Article(id=1327706556793204943, tenantId=1146029695717560320, journalId=1327236479169368089, issueId=1327706386865173249, articleNumber=null, orderNo=null, doi=10.11720/wtyht.2026.1133, pmid=null, cstr=null, oa=null, hot=null, price=null, onlineType=0, articleFormat=0, articleType=null, articleTypeStr=null, receivedDate=1744560000000, receivedDateStr=2025-04-14, revisedDate=1762185600000, revisedDateStr=2025-11-04, acceptedDate=null, acceptedDateStr=null, onlineDate=1794479510733, onlineDateStr=2026-11-12, pubDate=1786291200000, pubDateStr=2026-08-10, doiRegisterDate=null, doiRegisterDateStr=null, onlineIssueDate=1794479510733, onlineIssueDateStr=2026-11-12, onlineJustAcceptDate=null, onlineJustAcceptDateStr=null, onlineFirstDate=null, onlineFirstDateStr=null, sourceXml=null, magXml=null, createTime=1794479510733, creator=13701087609, updateTime=1794479510733, updator=13701087609, issue=Issue{id=1327706386865173249, tenantId=1146029695717560320, journalId=1327236479169368089, year='2026', volume='50', issue='4', pageStart='615', pageEnd='822', issueExtLink='null', onlineDate='null', pubDate='1786291200000', pubDateStr='2026-08-10', beforeIssueId=null, nextIssueId=null, price=null, status=1, issueComplete=1, articleOrder=1, issueType=-1, specialIssue=null, createTime=1794479470220, creator='13701087609', updateTime=1794532608177, updator='13701087609', preIssue=null, nextIssue=null, articleTotal=null, ext={EN=IssueExt(id=1327929264084725890, tenantId=1146029695717560320, journalId=1327236479169368089, issueId=1327706386865173249, language=EN, specialIssueTitle=, coverIllustrator=null, specialIssueEditor=, specialIssueAbout=), CN=IssueExt(id=1327929264084725891, tenantId=1146029695717560320, journalId=1327236479169368089, issueId=1327706386865173249, language=CN, specialIssueTitle=, coverIllustrator=null, specialIssueEditor=, specialIssueAbout=)}, issueFiles=null, downloadFileDto=null}, startPage=750, endPage=761, ext={EN=ArticleExt(id=1327706556956782800, articleId=1327706556793204943, tenantId=1146029695717560320, journalId=1327236479169368089, language=EN, title=A three-dimensional gravity inversion method based on a multi-task FKANET framework, columnId=null, journalTitle=Geophysical and Geochemical Exploration, columnName=null, runingTitle=null, highlight=null, articleAbstract=

To achieve high-quality mapping between surface 2D gravity anomaly observations and subsurface three-dimensional (3D) density geological bodies, this paper proposes a novel deep learning network architecture—Fusion Kernel Attention Network (FKANET). Compared to conventional neural network models previously used for 3D gravity inversion, FKANET can improve the inversion accuracy of both geometry and density amplitude of subsurface 3D density geological bodies while significantly reducing the number of network parameters. Furthermore, to address the widespread noise in actual gravity anomaly observations, this study reconstructs an improved network model based on a multi-task framework using the FKANET architecture, aiming to achieve simultaneous inversion and denoising of noisy data. Both tasks share the same encoder to extract gravity anomaly features from the raw gravity anomaly observations. In contrast, two independent decoders are used to perform denoising and 3D geological body inversion individually. Besides the adaptive denoising capability to effectively suppress the noise interferences in gravity anomaly observations, this multi-task framework poses sufficient constraints on the inversion process to enhance the model's ability to resolve subsurface 3D density geological bodies. These characteristics collectively enhance the stability and accuracy of inversion results. This method was applied to the inversion of field data from the Sannicolas region in central Mexico, demonstrating that the multi-task FKANET can achieve high-precision, real-time inversion of geological boundaries and density while also exhibiting excellent noise immunity.

, authors=Qi JING, Yu-Jie ZHANG*, Shun WANG, authorsList=Qi JING, Yu-Jie ZHANG, Shun WANG, authorCompany=null, correspAuthors=Yu-Jie ZHANG, authorNote=null, correspAuthorsNote=null, copyrightStatement=Copyright © 2026 Geophysical and Geochemical Exploration. All rights reserved., 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, fund=null), CN=ArticleExt(id=1327706560366751976, articleId=1327706556793204943, tenantId=1146029695717560320, journalId=1327236479169368089, language=CN, title=基于多任务FKANET框架的三维重力反演方法, columnId=1327706089249943952, journalTitle=物探与化探, columnName=方法研究信息处理仪器研制, runingTitle=null, highlight=null, articleAbstract=

为实现地表二维重力异常观测数据与地下三维密度地质体的高质量映射,本文采用了一种新型深度学习网络架构——FKANET(fusion kernel attention network),相较于以往三维重力反演常用的神经网络模型,FKANET在大幅减少网络模型参数量的同时,也提升了对地下三维密度地质体几何形态和密度幅值的反演精度。此外,针对实际重力异常观测数据中普遍存在的噪声问题,本文在FKANET网络架构上重新构建了一个基于多任务框架的改进型网络模型,实现对含噪数据的同步反演和去噪。两个任务共享相同编码器,从原始重力异常观测数据中提取重力异常特征,并分别采用两个独立的解码器完成重力异常数据的去噪任务和三维地质体的反演任务。该多任务框架不仅具备自适应去噪能力,能有效抑制重力异常观测数据中的噪声干扰,而且可以充分约束反演过程,增强模型对地下三维密度地质体的解析能力,从而提升反演结果的稳定性和精度。应用此方法于墨西哥中部SanNicolas地区进行实测数据的反演,实验结果表明,多任务FKANET可以实现对地质体边界和密度的高精度实时反演,并具有良好的抗噪能力。

, authors=景旗, 张玉洁*, 王顺, authorsList=景旗, 张玉洁, 王顺, authorCompany=null, correspAuthors=张玉洁, authorNote=

景旗(2001−),男,硕士,主要研究方向为重磁数据处理、深度学习。Email:

, correspAuthorsNote=
张玉洁(1981−),女,博士,副教授,主要研究方向为重磁数据处理、深度学习。Email:
, copyrightStatement=版权所有 © 2026 物探与化探编辑部, copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=wzaBxiEjVc6LHrKBe4wsoQ==, magXml=gfkRer1wNm06SvPBW5m6eQ==, pdfUrl=null, pdf=/OJvChFRrdfTlJAxRakE2w==, pdfFileSize=7906732, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=LTiAfx2sKBqiAaQ/mhpjvA==, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=esfRgKxu6qt/Vl2Ee5X35A==, mapNumber=null, fund=null)}, authors=[Author(id=1327706560685519085, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, orderNo=0, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=2326671705@qq.com, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1327706560815542511, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, authorId=1327706560685519085, language=EN, stringName=Qi JING, firstName=Qi, middleName=null, lastName=JING, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=School of Mathematics and Physics, China University of Geosciences, Wuhan 430074, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1327706560920400112, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, authorId=1327706560685519085, language=CN, stringName=景旗, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=中国地质大学(武汉) 数学与物理学院,湖北 武汉 430074, bio={"content":"

景旗(2001−),男,硕士,主要研究方向为重磁数据处理、深度学习。Email:

"}, bioImg=null, bioContent=

景旗(2001−),男,硕士,主要研究方向为重磁数据处理、深度学习。Email:

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1327706560589050089, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, xref=null, ext=[AuthorCompanyExt(id=1327706560601633002, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, companyId=1327706560589050089, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=School of Mathematics and Physics, China University of Geosciences, Wuhan 430074, China), AuthorCompanyExt(id=1327706560618410219, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, companyId=1327706560589050089, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=中国地质大学(武汉) 数学与物理学院,湖北 武汉 430074)])]), Author(id=1327706560991703282, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, orderNo=1, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=zhangyujie@cug.edu.cn, emailSecond=null, emailThird=null, correspondingAuthor=1, authorType=1, ext={EN=AuthorExt(id=1327706561054617844, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, authorId=1327706560991703282, language=EN, stringName=Yu-Jie ZHANG, firstName=Yu-Jie, middleName=null, lastName=ZHANG, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=*, address=School of Mathematics and Physics, China University of Geosciences, Wuhan 430074, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1327706561117532405, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, authorId=1327706560991703282, language=CN, stringName=张玉洁, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=*, address=中国地质大学(武汉) 数学与物理学院,湖北 武汉 430074, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1327706560589050089, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, xref=null, ext=[AuthorCompanyExt(id=1327706560601633002, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, companyId=1327706560589050089, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=School of Mathematics and Physics, China University of Geosciences, Wuhan 430074, China), AuthorCompanyExt(id=1327706560618410219, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, companyId=1327706560589050089, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=中国地质大学(武汉) 数学与物理学院,湖北 武汉 430074)])]), Author(id=1327706561176252663, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, orderNo=2, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1327706561260138745, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, authorId=1327706561176252663, language=EN, stringName=Shun WANG, firstName=Shun, middleName=null, lastName=WANG, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=School of Mathematics and Physics, China University of Geosciences, Wuhan 430074, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1327706561331441914, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, authorId=1327706561176252663, language=CN, stringName=王顺, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=中国地质大学(武汉) 数学与物理学院,湖北 武汉 430074, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1327706560589050089, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, xref=null, ext=[AuthorCompanyExt(id=1327706560601633002, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, companyId=1327706560589050089, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=School of Mathematics and Physics, China University of Geosciences, Wuhan 430074, China), AuthorCompanyExt(id=1327706560618410219, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, companyId=1327706560589050089, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=中国地质大学(武汉) 数学与物理学院,湖北 武汉 430074)])])], keywords=[Keyword(id=1327706561432105211, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, language=EN, orderNo=1, keyword=gravity anomaly), Keyword(id=1327706561490825468, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, language=EN, orderNo=2, keyword=denoising inversion), Keyword(id=1327706561616654589, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, language=EN, orderNo=3, keyword=Fusion Kernel Attention Network (FKANET)), Keyword(id=1327706561675374846, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, language=EN, orderNo=4, keyword=multi-task learning), Keyword(id=1327706561738289407, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, language=CN, orderNo=1, keyword=重力异常), Keyword(id=1327706561822175488, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, language=CN, orderNo=2, keyword=去噪反演), Keyword(id=1327706561897672961, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, language=CN, orderNo=3, keyword=FKANET), Keyword(id=1327706561956393218, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, language=CN, orderNo=4, keyword=多任务学习)], refs=[Reference(id=1327706565467025693, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=1, rfOrder=0, authorNames=null, journalName=null, refType=null, unstructuredReference=Li Y G, Oldenburg D W. 3-D inversion of gravity data[J]. Geophysics, 1998, 63(1): 109−119., articleTitle=null, refAbstract=null), Reference(id=1327706565534134558, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=2, rfOrder=1, authorNames=null, journalName=null, refType=null, unstructuredReference=郭信, 兰学毅, 严加永, 等. 大比例尺重力调查与成矿预测: 以江西朱溪钨矿外围为例[J]. 地质与勘探, 2020, 56(5): 985−1004., articleTitle=null, refAbstract=null), Reference(id=1327706565601243423, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, 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=null, journalName=null, refType=null, unstructuredReference=Guo X, Lan X Y, Yan J Y, et al. Large-scale gravity survey and metallogenic prediction: An example of the Zhuxi tungsten deposit in Jiangxi Province[J]. Geology and Exploration, 2020, 56(5): 985−1004., articleTitle=null, refAbstract=null), Reference(id=1327706565676740896, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=3, rfOrder=3, authorNames=null, journalName=null, refType=null, unstructuredReference=Omollo P, Nishijima J. Analysis and interpretation of the gravity data to delineate subsurface structural geometry of the Olkaria geothermal reservoir, Kenya[J]. Geothermics, 2023, 110: 102663., articleTitle=null, refAbstract=null), Reference(id=1327706565752238369, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=4, rfOrder=4, authorNames=null, journalName=null, refType=null, unstructuredReference=Martyshko P, Ladovskii I, Byzov D, et al. Structural tectonic scheme creation based on seismic-gravity models and isostasy usage: Ural case[J]. Russian Journal of Earth Sciences, 2024: 1−11., articleTitle=null, refAbstract=null), Reference(id=1327706565819347234, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=5, rfOrder=5, authorNames=null, journalName=null, refType=null, unstructuredReference=Altinoğlu F F. Mapping of the structural lineaments and sedimentary basement relief using gravity data to guide mineral exploration in the Denizli Basin[J]. Minerals, 2023, 13(10): 1276., articleTitle=null, refAbstract=null), Reference(id=1327706565920010531, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=6, rfOrder=6, authorNames=null, journalName=null, refType=null, unstructuredReference=Li Y G, Oldenburg D W. 3-D inversion of magnetic data[J]. Geophysics, 1996, 61(2): 394−408., articleTitle=null, refAbstract=null), Reference(id=1327706565991313700, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=7, rfOrder=7, authorNames=null, journalName=null, refType=null, unstructuredReference=Nagihara S, Hall S A. Three-dimensional gravity inversion using simulated annealing: Constraints on the diapiric roots of allochthonous salt structures[J]. Geophysics, 2001, 66(5): 1438−1449., articleTitle=null, refAbstract=null), Reference(id=1327706566083588389, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=8, rfOrder=8, authorNames=null, journalName=null, refType=null, unstructuredReference=陈晓, 曾志文, 邓居智, 等. 基于不等式和Gramian约束的MT和重力正则化联合反演[J]. 物探与化探, 2023, 47(3): 575−583., articleTitle=null, refAbstract=null), Reference(id=1327706566150697254, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=8, rfOrder=9, authorNames=null, journalName=null, refType=null, unstructuredReference=Chen X, Zeng Z W, Deng J Z, et al. Regularized joint inversion of magnetotelluric and gravity data based on inequality and Gramian constraints[J]. Geophysical and Geochemical Exploration, 2023, 47(3): 575−583., articleTitle=null, refAbstract=null), Reference(id=1327706566251360551, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=9, rfOrder=10, authorNames=null, journalName=null, refType=null, unstructuredReference=Li Y G, Oldenburg D W. Fast inversion of large-scale magnetic data using wavelet transforms and a logarithmic barrier method[J]. Geophysical Journal International, 2003, 152(2): 251−265., articleTitle=null, refAbstract=null), Reference(id=1327706566339440936, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=10, rfOrder=11, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhu D, Hu X Y, Liu S, et al. Gravity inversion usingL0 norm for sparse constraints[J]. Geophysical Journal International, 2023, 236(2): 904−923., articleTitle=null, refAbstract=null), Reference(id=1327706566414938409, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=11, rfOrder=12, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhdanov M S. New advances in regularized inversion of gravity and electromagnetic data[J]. Geophysical Prospecting, 2009, 57(4): 463−478., articleTitle=null, refAbstract=null), Reference(id=1327706566482047274, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=12, rfOrder=13, authorNames=null, journalName=null, refType=null, unstructuredReference=陈闫, 李桐林, 范翠松, 等. 重力梯度全张量数据三维共轭梯度聚焦反演[J]. 地球物理学进展, 2014, 29(3): 1133−1142., articleTitle=null, refAbstract=null), Reference(id=1327706566561739052, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=12, rfOrder=14, authorNames=null, journalName=null, refType=null, unstructuredReference=Chen Y, Li T L, Fan C S, et al. The 3D focusing inversion of full tensor gravity gradient data based on conjugate gradient[J]. Progress in Geophysics, 2014, 29(3): 1133−1142., articleTitle=null, refAbstract=null), Reference(id=1327706566670790957, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=13, rfOrder=15, authorNames=null, journalName=null, refType=null, unstructuredReference=Li J, Xu Z W, Jian X X, et al. Gravity and magnetic focusing inversion in revealing the metallogenic pattern of Dahongshan copper–iron deposit in the Kangdian area, China[J]. IEEE Transactions on Geoscience and Remote Sensing, 2023, 61: 1−10., articleTitle=null, refAbstract=null), Reference(id=1327706566767259950, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=14, rfOrder=16, authorNames=null, journalName=null, refType=null, unstructuredReference=李丽丽, 何照坤, 马国庆, 等. 基于物性一致性的重力及其梯度数据协同反演方法[J]. 