logging image), Keyword(id=1245407866685866870, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, language=EN, orderNo=4, keyword=MobileNetV3), Keyword(id=1245407866815890312, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, language=EN, orderNo=5, keyword=asymptotic feature pyramid network(AFPN)), Keyword(id=1245407867050771356, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, language=CN, orderNo=1, keyword=PSPNet), Keyword(id=1245407867210154922, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, language=CN, orderNo=2, keyword=裂缝识别), Keyword(id=1245407867315012536, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, language=CN, orderNo=3, keyword=电成像测井图像), Keyword(id=1245407867432453060, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, language=CN, orderNo=4, keyword=MobileNetV3), Keyword(id=1245407867524727759, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, language=CN, orderNo=5, keyword=AFPN)], refs=[Reference(id=1245407872293650852, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=https://link.cnki.net/urlid/62.1177.te.20240509.2130.002, language=null, rfNumber=[1], rfOrder=0, authorNames=徐诗雨, 祝怡, 曾乙洋, journalName=天然气地球科学, refType=null, unstructuredReference=徐诗雨, 祝怡, 曾乙洋, . 蓬莱气田茅口组储层孔隙结构特征及分类评价[J/OL]. 天然气地球科学. https://link.cnki.net/urlid/62.1177.te.20240509.2130.002., articleTitle=蓬莱气田茅口组储层孔隙结构特征及分类评价, refAbstract=null), Reference(id=1245407872419479984, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=https://link.cnki.net/urlid/62.1177.te.20240509.2130.002, language=null, rfNumber=[1], rfOrder=1, authorNames=Xu Shiyu, Zhu Yi, Zeng Yiyang, journalName=Natural Gas Geoscience, refType=null, unstructuredReference=Xu Shiyu, Zhu Yi, Zeng Yiyang, et al. Pore structure characteristics and classification evaluation of Maokou Formation reservoir in Penglai Gas Field[J/OL]. Natural Gas Geoscience. https://link.cnki.net/urlid/62.1177.te.20240509.2130.002., articleTitle=Pore structure characteristics and classification evaluation of Maokou Formation reservoir in Penglai Gas Field, refAbstract=null), Reference(id=1245407872528531900, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=[2], rfOrder=2, authorNames=陈泓位, 王时林, 和源, journalName=油气地质与采收率, refType=null, unstructuredReference=陈泓位, 王时林, 和源, . 四川盆地中北部 DB1井区灯影组四段沉积相及储层特征[J/OL]. 油气地质与采收率. https://doi.org/10.13673/j.pgre.202308035., articleTitle=四川盆地中北部 DB1井区灯影组四段沉积相及储层特征, refAbstract=null), Reference(id=1245407872671138248, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=[2], rfOrder=3, authorNames=Chen Hongwei, Wang Shilin, He Yuan, journalName=Oil and Gas Geology and Recovery, refType=null, unstructuredReference=Chen Hongwei, Wang Shilin, He Yuan, et al. Sedimentary facies and reservoir characteri sticsof the fourth member of Dengying Formation, DB1 well area, north-central Sichuan Basin[J/OL]. Oil and Gas Geology and Recovery. https://doi.org/10.13673/j.pgre.202308035., articleTitle=Sedimentary facies and reservoir characteri sticsof the fourth member of Dengying Formation, DB1 well area, north-central Sichuan Basin, refAbstract=null), Reference(id=1245407872809550290, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, doi=null, pmid=null, pmcid=null, year=2022, volume=22, issue=21, pageStart=9095, pageEnd=9104, url=null, language=null, rfNumber=[3], rfOrder=4, authorNames=苏可嘉, 秦臻, 邓呈祥, journalName=科学技术与工程, refType=null, unstructuredReference=苏可嘉, 秦臻, 邓呈祥, . 致密砂岩裂缝填充识别及其测井响应特征——以鄂尔多斯盆地镇泾油田延长组为例[J]. 科学技术与工程, 2022, 22(21): 9095-9104., articleTitle=致密砂岩裂缝填充识别及其测井响应特征——以鄂尔多斯盆地镇泾油田延长组为例, refAbstract=null), Reference(id=1245407872935379420, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, doi=null, pmid=null, pmcid=null, year=2022, volume=22, issue=21, pageStart=9095, pageEnd=9104, url=null, language=null, rfNumber=[3], rfOrder=5, authorNames=Su Kejia, Qin Zhen, Deng Chengxiang, journalName=null, refType=null, unstructuredReference=Su Kejia, Qin Zhen, Deng Chengxiang, et al. Fracture-filling identification and logging response characteristics of tight sandstone: a casestudy of Yanchang Formation, Zhenjing Oilfield, Ordos Basin[J]. Science Technology and Engineering, 2022, 22(21): 9095-9104., articleTitle=Fracture-filling identification and logging response characteristics of tight sandstone: a casestudy of Yanchang Formation, Zhenjing Oilfield, Ordos Basin[J]. Science Technology and Engineering, refAbstract=null), Reference(id=1245407873086374377, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, doi=null, pmid=null, pmcid=null, year=2024, volume=24, issue=7, pageStart=2673, pageEnd=2682, url=null, language=null, rfNumber=[4], rfOrder=6, authorNames=侯筱晓, 刘成林, 陈茜, journalName=科学技术与工程, refType=null, unstructuredReference=侯筱晓, 刘成林, 陈茜, . 碳酸盐岩构造裂缝特征与形成期次分析: 以阿姆河盆地右岸东部为例[J]. 科学技术与工程, 2024, 24(7): 2673-2682., articleTitle=碳酸盐岩构造裂缝特征与形成期次分析: 以阿姆河盆地右岸东部为例, refAbstract=null), Reference(id=1245407873220592115, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, doi=null, pmid=null, pmcid=null, year=2024, volume=24, issue=7, pageStart=2673, pageEnd=2682, url=null, language=null, rfNumber=[4], rfOrder=7, authorNames=Hou Xiaoxiao, Liu Chenglin, Chen Qian, journalName=Science Technology and Engineering, refType=null, unstructuredReference=Hou Xiaoxiao, Liu Chenglin, Chen Qian, et al. Characteristics and formation stages of carbonate tectonic fractures: a case study of the eastern right bank of the Amu Darya Basin[J]. Science Technology and Engineering, 2024, 24(7): 2673-2682., articleTitle=Characteristics and formation stages of carbonate tectonic fractures: a case study of the eastern right bank of the Amu Darya Basin, refAbstract=null), Reference(id=1245407873325449725, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, doi=null, pmid=null, pmcid=null, year=2022, volume=46, issue=5, pageStart=592, pageEnd=598, url=null, language=null, rfNumber=[5], rfOrder=8, authorNames=刘浩杰, 盖姗姗, 于文政, journalName=测井技术, refType=null, unstructuredReference=刘浩杰, 盖姗姗, 于文政, . 准噶尔盆地永进油田超深层致密砂岩裂缝测井识别[J]. 测井技术, 2022, 46(5): 592-598., articleTitle=准噶尔盆地永进油田超深层致密砂岩裂缝测井识别, refAbstract=null), Reference(id=1245407873463861771, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, doi=null, pmid=null, pmcid=null, year=2022, volume=46, issue=5, pageStart=592, pageEnd=598, url=null, language=null, rfNumber=[5], rfOrder=9, authorNames=Liu Haojie, Gai Shanshan, Yu Wenzheng, journalName=Logging Technology, refType=null, unstructuredReference=Liu Haojie, Gai Shanshan, Yu Wenzheng, et al. Logging identification of ultra-deep tight sandstone fractures in Yongjin Oilfield, Junggar Basin[J]. Logging Technology, 2022, 46(5): 592-598., articleTitle=Logging identification of ultra-deep tight sandstone fractures in Yongjin Oilfield, Junggar Basin, refAbstract=null), Reference(id=1245407873593885200, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, doi=null, pmid=null, pmcid=null, year=2022, volume=9, issue=4, pageStart=16, pageEnd=22, url=null, language=null, rfNumber=[6], rfOrder=10, authorNames=刘双莲, journalName=非常规油气, refType=null, unstructuredReference=刘双莲. 常规测井技术识别火山岩裂缝方法研究——以松南火山岩为例[J]. 非常规油气, 2022, 9(4): 16-22., articleTitle=常规测井技术识别火山岩裂缝方法研究——以松南火山岩为例, refAbstract=null), Reference(id=1245407873728102940, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, doi=null, pmid=null, pmcid=null, year=2022, volume=9, issue=4, pageStart=16, pageEnd=22, url=null, language=null, rfNumber=[6], rfOrder=11, authorNames=Liu Shuanglian, journalName=Unconventional Oil and Gas, refType=null, unstructuredReference=Liu Shuanglian. Research on conventional logging techniques to identify volcanic fractures: a case study of Songnan volcanic rocks[J]. Unconventional Oil and Gas, 2022, 9(4): 16-22., articleTitle=Research on conventional logging techniques to identify volcanic fractures: a case study of Songnan volcanic rocks, refAbstract=null), Reference(id=1245407873858126375, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, doi=null, pmid=null, pmcid=null, year=2024, volume=2024, issue=null, pageStart=3380390, pageEnd=null, url=null, language=null, rfNumber=[7], rfOrder=12, authorNames=Li H, Li H, Li C, journalName=IEEE Transactions on Geo-science and Remote Sensing, refType=null, unstructuredReference=Li H, Li H, Li C, et al. Hybrid swin transformer-CNN model for pore-crack structureidentification[J]. IEEE Transactions on Geo-science and Remote Sensing, 2024, 2024: 3380390., articleTitle=Hybrid swin transformer-CNN model for pore-crack structureidentification, refAbstract=null), Reference(id=1245407874021704243, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, doi=null, pmid=null, pmcid=null, year=2024, volume=12, issue=2, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=[8], rfOrder=13, authorNames=Olya B A M, Mohebian R, Bagheri H, journalName=Interpretation, refType=null, unstructuredReference=Olya B A M, Mohebian R, Bagheri H, et al. Toward real-time fracture detection on image logs using deep convolutional neural network YOLOv5[J]. Interpretation, 2024, 12(2): SB9-SB18., articleTitle=Toward real-time fracture detection on image logs using deep convolutional neural network YOLOv5, refAbstract=null), Reference(id=1245407874122367546, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, doi=null, pmid=null, pmcid=null, year=2023, volume=351, issue=null, pageStart=128992, pageEnd=null, url=null, language=null, rfNumber=[9], rfOrder=14, authorNames=Du L, Lu X, Li H, journalName=Fuel, refType=null, unstructuredReference=Du L, Lu X, Li H. Automatic fracture detection from the images of electrical image logsusing Mask R-CNN[J]. Fuel, 2023, 351: 128992., articleTitle=Automatic fracture detection from the images of electrical image logsusing Mask R-CNN, refAbstract=null), Reference(id=1245407874223030854, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, doi=null, pmid=null, pmcid=null, year=2024, volume=39, issue=2, pageStart=727, pageEnd=736, url=null, language=null, rfNumber=[10], rfOrder=15, authorNames=马同乐, 刘红岐, 廖海博, journalName=地球物理学进展, refType=null, unstructuredReference=马同乐, 刘红岐, 廖海博, . 融合通道和空间交叉注意力的裂缝识别方法[J]. 地球物理学进展, 2024, 39(2): 727-736., articleTitle=融合通道和空间交叉注意力的裂缝识别方法, refAbstract=null), Reference(id=1245407874344665685, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, doi=null, pmid=null, pmcid=null, year=2019, volume=39, issue=2, pageStart=727, pageEnd=736, url=null, language=null, rfNumber=[10], rfOrder=16, authorNames=Ma Tongle, Liu Hongqi, Liao Haibo, journalName=Progress in Geophysics, refType=null, unstructuredReference=Ma Tongle, Liu Hongqi, Liao Haibo, et al. Fusingchannels and spatial cross attentionfor crack identification[J]. Progress in Geophysics, 2019, 39(2): 727-736., articleTitle=Fusingchannels and spatial cross attentionfor crack identification, refAbstract=null), Reference(id=1245407874466300506, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, doi=null, pmid=null, pmcid=null, year=2017, volume=null, issue=null, pageStart=2881, pageEnd=2890, url=null, language=null, rfNumber=[11], rfOrder=17, authorNames=Zhao H, Shi J, Qi X, journalName=Proceedings of the IEEEConference on Computer Vision and PatternRecognition, refType=null, unstructuredReference=Zhao H, Shi J, Qi X, et al. Pyramid sceneparsing network[C]// Proceedings of the IEEEConference on Computer Vision and PatternRecognition. New York: IEEE, 2017: 2881-2890., articleTitle=Pyramid sceneparsing network, refAbstract=null), Reference(id=1245407874558575205, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, doi=null, pmid=null, pmcid=null, year=2019, volume=null, issue=null, pageStart=1314, pageEnd=1324, url=null, language=null, rfNumber=[12], rfOrder=18, authorNames=Howard A, Sandler M, Chu G, journalName=Proceedings of the IEEE/CVF International Conference on Computer Vision, refType=null, unstructuredReference=Howard A, Sandler M, Chu G, et al. Searching for mobilenetv3[C]// Proceedings of the IEEE/CVF International Conference on Computer Vision. New York: IEEE, 2019: 1314-1324., articleTitle=Searching for mobilenetv3, refAbstract=null), Reference(id=1245407874663432817, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=[13], rfOrder=19, authorNames=Yang G, Lei J, Zhu Z, journalName=arXiv preprint arXiv: 2306. 