Article(id=1327011715012735599, tenantId=1146029695717560320, journalId=1324759914146529478, issueId=1327011511924535414, articleNumber=null, orderNo=null, doi=10.7544/issn1000-1239.202330491, pmid=null, cstr=null, oa=null, hot=null, price=null, onlineType=0, articleFormat=0, articleType=null, articleTypeStr=research-article, receivedDate=1686499200000, receivedDateStr=2023-06-12, revisedDate=1707148800000, revisedDateStr=2024-02-06, acceptedDate=null, acceptedDateStr=null, onlineDate=1794313847544, onlineDateStr=2026-11-10, pubDate=1735660800000, pubDateStr=2025-01-01, doiRegisterDate=null, doiRegisterDateStr=null, onlineIssueDate=1794313847544, onlineIssueDateStr=2026-11-10, onlineJustAcceptDate=null, onlineJustAcceptDateStr=null, onlineFirstDate=null, onlineFirstDateStr=null, sourceXml=null, magXml=null, createTime=1794313847544, creator=13701087609, updateTime=1794313847544, updator=13701087609, issue=Issue{id=1327011511924535414, tenantId=1146029695717560320, journalId=1324759914146529478, year='2025', volume='62', issue='1', pageStart='1', pageEnd='283', issueExtLink='null', onlineDate='null', pubDate='1735660800000', pubDateStr='2025-01-01', beforeIssueId=null, nextIssueId=null, price=null, status=1, issueComplete=1, articleOrder=1, issueType=-1, specialIssue=null, createTime=1794313799123, creator='13701087609', updateTime=1794376660021, updator='13701087609', preIssue=null, nextIssue=null, articleTotal=null, ext={EN=IssueExt(id=1327275170160554491, tenantId=1146029695717560320, journalId=1324759914146529478, issueId=1327011511924535414, language=EN, specialIssueTitle=, coverIllustrator=null, specialIssueEditor=, specialIssueAbout=), CN=IssueExt(id=1327275170160554492, tenantId=1146029695717560320, journalId=1324759914146529478, issueId=1327011511924535414, language=CN, specialIssueTitle=, coverIllustrator=null, specialIssueEditor=, specialIssueAbout=)}, issueFiles=null, downloadFileDto=null}, startPage=132, endPage=143, ext={EN=ArticleExt(id=1327011715318919792, articleId=1327011715012735599, tenantId=1146029695717560320, journalId=1324759914146529478, language=EN, title=Point Cloud Segmentation Algorithm Based on Contrastive Learning and Label Mining, columnId=1325489891789910943, journalTitle=Journal of Computer Research and Development, columnName=Artificial Intelligence, runingTitle=null, highlight=null, articleAbstract=

Point cloud segmentation algorithm based on deep learning can effectively segment point clouds in high-dimensional space by designing complex feature extraction modules. However, the lack of feature mining for boundary point set results in suboptimal accuracy in boundary segmentation. Some studies have applied the idea of contrastive learning to point cloud segmentation to solve the problem of insufficient boundary region segmentation performance, but the disorder and sparse characteristics of point cloud have not been fully utilized, and the feature extraction is not accurate enough. To solve these problems, we propose CL2M to learn more accurate features of point clouds at different locations through the self-attention mechanism, and the contrastive learning method is introduced to improve the segmentation accuracy of point cloud boundaries. In the process of contrastive boundary learning, labels in semantic space are deeply mined and a contrastive boundary learning module based on label distribution is designed to make the label distribution of point cloud in high-dimensional space contain more semantic information. The model makes full use of the label distribution law to calculate the distance between distributions, and can accurately divide positive and negative samples, reducing the cumulative errors caused by conventional hard partition. The results on two public data sets show that CL2M is superior to the existing point cloud segmentation model on several evaluation indexes, which verifies the effectiveness of the model.

, authors=Hua Huang1, 2, 4, Yifan Bu1, 2, Hongli Xu1, 3, Xiaorong Wang1, 2, authorsList=Hua Huang, Yifan Bu, Hongli Xu, Xiaorong Wang, authorCompany=null, correspAuthors=null, authorNote=

born in 1977. PhD, associate professor. Member of CCF. His main research interests include computer vision, machine learning, and the applications in the railway field

born in 1998. Master. Her main research interests include 3D vision and point cloud segmentation

born in 1963. PhD, professor. Her main research interests include machine learning and machine vision

born in 1999. Master candidate. Student member of CCF. His main research interests include 3D vision and point cloud segmentation