世界地质, 2024, 43(2): 271−281., articleTitle=null, refAbstract=null), Reference(id=1327706566834368815, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=14, rfOrder=17, authorNames=null, journalName=null, refType=null, unstructuredReference=Li L L, He Z K, Ma G Q, et al. Cooperative inversion method of gravity and its gradient data based on physical consistency[J]. Global Geology, 2024, 43(2): 271−281., articleTitle=null, refAbstract=null), Reference(id=1327706568503701808, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=15, rfOrder=18, authorNames=null, journalName=null, refType=null, unstructuredReference=Chasseriau P, Chouteau M. 3D gravity inversion using a model of parameter covariance[J]. Journal of Applied Geophysics, 2003, 52(1): 59−74., articleTitle=null, refAbstract=null), Reference(id=1327706568595976497, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=16, rfOrder=19, authorNames=null, journalName=null, refType=null, unstructuredReference=Pratt G, Shin C, Hicks. Gauss-Newton and full Newton methods in frequency-space seismic waveform inversion[J]. Geophysical Journal International, 1998, 133(2): 341−362., articleTitle=null, refAbstract=null), Reference(id=1327706568663085362, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=17, rfOrder=20, authorNames=null, journalName=null, refType=null, unstructuredReference=Wei X L, Sun J J, Sen M K. Quantifying uncertainty of salt body shapes recovered from gravity data using trans-dimensional Markov chain Monte Carlo sampling[J]. Geophysical Journal International, 2022, 232(3): 1957−1978., articleTitle=null, refAbstract=null), Reference(id=1327706568734388531, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=18, rfOrder=21, authorNames=null, journalName=null, refType=null, unstructuredReference=Toushmalani R. Comparison result of inversion of gravity data of a fault by particle swarm optimization and Levenberg-Marquardt methods[J]. SpringerPlus, 2013, 2: 462., articleTitle=null, refAbstract=null), Reference(id=1327706568814080308, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=19, rfOrder=22, authorNames=null, journalName=null, refType=null, unstructuredReference=Kaftan İ. Interpretation of magnetic anomalies using a genetic algorithm[J]. Acta Geophysica, 2017, 65(4): 627−634., articleTitle=null, refAbstract=null), Reference(id=1327706568876994869, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=20, rfOrder=23, authorNames=null, journalName=null, refType=null, unstructuredReference=Liu S, Hu X Y, Liu T Y. A stochastic inversion method for potential field data: Ant colony optimization[J]. Pure and Applied Geophysics, 2014, 171(7): 1531−1555., articleTitle=null, refAbstract=null), Reference(id=1327706568935715126, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=21, rfOrder=24, authorNames=null, journalName=null, refType=null, unstructuredReference=Huang R, Liu S, Qi R, et al. Deep learning 3D sparse inversion of gravity data[J]. Journal of Geophysical Research: Solid Earth, 2021, 126(11): e2021JB022476., articleTitle=null, refAbstract=null), Reference(id=1327706569011212599, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=22, rfOrder=25, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhou S, Wei Y, Lu P Y, et al. Deep-learning gravity inversion method with depth-weighting constraints and its application in geothermal exploration[J]. Remote Sensing, 2024, 16(23): 4467., articleTitle=null, refAbstract=null), Reference(id=1327706569078321464, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=23, rfOrder=26, authorNames=null, journalName=null, refType=null, unstructuredReference=张志厚, 廖晓龙, 曹云勇, 等. 基于深度学习的重力异常与重力梯度异常联合反演[J]. 地球物理学报, 2021, 64(4): 1435−1452., articleTitle=null, refAbstract=null), Reference(id=1327706569158013241, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=23, rfOrder=27, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhang Z H, Liao X L, Cao Y Y, et al. Joint gravity and gravity gradient inversion based on deep learning[J]. Chinese Journal of Geophysics, 2021, 64(4): 1435−1452., articleTitle=null, refAbstract=null), Reference(id=1327706569241899322, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=24, rfOrder=28, authorNames=null, journalName=null, refType=null, unstructuredReference=Chen X H, Du Y K, Liu Z, et al. Inversion of density interfaces using the pseudo-backpropagation neural network method[J]. Pure and Applied Geophysics, 2018, 175(12): 4427−4447., articleTitle=null, refAbstract=null), Reference(id=1327706569309008187, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=25, rfOrder=29, authorNames=null, journalName=null, refType=null, unstructuredReference=He S Y, Cai H Z, Liu S, et al. Recovering 3D basement relief using gravity data through convolutional neural networks[J]. Journal of Geophysical Research: Solid Earth, 2021, 126(10): e2021JB022611., articleTitle=null, refAbstract=null), Reference(id=1327706569384505660, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=26, rfOrder=30, authorNames=null, journalName=null, refType=null, unstructuredReference=Chen Z X, Meng X H, Guo L H, et al. Three-dimensional fast forward modeling and the inversion strategy for large scale gravity and gravimetry data based on GPU[J]. Chinese Journal of Geophysics, 2012, 55(12): 4069−4077., articleTitle=null, refAbstract=null), Reference(id=1327706569472586045, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=27, rfOrder=31, authorNames=null, journalName=null, refType=null, unstructuredReference=Milletari F, Navab N, Ahmadi S A. V-net: Fully convolutional neural networks for volumetric medical image segmentation[C]//2016 Fourth International Conference on 3D Vision (3DV), IEEE, 2016: 565−571., articleTitle=null, refAbstract=null)], funds=null, companyList=[AuthorCompany(id=1327706560589050089, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, xref=null, ext=[AuthorCompanyExt(id=1327706560601633002, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, companyId=1327706560589050089, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=School of Mathematics and Physics, China University of Geosciences, Wuhan 430074, China), AuthorCompanyExt(id=1327706560618410219, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, companyId=1327706560589050089, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=中国地质大学(武汉) 数学与物理学院,湖北 武汉 430074)])], figs=[ArticleFig(id=1327706562078028035, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, language=EN, label=Figure 1, caption=Schematic of prism model for underground space, figureFileSmall=LCJRybMIT9uJ8BjqID+N6Q==, figureFileBig=LTiAfx2sKBqiAaQ/mhpjvA==, tableContent=null), ArticleFig(id=1327706562170302724, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, language=CN, label=图1, caption=地下空间长方体单元模型示意, figureFileSmall=LCJRybMIT9uJ8BjqID+N6Q==, figureFileBig=LTiAfx2sKBqiAaQ/mhpjvA==, tableContent=null), ArticleFig(id=1327706562363240709, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, language=EN, label=Figure 2, caption=Two representative density models in the training set, figureFileSmall=KAZrajZtOfncps+NdlvD/w==, figureFileBig=cGZcU9pXykqFyX0h23wu0w==, tableContent=null), ArticleFig(id=1327706564032573703, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, language=CN, label=图2, caption=训练集中两个具有代表性的密度模型, figureFileSmall=KAZrajZtOfncps+NdlvD/w==, figureFileBig=cGZcU9pXykqFyX0h23wu0w==, tableContent=null), ArticleFig(id=1327706564141625608, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, language=EN, label=Figure 3, caption=Five conventional density models, figureFileSmall=J02ijk9vJw5545fslz1e0g==, figureFileBig=w8BJLD6MYzoa0buupDGM5A==, tableContent=null), ArticleFig(id=1327706564208734473, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, language=CN, label=图3, caption=5类常规密度模型, figureFileSmall=J02ijk9vJw5545fslz1e0g==, figureFileBig=w8BJLD6MYzoa0buupDGM5A==, tableContent=null), ArticleFig(id=1327706564275843338, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, language=EN, label=Figure 4, caption=Multi task FKANET model structure, figureFileSmall=ABOdKQXz4s/g2RSZT+IdJw==, figureFileBig=37vZUw6RBpygz3+4O3zJrA==, tableContent=null), ArticleFig(id=1327706564351340811, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, language=CN, label=图4, caption=多任务FKANET模型结构, figureFileSmall=ABOdKQXz4s/g2RSZT+IdJw==, figureFileBig=37vZUw6RBpygz3+4O3zJrA==, tableContent=null), ArticleFig(id=1327706564401672460, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, language=EN, label=Figure 5, caption=RSAU-7 structure, figureFileSmall=y5SD6umoRZd8sOm0BlWgyw==, figureFileBig=Kv/jlNzlUogvycqx76o96Q==, tableContent=null), ArticleFig(id=1327706564456198413, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, language=CN, label=图5, caption=RSAU-7结构, figureFileSmall=y5SD6umoRZd8sOm0BlWgyw==, figureFileBig=Kv/jlNzlUogvycqx76o96Q==, tableContent=null), ArticleFig(id=1327706564519112974, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, language=EN, label=Figure 6, caption=RGHPA structure, figureFileSmall=6NOAzFBrV3d958cfxJS6oA==, figureFileBig=JiuGRoKRIwwzzPbrMEzrcw==, tableContent=null), ArticleFig(id=1327706564590416143, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, language=CN, label=图6, caption=RGHPA结构, figureFileSmall=6NOAzFBrV3d958cfxJS6oA==, figureFileBig=JiuGRoKRIwwzzPbrMEzrcw==, tableContent=null), ArticleFig(id=1327706564657525008, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, language=EN, label=Figure 7, caption=CAB structure, figureFileSmall=irz0IpWy8nCR0s/QgiDngw==, figureFileBig=dr1ZlFLECgNXLjCaxOMIaQ==, tableContent=null), ArticleFig(id=1327706564724633873, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, language=CN, label=图7, caption=CAB结构, figureFileSmall=irz0IpWy8nCR0s/QgiDngw==, figureFileBig=dr1ZlFLECgNXLjCaxOMIaQ==, tableContent=null), ArticleFig(id=1327706564791742738, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, language=EN, label=Figure 8, caption=3Dmodel views and density model cross-sections at depths of 300 m and 600 m, figureFileSmall=QSpvDlEB/roaO477M857SA==, figureFileBig=dVMYAqs6NzQQ8J1pErlJKg==, tableContent=null), ArticleFig(id=1327706564846268691, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, language=CN, label=图8, caption=模型3D视图以及密度模型在深度300m和600m的横截面, figureFileSmall=QSpvDlEB/roaO477M857SA==, figureFileBig=dVMYAqs6NzQQ8J1pErlJKg==, tableContent=null), ArticleFig(id=1327706564909183252, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, language=EN, label=Figure 9, caption=Comparison of inversion results of 10% Gaussian noise data, figureFileSmall=IOwdBJiDWiSmaZpRcN+zkQ==, figureFileBig=OW6V0472Dp5vA6I7mjhnBg==, tableContent=null), ArticleFig(id=1327706564980486421, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, language=CN, label=图9, caption=10%的高斯噪声数据反演结果对比, figureFileSmall=IOwdBJiDWiSmaZpRcN+zkQ==, figureFileBig=OW6V0472Dp5vA6I7mjhnBg==, tableContent=null), ArticleFig(id=1327706565035012374, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, language=EN, label=Figure 10, caption=Remaining gravity anomaly of the Saint Nicholas deposit (a) and residual gravity anomaly after denoising of actual data (b), figureFileSmall=mQoNuOgkXpyaajfugRCGmA==, figureFileBig=6uklzM+Nhvc4NeN1BBzqSg==, tableContent=null), ArticleFig(id=1327706565102121239, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, language=CN, label=图10, caption=圣尼古拉斯矿床的剩余重力异常(a)和对实际数据去噪后的剩余重力异常(b), figureFileSmall=mQoNuOgkXpyaajfugRCGmA==, figureFileBig=6uklzM+Nhvc4NeN1BBzqSg==, tableContent=null), ArticleFig(id=1327706565156647192, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, language=EN, label=Figure 11, caption=cross-sectional views of the inversion results of the A–A' and B–B' lines of UNet (a), UNet3+ (b), U2Net (c), multi task FKANET (d) and PCG (e) at the SanNicolas deposit, figureFileSmall=aNV+z1GFSHAPocwIxdM1ug==, figureFileBig=vwEYO/w02sAXxaBbgIK3tA==, tableContent=null), ArticleFig(id=1327706565206978841, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, language=CN, label=图11, caption=UNet(a)、UNet3+(b)、U2Net(c)、多任务FKANET(d)和PCG(e)在圣尼古拉斯矿床的A–A'线和B-B'线反演结果截面, figureFileSmall=aNV+z1GFSHAPocwIxdM1ug==, figureFileBig=vwEYO/w02sAXxaBbgIK3tA==, tableContent=null), ArticleFig(id=1327706565261504794, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, language=EN, label=Table 1, caption=