15988, 2023, refType=null, unstructuredReference=Yang G, Lei J, Zhu Z, et al. AFPN: asymptotic feature pyramid network for object detection[J]. arXiv preprint arXiv: 2306. 15988, 2023., articleTitle=AFPN: asymptotic feature pyramid network for object detection, refAbstract=null), Reference(id=1245407874772484724, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, doi=null, pmid=null, pmcid=null, year=2022, volume=null, issue=null, pageStart=5728, pageEnd=5739, url=null, language=null, rfNumber=[14], rfOrder=20, authorNames=Zamir S W, Arora A, Khan S, journalName=Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, refType=null, unstructuredReference=Zamir S W, Arora A, Khan S, et al. Restormer: efficient transformer for high-resolutionimage restoration[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2022: 5728-5739., articleTitle=Restormer: efficient transformer for high-resolutionimage restoration, refAbstract=null), Reference(id=1245407874944451197, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, doi=null, pmid=null, pmcid=null, year=2016, volume=null, issue=3DV, pageStart=565, pageEnd=571, url=null, language=null, rfNumber=[15], rfOrder=21, authorNames=Milletari F, Navab N, Ahmadi S A, journalName=null, refType=null, unstructuredReference=Milletari F, Navab N, Ahmadi S A, et al.Fully convolutional neural networks for volumetric medical image segmentation[C]//Proceedings of the 2016 Fourth International Conference on 3D Vision(3DV). New York: IEEE, 2016: 565-571., articleTitle=null, refAbstract=null), Reference(id=1245407875070280328, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, doi=null, pmid=null, pmcid=null, year=2017, volume=42, issue=2, pageStart=318, pageEnd=327, url=null, language=null, rfNumber=[16], rfOrder=22, authorNames=Lin T Y, Goyal P, Girshick R, journalName=IEEE Transactions on Pattern Analysis and Machine Intelligence, refType=null, unstructuredReference=Lin T Y, Goyal P, Girshick R, et al. Focal lossfor dense object detection[C]// IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017: 42(2): 318-327., articleTitle=Focal lossfor dense object detection, refAbstract=null), Reference(id=1245407875229663893, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, doi=null, pmid=null, pmcid=null, year=2020, volume=null, issue=null, pageStart=11531, pageEnd=11539, url=null, language=null, rfNumber=[17], rfOrder=23, authorNames=Wang Q, Wu B, Zhu P, journalName=Proceedings of the 2020IEEE/CVF Conference on Computer Visionand Pattern Recognition, refType=null, unstructuredReference=Wang Q, Wu B, Zhu P, et al. ECA-Net: efficient channel attention for deep convolutional neural networks[C]// Proceedings of the 2020IEEE/CVF Conference on Computer Visionand Pattern Recognition. Piscataway: IEEE, 2020: 11531-11539., articleTitle=ECA-Net: efficient channel attention for deep convolutional neural networks, refAbstract=null), Reference(id=1245407875326132894, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, doi=null, pmid=null, pmcid=null, year=2022, volume=35, issue=null, pageStart=6789, pageEnd=6801, url=null, language=null, rfNumber=[18], rfOrder=24, authorNames=Goyal A, Bochkovskiy A, Deng J, journalName=Advances in Neural Information Processing Systems, refType=null, unstructuredReference=Goyal A, Bochkovskiy A, Deng J, et al. Nondeep networks[J]. Advances in Neural Information Processing Systems, 2022, 35: 6789-6801., articleTitle=Nondeep networks, refAbstract=null), Reference(id=1245407875410018980, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=[19], rfOrder=25, authorNames=Li Y, Hou Q, Zheng Z, journalName=arXiv preprint arXiv: 2303. 09030, 2023, refType=null, unstructuredReference=Li Y, Hou Q, Zheng Z, et al. Large selectivekernel network for remote sensing object detection[J]. arXiv preprint arXiv: 2303. 09030, 2023., articleTitle=Large selectivekernel network for remote sensing object detection, refAbstract=null), Reference(id=1245407875514876591, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, doi=null, pmid=null, pmcid=null, year=2021, volume=null, issue=null, pageStart=11863, pageEnd=11874, url=null, language=null, rfNumber=[20], rfOrder=26, authorNames=Yang L, Zhang R Y, Li L, journalName=Proceedings of the 38th InternationalConference on Machine Learning, refType=null, unstructuredReference=Yang L, Zhang R Y, Li L, et al. SimAM:a simple, parameter-free attention module for convolutional neural networks[C]// Proceedings of the 38th InternationalConference on Machine Learning. Online: PMLR, 2021: 11863-11874., articleTitle=SimAM:a simple, parameter-free attention module for convolutional neural networks, refAbstract=null), Reference(id=1245407875636511414, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, doi=null, pmid=null, pmcid=null, year=2015, volume=null, issue=null, pageStart=3431, pageEnd=3440, url=null, language=null, rfNumber=[21], rfOrder=27, authorNames=Long J, Shelhamer E, Darrell T, journalName=Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, refType=null, unstructuredReference=Long J, Shelhamer E, Darrell T. Fully convolutional networks for semantic segmentation[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2015: 3431-3440., articleTitle=Fully convolutional networks for semantic segmentation, refAbstract=null), Reference(id=1245407875774923458, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=[22], rfOrder=28, authorNames=Chen L C, Papandreou G, Schroff F, journalName=arXiv preprint arXiv: 1706. 05587, 2017, refType=null, unstructuredReference=Chen L C, Papandreou G, Schroff F, et al. Rethinking atrous convolution for semanticimage segmentation[J]. arXiv preprint arXiv: 1706. 05587, 2017., articleTitle=Rethinking atrous convolution for semanticimage segmentation, refAbstract=null), Reference(id=1245407875879781071, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, doi=null, pmid=null, pmcid=null, year=2018, volume=null, issue=null, pageStart=801, pageEnd=818, url=null, language=null, rfNumber=[23], rfOrder=29, authorNames=Chen L C, Zhu Y, Papandreou G, journalName=Proceedings of the European Conference onComputer Vision(ECCV), refType=null, unstructuredReference=Chen L C, Zhu Y, Papandreou G, et al. En-coder-decoder with atrous separable convolu-tion for semantic image segmentation[C]// Proceedings of the European Conference onComputer Vision(ECCV). New York: IEEE, 2018: 801-818., articleTitle=En-coder-decoder with atrous separable convolu-tion for semantic image segmentation, refAbstract=null), Reference(id=1245407875993027286, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, doi=null, pmid=null, pmcid=null, year=2013, volume=34, issue=1, pageStart=192, pageEnd=200, url=null, language=null, rfNumber=[24], rfOrder=30, authorNames=李宁, 王才志, 刘英明, journalName=石油学报, refType=null, unstructuredReference=李宁, 王才志, 刘英明, . 基于Java-NetBeans的第三代测井软件CIFLog[J]. 石油学报, 2013, 34(1): 192-200., articleTitle=基于Java-NetBeans的第三代测井软件CIFLog, refAbstract=null), Reference(id=1245407876102079201, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, doi=null, pmid=null, pmcid=null, year=2013, volume=34, issue=1, pageStart=192, pageEnd=200, url=null, language=null, rfNumber=[24], rfOrder=31, authorNames=Li Ning, Wang Caizhi, Liu Yingming, journalName=Acta Petrolei Sinica, refType=null, unstructuredReference=Li Ning, Wang Caizhi, Liu Yingming, et al. The third generation logging software CIFLog based on Java-NetBeans[J]. Acta Petrolei Sinica, 2013, 34(1): 192-200., articleTitle=The third generation logging software CIFLog based on Java-NetBeans, refAbstract=null)], funds=[Fund(id=1245407872104907154, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, awardId=41674136, language=CN, fundingSource=国家自然科学基金(41674136), fundOrder=null, country=null)], companyList=[AuthorCompany(id=1245407864328667776, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, xref=1, ext=[AuthorCompanyExt(id=1245407864341250690, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, companyId=1245407864328667776, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1 School of Computer Science, Yangtze University, Jingzhou 434023, China), AuthorCompanyExt(id=1245407864353833603, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, companyId=1245407864328667776, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1 长江大学计算机科学学院, 荆州 434023)]), AuthorCompany(id=1245407864475468427, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, xref=2, ext=[AuthorCompanyExt(id=1245407864496439950, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, companyId=1245407864475468427, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2 Oil and Gas Resources and Exploration Technology, Ministry of Education, Yangtze University, Wuhan 430100, China), AuthorCompanyExt(id=1245407864513217168, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, companyId=1245407864475468427, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2 油气资源与勘探技术教育部重点实验室(长江大学), 武汉 430100)]), AuthorCompany(id=1245407864639046298, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, xref=3, ext=[AuthorCompanyExt(id=1245407864655823516, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, companyId=1245407864639046298, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3 Tuha Branch, China National Logging Corporation, Hami 839000, China), AuthorCompanyExt(id=1245407864672600734, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, companyId=1245407864639046298, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3 中国石油集团测井有限公司吐哈分公司, 哈密 839000)])], figs=[ArticleFig(id=1245407867684111334, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, language=EN, label=Fig.1, caption=PSPNet network structure, figureFileSmall=jvZr7DenkHMV5LisHeD8bQ==, figureFileBig=ewL3af8OgPjMoV4Qj1JvBw==, tableContent=null), ArticleFig(id=1245407867805746162, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, language=CN, label=图1, caption=PSPNet网络结构, figureFileSmall=jvZr7DenkHMV5LisHeD8bQ==, figureFileBig=ewL3af8OgPjMoV4Qj1JvBw==, tableContent=null), ArticleFig(id=1245407868053209098, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, language=EN, label=Fig.2, caption=MobileNetV3 bneck, figureFileSmall=g+NjBeSZ8OtTr5YstmrX2A==, figureFileBig=jHdkdZEXVEMKRQ4VXitEvg==, tableContent=null), ArticleFig(id=1245407868141289491, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, language=CN, label=图2, caption=MobileNetV3 bneck, figureFileSmall=g+NjBeSZ8OtTr5YstmrX2A==, figureFileBig=jHdkdZEXVEMKRQ4VXitEvg==, tableContent=null), ArticleFig(id=1245407868220981281, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, language=EN, label=Fig.3, caption=Dilated convolution, figureFileSmall=roF8jkykvku085MTWs84Rw==, figureFileBig=mXMcpJvDQML5lnluAY4/Xg==, tableContent=null), ArticleFig(id=1245407868330033193, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, language=CN, label=图3, caption=空洞卷积