, correspAuthorsNote=null, copyrightStatement=Copyright © 2025 Journal of Computer Research and Development. 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=1327011719215428232, articleId=1327011715012735599, tenantId=1146029695717560320, journalId=1324759914146529478, language=CN, title=基于对比学习和标签挖掘的点云分割算法, columnId=1325489892104483745, journalTitle=计算机研究与发展, columnName=人工智能, runingTitle=null, highlight=null, articleAbstract=

基于深度学习的点云分割算法通过设计复杂的特征提取模块,可以对高维空间点云进行有效的分割. 但由于缺乏对边界点集的特征挖掘,使得其对边界分割的精度欠佳. 已有将对比学习思想用于点云分割以解决边界区域分割性能不足问题的研究中,忽略了点云无序和稀疏特性,特征提取不够准确. 对此,设计了基于对比学习和标签挖掘的点云分割模型CL2M (contrastive learning label mining),通过自注意力机制学习不同位置处点云更为精准的特征,并引入对比学习方法,提高了点云边界处的分割精度. 在对比边界学习过程中通过深入挖掘语义空间中的标签并设计了基于标签分布的对比边界学习模块,使得高维空间点云标签分布包含更多的语义信息. CL2M充分利用标签的分布规律计算分布间的距离,可准确划分正负样本,减少了常规硬划分带来的累计错误. 在2个公开数据集上进行的实验结果表明,CL2M在多个评价指标上优于既有的点云分割模型,验证了模型的有效性.

, authors=黄华1, 2, 4, 卜一凡1, 2, 许宏丽1, 3, 王晓荣1, 2, authorsList=黄华, 卜一凡, 许宏丽, 王晓荣, authorCompany=null, correspAuthors=null, authorNote=

1977年生. 博士,副教授. CCF会员. 主要研究方向为计算机视觉、机器学习及其在铁路领域中的应用

1998年生. 硕士. 主要研究方向为3D视觉、点云分割

1963年生. 博士,教授. 主要研究方向为机器学习、机器视觉

1999年生. 硕士研究生. CCF学生会员. 主要研究方向为3D视觉、点云分割

, correspAuthorsNote=null, copyrightStatement=版权所有 © 2025 计算机研究与发展编辑部, copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=IqhpIwdl6LXJIAeJemFlPQ==, magXml=S/pHUmNp1DY6r3wZ+7odaQ==, pdfUrl=null, pdf=V7Rk8HLFE8dzEhgHW2sCNA==, pdfFileSize=1957860, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=ugHfQpqV5xfv9JiW++dXEA==, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=LeyMeGIh/ei0VbeY7MJPDw==, mapNumber=null, fund=null)}, authors=[Author(id=1327012209328238972, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, orderNo=0, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=hhua@bjtu.edu.cn, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1327012211068875136, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, authorId=1327012209328238972, language=EN, stringName=Hua Huang, firstName=Hua, middleName=null, lastName=Huang, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, 4, address=1School of Computer and Information Technology, Beijing Jiaotong University, Beijing 100044
2Beijing Key Lab of Traffic Data Analysis and Mining (Beijing Jiaotong University), Beijing 100044
4Frontiers Science Center for Smart High-speed Railway System (Beijing Jiaotong University), Beijing 100044, bio={"content":"

born in 1977. PhD, associate professor. Member of CCF. His main research interests include computer vision, machine learning, and the applications in the railway field

"}, bioImg=null, bioContent=

born in 1977. PhD, associate professor. Member of CCF. His main research interests include computer vision, machine learning, and the applications in the railway field

, aboutCorrespAuthor=null), CN=AuthorExt(id=1327012211135984001, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, authorId=1327012209328238972, language=CN, stringName=黄华, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, 4, address=1北京交通大学计算机与信息技术学院 北京 100044
2交通数据分析与挖掘北京市重点实验室(北京交通大学) 北京 100044
4智慧高铁系统前沿科学中心(北京交通大学) 北京 100044, bio={"img":"cTMd2Qm+tWqRhe0b6wF+FA==","content":"