Comparison of FKANET and multi task FKANET inversion results

, figureFileSmall=null, figureFileBig=null, tableContent=
评价指标噪声水平UNetUNet3+U2NetFKANET多任务FKANET
MAE3%0.02150.02130.02060.02040.0196
5%0.02160.02140.02100.02090.0200
10%0.02300.02230.02290.02180.0216
MIoU3%0.54260.56090.57570.57990.5910
5%0.54720.55230.55320.55720.5735
10%0.50520.50110.50970.51160.5209
模型大小/(MB)1181031682427
), ArticleFig(id=1327706565345390875, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, language=CN, label=表1, caption=

FKANET与多任务FKANET反演结果的比较

, figureFileSmall=null, figureFileBig=null, tableContent=
评价指标噪声水平UNetUNet3+U2NetFKANET多任务FKANET
MAE3%0.02150.02130.02060.02040.0196
5%0.02160.02140.02100.02090.0200
10%0.02300.02230.02290.02180.0216
MIoU3%0.54260.56090.57570.57990.5910
5%0.54720.55230.55320.55720.5735
10%0.50520.50110.50970.51160.5209
模型大小/(MB)1181031682427
)], attaches=null, journal=Journal(id=1327236024032858132, delFlag=0, nameCn=物探与化探, nameEn=Geophysical and Geochemical Exploration, nameHistory1=null, nameHistory2=null, issn=1000-8918, eissn=null, cn=11-1906/P, coden=null, periodic=1, language=CN, oaType=null, 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=Q9arnnRH1j311w2tJP+SLw==, journalPrice=null, startedYear=null, abbrevIsoEn=Geophysical and Geochemical Exploration, journalRemark=null, publicationField=null, createdTime=1794367326984, updatedTime=1794367816166, createdBy=18614031015, updatedBy=13701087609, firstLetterCn=W, firstLetterEn=W, subjectCode=Engineering, subjectName=null, subjectCodeEn=Engineering, subjectNameEn=null, picCn=Q9arnnRH1j311w2tJP+SLw==, picEn=83mUIs+38HfkGogm/M2tdQ==, jcr=null, cjcr=null, exts=[JournalExt(id=1327238075878584575, language=CN, name=物探与化探, nameHistory1=null, nameHistory2=null, managedBy=, sponsoredBy=, publishedBy=, editorOffice=, officeProv=null, officeCity=null, officeAddr=, officeZip=, editDirector=, officeDirector=null, officePhone=null, coverPicUrl=null, journalRemark=, submitArticleUrl=null, websiteUrl=, createdTime=1794367816180, updatedTime=1794367816180, createdBy=13701087609, updatedBy=13701087609, submissionGuidelinesUrl=, submissionAuthorUrl=https://www.wutanyuhuatan.com/journalx_wt/authorLogOn.action, submissionEditorUrl=https://www.wutanyuhuatan.com/journalx_wt/editorLogOn.action, submissionReviewUrl=https://www.wutanyuhuatan.com/journalx_wt/expertLogOn.action, submissionCeEditorUrl=https://www.wutanyuhuatan.com/journalx_wt/editorInChiefLogOn.action, submissionAeEditorUrl=, option={"copyright":""}), JournalExt(id=1327238075937304832, language=EN, name=Geophysical and Geochemical Exploration, nameHistory1=null, nameHistory2=null, managedBy=, sponsoredBy=, publishedBy=, editorOffice=, officeProv=null, officeCity=null, officeAddr=, officeZip=, editDirector=, officeDirector=null, officePhone=null, coverPicUrl=null, journalRemark=, submitArticleUrl=null, websiteUrl=, createdTime=1794367816194, updatedTime=1794367816194, createdBy=13701087609, updatedBy=13701087609, submissionGuidelinesUrl=, submissionAuthorUrl=https://www.wutanyuhuatan.com/journalx_wt/authorLogOn.action, submissionEditorUrl=https://www.wutanyuhuatan.com/journalx_wt/editorLogOn.action, submissionReviewUrl=https://www.wutanyuhuatan.com/journalx_wt/expertLogOn.action, submissionCeEditorUrl=https://www.wutanyuhuatan.com/journalx_wt/editorInChiefLogOn.action, submissionAeEditorUrl=, option={"copyright":""})], databaseList=null, tenantJournalId=1327236479169368089, websiteList=[Website(id=1327238745817985575, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1327236479169368089, 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/wtyht/CN, language=CN, createTime=1794367975906, createBy=18614031015, updateTime=1794368283036, updateBy=18614031015, name=物探与化探-中文, tplId=1146099689490845704, title=物探与化探, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1327240131204984887, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1327238745817985575, code=articleTextType, value=kx, createTime=1794368306208, updateTime=1794368306208, creator=18614031015, updator=18614031015), WebsiteProps(id=1327240131171430452, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1327238745817985575, code=banner, value=null, createTime=1794368306200, updateTime=1794368306200, creator=18614031015, updator=18614031015), WebsiteProps(id=1327240131234345018, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1327238745817985575, code=grayFlag, value=0, createTime=1794368306215, updateTime=1794368306215, creator=18614031015, updator=18614031015), WebsiteProps(id=1327240131163041843, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1327238745817985575, code=logo, value=https://castjournals.cast.org.cn/joweb/wtyht/CN/file/pic?fileId=vZZWCUTLvEhr0THTPpfgig==, createTime=1794368306198, updateTime=1794368306198, creator=18614031015, updator=18614031015), WebsiteProps(id=1327240131255316540, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1327238745817985575, code=minRunFlag, value=0, createTime=1794368306220, updateTime=1794368306220, creator=18614031015, updator=18614031015), WebsiteProps(id=1327240131196596278, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1327238745817985575, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/wtyht/CN/file/pic, createTime=1794368306206, updateTime=1794368306206, creator=18614031015, updator=18614031015), WebsiteProps(id=1327240131246927931, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1327238745817985575, code=silenceFlag, value=0, createTime=1794368306218, updateTime=1794368306218, creator=18614031015, updator=18614031015), WebsiteProps(id=1327240131184013365, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1327238745817985575, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_cn_619/, createTime=1794368306203, updateTime=1794368306203, creator=18614031015, updator=18614031015), WebsiteProps(id=1327240131213373496, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1327238745817985575, code=themeColor, value=null, createTime=1794368306210, updateTime=1794368306210, creator=18614031015, updator=18614031015), WebsiteProps(id=1327240131225956409, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1327238745817985575, code=themeStyle, value=null, createTime=1794368306213, updateTime=1794368306213, creator=18614031015, updator=18614031015)]), Website(id=1327238745885094441, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1327236479169368089, 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/wtyht/EN, language=EN, createTime=1794367975922, createBy=18614031015, updateTime=1794368278613, updateBy=18614031015, name=物探与化探-英文, tplId=1146101810881728533, title=Geophysical and Geochemical Exploration, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1327240160451866691, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1327238745885094441, code=articleTextType, value=kx, createTime=1794368313181, updateTime=1794368313181, creator=18614031015, updator=18614031015), WebsiteProps(id=1327240160430895168, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1327238745885094441, code=banner, value=null, createTime=1794368313177, updateTime=1794368313177, creator=18614031015, updator=18614031015), WebsiteProps(id=1327240160468643910, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1327238745885094441, code=grayFlag, value=0, createTime=1794368313185, updateTime=1794368313185, creator=18614031015, updator=18614031015), WebsiteProps(id=1327240160422506559, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1327238745885094441, code=logo, value=https://castjournals.cast.org.cn/joweb/wtyht/EN/file/pic?fileId=vZZWCUTLvEhr0THTPpfgig==, createTime=1794368313174, updateTime=1794368313174, creator=18614031015, updator=18614031015), WebsiteProps(id=1327240160481226824, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1327238745885094441, code=minRunFlag, value=0, createTime=1794368313188, updateTime=1794368313188, creator=18614031015, updator=18614031015), WebsiteProps(id=1327240160443478082, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1327238745885094441, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/wtyht/EN/file/pic, createTime=1794368313180, updateTime=1794368313180, creator=18614031015, updator=18614031015), WebsiteProps(id=1327240160477032519, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1327238745885094441, code=silenceFlag, value=0, createTime=1794368313187, updateTime=1794368313187, creator=18614031015, updator=18614031015), WebsiteProps(id=1327240160439283777, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1327238745885094441, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_en_623/, createTime=1794368313178, updateTime=1794368313178, creator=18614031015, updator=18614031015), WebsiteProps(id=1327240160456060996, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1327238745885094441, code=themeColor, value=null, createTime=1794368313182, updateTime=1794368313182, creator=18614031015, updator=18614031015), WebsiteProps(id=1327240160464449605, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1327238745885094441, code=themeStyle, value=null, createTime=1794368313184, updateTime=1794368313184, creator=18614031015, updator=18614031015)])], journalTitle=物探与化探, weixinUrl=null, journalUrl=https://www.wutanyuhuatan.com/, iacademicId=null, status=1, seqNo=null, journalTitleEn=Geophysical and Geochemical Exploration, journalPhotoCn=Q9arnnRH1j311w2tJP+SLw==, journalPhotoEn=83mUIs+38HfkGogm/M2tdQ==, journalFirstLetter=W, 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=, provinceCode=null, provinceName=null, collectFlag=false, interPubPlatform=, interPubPlatformUrl=null), detailUrlCn=https://castjournals.cast.org.cn/joweb/wtyht/CN/10.11720/wtyht.2026.1133, detailUrlEn=https://castjournals.cast.org.cn/joweb/wtyht/EN/10.11720/wtyht.2026.1133, pdfUrlCn=https://castjournals.cast.org.cn/joweb/wtyht/CN/PDF/10.11720/wtyht.2026.1133, pdfUrlEn=https://castjournals.cast.org.cn/joweb/wtyht/EN/PDF/10.11720/wtyht.2026.1133, aliStartDate=0, aliEndDate=0, collectionFlag=false, citedCount=null, citedUrl=null, previewStatus=0, delFlag=0, hasFullText=1, orderTime=1786291200000, fullTextJson=null, articleText=null, reference=null)
收藏切换
基于多任务FKANET框架的三维重力反演方法
景旗 , 张玉洁 * , 王顺
物探与化探 | 方法研究信息处理仪器研制 2026,50(4): 750-761
收起
收藏切换
物探与化探 |方法研究信息处理仪器研制 2026 , 50 (4) : 750 -761
基于多任务FKANET框架的三维重力反演方法
全屏
[Author(id=1327706560685519085, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, orderNo=0, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=2326671705@qq.com, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1327706560815542511, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, authorId=1327706560685519085, language=EN, stringName=Qi JING, firstName=Qi, middleName=null, lastName=JING, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=School of Mathematics and Physics, China University of Geosciences, Wuhan 430074, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1327706560920400112, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, authorId=1327706560685519085, language=CN, stringName=景旗, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=中国地质大学(武汉) 数学与物理学院,湖北 武汉 430074, bio={"content":"