黄色圆点表示卷积核;绿色部分表示卷积操作后的感受野范围

, figureFileSmall=roF8jkykvku085MTWs84Rw==, figureFileBig=mXMcpJvDQML5lnluAY4/Xg==, tableContent=null), ArticleFig(id=1245407868409724978, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, language=EN, label=Fig.4, caption=AFPN module, figureFileSmall=VwXHJBLvZjLQSdzl316dNQ==, figureFileBig=g17CW5lxTuBooIOJu46JEw==, tableContent=null), ArticleFig(id=1245407868518776889, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, language=CN, label=图4, caption=AFPN模块, figureFileSmall=VwXHJBLvZjLQSdzl316dNQ==, figureFileBig=g17CW5lxTuBooIOJu46JEw==, tableContent=null), ArticleFig(id=1245407868615245896, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, language=EN, label=Fig.5, caption=MDTA attention mechanism, figureFileSmall=5LDlYUXZ8yX16BF+feemuQ==, figureFileBig=/OMGOII3vQqUtirkBNvw2Q==, tableContent=null), ArticleFig(id=1245407868736880724, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, language=CN, label=图5, caption=MDTA注意力机制, figureFileSmall=5LDlYUXZ8yX16BF+feemuQ==, figureFileBig=/OMGOII3vQqUtirkBNvw2Q==, tableContent=null), ArticleFig(id=1245407868845932639, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, language=EN, label=Fig. 6, caption=Improve the overall structure of PSPNet network, figureFileSmall=gikp4aK3e43Arg+JjdT11A==, figureFileBig=Ya0zIgDnRI3PM2S/9D7fIw==, tableContent=null), ArticleFig(id=1245407868963373164, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, language=CN, label=图6, caption=改进PSPNet 网络总体结构, figureFileSmall=gikp4aK3e43Arg+JjdT11A==, figureFileBig=Ya0zIgDnRI3PM2S/9D7fIw==, tableContent=null), ArticleFig(id=1245407869080813689, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, language=EN, label=Fig.7, caption=Number of pixels in different categories, figureFileSmall=DsvEXFSV+JPVr/dAb0NzaA==, figureFileBig=CpUTWhOgxPyynl+kTtnXaw==, tableContent=null), ArticleFig(id=1245407869194059906, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, language=CN, label=图7, caption=不同类别像素个数, figureFileSmall=DsvEXFSV+JPVr/dAb0NzaA==, figureFileBig=CpUTWhOgxPyynl+kTtnXaw==, tableContent=null), ArticleFig(id=1245407869336666259, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, language=EN, label=Fig.8, caption=Comparison of segmentation effect of ablation experiment, figureFileSmall=MX0xYE9TA21OF2EA0lwhSg==, figureFileBig=7uaYOrgak27dZuEI29GwVw==, tableContent=null), ArticleFig(id=1245407869449912479, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, language=CN, label=图8, caption=消融实验分割效果对比图, figureFileSmall=MX0xYE9TA21OF2EA0lwhSg==, figureFileBig=7uaYOrgak27dZuEI29GwVw==, tableContent=null), ArticleFig(id=1245407869563158696, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, language=EN, label=Fig.9, caption=Comparison of mIoU and mPA of various networks, figureFileSmall=0fYzrvQOJPPq68fZJJ6Jhw==, figureFileBig=/PlZ+zwgIQU+0u94ZAJYLg==, tableContent=null), ArticleFig(id=1245407869672210609, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, language=CN, label=图9, caption=各种网络的mIoU和mPA对比, figureFileSmall=0fYzrvQOJPPq68fZJJ6Jhw==, figureFileBig=/PlZ+zwgIQU+0u94ZAJYLg==, tableContent=null), ArticleFig(id=1245407869785456829, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, language=EN, label=Fig.10, caption=Comparison of parameters,weights and computations of various network, figureFileSmall=ROLZjoEtpKk31jlNqg5U+A==, figureFileBig=JOQmgH7905MnManZLdIADg==, tableContent=null), ArticleFig(id=1245407869928063170, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, language=CN, label=图10, caption=各种网络的参数量、权重和计算量对比图, figureFileSmall=ROLZjoEtpKk31jlNqg5U+A==, figureFileBig=JOQmgH7905MnManZLdIADg==, tableContent=null), ArticleFig(id=1245407870070669523, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, language=EN, label=Fig.11, caption=Different network recognition results, figureFileSmall=R3A6CppDbS3zUzrmd7Rwxw==, figureFileBig=Kcdw9cTtMY5Aw7nfWDLBhA==, tableContent=null), ArticleFig(id=1245407870192304349, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, language=CN, label=图11, caption=不同网络识别结果, figureFileSmall=R3A6CppDbS3zUzrmd7Rwxw==, figureFileBig=Kcdw9cTtMY5Aw7nfWDLBhA==, tableContent=null), ArticleFig(id=1245407870288773354, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, language=EN, label=Fig.12, caption=System operation interface, figureFileSmall=JhJW0daM8ptQYQH3vcENYQ==, figureFileBig=pLxcGxq31gHKWTmFCa5mNg==, tableContent=null), ArticleFig(id=1245407870427185400, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, language=CN, label=图12, caption=系统操作界面, figureFileSmall=JhJW0daM8ptQYQH3vcENYQ==, figureFileBig=pLxcGxq31gHKWTmFCa5mNg==, tableContent=null), ArticleFig(id=1245407870603346184, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, language=EN, label=Fig.13, caption=Crack identification flow chart, figureFileSmall=pOWylLWCCRNTDm8SIpONQA==, figureFileBig=e/TkjaGUF6FwWD2ly+VqHQ==, tableContent=null), ArticleFig(id=1245407870704009492, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, language=CN, label=图13, caption=裂缝识别流程图, figureFileSmall=pOWylLWCCRNTDm8SIpONQA==, figureFileBig=e/TkjaGUF6FwWD2ly+VqHQ==, tableContent=null), ArticleFig(id=1245407870829838622, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, language=EN, label=Table 1, caption=