1977年生. 博士,副教授. CCF会员. 主要研究方向为计算机视觉、机器学习及其在铁路领域中的应用

"}, bioImg=cTMd2Qm+tWqRhe0b6wF+FA==, bioContent=

1977年生. 博士,副教授. CCF会员. 主要研究方向为计算机视觉、机器学习及其在铁路领域中的应用

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1327012208984306031, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, xref=1, ext=[AuthorCompanyExt(id=1327012208992694640, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, companyId=1327012208984306031, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1School of Computer and Information Technology, Beijing Jiaotong University, Beijing 100044), AuthorCompanyExt(id=1327012209001083249, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, companyId=1327012208984306031, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1北京交通大学计算机与信息技术学院 北京 100044)]), AuthorCompany(id=1327012209059803506, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, xref=2, ext=[AuthorCompanyExt(id=1327012209068192115, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, companyId=1327012209059803506, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2Beijing Key Lab of Traffic Data Analysis and Mining (Beijing Jiaotong University), Beijing 100044), AuthorCompanyExt(id=1327012209072386420, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, companyId=1327012209059803506, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2交通数据分析与挖掘北京市重点实验室(北京交通大学) 北京 100044)]), AuthorCompany(id=1327012209256935800, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, xref=4, ext=[AuthorCompanyExt(id=1327012209265324409, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, companyId=1327012209256935800, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=4Frontiers Science Center for Smart High-speed Railway System (Beijing Jiaotong University), Beijing 100044), AuthorCompanyExt(id=1327012209273713018, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, companyId=1327012209256935800, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=4智慧高铁系统前沿科学中心(北京交通大学) 北京 100044)])]), Author(id=1327012211236647299, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, 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=1327012211349893510, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, authorId=1327012211236647299, language=EN, stringName=Yifan Bu, firstName=Yifan, middleName=null, lastName=Bu, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, address=1School of Computer and Information Technology, Beijing Jiaotong University, Beijing 100044
2Beijing Key Lab of Traffic Data Analysis and Mining (Beijing Jiaotong University), Beijing 100044, bio={"content":"

born in 1998. Master. Her main research interests include 3D vision and point cloud segmentation

"}, bioImg=null, bioContent=

born in 1998. Master. Her main research interests include 3D vision and point cloud segmentation

, aboutCorrespAuthor=null), CN=AuthorExt(id=1327012211429585287, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, authorId=1327012211236647299, language=CN, stringName=卜一凡, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, address=1北京交通大学计算机与信息技术学院 北京 100044
2交通数据分析与挖掘北京市重点实验室(北京交通大学) 北京 100044, bio={"img":"QJULnevITJkz6Qm5SVSfiQ==","content":"

1998年生. 硕士. 主要研究方向为3D视觉、点云分割

"}, bioImg=QJULnevITJkz6Qm5SVSfiQ==, bioContent=

1998年生. 硕士. 主要研究方向为3D视觉、点云分割

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1327012208984306031, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, xref=1, ext=[AuthorCompanyExt(id=1327012208992694640, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, companyId=1327012208984306031, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1School of Computer and Information Technology, Beijing Jiaotong University, Beijing 100044), AuthorCompanyExt(id=1327012209001083249, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, companyId=1327012208984306031, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1北京交通大学计算机与信息技术学院 北京 100044)]), AuthorCompany(id=1327012209059803506, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, xref=2, ext=[AuthorCompanyExt(id=1327012209068192115, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, companyId=1327012209059803506, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2Beijing Key Lab of Traffic Data Analysis and Mining (Beijing Jiaotong University), Beijing 100044), AuthorCompanyExt(id=1327012209072386420, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, companyId=1327012209059803506, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2交通数据分析与挖掘北京市重点实验室(北京交通大学) 北京 100044)])]), Author(id=1327012211484111241, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, 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=1327012211559608716, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, authorId=1327012211484111241, language=EN, stringName=Hongli Xu, firstName=Hongli, middleName=null, lastName=Xu, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 3, address=1School of Computer and Information Technology, Beijing Jiaotong University, Beijing 100044
3Key Laboratory of Beijing for Railway Engineering (Beijing Jiaotong University), Beijing 100044, bio={"content":"

born in 1963. PhD, professor. Her main research interests include machine learning and machine vision

"}, bioImg=null, bioContent=

born in 1963. PhD, professor. Her main research interests include machine learning and machine vision

, aboutCorrespAuthor=null), CN=AuthorExt(id=1327012211668660621, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, authorId=1327012211484111241, language=CN, stringName=许宏丽, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 3, address=1北京交通大学计算机与信息技术学院 北京 100044
3轨道工程北京市重点实验室(北京交通大学) 北京 100044, bio={"img":"zPg9nj+uzosLm1GKyMwxLg==","content":"