景旗(2001−),男,硕士,主要研究方向为重磁数据处理、深度学习。Email:

"}, bioImg=null, bioContent=

景旗(2001−),男,硕士,主要研究方向为重磁数据处理、深度学习。Email:

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1327706560589050089, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, xref=null, ext=[AuthorCompanyExt(id=1327706560601633002, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, companyId=1327706560589050089, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=School of Mathematics and Physics, China University of Geosciences, Wuhan 430074, China), AuthorCompanyExt(id=1327706560618410219, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, companyId=1327706560589050089, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=中国地质大学(武汉) 数学与物理学院,湖北 武汉 430074)])]), Author(id=1327706560991703282, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, orderNo=1, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=zhangyujie@cug.edu.cn, emailSecond=null, emailThird=null, correspondingAuthor=1, authorType=1, ext={EN=AuthorExt(id=1327706561054617844, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, authorId=1327706560991703282, language=EN, stringName=Yu-Jie ZHANG, firstName=Yu-Jie, middleName=null, lastName=ZHANG, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=*, address=School of Mathematics and Physics, China University of Geosciences, Wuhan 430074, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1327706561117532405, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, authorId=1327706560991703282, language=CN, stringName=张玉洁, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=*, address=中国地质大学(武汉) 数学与物理学院,湖北 武汉 430074, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1327706560589050089, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, xref=null, ext=[AuthorCompanyExt(id=1327706560601633002, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, companyId=1327706560589050089, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=School of Mathematics and Physics, China University of Geosciences, Wuhan 430074, China), AuthorCompanyExt(id=1327706560618410219, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, companyId=1327706560589050089, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=中国地质大学(武汉) 数学与物理学院,湖北 武汉 430074)])]), Author(id=1327706561176252663, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, orderNo=2, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1327706561260138745, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, authorId=1327706561176252663, language=EN, stringName=Shun WANG, firstName=Shun, middleName=null, lastName=WANG, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=School of Mathematics and Physics, China University of Geosciences, Wuhan 430074, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1327706561331441914, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, authorId=1327706561176252663, language=CN, stringName=王顺, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=中国地质大学(武汉) 数学与物理学院,湖北 武汉 430074, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1327706560589050089, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, xref=null, ext=[AuthorCompanyExt(id=1327706560601633002, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, companyId=1327706560589050089, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=School of Mathematics and Physics, China University of Geosciences, Wuhan 430074, China), AuthorCompanyExt(id=1327706560618410219, tenantId=1146029695717560320, journalId=1327236479169368089, articleId=1327706556793204943, companyId=1327706560589050089, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=中国地质大学(武汉) 数学与物理学院,湖北 武汉 430074)])])]
景旗 , 张玉洁* , 王顺
作者信息
  • 中国地质大学(武汉) 数学与物理学院,湖北 武汉 430074
通讯作者:
张玉洁(1981−),女,博士,副教授,主要研究方向为重磁数据处理、深度学习。Email:
作者简介:

景旗(2001−),男,硕士,主要研究方向为重磁数据处理、深度学习。Email:

A three-dimensional gravity inversion method based on a multi-task FKANET framework
Qi JING , Yu-Jie ZHANG* , Shun WANG
Affiliations
  • School of Mathematics and Physics, China University of Geosciences, Wuhan 430074, China
出版时间: 2026-08-10 doi: 10.11720/wtyht.2026.1133
文章导航
收藏切换

为实现地表二维重力异常观测数据与地下三维密度地质体的高质量映射,本文采用了一种新型深度学习网络架构——FKANET(fusion kernel attention network),相较于以往三维重力反演常用的神经网络模型,FKANET在大幅减少网络模型参数量的同时,也提升了对地下三维密度地质体几何形态和密度幅值的反演精度。此外,针对实际重力异常观测数据中普遍存在的噪声问题,本文在FKANET网络架构上重新构建了一个基于多任务框架的改进型网络模型,实现对含噪数据的同步反演和去噪。两个任务共享相同编码器,从原始重力异常观测数据中提取重力异常特征,并分别采用两个独立的解码器完成重力异常数据的去噪任务和三维地质体的反演任务。该多任务框架不仅具备自适应去噪能力,能有效抑制重力异常观测数据中的噪声干扰,而且可以充分约束反演过程,增强模型对地下三维密度地质体的解析能力,从而提升反演结果的稳定性和精度。应用此方法于墨西哥中部SanNicolas地区进行实测数据的反演,实验结果表明,多任务FKANET可以实现对地质体边界和密度的高精度实时反演,并具有良好的抗噪能力。

重力异常  /  去噪反演  /  FKANET  /  多任务学习

To achieve high-quality mapping between surface 2D gravity anomaly observations and subsurface three-dimensional (3D) density geological bodies, this paper proposes a novel deep learning network architecture—Fusion Kernel Attention Network (FKANET). Compared to conventional neural network models previously used for 3D gravity inversion, FKANET can improve the inversion accuracy of both geometry and density amplitude of subsurface 3D density geological bodies while significantly reducing the number of network parameters. Furthermore, to address the widespread noise in actual gravity anomaly observations, this study reconstructs an improved network model based on a multi-task framework using the FKANET architecture, aiming to achieve simultaneous inversion and denoising of noisy data. Both tasks share the same encoder to extract gravity anomaly features from the raw gravity anomaly observations. In contrast, two independent decoders are used to perform denoising and 3D geological body inversion individually. Besides the adaptive denoising capability to effectively suppress the noise interferences in gravity anomaly observations, this multi-task framework poses sufficient constraints on the inversion process to enhance the model's ability to resolve subsurface 3D density geological bodies. These characteristics collectively enhance the stability and accuracy of inversion results. This method was applied to the inversion of field data from the Sannicolas region in central Mexico, demonstrating that the multi-task FKANET can achieve high-precision, real-time inversion of geological boundaries and density while also exhibiting excellent noise immunity.