Comparison of network structure before and after MobileNetV3 optimization

, figureFileSmall=null, figureFileBig=null, tableContent=
Stage 优化前 优化后
Input Operator S Input Operator S
1 473×473×3 conv2d 2 473×473×3 conv2d 2
2 237×237×16 bneck,3×3 1 237×237×16 bneck,3×3 1
3 237×237×16 bneck,3×3 2 237×237×16 bneck,3×3 2
4 119×119×24 bneck,3×3 1 119×119×24 bneck,3×3 1
5 119×119×24 bneck,5×5 2 119×119×24 bneck,5×5 2
6 60×60×40 bneck,5×5 1 60×60×40 bneck,5×5 1
7 60×60×40 bneck,5×5 1 60×60×40 bneck,5×5 1
8 60×60×40 bneck,3×3 2 60×60×40 bneck,3×3 1
9 30×30×80 bneck,3×3 1 60×60×80 bneck,3×3,dilate=2 1
10 30×30×80 bneck,3×3 1 60×60×80 bneck,3×3,dilate=2 1
11 30×30×80 bneck,3×3 1 60×60×80 bneck,3×3,dilate=2 1
12 30×30×80 bneck,3×3 1 60×60×80 bneck,3×3,dilate=2 1
13 30×30×112 bneck,3×3 1 60×60×112 bneck,3×3,dilate=2 1
14 30×30×112 bneck,5×5 2 60×60×112 bneck,5×5 1
15 15×15×160 bneck,5×5 1 60×60×160 bneck,5×5 1
16 15×15×160 bneck,5×5 1 60×60×160 bneck,5×5 1
), ArticleFig(id=1245407870959862058, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, language=CN, label=表1, caption=

MobileNetV3优化前后网络结构对比

, figureFileSmall=null, figureFileBig=null, tableContent=
Stage 优化前 优化后
Input Operator S Input Operator S
1 473×473×3 conv2d 2 473×473×3 conv2d 2
2 237×237×16 bneck,3×3 1 237×237×16 bneck,3×3 1
3 237×237×16 bneck,3×3 2 237×237×16 bneck,3×3 2
4 119×119×24 bneck,3×3 1 119×119×24 bneck,3×3 1
5 119×119×24 bneck,5×5 2 119×119×24 bneck,5×5 2
6 60×60×40 bneck,5×5 1 60×60×40 bneck,5×5 1
7 60×60×40 bneck,5×5 1 60×60×40 bneck,5×5 1
8 60×60×40 bneck,3×3 2 60×60×40 bneck,3×3 1
9 30×30×80 bneck,3×3 1 60×60×80 bneck,3×3,dilate=2 1
10 30×30×80 bneck,3×3 1 60×60×80 bneck,3×3,dilate=2 1
11 30×30×80 bneck,3×3 1 60×60×80 bneck,3×3,dilate=2 1
12 30×30×80 bneck,3×3 1 60×60×80 bneck,3×3,dilate=2 1
13 30×30×112 bneck,3×3 1 60×60×112 bneck,3×3,dilate=2 1
14 30×30×112 bneck,5×5 2 60×60×112 bneck,5×5 1
15 15×15×160 bneck,5×5 1 60×60×160 bneck,5×5 1
16 15×15×160 bneck,5×5 1 60×60×160 bneck,5×5 1
), ArticleFig(id=1245407871173771572, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, language=EN, label=Table 2, caption=

Ablation experiment before and after MobileNetV3 optimization

, figureFileSmall=null, figureFileBig=null, tableContent=
组别 mIoU/% mPA/% 计算量/G
MobileNetV3 68.1 78.55 0.96
优化的MobileNetV3 77.88 85.23 4.2
), ArticleFig(id=1245407871299600700, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, language=CN, label=表2, caption=

MobileNetV3优化前后的消融实验

, figureFileSmall=null, figureFileBig=null, tableContent=
组别 mIoU/% mPA/% 计算量/G
MobileNetV3 68.1 78.55 0.96
优化的MobileNetV3 77.88 85.23 4.2
), ArticleFig(id=1245407871417041225, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, language=EN, label=Table 3, caption=

Ablation experiment

, figureFileSmall=null, figureFileBig=null, tableContent=
组别 优化的
MobileNetV3
AFPN MDTA Dice Loss+
Focal Loss
mIoU/% mPA/% F1/% 参数量/M 权重/MB 计算量/G
A × × × × 78.25 84.85 86.72 46.70 182.00 162.10
B 77.88 85.23 86.41 2.16 8.60 4.20
C 79.40 85.48 87.60 2.49 10.50 6.47
D 79.85 85.89 87.91 2.66 11.20 6.92
E 78.86 88.18 87.16 2.16 8.60 4.20
F 80.66 89.89 88.49 2.49 10.50 6.47
G 81.42 91.23 89.08 2.66 11.20 6.92
), ArticleFig(id=1245407871517704531, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, language=CN, label=表3, caption=

消融实验

, figureFileSmall=null, figureFileBig=null, tableContent=
组别 优化的
MobileNetV3
AFPN MDTA Dice Loss+
Focal Loss
mIoU/% mPA/% F1/% 参数量/M 权重/MB 计算量/G
A × × × × 78.25 84.85 86.72 46.70 182.00 162.10
B 77.88 85.23 86.41 2.16 8.60 4.20
C 79.40 85.48 87.60 2.49 10.50 6.47
D 79.85 85.89 87.91 2.66 11.20 6.92
E 78.86 88.18 87.16 2.16 8.60 4.20
F 80.66 89.89 88.49 2.49 10.50 6.47
G 81.42 91.23 89.08 2.66 11.20 6.92
), ArticleFig(id=1245407871668699489, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, language=EN, label=Table 4, caption=

Attention ablation experiment

, figureFileSmall=null, figureFileBig=null, tableContent=
组别 mIoU/% mPA/%
C 79.40 85.48
C+LSKA 78.92 84.24
C+ECA 79.78 85.79
C+PNA 79.69 85.68
C+SimAM 78.79 84.20
C+MDTA 79.85 85.89
), ArticleFig(id=1245407871794528628, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, language=CN, label=表4, caption=

注意力消融实验

, figureFileSmall=null, figureFileBig=null, tableContent=
组别 mIoU/% mPA/%
C 79.40 85.48
C+LSKA 78.92 84.24
C+ECA 79.78 85.79
C+PNA 79.69 85.68
C+SimAM 78.79 84.20
C+MDTA 79.85 85.89
), ArticleFig(id=1245407871911969152, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, language=EN, label=Table 5, caption=

Contrast experiment

, figureFileSmall=null, figureFileBig=null, tableContent=
组别 mIoU/% mPA/% 参数量/
M
权重/
MB
计算
量/G
FCN-ResNet50 69.50 75.75 35.30 276.0 172.6
FCN-ResNet101 68.20 74.45 54.29 424.9 263.3
DeeplabV3-ResNet50 70.60 81.90 41.90 328.0 152.6
DeeplabV3-MobileNetV3 66.70 75.85 11.02 86.4 8.7
LRASPP-MobileNetV3 69.50 75.35 3.20 25.3 2.1
Deeplabv3Plus-Xception 78.28 84.74 54.70 214.7 83.4
PSPNet-ResNet50 78.25 84.85 46.70 182.0 162.1
本文改进算法 81.42 91.23 2.66 11.2 6.9
), ArticleFig(id=1245407872000049543, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1245407859320668488, language=CN, label=表5, caption=

对比实验

, figureFileSmall=null, figureFileBig=null, tableContent=
组别 mIoU/% mPA/% 参数量/
M
权重/
MB
计算
量/G
FCN-ResNet50 69.50 75.75 35.30 276.0 172.6
FCN-ResNet101 68.20 74.45 54.29 424.9 263.3
DeeplabV3-ResNet50 70.60 81.90 41.90 328.0 152.6
DeeplabV3-MobileNetV3 66.70 75.85 11.02 86.4 8.7
LRASPP-MobileNetV3 69.50 75.35 3.20 25.3 2.1
Deeplabv3Plus-Xception 78.28 84.74 54.70 214.7 83.4
PSPNet-ResNet50 78.25 84.85 46.70 182.0 162.1
本文改进算法 81.42 91.23 2.66 11.2 6.9
)], 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.1671-1815.2309833, detailUrlEn=https://castjournals.cast.org.cn/joweb/kxjsygc/EN/10.12404/j.issn.1671-1815.2309833, pdfUrlCn=https://castjournals.cast.org.cn/joweb/kxjsygc/CN/PDF/10.12404/j.issn.1671-1815.2309833, pdfUrlEn=https://castjournals.cast.org.cn/joweb/kxjsygc/EN/PDF/10.12404/j.issn.1671-1815.2309833, aliStartDate=null, aliEndDate=null, collectionFlag=false, citedCount=null, citedUrl=null, reference=null)
收藏切换
改进PSPNet的电成像测井裂缝自动识别
收藏切换
PDF下载
申科 1 , 肖小玲 1, 2 , 张翔 2, * , 林茂山 3
科学技术与工程 | 论文·天文学、地球科学 2025,25(7): 2691-2702
收起
收藏切换
科学技术与工程 | 论文·天文学、地球科学 2025, 25(7): 2691-2702
改进PSPNet的电成像测井裂缝自动识别
全屏
申科1 , 肖小玲1, 2, 张翔2, * , 林茂山3
作者信息
  • 1 长江大学计算机科学学院, 荆州 434023
  • 2 油气资源与勘探技术教育部重点实验室(长江大学), 武汉 430100
  • 3 中国石油集团测井有限公司吐哈分公司, 哈密 839000
  • 申科(1998—),男,汉族,湖北黄冈人,硕士研究生。研究方向:基于深度学习的测井数据处理。E-mail:

通讯作者:

* 张翔(1969—),男,汉族,湖北蕲春人,博士,教授。研究方向:成像测井与地球物理信息处理。E-mail:
Improved Automatic Crack Identification for Electrical Imaging Logging Using PSPNet
Ke SHEN1 , Xiao-ling XIAO1, 2, Xiang ZHANG2, * , Mao-shan LIN3
Affiliations
  • 1 School of Computer Science, Yangtze University, Jingzhou 434023, China
  • 2 Oil and Gas Resources and Exploration Technology, Ministry of Education, Yangtze University, Wuhan 430100, China
  • 3 Tuha Branch, China National Logging Corporation, Hami 839000, China
出版时间: 2025-03-08 doi: 10.12404/j.issn.1671-1815.2309833
文章导航
收藏切换

针对裂缝特征提取困难导致裂缝分割精度低、网络参数量计算量大的问题,提出一种改进的PSPNet(pyramid scene parseing network)网络用于自动识别电成像测井图像中的裂缝。首先将PSPNet中的骨干网络替换为优化的MobileNetV3网络,减少网络参数量和计算量;其次,引入渐进特征金字塔(asymptotic feature pyramid network,AFPN),用于增加多尺度信息的交互,增强对细小裂缝的识别能力;接着,引入多深度卷积头转置注意力(multi-depthwise Conv head transposed attention,MDTA)进行全局特征的提取,提升关键信息的提取能力;最后,采用Focal Loss和Dice Loss组合相加作为损失函数,以解决数据集类别占比不平衡的问题。实验结果表明,改进的PSPNet网络对电成像测井裂缝具有较好的分割效果。与PSPNet网络相比,mIoU(mean intersection over union)提升了3.17%,mPA(mean pixel accuracy)提升了6.38%。此外,研究成果的参数量、计算量、权重分别比原模型减少94.3%、95.7%和93.8%。同时,开发了基于CIFLog的裂缝识别系统,该系统能够满足对电成像测井的实际需要。

PSPNet  /  裂缝识别  /  电成像测井图像  /  MobileNetV3  /  AFPN

An improved PSPNet(pyramid scene parseing network) network was proposed to automatically identify fractures in electrical imaging logging images, which was difficult to extract fracture features and led to low segmentation accuracy and large calculation of network parameters. Firstly, the backbone network in PSPNet was replaced with the optimized MobileNetV3 network, which could significantly reduce the number of network parameters and the amount of computation. Secondly, the asymptotic feature pyramid network(AFPN) was introduced to increase the interaction of multi-scale information and enhance the recognition ability of small cracks. Then, multi-depthwise Conv head transposed attention(MDTA) was introduced to extract global features and improve the extraction ability of key information. Finally, the combination of Focal Loss and Dice Loss were used as a loss function to solve the problem of unbalanced proportion of data sets. The experimental results show that the improved PSPNet network has a good segmentation effect on the fracture in the electrical imaging logging. Compared with the PSPNet network, mIoU(mean intersection over union) improved by 3.17% and mPA(mean pixel accuracy) improved by 6.38%. In addition, the number of parameters, calculation amount and weight of the proposed algorithm are reduced by 94.3%, 95.7% and 93.8% respectively compared with the original model. At the same time, the crack identification system based on CIFLog is developed, which can meet the practical needs of the electrical imaging logging.