1963年生. 博士,教授. 主要研究方向为机器学习、机器视觉

"}, bioImg=zPg9nj+uzosLm1GKyMwxLg==, bioContent=

1963年生. 博士,教授. 主要研究方向为机器学习、机器视觉

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1327012208984306031, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, xref=1, ext=[AuthorCompanyExt(id=1327012208992694640, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, companyId=1327012208984306031, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1School of Computer and Information Technology, Beijing Jiaotong University, Beijing 100044), AuthorCompanyExt(id=1327012209001083249, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, companyId=1327012208984306031, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1北京交通大学计算机与信息技术学院 北京 100044)]), AuthorCompany(id=1327012209147883893, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, xref=3, ext=[AuthorCompanyExt(id=1327012209173049718, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, companyId=1327012209147883893, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3Key Laboratory of Beijing for Railway Engineering (Beijing Jiaotong University), Beijing 100044), AuthorCompanyExt(id=1327012209177244023, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, companyId=1327012209147883893, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3轨道工程北京市重点实验室(北京交通大学) 北京 100044)])]), Author(id=1327012211748352399, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, 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=1327012211823849874, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, authorId=1327012211748352399, language=EN, stringName=Xiaorong Wang, firstName=Xiaorong, middleName=null, lastName=Wang, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, address=1School of Computer and Information Technology, Beijing Jiaotong University, Beijing 100044
2Beijing Key Lab of Traffic Data Analysis and Mining (Beijing Jiaotong University), Beijing 100044, bio={"content":"

born in 1999. Master candidate. Student member of CCF. His main research interests include 3D vision and point cloud segmentation

"}, bioImg=null, bioContent=

born in 1999. Master candidate. Student member of CCF. His main research interests include 3D vision and point cloud segmentation

, aboutCorrespAuthor=null), CN=AuthorExt(id=1327012211890958739, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, authorId=1327012211748352399, language=CN, stringName=王晓荣, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, address=1北京交通大学计算机与信息技术学院 北京 100044
2交通数据分析与挖掘北京市重点实验室(北京交通大学) 北京 100044, bio={"img":"xRyHRB9THrxS/RsrNh9xjQ==","content":"