gravity anomaly  /  denoising inversion  /  Fusion Kernel Attention Network (FKANET)  /  multi-task learning
景旗, 张玉洁, 王顺. 基于多任务FKANET框架的三维重力反演方法. 物探与化探, 2026 , 50 (4) : 750 -761 . DOI: 10.11720/wtyht.2026.1133
Qi JING, Yu-Jie ZHANG, Shun WANG. A three-dimensional gravity inversion method based on a multi-task FKANET framework[J]. Geophysical and Geochemical Exploration, 2026 , 50 (4) : 750 -761 . DOI: 10.11720/wtyht.2026.1133
三维重力反演是揭示地球内部物质分布与构造演化的核心地球物理手段,而重力勘探作为深部探测的关键技术,凭借其对密度差异的敏感性高、观测范围广及成本效率高等优势,在矿产资源勘查[1-2]、地壳结构解析[3-4]及刻画断裂带几何形态[5]中发挥不可替代的作用。随着地球物理学的不断发展和技术的革新,三维重力反演在地质勘探、资源勘查、环境监测等领域的应用越来越重要。
传统的三维重力反演通常采用对地下空间进行网格化离散的正则化反演方法[6-8],通过优化数据拟合项和正则化项的目标函数来反演密度的分布。常见的正则化约束包括深度加权[9]、平滑约束[10-11]、聚焦约束[12-13]和物理边界约束[14]等。为了求解这些约束条件下的反演问题,通常会结合不同的优化算法,如共轭梯度[15]、牛顿法[16]和贝叶斯方法[17]等,这些方法可以有效反演出地下介质的初步特征,但往往其反演结果的空间分辨率不高,难以充分挖掘重力异常数据的有效信息。为应对此类问题,一些启发式优化算法包括粒子群算法[18]、遗传算法[19]和蚁群算法[20]等,逐渐成为替代或补充传统方法的有效手段,这类算法借助模拟自然界或社会系统的搜索机制,在应对非线性、多解性和复杂约束问题上展现出显著优势。然而,随着地质条件的日益复杂化和观测数据规模的持续增长,传统方法在应对复杂地质结构和高维数据时暴露出一定局限性,例如对一些非线性约束条件和优化策略的初值依赖性可能导致结果收敛至局部最优解,且整体计算效率难以满足大规模高精度反演的需求。
近年来,随着人工智能技术的快速发展,深度学习方法为解决传统三维重力反演中的关键问题提供了新的思路。与传统的地球物理反演不同,基于深度学习的反演方法依赖于数据驱动的方法,将二维观测数据映射到三维物理模型中[21-23]。在三维重力反演中,深度学习能够利用已有的重力数据及其对应的地下密度分布数据,训练出准确映射两者关系的模型,从而显著改进反演流程。例如Chen等[24]将重力异常反演理论与BP神经网络(back propagation neural network,BP)相结合,将测量的重力异常作为输入,以恢复多密度界面的三维分布。He等[25]使用卷积神经网络(convolutional neural networks, CNN)直接基于重力数据进行深度到基底的反演,以恢复沉积物基底的地形。相较于传统方法,尽管深度学习方法在重力数据反演中展现出优越的全局优化能力和较高的计算效率,但仍存在如模型的泛化能力受限于训练数据的质量和覆盖范围,深度学习方法缺乏对实际数据噪声的鲁棒性和计算资源消耗较大等问题。
因此,本文采用一种新型U型编码器—解码器框架用于三维重力反演,在编码器阶段,模型通过输入二维重力异常数据,并利用特定的编码器结构对数据进行高效编码,在保证反演精度的同时降低计算资源与时间消耗;在解码器阶段,通过体素网格重建技术还原出三维地下结构。针对实际勘探中数据可能受到噪声干扰的问题,本文在网络结构中引入一种多任务学习机制,通过联合优化任务间相关性对重力数据进行反演与去噪处理,从而提升模型的鲁棒性与实用性。
正演模拟是反演问题求解的基础,其核心在于利用正演理论计算地下质量分布对重力场的影响,从而获得观测点处的重力异常。在重力场正演建模过程中,地下介质通常被离散化为多个具有特定密度的基本单元,并通过对这些单元的引力贡献进行积分,以求解目标区域的总引力势和重力异常分布。
为满足复杂地质体建模的需求,本文采用长方体单元对地下质量进行离散化(图1),并基于图1所示的坐标系进行建模,参考Chen等[26]在该坐标系下推导的引力矢量计算表达式:
$ \begin{split}& g_z= \\ &\mathrm{G}\rho\left|\int_{x_1}^{x_2} \int_{y_1}^{y_2} \int_{z_1}^{z_2} x\ln\left(y+r\right)+y\ln\left(x+r\right)-z\tan^{-1}\frac{xy}{zr}\mathrm{d}z\mathrm{d}y\mathrm{d}x\right|,\end{split}$
式中:$ (\xi ,\eta ,\zeta ) $为观测点坐标,($ {x}_{1} $,$ {x}_{2} $),($ {y}_{1} $,$ {y}_{2} $),($ {z}_{1} $,$ {z}_{2} $)为每个场源点$ (x,y,z) $所对应的积分范围,r= $ \sqrt{{\left(\xi -x\right)}^{2}+{\left(\eta -y\right)}^{2}+{\left(\zeta -z\right)}^{2}} $是计算场源点和观测点之间的欧几里得距离。假设地下被离散为${M} $个长方体单元并且地上有$ N $个观测点,则观测方程组为如下向量矩阵形式:
$ {{\boldsymbol{d}}={\boldsymbol{Am}}} \text{,} $
式中:d是维数为$ N\times 1 $的观测数据向量;A是维数为$ N\times M $的核矩阵;$ \boldsymbol{m} $是维数为$ M\times 1 $的密度参数向量。
本文将地下空间均匀划分为32$ \times $32$ \times $16的长方体单元,地上划分为32$ \times $32个观测点网格,对应实际边长大小皆为$ 50\text{ m}\times 50\text{ m}\times 50\text{ m} $,每个长方体单元视为一个地下异常体,剩余密度均取固定值1 $ \mathrm{g}\cdot \mathrm{c}{\mathrm{m}}^{-3} $,背景密度为0 $ \mathrm{g}\cdot \mathrm{c}{\mathrm{m}}^{-3} $。通过这种划分可将复杂的地下密度分布离散化,使其能够在数值计算中进行处理,同时提高反演的精度和稳定性。
训练一个良好的神经网络模型需要大量的数据集,本文基于随机游走的方法生成了22 000个多样化的三维密度模型作为数据集,其中20 000个样本作为训练集,2 000个样本作为测试集。样本模型示意如图2所示,包括两种:由一个随机起点生成和两个随机起点生成的模型。
为了验证所提反演方法的泛化性,在验证集中生成5类常规模型,每类常规模型包含100个样本,如图3a~e所示。
这些代表性的验证模型涵盖了实际地质勘探中常见的异常体类型,能有效评估神经网络在处理不同地质构造时的鲁棒性与泛化能力,增强网络对不同地下结构的识别与适应能力,为其在真实场景中的应用提供有力支持。
实际中重力观测数据不可避免地包含噪声信息,传统正则化反演方法如Tikhonov正则化或总变差(total variation,TV)正则化等本身具有一定的稳健性,能够在反演过程中抑制噪声的影响。然而,在实际应用中仍有研究倾向于在反演前对观测数据进行滤波或光滑处理,以提高稳定性,这在某些情况下可能导致信号的过度平滑,从而影响异常体边界的解析能力。为提高对复杂噪声的适应能力并保留更多有效地质信息,本文采用了一种将去噪和反演任务协同建模的FKANET多任务框架,如图4所示,整个网络由残差U型模块(residual squeeze and excitation U-shaped network large,RSAU-L)、残差分组多轴哈达玛积注意力机制模块(residual group multi-axis Hadamard product attention,RGHPA)和通道注意力模块(channel attention block,CAB)3个模块组合而成,并采用共享编码器结构,以同时实现反演与去噪任务。其中,去噪任务有助于提取无噪声场特征,从而提升反演精度,而两个不同的解码器分别用于去噪处理和地下三维地质体的反演。
多任务FKANET的左侧为编码器部分,该部分接收带噪的二维重力数据作为输入,并输出去噪后的重力数据。编码器由3个残差U型模块(RSAU-7、RSAU-6、RSAU-5)和两个RGHPA模块组成,最终一层采用CAB模块。每个编码器模块的输出经过上采样后,与上层特征图拼接并进行卷积操作。最终,通过1×1卷积层与Sigmoid函数生成去噪后的重力数据。在解码器阶段,模型生成6个侧输出的显著性反演结果,并通过上采样将其调整至与标签相同的大小,再通过连接操作融合多级特征,最终利用1×1卷积层和Sigmoid函数生成反演模型。与传统的重力反演神经网络相比,这一结构的设计避免了为不同任务分别训练多个独立网络,并且使得阶段内多尺度特征的提取和阶段间多层次特征的聚合更加高效。通过集成这些先进的模块,与之前的重力反演常用的网络相比,FKANET显著减少了参数和计算负荷,同时提高了分割性能。
残差U型模块(RSAU-L)是受UNet网络启发而提出的一种新的模块,旨在提取阶段内多尺度特征。RSAU-L采用逐级下采样提取多尺度特征,并通过逐级上采样、级联与卷积操作将信息编码成高分辨率特征图,有效缓解了大尺度直接上采样造成的细节损失,同时通过配置不同参数从任意空间分辨率的输入特征中提取多尺度特征,减少信息丢失。以RSAU-7为例(图5),其中L表示编码器层数,$ {C}_\mathrm{in} $、$ {C}_\mathrm{out} $、$ {C}_\mathrm{mid} $分别为输入、输出通道和内层通道数。模型前向传播时,首先对输入的重力数据($ {H\times W\times}{C}_\mathrm{in} $)利用conv2d+BN+GeLU组合进行特征提取,随后,通过5次下采样使得特征图尺寸逐步减半,通道数增加,并在每层再次应用conv2d+BN+GeLU以对特征进行提取,并逐级上采样以恢复特征图尺寸。最终,将上采样特征与初始的特征合并,并引入全局注意力机制进行融合,得到最终的输出($ H\times W\times {C}_\mathrm{{out}} $)。
为了提高三维重力反演模型的分割性能,本文采用一种基于哈达玛积的残差分组多轴哈达玛积注意力机制(RGHPA),以减少计算复杂度并增强特征提取能力(图6)。该方法首先沿通道维度将输入特征图划分为4组,每组特征映射通过双线性插值(bilinear interpolation,BI)调整至特定尺度,以匹配注意力计算需求并增强空间特征的表达能力,其中前3组分别在高度—宽度(H-W)、通道—高度(C-H)和通道—宽度(C-W)方向上执行哈达玛积注意力(Hadamard product attention,HPA)计算,以捕获多方向的全局依赖关系,第4组仅采用深度可分离卷积(depthwise separable convolution,DW),以增强高级特征表达能力并降低计算量。随后,所有分组特征在通道维度上拼接,并通过另一层DW进行信息融合。