PSPNet  /  fracture identification  /  electrical imaging logging image  /  MobileNetV3  /  asymptotic feature pyramid network(AFPN)
申科, 肖小玲, 张翔, 林茂山. 改进PSPNet的电成像测井裂缝自动识别. 科学技术与工程, 2025 , 25 (7) : 2691 -2702 . DOI: 10.12404/j.issn.1671-1815.2309833
Ke SHEN, Xiao-ling XIAO, Xiang ZHANG, Mao-shan LIN. Improved Automatic Crack Identification for Electrical Imaging Logging Using PSPNet[J]. Science Technology and Engineering, 2025 , 25 (7) : 2691 -2702 . DOI: 10.12404/j.issn.1671-1815.2309833
裂缝是石油和天然气储层的重要特征,对储层的渗透性和储集性有显著影响。裂缝的发育可以大幅提高油气产量[1]。使得利用测井数据对裂缝进行识别成为油气勘探领域的研究热点。
裂缝识别技术在油气勘探中具有至关重要的作用。传统的裂缝识别方法主要结合岩心观察、成像测井和常规测井等资料进行裂缝的识别[2]。苏可嘉等[3]利用岩心资料、成像测井资料和常规测井资料统计裂缝的产状、填充矿物和填充情况。通过交汇图和正态分布,总结了有填充和未填充裂缝的测井响应特征,构建了曲线变化幅度参数和曲线变化率参数,准确识别有无填充的裂缝。侯筱晓等[4]综合利用岩心、测井解释等多种资料描述裂缝发育因素,研究了多期次裂缝对天然气运移、聚集的促进作用,证明了晚期裂缝是重要的天然气运移通道。该研究为天然气开发提供了科学依据。刘浩杰等[5]提出一种新的裂缝识别方法,即通过曲线重构确定裂缝发育敏感曲线,并结合小波分解和支持向量机回归法开展裂缝预测,该方法在裂缝识别方面表现出较高的精度和效率,能够同时识别大段水平缝以及薄层裂缝。刘双莲[6]利用常规测井原理,分析了火山岩不同裂缝类型及裂缝与气体共存时的测井曲线响应特征,为在开发井中识别裂缝提供了新方法。
上述传统的裂缝识别方法需要结合岩心及多种测井数据进行综合分析,增加了研究的复杂性和成本。随着深度学习的发展以及在各个领域成功应用,基于深度学习的裂缝识别方法逐渐崭露头角。基于深度学习的方法能够自动学习数据特征,通过深度神经网络模型,从原始数据中逐层抽象出更高层次、更抽象的特征,这些特征对于后续的任务(如分类、检测、分割等)具有更好的代表性。目前基于深度学习的测井裂缝识别研究有很多,Li等[7]设计了一种结合Swin Transformer和卷积神经网络(convolutional neural networks, CNN)的分类网络模型,其中Swin Transformer用于图像全局特征提取,CNN用于边缘特征提取,然后融合这些特征进行分类预测,使“孔隙-裂缝”类型的图像预测准确率更高,准确率达到95.92%。然而,该模型的参数量和计算量较高,且分类任务只能输出一个包含每个类别标签的概率的特征向量,不能反映空间上的位置信息。Olya等[8]使用YOLOv5网络模型对裂缝进行目标检测,在少量的训练数据集下能够完全检测出具有正弦和半正弦的电成像测井图像。但目标检测任务只能用边界框进行定位,粗略地勾勒对象的范围,无法精确描述对象的形状和边界。因此,现利用语义分割网络模型对图像中的裂缝区域进行分割。由于语义分割是端到端的模型,输入和输出图像尺寸相同,能够准确表示裂缝在图像中的位置,并精确描述其形状和边界。Du等[9]使用Mask R-CNN模型实现电成像测井裂缝的检测和分割,总体识别精度较高,但在模糊和噪声的图像中会影响对裂缝的完整识别,且Mask R-CNN的参数量和计算量较大,且并未对模型进行轻量化处理。马同乐等[10]提出了融合通道和空间交叉注意力模块,该模块结合通道注意力机制和空间交叉注意力机制增强语义信息的提取来提高裂缝的识别准确率。改进模型具有更高的精度且平均交并比(mean intersection over union, mIoU)达到了74.61%。
针对上述研究在语义分割领域分割裂缝时存在分割精度低的问题,并且没有在压缩网络参数量和计算量方面做优化处理,现致力于在PSPNett(pyramid scene parseing network)网络的基础上进行改进:首先为了轻量化模型,将原始骨干网络ResNet50替换为MobileNetV3,使网络整体参数量和计算量大幅降低,并且由于直接将MobileNetV3替换到PSPNet中导致模型识别精度降低,因此将MobileNetV3进一步优化,将深度卷积的步长和空洞率做处理,达到与原始PSPNet接近的精度;其次,为了更好地学习小目标,使细小裂缝分割更完整,引入渐进特征金字塔网络(asymptotic feature pyramid network, AFPN);为了进一步提升全局特征提取能力,引入多深度卷积头转置注意力(multi-depthwise conv head transposed attention, MDTA);为改善数据集中裂缝与背景像素数量分配不均衡的问题,引入Dice Loss+Focal Loss组合损失函数。通过以上改进实现对电成像测井裂缝的分割。最后,设计CIFLog平台的裂缝识别系统,将改进算法应用于该系统当中,实现在真实应用场景中对裂缝的自动识别。
基于深度学习的语义分割网络自动化程度高,无需复杂的人工参与,主要有以下几个原因:①语义分割模型采用端到端的训练方式,从输入图像直接输出每个像素的分类结果,无需人为干预;②不需要手动设计和提取特征。模型通过大量数据的训练,自主识别出最有利于分割任务的特征;③模型训练完毕后,预测过程中只需输入图像,模型自动输出每个像素分类结果。
PSPNet语义分割网络[11]的结构图如图1所示。图1(a)为输入图像,图1(b)为通过卷积神经网络(CNN)进行特征提取后的特征图。在图1(c)处,对特征图进行4种不同尺度的平均池化(Pool):第一层将特征图划分为1×1的区域并进行平均池化,第二层划分为2×2的子区域并分别进行平均池化,第三层划分为3×3的子区域并分别进行平均池化,第四层划分为6×6的子区域并分别进行平均池化。随后通过1×1卷积(Conv)调整通道数,再采用双线性插值上采样使每个特征层的尺寸与原始图像相同。最后通过拼接操作将4个特征层与输入特征在通道维度进行特征拼接,以融合不同尺度的上下文信息,并通过1×1卷积调整通道数,使其与数据集类别一致,最后通过逐像素分类得到预测掩码图像。PSPNet网络通过金字塔池化模块捕获从局部到全局的不同尺度的上下文信息,使其实现更加高效的语义分割。
由于PSPNet网络主要的参数量和计算量是骨干网络ResNet50带来的,因此为了减少网络参数量和计算量,现引入轻量化网络MobileNetV3[12]来替换ResNet50,并对其进行优化处理。
MobileNetV3有如下特点:①使用深度可分离卷积代替传统标准卷积,在保证模型精度不下降的情况下降低计算量;②使用倒残差结构,与传统残差连接不同,倒残差结构在深度可分离卷积的基础上,通过先升维后降维的设计,进一步提升计算效率并获取更多特征信息;③引入SE(sequeeze and excitation)注意力机制,通过全局平均池化操作,计算得到每个二维特征通道的重要性分数,使网络自适应地决定每个通道在特征表示中的贡献程度。
MobileNetV3是由特征提取层和分类层组成,如表1所示的优化前网络结构展示了MobileNetV3-Large的前16个操作阶段,Stage表示操作的阶段,由于语义分割网络不需要分类层,现仅取MobileNetV3-large版本前16个阶段的特征提取层作为改进算法的骨干网络并将其进行优化。其中第1阶段是标准卷积层(conv2d),第2~第16阶段是15个瓶颈结构(bneck)堆叠而成。bneck模块如图2所示,包括1×1卷积层(Conv)、激活函数(NL)、归一化(BN)、深度卷积(Dwise)和SE注意力机制。
由于bneck模块堆叠导致特征图的空间分辨率降低,丢失了重要的空间信息。为了解决此问题,在原有的MobileNetV3结构上进行了优化,通过调整深度可分离卷积中的深度卷积步长来减少下采样的次数,并将深度卷积的空洞率改为2(dilate=2)来增大卷积的感受野。在图3中,黄色圆点表示卷积核,展示了具有空洞率为1的普通卷积和空洞率为2的空洞卷积,两种卷积的卷积核大小都为3×3。绿色部分表示卷积操作后的感受野范围,感受野的计算公式为
n=k+(k-1)(d-1)
式(1)中:k为卷积核大小;d为空洞率;n为感受野大小。
根据式(1)可知,增大空洞率可以扩大感受野,从而可以更好地获取全局特征信息。具体优化细节如表1所示。
表1优化后的网络结构所示,在第8个Stage的bneck模块中,将3×3大小的深度卷积步长由2改为1,从而保持输出特征图的长和宽为60×60。为了增大卷积核的感受野,将第9~13个Stage中的bneck模块中3×3大小的深度卷积空洞率设置为2(dilate=2),为的是在不改变输入特征图大小的同时增大卷积的感受野;第14个Stage的bneck模块中,将5×5的卷积步长由2改为1,不改变特征图的大小。
通过这些处理,优化后的MobileNetV3网络能够更好地提取全局特征并保留更多的空间位置信息,从而在裂缝识别任务中获得更好的分割性能。
由于裂缝在图像中呈现类似正弦形状的线性分布特点,既包含小目标又包含大目标,而小目标相较于大目标更难识别。在PSPNet模型中随着卷积层的不断深入,小目标信息会逐渐丢失,导致细小裂缝分割不连续。为了提高裂缝分割精度,引入AFPN[13],通过对多尺度特征(大、中、小目标特征)的充分融合来解决小目标特征小,纹理不清晰的问题。
AFPN将非相邻层级的特征利用渐进融合的方式进行充分融合。首先将两个不同分辨率的特征进行融合,随后依次融合更高级特征,以提高非相邻特征层之间的语义相关性。为防止不同特征图之间存在信息冲突的问题,AFPN中采用了自适应空间特征融合模块(adaptively spatial feature fusion,ASFF)来自适应分配空间权重,增强了关键特征层的重要性。具体表达式为
${y}_{ij}^{l}$=${\alpha }_{ij}^{l}{x}_{ij}^{1\to l}$+${\beta }_{ij}^{l}{x}_{ij}^{2\to l}$+${\gamma }_{ij}^{l}{x}_{ij}^{3\to l}$
式(2)中:${x}_{ij}^{n\to l}$为位置(i,j)处从第n层到第l层的特征向量;${\alpha }_{ij}^{l}$${\beta }_{ij}^{l}$${\gamma }_{ij}^{l}$为不同层的特征在l层的空间权重,其中${\alpha }_{ij}^{l}$+${\beta }_{ij}^{l}$+${\gamma }_{ij}^{l}$=1;${y}_{ij}^{l}$为通过自适应融合得到的特征向量。
图4展示了AFPN的网络结构,其中绿色箭头表示自适应空间特征融合过程。优化的MobileNetV3在第5、12、14阶段的输出特征,经过双线性插值及MDTA注意力机制编码后,得到F1、F2、F3,这些特征与AFPN的输入特征对齐。PSP_Feature是经过PSP模块编码,然后通过两个卷积层和双线性差值操作后得到的特征。浅层特征F1、F2、F3与深层PSP_Feature作为AFPN的不同尺度的输入特征,通过渐进融合,充分融合低层细节信息与高层语义信息,得到丰富细节信息并提升对小目标的识别能力。
注意力机制已在图像识别领域取得了重要成果,可以灵活应用于其他网络模型结构中,且通过注意力能够关注更有用的像素信息。由于裂缝像素信息提取困难,导致分割出的裂缝完整性欠佳,故引入多深度卷积头部转置注意力(multi-depthwise Conv head transposed attention, MDTA)[14]模块,利用MDTA注意力的全局特征信息提取能力,从而更准确地定位裂缝的区域。MDTA网络结构如图5所示。首先,输入的特征图X${R}^{\stackrel{\wedge }{H}\times \stackrel{\wedge }{W}\times \stackrel{\wedge }{C}}$经过归一化处理得到Y${R}^{\stackrel{\wedge }{H}\times \stackrel{\wedge }{W}\times \stackrel{\wedge }{C}}$,随后使用1×1的卷积进行跨通道信息融合,然后经过3×3的深度卷积编码局部空间上下文信息,得到query(Q)、key(K)、value(V),用于丰富局部上下文信息,随后分别进行Reshape操作,继续计算QK的跨通道交叉协方差得到全局上下文的转置注意力图(transposed-attention map)A${R}^{\stackrel{\wedge }{C}\times \stackrel{\wedge }{C}}$,最后通过聚合局部和全局的信息交互有效提升裂缝精度,QKV的计算公式分别为
$\boldsymbol{Q}=\boldsymbol{W}_{\mathrm{d}}^{Q} \boldsymbol{W}_{\mathrm{p}}^{Q} \boldsymbol{Y}$
$\boldsymbol{K}=\boldsymbol{W}_{\mathrm{d}}^{K} \boldsymbol{W}_{\mathrm{p}}^{K} \boldsymbol{Y}$
$\boldsymbol{V}=\boldsymbol{W}_{\mathrm{d}}^{V} \boldsymbol{W}_{\mathrm{p}}^{V} \boldsymbol{Y}$
式中:${W}_{p}^{(·)}$为1×1的卷积;${W}_{d}^{(·)}$为深度卷积。
MDTA注意力机制实现过程表达式分别为
$\stackrel{\wedge }{X}$=WpAttention($\stackrel{\wedge }{Q}$,$\stackrel{\wedge }{K}$,$\stackrel{\wedge }{V}$)+X