1999年生. 硕士研究生. CCF学生会员. 主要研究方向为3D视觉、点云分割

"}, bioImg=xRyHRB9THrxS/RsrNh9xjQ==, bioContent=

1999年生. 硕士研究生. CCF学生会员. 主要研究方向为3D视觉、点云分割

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1327012208984306031, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, xref=1, ext=[AuthorCompanyExt(id=1327012208992694640, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, companyId=1327012208984306031, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1School of Computer and Information Technology, Beijing Jiaotong University, Beijing 100044), AuthorCompanyExt(id=1327012209001083249, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, companyId=1327012208984306031, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1北京交通大学计算机与信息技术学院 北京 100044)]), AuthorCompany(id=1327012209059803506, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, xref=2, ext=[AuthorCompanyExt(id=1327012209068192115, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, companyId=1327012209059803506, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2Beijing Key Lab of Traffic Data Analysis and Mining (Beijing Jiaotong University), Beijing 100044), AuthorCompanyExt(id=1327012209072386420, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, companyId=1327012209059803506, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2交通数据分析与挖掘北京市重点实验室(北京交通大学) 北京 100044)])])], keywords=[Keyword(id=1327012212079702420, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, language=EN, orderNo=1, keyword=computer vision), Keyword(id=1327012212167782805, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, language=EN, orderNo=2, keyword=point cloud segmentation), Keyword(id=1327012212272640406, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, language=EN, orderNo=3, keyword=contrastive learning), Keyword(id=1327012212364915095, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, language=EN, orderNo=4, keyword=self-attention mechanism), Keyword(id=1327012212427829656, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, language=EN, orderNo=5, keyword=boundary mining), Keyword(id=1327012212515910041, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, language=CN, orderNo=1, keyword=计算机视觉), Keyword(id=1327012212616573338, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, language=CN, orderNo=2, keyword=点云分割), Keyword(id=1327012212700459419, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, language=CN, orderNo=3, keyword=对比学习), Keyword(id=1327012212784345500, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, language=CN, orderNo=4, keyword=自注意力机制), Keyword(id=1327012212880814493, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, language=CN, orderNo=5, keyword=边界挖掘)], refs=[Reference(id=1327012216924123575, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, 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=Gong Zheng, Sun Junhua, Zhang Guangjun. Dynamic structured-light measurement for wheel diameter based on the cycloid constraint[J]. Applied Optics, 2016, 55(1): 198−207, articleTitle=null, refAbstract=null), Reference(id=1327012217041564088, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, 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=秦彩杰,管强. 三维点云数据分割研究现状[J]. 宜宾学院学报,2017,17(6):30−35, articleTitle=null, refAbstract=null), Reference(id=1327012217125450169, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, 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=Qin Caijie, Guan Qiang. Research status of 3D point cloud data segmentation[J]. Journal of Yibin University, 2017, 17(6): 30−35 (in Chinese), articleTitle=null, refAbstract=null), Reference(id=1327012217217724858, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, 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=Bhanu B, Lee S, Ho C C, et al. Range data processing: Representation of surfaces by edges [C] //Proc of the 8th Int Conf on Pattern Recognition. Los Alamitos, CA: IEEE Computer Society, 1986: 236−238, articleTitle=null, refAbstract=null), Reference(id=1327012217301610939, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, 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=Jiang Xiaoyi, Bunke H. Edge detection in range images based on scan line approximation[J]. Computer Vision and Image Understanding, 1999, 73(2): 183−99, articleTitle=null, refAbstract=null), Reference(id=1327012217452605884, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, 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=Wang Zhe, Hong Liu, Yueliang Qian, et al. Real-time plane segmentation and obstacle detection of 3D point clouds for indoor scenes [C] //Proc of the Computer Vision–ECCV. Berlin: Springer, 2012: 22−31, articleTitle=null, refAbstract=null), Reference(id=1327012217570046397, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, 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=Papon J, Alexey A, Markus S, et al. Voxel cloud connectivity segmentation-supervoxels for point clouds [C] //Proc of the 31st IEEE Conf on Computer Vision and Pattern Recognition. Piscataway, NJ: IEEE, 2013: 2027−2034, articleTitle=null, refAbstract=null), Reference(id=1327012217666515390, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, 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=Zhan Qingming, Liang Yu, Liang Yubing. A point cloud segmentation method based on vector estimation and color clustering [C] //Proc of the 2nd Int Conf on Information Science and Engineering. Piscataway, NJ: IEEE, 2010: 3463−3466, articleTitle=null, refAbstract=null), Reference(id=1327012217758790079, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, 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=Zhan Qingming, Liang Yu. Segmentation of LiDAR point cloud based on similarity measures in multi-dimension Euclidean space [C] //Proc of the 2nd Advances in Computer