本文采用的CAB模块(channel attention block)基于通道注意力的方式,旨在提升重力反演任务中的高分辨率重建能力。如图7所示,CAB模块由两部分组成:一是由普通卷积层、GeLU激活函数与空洞卷积组合形成的特征提取结构;二是通道注意力模块。其中对卷积层通道数进行压缩和扩展以降低计算成本,并自适应调整通道特征,而通道注意力机制可增强特征表示能力,提高特征区分度,同时减少参数量和计算开销,加快训练与推理速度。CAB模块可通过调整压缩比灵活适应不同任务需求,有效聚焦于关键特征通道,并提升反演结果的分割精度和边界细节捕获能力。
在三维重力反演中,损失函数是用来衡量模型反演结果与实际观测数据之间差异程度的指标,选择适当的损失函数对于训练出准确的反演模型至关重要。Milletari等[27]通过使用Dice损失函数在目标体和背景体素数量不平衡的情况下,清晰地刻画出了小目标体的形状和位置信息(式(3)),这种函数特别适用于重力反演领域中的小型目标体检测,能够在大范围内准确捕捉小型矿体的形状和位置信息。Dice损失函数如下:
$ \text{Dice Loss}=1-\sum\limits_{i=1}^{N}\,\frac{2\left| {\boldsymbol{m}}_{i}\cap {\hat{\boldsymbol{m}}}_{i}\right| +\epsilon }{\left| {\boldsymbol{m}}_{i}\right| +\left| {\hat{\boldsymbol{m}}}_{i}\right| +\epsilon } \text{,} $
式中:$ {\boldsymbol{m}}_{i}\cap {\hat{\boldsymbol{m}}}_{i} $表示第$ i $个真实模型和重建模型重合的体积大小;$ \mathrm{\epsilon } $是一个取值为1的常数,而在多任务反演中,相应的损失函数$ L $可设计为:
$ \begin{cases}{{L}}_{\text{inv}}=\mathrm{Dice}~\mathrm{Loss} \left({\boldsymbol{m}}_{i},{\hat{\boldsymbol{m}}}_{i}\right)\\{{L}}_{\text{denoise}}=\mathrm{MSE} \left({\boldsymbol{d}}_{i},{\hat{\boldsymbol{d}}}_{i}\right)\\{L}={{L}}_{\text{inv}}+\lambda {{L}}_{\text{denoise}}\end{cases} $
式中:$ {\boldsymbol{d}}_{i} $和$ {\hat{\boldsymbol{d}}}_{i} $表示第$ i $个带噪重力数据和去噪重力数据,并取两者均方误差(mean squared error,MSE);$ {L}_{\text{inv}} $为反演损失;$ {L}_{\text{denoise}} $为去噪损失;$ \lambda $为权重系数;设为0.1。
为了进一步提升模型的反演精度和稳定性,本文引入平均绝对误差(mean absolute error,MAE)和平均交并比(mean intersection over union,MIoU)两种评价指标,在反演任务中计算如式(5)~(6):
$ \text{MAE}=\frac{1}{K}\sum\limits_{i=1}^{K}\left| {\boldsymbol{m}}_{i}-{\hat{\boldsymbol{m}}}_{i}\right| , $
$ \text{MIoU}=\frac{1}{K}\sum\limits_{i=1}^{K}\,\frac{\left| {\boldsymbol{m}}_{i}\cap {\hat{\boldsymbol{m}}}_{i}\right| }{\left| {\boldsymbol{m}}_{i}\cup {\hat{\boldsymbol{m}}}_{i}\right| } , $
式中:$ {\boldsymbol{m}}_{i}\cup {\hat{\boldsymbol{m}}}_{i} $表示第i个真实模型和反演模型所覆盖的总体积大小;K为训练样本的数量。通过MAE和MIoU两种评价指标,可以更加全面地优化和评估神经网络模型对地下三维重力反演任务的性能。
模型实验将随机游走生成的20 000个多样化三维密度模型作为神经网络的训练集,2 000个随机模型用于测试和500个常规模型用于验证,并使用正演模型作为反演标签,正演数据作为去噪标签,带噪数据作为输入,输出为反演结果。在训练阶段,设置批量大小为32,学习率为3×$ {10}^{-4} $,网络参数更新选择Adam优化器(PC配置:AMDRyzen 7 5800H 8-Core Processor 3.60 GHz,GPU:RTX3070,RAM:32 GB,PyTorch版本:1.2.0,以下实验均在此设备上进行)。
多任务FKANET的公式可以分为两个部分:
$ \begin{array}{c}\boldsymbol{m}=F_{\mathrm{inv}}\left(\boldsymbol{d}_{\mathrm{noise}}\right)=F_{\mathrm{DI}}\left[F_{\mathrm{E}}\left(\boldsymbol{d}_{\mathrm{noise}}\right)\right] \\ \boldsymbol{d}=F_{\mathrm{denoise}}\left(\boldsymbol{d}_{\mathrm{noise}}\right)=F_{\mathrm{DD}}\left[F_{\mathrm{E}}\left(\boldsymbol{d}_{\mathrm{noise}}\right)\right]\end{array}, $
式中:$ {F}_{\text{inv}} $为反演函数;$ {F}_{\text{denoise}} $为去噪函数;$ {\boldsymbol{d}}_{\text{noise}} $表示带噪声的输入重力数据;$ {F}_{\text{E}} $为编码器;$ {F}_{\text{DI}} $为解码器;$ \boldsymbol{d} $表示去噪输出;$ \boldsymbol{m} $表示反演输出。当网络使用编码器函数和去噪解码器函数学习去噪知识时,编码器和解码器都会有去噪的信息。当网络执行反演任务时,编码器将降噪后的信息传递给密度解码器函数,以获得更好的带噪数据反演结果。
在实际数据反演过程中,噪声水平具有随机性。因此,为了确保模型的泛化能力和抗噪性能,本实验在训练集和验证集的重力异常数据中分别加入2%、4%和8%的高斯噪声进行训练与验证,而在测试阶段,在测试集数据中分别加入3%、5%和10%的高斯噪声,以验证网络是否能够在训练未涉及的噪声条件下仍保持良好性能。
图8给出了类型Ⅳ和类型Ⅴ理论模型,以及在3%、5%和10%高斯噪声条件下多任务FKANET的反演结果;表1进一步列出了不同方法的定量评价指标及模型规模。在面对不同噪声水平的数据时,多任务FKANET表现出更佳的反演能力,不仅在保持较少模型参数的情况下提升了反演效果,还在处理噪声时展现出更好的鲁棒性,这对实际应用尤为重要。试验表明,多任务FKANET在各噪声水平下均优于单任务模型,尤其在处理复杂形状源时表现突出。以类型Ⅳ与类型Ⅴ模型为例,多任务FKANET能够较好识别不同埋深及复杂形态的密度异常体,能够从叠加的噪声重力场数据中精确区分出不同的密度源。
另外,为了更贴近实际应用场景,本研究利用含噪声的数据进行了实验,并将多任务FKANET与单任务FKANET方法进行了对比。在相同的数据集上引入10%的噪声水平进行测试,以检验在较强干扰或测量误差较大的条件下模型的适应能力,确保其在实际应用中的可靠性。实验结果如图9所示,图9b展示了单任务FKANET在无噪声数据集上训练所得的反演结果,尽管该方法能够近似反演异常源的形状和埋藏深度,但物理性质信息存在偏差。相比之下,图9c显示了多任务FKANET在噪声数据集上训练后的反演结果,其反演结果在水平和垂直方向上与综合模型的块体边缘吻合良好,并且展现出优异的抗噪性能,进一步验证了多任务FKANET在噪声环境下的深度学习反演方法的有效性。
SanNicolas矿床位于墨西哥萨卡特卡斯州,是一个含Au、Ag、Cu和Zn的火山块状硫化物矿床,赋存于一系列交错分布的基性火山岩和长英质火山岩里。新近纪角砾岩厚度从50~150 m不等,覆盖在包裹矿床的火山岩上。硫化物矿床密度为3.5 g$ \cdot $cm−3,产于上盘流纹岩穹丘与下盘火山—沉积层序的接触处,其余岩矿密度在2.3~2.7 g$ \cdot $cm−3之间,因此,硫化物矿体和其围岩的密度差为0.8~1.2 g$ \cdot $cm−3。从异常密度模型信息图10可以看出,图10a中黑色散点标记的是重力观测点的位置,A–A'线和B–B'线分别是北向和东向的地质剖面的位置,块状硫化物导致了图中央的主要重力异常,且其顶部位于地表以下120~200 m,底部位于地表以下400~450 m,最大重力异常高达2.2 mGal。
利用多任务FKANET方法、UNet、UNet3+、U2Net以及传统的预条件共轭梯度方法(preconditioned conjugate gradient algorithm,PCG),对研究区域的剩余重力异常数据进行了三维密度反演,图10b展示了多任务网络对实测数据的去噪结果,在去除噪声干扰的同时,有效保留了原始地质构造的主要异常特征。
图11为不同网络所重建的地下模型在A–A'线和B–B'线两个剖面上的反演结果,绿框勾勒的是硫化矿体的轮廓。从UNet、UNet3+、U2Net以及PCG的反演结果可以看出,这些方法在浅层区域的特征较为明显,但整体形状相对平坦,密度分布较为离散。相比之下,多任务FKANET反演结果在深度上表现更为精确,尤其在中心区域,其密度分布更接近实际边界,所得密度值更为精确,同时反演结果的边界更加清晰。
本文针对三维重力反演中二维观测数据向三维密度结构的精确映射及噪声干扰对反演结果的影响问题,采用一种新型深度学习多任务FKANET架构。网络在结构设计上引入RSAU-L模块与全局注意力机制,能够动态聚焦输入数据的关键信息,增强了局部与全局特征的捕捉能力,从而实现对地下密度结构的全面解析。同时,RGHPA模块和CAB注意力模块的引入,可多角度获取和整合数据特征信息,在细节与边界层面进一步提升了对密度分布的精确还原。实验结果表明,与多种现有神经网络方法相比,多任务FKANET在合成模型上的反演效果最佳,能够准确重建地下模型的结构和边界。在实际数据应用中,将该方法应用于墨西哥萨卡特卡斯州SanNicolas矿床带噪声干扰数据的反演时,所得结果与地质资料及前人研究成果一致,并表现出更小的误差,密度值的估算更为精确。
综上所述,多任务FKANET为三维重力反演提供了一种新思路,其在精确描述地下密度分布、应对噪声干扰、还原地质结构细节等方面具有显著优势,并在实际矿区地下密度异常体识别中展现出较高的应用价值。另外,需要指出的是,本文实验中密度值主要集中在正值范围,这与先验地质信息密切相关,但在某些复杂地质环境下可能无法充分反映负密度对比体的存在。未来研究可以在此基础上结合更多地球物理数据和多源信息作为约束与先验信息,同时拓展模型对正负密度变化的表达能力,以进一步提升反演结果的准确性和可靠性。