Attention($\stackrel{\wedge }{Q}$,$\stackrel{\wedge }{K}$,$\stackrel{\wedge }{V}$)=VSoftmax($\stackrel{\wedge }{K}\stackrel{\wedge }{Q}$)
式中:$\stackrel{\wedge }{X}$X分别为示输出特征图和输入特征图;α为可学习的缩放参数。
本文中将优化的MobileNetV3中3个浅层特征作为MDTA注意力机制的输入特征,挖掘浅层特征中的全局特征,提升裂缝特征的表达能力。
在电成像裂缝图像当中,裂缝的像素数量只占整个图像像素数量的少部分,在进行模型训练时所贡献的梯度较小,神经网络在反向传播过程中难以学习到裂缝特征。这种背景和前景目标的样本占比不平衡性,导致模型更难以提取裂缝区域。因此,提出将Dice Loss[15]函数作为网络训练损失函数,Dice Loss损失函数的计算公式为
Dice_Loss=1-$\frac{2\sum _{i=1}^{N}{p}_{i}{g}_{i}}{\sum _{i=1}^{N}{p}_{i}^{2}+\sum _{i=1}^{N}{g}_{i}^{2}}$
式(8)中:N为像素总数;g为真实值;p为预测值。
为更有效的应对样本不平衡的问题,引入Focal Loss[16]损失函数,表达式为
Focal_Loss(pt)=-(1-pt)γlgpt
式(9)中:lgpt为交叉熵损失函数,pt为模型预测该像素点为裂缝像素的概率值;γ为聚焦系数,取值范围为[0,+∞]。当γ=0时退化为普通交叉熵损失函数,γ越大,越能提高困难样本的学习能力,以此来解决样本类别不平衡的问题。
Focal Loss能够缓解训练过程中损失值震荡幅度大导致的预测精度不稳定的问题,因此使用Focal Loss与Dice Loss的相加的组合损失函数作为总损失函数,其计算公式为
DF_Loss=Dice_Loss+Focal_Loss
图6所示,本文算法在PSPNet的基础上进行网络改进,由于原始PSPNet的ResNet50骨干提取网络参数量和计算量过大导致很难训练,因此替换ResNet50引入优化的MobileNetV3轻量级网络结构,大幅减少参数量和计算量,并且将第9~13阶段bneck结构中的3×3的深度卷积空洞率设置为2来增大网络的感受野,并将第8和第14阶段bneck结构中步长为2的深度卷积的步长改为1,以此减少两次特征长和宽的压缩,在轻量化的基础上与原PSPNet的精度基本一致。此外,引出改进MobileNetV3的第5、第12和第14阶段的输出特征图,其特征图尺寸分别为60×60×40、60×60×112和60×60×160,随后进行双线性插值(interpolate)调整特征图长和宽得到特征图大小为128×128×40、64×64×112和32×32×160,然后将这3个特征图作为MDTA注意力机制的输入特征,充分提取浅层特征的全局特征信息。MDTA注意力编码完成后分别得到F1∈128×128×40、F2∈64×64×112和F3∈32×32×160。优化的MobileNetV3输出的特征图经过PSP模块编码后得到尺寸为60×60×320的特征图,随后经过1×1卷积得到尺寸为60×60×2的特征图,再经过线性插值缩放和1×1卷积调整特征图尺寸,得到PSP_Feature∈256×256×4。将PSP_Feature、F1、F2、F3作为AFPN模块的输入特征图,通过AFPN模块对多尺度特征的提取,提升小目标分割精度,AFPN输出的特征图通过线性插值对多级特征图进行对齐,然后通过通道拼接将多级特征图融合在一起,将融合在一起的多级特征继续与PSP模块编码和1×1卷积调整通道后的特征沿通道进行特征拼接得到尺寸为360×360×402的特征图,最后通过1×1卷积和线性插值缩放,将特征图尺寸变为473×473×2,以便进行像素的分类预测,最后得到预测出的mask分割图。
实验使用的显卡为NVDIA GeForce RTX3060,显存为12 GB。CPU型号为Intel Core i5-12400F。操作系统使用Windows11,编程语言及版本为Python3.8.2。使用PyTorch2.0+CUDA1.13.0进行模型的搭建,编程软件为PyCharm。本实验训练使用的初始学习率为0.01,使用余弦退火算法进行学习率的调整,最后衰减致0。输入图片分辨率为473×473,批次为6,训练轮数为150轮。
本文中数据集使用了402张电成像测井图像,图像经过VOC(视觉目标分类)格式的标注,标注包括背景和裂缝两类目标。将402张图像划分为320张训练集,82张验证集。分别将训练集和验证集图像进行缩放和旋转,得到1 280张训练集图像、328张验证集图像,最后统计了训练集和验证集标签中不同类别的像素个数,如图7所示。
为验证改进网络模型的有效性,使用平均像素准确率(mean pixel accuracy, mPA)、平均交并比(mean intersection over union, mIoU)、准确率P(precision)、召回率R(recall)、参数量(Parameter)、计算量(FLOPs/G)、权重、F1分数指标对网络性能进行比较。
mIoU表示的是模型预测得到的分割掩码和标注的真实掩码之间的重合程度,当预测掩码和真实掩码完全重合,表示模型预测精度已经达到最理想的结果,其计算公式为
mIoU=$\frac{1}{k+1}\sum _{i=0}^{k}\frac{{p}_{ii}}{\sum _{i=0}^{k}{p}_{ij}+\sum _{i=0}^{k}{p}_{ji}-{p}_{ii}}$×100%
式(11)中:pii为真实值为i、预测值为i的像素数量;pij为真实值为i、预测值为j的像素数量。
mPA表示每个类别的正确分类像素数与该类别真实像素数的比值的平均值,表达式为
mPA=$\frac{1}{k+1}\sum _{i=0}^{k}\frac{{p}_{ii}}{\sum _{i=0}^{k}{p}_{ij}}$×100%
准确率表示网络预测为正例的样本中真正预测正确的样本所占的比例。即
P=$\frac{TP}{TP+FP}$
召回率表示模型正确预测正样本占所有正样本的比例。即
R=$\frac{TP}{TP+FN}$
F1为准确率和召回率的调和平均数,数值越大性能越好。即
F1=$\frac{2PR}{R+P}$×100%
为证明使用空洞率和修改卷积步长比未做任何改进的MobileNetV3网络更加能够提升PSPNet的分割性能,将原始PSPNet中的ResNet50分别替换成原始MobileNetV3和优化后的MobileNetV3,训练结果如表2所示,相比于MobileNetV3,优化后MobileNetV3的mIoU和mPA分别提升了9.78%和6.68%,虽然优化后的计算量有所增加,但要远小于PSPNet的计算量。
为进一步验证改进算法的有效性,一共做了7组消融实验,A组为未改进的以ResNet50作为骨干网络的PSPNet网络,其余几组为改进方案,结果如表3所示。
通过表3可知,替换骨干网络为优化的MobileNetV3、引入AFPN、引入MDTA、使用Dice Loss+Facal Loss组合损失函数,都能有效提升裂缝分割精度。B组引入优化的MobileNetV3轻量级骨干网络,除mIoU和F1分数比原模型低0.37和0.31个百分点以外,mPA比原模型提高了0.38个百分点,优化的MobileNetV3主要优点还是在参数量、权重和计算量方面远低于PSPNet网络,能够部署在性能不高的边缘移动设备上并且得到较好的精度;C组引入优化的MobileNetV3和AFPN,mIoU、mPA、F1分数比二组提高了1.52、0.25、1.19个百分点,表明AFPN渐进式融合特征的方式能够进一步提升模型性能;D组引入优化的MobileNetV3、AFPN和MDTA,mIoU、mPA、F1分数比C组提高了0.45、0.41、0.31个百分点;E组引入优化的MobileNetV3和Dice Loss+Focal Loss,相较于B组只引入优化的MobileNetV3,mIoU、mPA、F1分数提升了0.98和2.95、0.75个百分点,说明对于样本占比不平衡的数据集,组合损失函数能够一定程度上提高模型的精度;F组引入优化的MobileNetV3、AFPN和Dice Loss+Focal Loss,相较于D组的3个改进点组合,mIoU、mPA和F1分数提高了0.81、4和0.58个百分点;G组引入优化的MobileNetV3、AFPN、MDTA和Dice+Focal,mIoU和mPA均高于其他几组,且相比于A组原模型而言,mIoU、mPA和F1分数分别提升了3.17、6.38和2.36个百分点。G组最终改进方案虽然相较于B组只引入优化的MobileNetV3而言,参数量、权重和计算量上都有略微增加,但相比于A组原网络模型,本文中改进网络的参数量、权重和计算量都远小于原网络模型。
为了更直观展现改进的效果,选择一张背景复杂且带有一条裂缝的电成像测井图像进行预测。图8中展示了消融实验中A、B、C、D、G组的分割结果图,从图8中可以看出,逐步加入各种改进方案都有不错的效果。A组使用未改进的PSPNet进行预测得到分割图,由于PSPNet没有对浅层特征充分利用,导致对裂缝分类不准确,在图8中存在大量噪声。B组在PSPNet的基础上引入优化的MobileNetV3网络结构,分割效果与PSPNet的效果差不多,但参数量和计算量分别降低95.3%、97.4%。C组在B组的基础上引入AFPN网络结构,通过渐进融合多尺度特征对狭小裂缝的分割更完整,从C组别分割结果图中可以看出裂缝断裂情况比B组要好,同时对裂缝分类更准确,噪声大幅减小。D组在C组基础上引入MDTA注意力机制,通过进一步提升浅层特征的全局特征提取能力,使裂缝整体分割更完整。G组在D组基础上引入组合损失函数,解决数据集中像素类别数量不平衡的问题,有效提升对裂缝的分类精度,减少了错分类、漏分类情况,分割出的裂缝整体更完整且噪声较少。
为验证所选取的注意力机制是最有效的,在引入优化的MobileNetV3、AFPN基础上(消融实验中的C组)引入不同的注意力机制,结果如表4所示。
引入ECA(efficient channel attention)[17]、PNA(parallel networks attention)[18]和MDTA的mIoU都有提升,具体而言,引入ECA的mIoU和mPA提升了0.38和0.31个百分点,引入PNA的mIoU和mPA提升了0.29和0.2个百分点,引入MDTA的mIoU和mPA提升了0.45和0.41个百分点;引入LSKA(large selective kernel attention)[19]和SimAM[20]的mIoU都略有下降。具体而言,引入LSKA的mIoU和mPA降低了0.48和1.24个百分点,引入SimAM的mIoU和mPA降低了0.61和1.28个百分点,总体来讲,相比于其他注意力机制,引入MDTA注意力机制的mIoU和mPA指标最高,能够有效关注裂缝信息。
为了验证改进之后的网络与其他网络在裂缝数据集上具有竞争力,使用与消融实验相同的配置环境,选用基于ResNet50和ResNet101的FCN网络[21]、基于ResNet50和MobileNetV3的DeeplabV3网络[22]、基于Xception的DeeplabV3Plus[23]网络、基于MobileNetV3的LRASPP[12]网络来进行对比实验,结果如表5所示,改进的PSPNet算法的mIoU为81.42%,比原模型提高了3.17个百分点,mPA为91.23%,比原模型提高了6.38个百分点。图9为各种网络的mIoU和mPA的对比气泡图,气泡越靠右上角表明综合精度越高,由图9可知,改进算法的mIoU和mPA均高于其他网络。图10为各种网络的参数量、权重和计算量的对比气泡图,气泡越小并且越靠左下角表明网络越轻量化,由图10可知,虽然改进算法的计算量比LRASPP多4.8 GFLOPs,但参数量和内存体积都比LRASPP小。改进后的网络参数量、权重和计算量也远小于其他网络,保证了在资源有限的终端中可以高效部署和应用。
图11中展示了基于ResNet50和ResNet101的FCN网络、基于ResNet50和MobileNetV3的Deeplabv3网络、基于Xception的DeeplabV3+网络、基于MobileNetV3的LRASPP网络,以及基于ResNet50的PSPNet和本文改进的语义分割网络(本文改进算法)在电成像测井裂缝数据集上的裂缝识别结果。
本节采用四张带有不同倾角的裂缝图片作为测试图片,图中白色虚线方框为网络识别裂缝出现断裂的情况。在水平缝的识别结果中,FCN-ResNet101和PSPNet-ResNet50识别的裂缝存在断裂,DeeplabV3-ResNet50虽然能够完全识别出裂缝,但误分割较严重,导致对噪声的屏蔽效果较差;在低角度缝的识别结果中,除DeeplabV3Plus-Xception和本文改进算法能够完整识别出裂缝以外,其他网络识别的裂缝均存在断裂;在两条中角度缝的识别结果中,只有本文算法识别的裂缝更连续,其他算法分割的裂缝都存在断裂;在高角度裂缝的识别结果中,DeeplabV3Plus-Xception和本文改进算法都能完整识别出裂缝。但本文算法还存在一些与裂缝特征相似的噪声,这些噪声导致模型误分割为裂缝。例如在本文算法分割出的低角度裂缝中,与标签对比后发现算法存在将噪声误分割为裂缝的情况。
总体而言,本文算法的误分割情况较少,裂缝连续性较好,与其他对比网络相比,本文算法的分割结果更接近标签值。
本文中利用CIFLog2.1[24]平台二次开发了电成像测井裂缝识别系统。使用NetBeans IDE 8.1和PyCharm 2020.1作为开发工具,编程语言为Java 1.8和Python 3.8,通过Pytorch搭建语义分割网络。所有开发工作均在Windows系统上完成。系统的主要功能是对电成像测井图像中的裂缝进行分割,并将分割结果显示在分割结果框中(图12)。
系统的流程图如图13所示,首先,选择一种语义分割网络,随后点击“选择图像”按钮从本地选择一张图像,选中的图像立即显示在“原图”框中。紧接着,点击“运行”按钮,系统将根据所选择的分割网络对原图进行精准的分割预测。预测完成后,分割出的图像将自动呈现于右侧的“分割结果”框中。最后点击“保存图像”按钮,将分割结果图保存到本地。整个操作流程简单高效。
针对PSPNet网络特征提取能力有限,通过引用优化的MobileNetV3网络、AFPN多尺度融合模块、MDTA注意力机制以及Dice Loss与Focal Loss的组合损失函数,提升了裂缝分割的准确性和完整性,大幅降低了模型参数量和计算量。通过对比实验,本文改进算法在轻量化和精度方面都具有显著优势。