Science and Engineering. Berlin: Springer, 2012: 349−357, articleTitle=null, refAbstract=null), Reference(id=1327012217888813504, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=9, rfOrder=9, authorNames=null, journalName=null, refType=null, unstructuredReference=Holz D, Holzer S, Rusu R B, et al. Real-Time plane segmentation using RGB-D cameras [C] //Proc of the 15th RoboCup. Berlin: Springer, 2011: 306−317, articleTitle=null, refAbstract=null), Reference(id=1327012217997865409, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=10, rfOrder=10, authorNames=null, journalName=null, refType=null, unstructuredReference=Li Yangyan, Wu Xiaokun, Chrysathou Y, et al. Globfit: Consistently fitting primitives by discovering global relations [C/OL] //Proc of the 11th ACM SIGGRAPH. New York: ACM, 2011[2023-01-20]. https://dl.acm.org/doi/abs/10.1145/1964921.1964947, articleTitle=null, refAbstract=null), Reference(id=1327012218064974274, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=11, rfOrder=11, authorNames=null, journalName=null, refType=null, unstructuredReference=Awadallah M, Lynn A, Sherin G. Segmentation of sparse noisy point clouds using active contour models [C]//Proc of the 21st IEEE Int Conf on Image Processing (ICIP). Piscataway, NJ: IEEE, 2014: 6061−6065, articleTitle=null, refAbstract=null), Reference(id=1327012218211774915, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=12, rfOrder=12, authorNames=null, journalName=null, refType=null, unstructuredReference=Wang Yanmin, Shi Hongbin. A segmentation method for point cloud based on local sample and statistic inference [C] //Proc of the 2nd Int Conf on Geo-Informatics in Resource Management and Sustainable Ecosystem(GRMSE). Berlin: Springer, 2015: 274−282, articleTitle=null, refAbstract=null), Reference(id=1327012218295660996, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=13, rfOrder=13, authorNames=null, journalName=null, refType=null, unstructuredReference=Boulch A, Guerry J, Le Saux B, et al. SnapNet: 3D point cloud semantic labeling with 2D deep segmentation networks[J]. Computers & Graphics, 2018, 71: 189−198, articleTitle=null, refAbstract=null), Reference(id=1327012218387935685, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=14, rfOrder=14, authorNames=null, journalName=null, refType=null, unstructuredReference=Wang Wei, Xu Yuan, Ren Yingchao, et al. Parsing of urban facades from 3D point clouds based on a novel multi-view domain[J]. Photogrammetric Engineering & Remote Sensing, 2021, 87(4): 283−293, articleTitle=null, refAbstract=null), Reference(id=1327012220057268679, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=15, rfOrder=15, authorNames=null, journalName=null, refType=null, unstructuredReference=Kundu A, Yin X, Fathi A, et al. Virtual multi-view fusion for 3D semantic segmentation [C] //Proc of the 16th European Conf on Computer Vision (ECCV). Berlin: Springer, 2020: 518−535, articleTitle=null, refAbstract=null), Reference(id=1327012220141154760, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=16, rfOrder=16, authorNames=null, journalName=null, refType=null, unstructuredReference=Chen Xiaozhi, Ma Huimin, Wan Ji, et al. Multi-view 3D object detection network for autonomous driving [C] //Proc of the 35th IEEE Conf on Computer Vision and Pattern Recognition. Piscataway, NJ: IEEE, 2017: 1907−1915, articleTitle=null, refAbstract=null), Reference(id=1327012220208263625, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=17, rfOrder=17, authorNames=null, journalName=null, refType=null, unstructuredReference=Qi C R, Su Hao, Mo K, et al. PointNet: Deep learning on point sets for 3D classification and segmentation [C] //Proc of the 35th IEEE Conf on Computer Vision and Pattern Recognition. Piscataway, NJ: IEEE, 2017: 652−660, articleTitle=null, refAbstract=null), Reference(id=1327012220279566794, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=18, rfOrder=18, authorNames=null, journalName=null, refType=null, unstructuredReference=Qi C R, Yi L, Su Hao, et al. PointNet++: Deep hierarchical feature learning on point sets in a metric space[J/OL]Advances in Neural Information Processing Systems, 2017, 30. [2024-01-20]. https://proceedings.neurips.cc/paper_files/paper/2017/hash/d8bf84be3800d12f74d8b05e9b89836f-Abstract.html, articleTitle=null, refAbstract=null), Reference(id=1327012220346675659, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=19, rfOrder=19, authorNames=null, journalName=null, refType=null, unstructuredReference=Thomas H, Qi C R, Deschaud J E, et al. KPConv: Flexible and deformable convolution for point clouds [C] //Proc of the 37th IEEE/CVF Int Conf on Computer Vision. Piscataway, NJ: IEEE, 2019: 6411−6420, articleTitle=null, refAbstract=null), Reference(id=1327012220426367436, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=20, rfOrder=20, authorNames=null, journalName=null, refType=null, unstructuredReference=Xu Mutian, Ding Runyu, Zhao Hengshuang, et al. PAConv: Position adaptive convolution with dynamic kernel assembling on point clouds [C] //Proc of the 39th IEEE/CVF Conf on Computer Vision and Pattern Recognition. Piscataway, NJ: IEEE, 2021: 3173−3182, articleTitle=null, refAbstract=null), Reference(id=1327012220631888333, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=21, rfOrder=21, authorNames=null, journalName=null, refType=null, unstructuredReference=Wu Felix, Fan A, Baevski A, et al. Pay less attention with lightweight and dynamic convolutions [J]. arXiv preprint, arXiv: 1901.10430, 2019, articleTitle=null, refAbstract=null), Reference(id=1327012220703191502, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=22, rfOrder=22, authorNames=null, journalName=null, refType=null, unstructuredReference=Yang Zhilin, Dai Zihang, Yang Yiming, et al. XLNet: Generalized autoregressive