参考文献 引证文献
排序方式:
1
Li Y G, Oldenburg D W. 3-D inversion of gravity data[J]. Geophysics, 1998, 63(1): 109−119.
2
郭信, 兰学毅, 严加永, 等. 大比例尺重力调查与成矿预测: 以江西朱溪钨矿外围为例[J]. 地质与勘探, 2020, 56(5): 985−1004.
Guo X, Lan X Y, Yan J Y, et al. Large-scale gravity survey and metallogenic prediction: An example of the Zhuxi tungsten deposit in Jiangxi Province[J]. Geology and Exploration, 2020, 56(5): 985−1004.
3
Omollo P, Nishijima J. Analysis and interpretation of the gravity data to delineate subsurface structural geometry of the Olkaria geothermal reservoir, Kenya[J]. Geothermics, 2023, 110: 102663.
4
Martyshko P, Ladovskii I, Byzov D, et al. Structural tectonic scheme creation based on seismic-gravity models and isostasy usage: Ural case[J]. Russian Journal of Earth Sciences, 2024: 1−11.
5
Altinoğlu F F. Mapping of the structural lineaments and sedimentary basement relief using gravity data to guide mineral exploration in the Denizli Basin[J]. Minerals, 2023, 13(10): 1276.
6
Li Y G, Oldenburg D W. 3-D inversion of magnetic data[J]. Geophysics, 1996, 61(2): 394−408.
7
Nagihara S, Hall S A. Three-dimensional gravity inversion using simulated annealing: Constraints on the diapiric roots of allochthonous salt structures[J]. Geophysics, 2001, 66(5): 1438−1449.
8
陈晓, 曾志文, 邓居智, 等. 基于不等式和Gramian约束的MT和重力正则化联合反演[J]. 物探与化探, 2023, 47(3): 575−583.
Chen X, Zeng Z W, Deng J Z, et al. Regularized joint inversion of magnetotelluric and gravity data based on inequality and Gramian constraints[J]. Geophysical and Geochemical Exploration, 2023, 47(3): 575−583.
9
Li Y G, Oldenburg D W. Fast inversion of large-scale magnetic data using wavelet transforms and a logarithmic barrier method[J]. Geophysical Journal International, 2003, 152(2): 251−265.
10
Zhu D, Hu X Y, Liu S, et al. Gravity inversion usingL0 norm for sparse constraints[J]. Geophysical Journal International, 2023, 236(2): 904−923.
11
Zhdanov M S. New advances in regularized inversion of gravity and electromagnetic data[J]. Geophysical Prospecting, 2009, 57(4): 463−478.
12
陈闫, 李桐林, 范翠松, 等. 重力梯度全张量数据三维共轭梯度聚焦反演[J]. 地球物理学进展, 2014, 29(3): 1133−1142.
Chen Y, Li T L, Fan C S, et al. The 3D focusing inversion of full tensor gravity gradient data based on conjugate gradient[J]. Progress in Geophysics, 2014, 29(3): 1133−1142.
13
Li J, Xu Z W, Jian X X, et al. Gravity and magnetic focusing inversion in revealing the metallogenic pattern of Dahongshan copper–iron deposit in the Kangdian area, China[J]. IEEE Transactions on Geoscience and Remote Sensing, 2023, 61: 1−10.
14
李丽丽, 何照坤, 马国庆, 等. 基于物性一致性的重力及其梯度数据协同反演方法[J]. 世界地质, 2024, 43(2): 271−281.
Li L L, He Z K, Ma G Q, et al. Cooperative inversion method of gravity and its gradient data based on physical consistency[J]. Global Geology, 2024, 43(2): 271−281.
15
Chasseriau P, Chouteau M. 3D gravity inversion using a model of parameter covariance[J]. Journal of Applied Geophysics, 2003, 52(1): 59−74.
16
Pratt G, Shin C, Hicks. Gauss-Newton and full Newton methods in frequency-space seismic waveform inversion[J]. Geophysical Journal International, 1998, 133(2): 341−362.
17
Wei X L, Sun J J, Sen M K. Quantifying uncertainty of salt body shapes recovered from gravity data using trans-dimensional Markov chain Monte Carlo sampling[J]. Geophysical Journal International, 2022, 232(3): 1957−1978.
18
Toushmalani R. Comparison result of inversion of gravity data of a fault by particle swarm optimization and Levenberg-Marquardt methods[J]. SpringerPlus, 2013, 2: 462.
19
Kaftan İ. Interpretation of magnetic anomalies using a genetic algorithm[J]. Acta Geophysica, 2017, 65(4): 627−634.
20
Liu S, Hu X Y, Liu T Y. A stochastic inversion method for potential field data: Ant colony optimization[J]. Pure and Applied Geophysics, 2014, 171(7): 1531−1555.
21
Huang R, Liu S, Qi R, et al. Deep learning 3D sparse inversion of gravity data[J]. Journal of Geophysical Research: Solid Earth, 2021, 126(11): e2021JB022476.
22
Zhou S, Wei Y, Lu P Y, et al. Deep-learning gravity inversion method with depth-weighting constraints and its application in geothermal exploration[J]. Remote Sensing, 2024, 16(23): 4467.
23
张志厚, 廖晓龙, 曹云勇, 等. 基于深度学习的重力异常与重力梯度异常联合反演[J]. 地球物理学报, 2021, 64(4): 1435−1452.
Zhang Z H, Liao X L, Cao Y Y, et al. Joint gravity and gravity gradient inversion based on deep learning[J]. Chinese Journal of Geophysics, 2021, 64(4): 1435−1452.
24
Chen X H, Du Y K, Liu Z, et al. Inversion of density interfaces using the pseudo-backpropagation neural network method[J]. Pure and Applied Geophysics, 2018, 175(12): 4427−4447.
25
He S Y, Cai H Z, Liu S, et al. Recovering 3D basement relief using gravity data through convolutional neural networks[J]. Journal of Geophysical Research: Solid Earth, 2021, 126(10): e2021JB022611.
26
Chen Z X, Meng X H, Guo L H, et al. Three-dimensional fast forward modeling and the inversion strategy for large scale gravity and gravimetry data based on GPU[J]. Chinese Journal of Geophysics, 2012, 55(12): 4069−4077.
27
Milletari F, Navab N, Ahmadi S A. V-net: Fully convolutional neural networks for volumetric medical image segmentation[C]//2016 Fourth International Conference on 3D Vision (3DV), IEEE, 2016: 565−571.
2026年第50卷第4期
80
36
0
引用本文
BibTeX
文章信息
doi: 10.11720/wtyht.2026.1133
  • 接收时间:2025-04-14
  • 首发时间:2026-11-12
  • 出版时间:2026-08-10
补充材料
相关文章
文章信息
作者
出版历史
  • 收稿日期:2025-04-14
  • 修回日期:2025-11-04
基金
作者信息
    中国地质大学(武汉) 数学与物理学院,湖北 武汉 430074

通讯作者:

张玉洁(1981−),女,博士,副教授,主要研究方向为重磁数据处理、深度学习。Email:
参考文献
分享链接
https://castjournals.cast.org.cn/joweb/wtyht/CN/10.11720/wtyht.2026.1133
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
科
Family
属数
Number of
genus
种数
Number of
species
占总种数比例
Percentage of
total species (%)
属
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