最后,使用CIFLog平台二次开发了电成像测井图像裂缝识别系统,实现了对裂缝区域的自动识别。该系统具有一定的实用性,通过直观的交互界面和高效的分割网络算法,大大提高了数据处理的效率。
本文改进算法仍有一些不足之处,对于一些与裂缝相似的噪声,仍然不好区分,下一步工作将针对这一问题,通过优化标签数据和改进网络结构进行处理。同时,计划利用将现有图像通过生成对抗网络去生成更多诱导缝、孔洞等类型的图像数据,进一步丰富数据集种类,提升网络泛化能力和实用性。
  • 国家自然科学基金(41674136)
参考文献 引证文献
排序方式:
[1]
徐诗雨, 祝怡, 曾乙洋, . 蓬莱气田茅口组储层孔隙结构特征及分类评价[J/OL]. 天然气地球科学. https://link.cnki.net/urlid/62.1177.te.20240509.2130.002. https://link.cnki.net/urlid/62.1177.te.20240509.2130.002
Xu Shiyu, Zhu Yi, Zeng Yiyang, et al. Pore structure characteristics and classification evaluation of Maokou Formation reservoir in Penglai Gas Field[J/OL]. Natural Gas Geoscience. https://link.cnki.net/urlid/62.1177.te.20240509.2130.002. https://link.cnki.net/urlid/62.1177.te.20240509.2130.002
[2]
陈泓位, 王时林, 和源, . 四川盆地中北部 DB1井区灯影组四段沉积相及储层特征[J/OL]. 油气地质与采收率. https://doi.org/10.13673/j.pgre.202308035.
Chen Hongwei, Wang Shilin, He Yuan, et al. Sedimentary facies and reservoir characteri sticsof the fourth member of Dengying Formation, DB1 well area, north-central Sichuan Basin[J/OL]. Oil and Gas Geology and Recovery. https://doi.org/10.13673/j.pgre.202308035.
[3]
苏可嘉, 秦臻, 邓呈祥, . 致密砂岩裂缝填充识别及其测井响应特征——以鄂尔多斯盆地镇泾油田延长组为例[J]. 科学技术与工程, 2022, 22(21): 9095-9104.
Su Kejia, Qin Zhen, Deng Chengxiang, et al. Fracture-filling identification and logging response characteristics of tight sandstone: a casestudy of Yanchang Formation, Zhenjing Oilfield, Ordos Basin[J]. Science Technology and Engineering, 2022, 22(21): 9095-9104.
[4]
侯筱晓, 刘成林, 陈茜, . 碳酸盐岩构造裂缝特征与形成期次分析: 以阿姆河盆地右岸东部为例[J]. 科学技术与工程, 2024, 24(7): 2673-2682.
Hou Xiaoxiao, Liu Chenglin, Chen Qian, et al. Characteristics and formation stages of carbonate tectonic fractures: a case study of the eastern right bank of the Amu Darya Basin[J]. Science Technology and Engineering, 2024, 24(7): 2673-2682.
[5]
刘浩杰, 盖姗姗, 于文政, . 准噶尔盆地永进油田超深层致密砂岩裂缝测井识别[J]. 测井技术, 2022, 46(5): 592-598.
Liu Haojie, Gai Shanshan, Yu Wenzheng, et al. Logging identification of ultra-deep tight sandstone fractures in Yongjin Oilfield, Junggar Basin[J]. Logging Technology, 2022, 46(5): 592-598.
[6]
刘双莲. 常规测井技术识别火山岩裂缝方法研究——以松南火山岩为例[J]. 非常规油气, 2022, 9(4): 16-22.
Liu Shuanglian. Research on conventional logging techniques to identify volcanic fractures: a case study of Songnan volcanic rocks[J]. Unconventional Oil and Gas, 2022, 9(4): 16-22.
[7]
Li H, Li H, Li C, et al. Hybrid swin transformer-CNN model for pore-crack structureidentification[J]. IEEE Transactions on Geo-science and Remote Sensing, 2024, 2024: 3380390.
[8]
Olya B A M, Mohebian R, Bagheri H, et al. Toward real-time fracture detection on image logs using deep convolutional neural network YOLOv5[J]. Interpretation, 2024, 12(2): SB9-SB18.
[9]
Du L, Lu X, Li H. Automatic fracture detection from the images of electrical image logsusing Mask R-CNN[J]. Fuel, 2023, 351: 128992.
[10]
马同乐, 刘红岐, 廖海博, . 融合通道和空间交叉注意力的裂缝识别方法[J]. 地球物理学进展, 2024, 39(2): 727-736.
Ma Tongle, Liu Hongqi, Liao Haibo, et al. Fusingchannels and spatial cross attentionfor crack identification[J]. Progress in Geophysics, 2019, 39(2): 727-736.
[11]
Zhao H, Shi J, Qi X, et al. Pyramid sceneparsing network[C]// Proceedings of the IEEEConference on Computer Vision and PatternRecognition. New York: IEEE, 2017: 2881-2890.
[12]
Howard A, Sandler M, Chu G, et al. Searching for mobilenetv3[C]// Proceedings of the IEEE/CVF International Conference on Computer Vision. New York: IEEE, 2019: 1314-1324.
[13]
Yang G, Lei J, Zhu Z, et al. AFPN: asymptotic feature pyramid network for object detection[J]. arXiv preprint arXiv: 2306. 15988, 2023.
[14]
Zamir S W, Arora A, Khan S, et al. Restormer: efficient transformer for high-resolutionimage restoration[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2022: 5728-5739.
[15]
Milletari F, Navab N, Ahmadi S A, et al.Fully convolutional neural networks for volumetric medical image segmentation[C]//Proceedings of the 2016 Fourth International Conference on 3D Vision(3DV). New York: IEEE, 2016: 565-571.
[16]
Lin T Y, Goyal P, Girshick R, et al. Focal lossfor dense object detection[C]// IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017: 42(2): 318-327.
[17]
Wang Q, Wu B, Zhu P, et al. ECA-Net: efficient channel attention for deep convolutional neural networks[C]// Proceedings of the 2020IEEE/CVF Conference on Computer Visionand Pattern Recognition. Piscataway: IEEE, 2020: 11531-11539.
[18]
Goyal A, Bochkovskiy A, Deng J, et al. Nondeep networks[J]. Advances in Neural Information Processing Systems, 2022, 35: 6789-6801.
[19]
Li Y, Hou Q, Zheng Z, et al. Large selectivekernel network for remote sensing object detection[J]. arXiv preprint arXiv: 2303. 09030, 2023.
[20]
Yang L, Zhang R Y, Li L, et al. SimAM:a simple, parameter-free attention module for convolutional neural networks[C]// Proceedings of the 38th InternationalConference on Machine Learning. Online: PMLR, 2021: 11863-11874.
[21]
Long J, Shelhamer E, Darrell T. Fully convolutional networks for semantic segmentation[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2015: 3431-3440.
[22]
Chen L C, Papandreou G, Schroff F, et al. Rethinking atrous convolution for semanticimage segmentation[J]. arXiv preprint arXiv: 1706. 05587, 2017.
[23]
Chen L C, Zhu Y, Papandreou G, et al. En-coder-decoder with atrous separable convolu-tion for semantic image segmentation[C]// Proceedings of the European Conference onComputer Vision(ECCV). New York: IEEE, 2018: 801-818.
[24]
李宁, 王才志, 刘英明, . 基于Java-NetBeans的第三代测井软件CIFLog[J]. 石油学报, 2013, 34(1): 192-200.
Li Ning, Wang Caizhi, Liu Yingming, et al. The third generation logging software CIFLog based on Java-NetBeans[J]. Acta Petrolei Sinica, 2013, 34(1): 192-200.
2025年第25卷第7期
PDF下载
141
63
引用本文
BibTeX
文章信息
doi: 10.12404/j.issn.1671-1815.2309833
  • 接收时间:2023-12-13
  • 首发时间:2026-03-30
  • 出版时间:2025-03-08
补充材料
相关文章
文章信息
作者
出版历史
  • 收稿日期:2023-12-13
  • 修回日期:2024-07-09
基金
国家自然科学基金(41674136)
作者信息
    1 长江大学计算机科学学院, 荆州 434023
    2 油气资源与勘探技术教育部重点实验室(长江大学), 武汉 430100
    3 中国石油集团测井有限公司吐哈分公司, 哈密 839000

通讯作者:

* 张翔(1969—),男,汉族,湖北蕲春人,博士,教授。研究方向:成像测井与地球物理信息处理。E-mail:
参考文献
分享链接
https://castjournals.cast.org.cn/joweb/kxjsygc/CN/10.12404/j.issn.1671-1815.2309833
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