pretraining for language understanding[J]. Neural Information Processing Systems, 2019, 33: 5753−5763, articleTitle=null, refAbstract=null), Reference(id=1327012220778688975, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=23, rfOrder=23, authorNames=null, journalName=null, refType=null, unstructuredReference=Devlin J, Chang Ming-Wei, Lee K, et al. BERT: Pre-training of deep bidirectional transformers for language understanding [J]. arXiv preprint, arXiv: 1810.04805, 2018, articleTitle=null, refAbstract=null), Reference(id=1327012220858380752, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=24, rfOrder=24, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhao Hengshuang, Li Jiang, Jia Jiaya, et al. Point Transformer [C] //Proc of the 39th IEEE/CVF Int Conf on Computer Vision. Piscataway, NJ: IEEE, 2021: 16259−16268, articleTitle=null, refAbstract=null), Reference(id=1327012220938072529, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=25, rfOrder=25, authorNames=null, journalName=null, refType=null, unstructuredReference=Chen Ting, Kornblith S, Norouzi M, et al. A simple framework for contrastive learning of visual representations [C] //Proc of the 37th Int Conf on Machine Learning. Washington, DC: ICLR, 2020: 1597−1607, articleTitle=null, refAbstract=null), Reference(id=1327012221068095954, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=26, rfOrder=26, authorNames=null, journalName=null, refType=null, unstructuredReference=Khosla P, Teterwak P, Wang Chen, et al. Supervised contrastive learning[J]. Neural Information Processing Systems, 2020, 33: 18661−18673, articleTitle=null, refAbstract=null), Reference(id=1327012221172953555, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=27, rfOrder=27, authorNames=null, journalName=null, refType=null, unstructuredReference=Xu Yufei, Zhang Qiming, Zhang Jing, et al. RegionCL: Can simple region swapping contribute to contrastive learning[J]. arXiv preprint, arXiv: 2111.12309, 2021, articleTitle=null, refAbstract=null), Reference(id=1327012221244256724, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=28, rfOrder=28, authorNames=null, journalName=null, refType=null, unstructuredReference=Tang Liyao, Zhan Yibing, Chen Zhe, et al. Contrastive boundary learning for point cloud segmentation[C]//Proc of the 40th IEEE/CVF Conf on Computer Vision and Pattern Recognition. Piscataway, NJ: IEEE, 2022: 8489−8499, articleTitle=null, refAbstract=null), Reference(id=1327012221319754197, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=29, rfOrder=29, authorNames=null, journalName=null, refType=null, unstructuredReference=Oord A, Li Yazhe, Vinyals O. Representation learning with contrastive predictive coding[J]. arXiv preprint, arXiv: 1807.03748, 2018, articleTitle=null, refAbstract=null), Reference(id=1327012221391057366, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=30, rfOrder=30, authorNames=null, journalName=null, refType=null, unstructuredReference=张重生,陈杰,李岐龙,等. 深度对比学习综述[J]. 自动化学报,2023,49(1):15−39, articleTitle=null, refAbstract=null), Reference(id=1327012221458166231, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=30, rfOrder=31, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhang Chongsheng, Chen Jie, Li Qilong, et al. A review of deep contrast learning[J]. Acta Automatica Sinica, 2023, 49(1): 15−39 (in Chinese), articleTitle=null, refAbstract=null), Reference(id=1327012221537858008, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=31, rfOrder=32, authorNames=null, journalName=null, refType=null, unstructuredReference=Hou Sikang, Shi Hongye, Cao Xianghai, et al. Hyperspectral imagery classification based on contrastive learning[J]. IEEE Transactions on Geoscience and Remote Sensing, 2021, 60: 1−13, articleTitle=null, refAbstract=null), Reference(id=1327012221596578265, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=32, rfOrder=33, authorNames=null, journalName=null, refType=null, unstructuredReference=Wang Xinlong, Zhang Rufeng, Shen Chunhua, et al. Dense contrastive learning for self-supervised visual pre-training[C] //Proc of the 39th IEEE/CVF Conf on Computer Vision and Pattern Recognition. Piscataway, NJ: IEEE, 2021: 3024−3033, articleTitle=null, refAbstract=null), Reference(id=1327012221676270042, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=33, rfOrder=34, authorNames=null, journalName=null, refType=null, unstructuredReference=Chaitanya K, Erdil E, Karani N, et al. Contrastive learning of global and local features for medical image segmentation with limited annotations[J]. Advances in Neural Information Processing Systems, 2020, 33: 12546−12558, articleTitle=null, refAbstract=null), Reference(id=1327012221739184603, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=34, rfOrder=35, authorNames=null, journalName=null, refType=null, unstructuredReference=康健,王智睿,祝若鑫,等. 基于监督对比学习正则化的高分辨率 SAR 图像建筑物提取方法[J]. 雷达学报,2022,11(1):157−167, articleTitle=null, refAbstract=null), Reference(id=1327012221814682076, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=34, rfOrder=36, authorNames=null, journalName=null, refType=null, unstructuredReference=Kang Jian, Wang Zhirui, Zhu Ruoxin, et al. High-resolution SAR image building extraction method based on supervised contrast learning regularization[J]. Journal of Radar, 2022, 11(1): 157−167 (in Chinese), articleTitle=null, refAbstract=null), Reference(id=1327012221885985245, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=35, rfOrder=37, authorNames=null, journalName=null, refType=null, unstructuredReference=侯琳琳. 基于几何特征的三维点云分割算法研究[D]. 北京:北京交通大学,2019, articleTitle=null, refAbstract=null), Reference(id=1327012221953094110, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=35, rfOrder=38, authorNames=null, journalName=null, refType=null, unstructuredReference=Hou Linlin. Research on 3D point cloud segmentation algorithm based on geometric features [D]. Beijing: Beijing Jiaotong University, 2019 (in Chinese), articleTitle=null, refAbstract=null), Reference(id=1327012222057951711, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=36, rfOrder=39, authorNames=null, journalName=null, refType=null, unstructuredReference=李娇娇,孙红岩,董雨,等. 基于深度学习的3维点云处理综述[J]. 计算机研究与发展,2022,59(5):1160−1179, articleTitle=null, refAbstract=null), Reference(id=1327012222120866272, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=36, rfOrder=40, authorNames=null, journalName=null, refType=null, unstructuredReference=Li Jiaojiao, Sun Hongyan, Dong Yu, et al. Survey of 3-dimensional point cloud processing based on deep learning[J]. Journal of Computer Research and Development, 2022, 59(5): 1160−1179 (in Chinese), articleTitle=null, refAbstract=null), Reference(id=1327012222187975137, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=37, rfOrder=41, authorNames=null, journalName=null, refType=null, unstructuredReference=Wu Wenxuan, Qi Zhongang, Li Fuxi. PointConv: Deep convolutional networks on 3D point clouds[C]//Proc of the 39th IEEE/CVF Conf on Computer Vision and Pattern Recognition. Piscataway, NJ: IEEE, 2019: 9621−9630, articleTitle=null, refAbstract=null), Reference(id=1327012222263472610, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=38, rfOrder=42, authorNames=null, journalName=null, refType=null, unstructuredReference=Ronneberger O, Fischer P, Brox T. U-Net: Convolutional networks for biomedical image segmentation[C] //Proc of the 18th Medical Image Computing and Computer-Assisted Intervention (MICCAI). Berlin: Springer, 2015: 234−241, articleTitle=null, refAbstract=null), Reference(id=1327012222330581475, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=39, rfOrder=43, authorNames=null, journalName=null, refType=null, unstructuredReference=Armeni I, Sax S, Zamir A R, et al. Joint 2D-3D-semantic data for indoor scene understanding[J]. arXiv preprint, arXiv: 1702.01105, 2017, articleTitle=null, refAbstract=null), Reference(id=1327012222389301732, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=40, rfOrder=44, authorNames=null, journalName=null, refType=null, unstructuredReference=Yi Li, Shao Lin, Manolis S, et al. Large-scale 3D shape reconstruction and segmentation from shapenet core55[J]. arXiv preprint, arXiv: 1710.06104, 2017, articleTitle=null, refAbstract=null)], funds=null, companyList=[AuthorCompany(id=1327012208984306031, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, xref=1, ext=[AuthorCompanyExt(id=1327012208992694640, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, companyId=1327012208984306031, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1School of Computer and Information Technology, Beijing Jiaotong University, Beijing 100044), AuthorCompanyExt(id=1327012209001083249, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, companyId=1327012208984306031, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1北京交通大学计算机与信息技术学院 北京 100044)]), AuthorCompany(id=1327012209059803506, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, xref=2, ext=[AuthorCompanyExt(id=1327012209068192115, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, companyId=1327012209059803506, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2Beijing Key Lab of Traffic Data Analysis and Mining (Beijing Jiaotong University), Beijing 100044), AuthorCompanyExt(id=1327012209072386420, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, companyId=1327012209059803506, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2交通数据分析与挖掘北京市重点实验室(北京交通大学) 北京 100044)]), AuthorCompany(id=1327012209147883893, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, xref=3, ext=[AuthorCompanyExt(id=1327012209173049718, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, companyId=1327012209147883893, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3Key Laboratory of Beijing for Railway Engineering (Beijing Jiaotong University), Beijing 100044), AuthorCompanyExt(id=1327012209177244023, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, companyId=1327012209147883893, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3轨道工程北京市重点实验室(北京交通大学) 北京 100044)]), AuthorCompany(id=1327012209256935800, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, xref=4, ext=[AuthorCompanyExt(id=1327012209265324409, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, companyId=1327012209256935800, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=4Frontiers Science Center for Smart High-speed Railway System (Beijing Jiaotong University), Beijing 100044), AuthorCompanyExt(id=1327012209273713018, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, companyId=1327012209256935800, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=4智慧高铁系统前沿科学中心(北京交通大学) 北京 100044)])], figs=[ArticleFig(id=1327012213216358814, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, language=EN, label=Figure 1, caption=Schematic diagram of CL2M model structure, figureFileSmall=OxNbMChPVkC3QnjfZTno9Q==, figureFileBig=ugHfQpqV5xfv9JiW++dXEA==, tableContent=null), ArticleFig(id=1327012213266690463, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, language=CN, label=图1, caption=CL2M模型结构示意图, figureFileSmall=OxNbMChPVkC3QnjfZTno9Q==, figureFileBig=ugHfQpqV5xfv9JiW++dXEA==, tableContent=null), ArticleFig(id=1327012213363159456, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, language=EN, label=Figure 2, caption=CL2M backbone network, figureFileSmall=xWLs8Q+nNSmkQl/iyleXFg==, figureFileBig=O3Z5+xJEzMJKC+9Z9aZx2g==, tableContent=null), ArticleFig(id=1327012213447045537, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, language=CN, label=图2, caption=CL2M骨干网络, figureFileSmall=xWLs8Q+nNSmkQl/iyleXFg==, figureFileBig=O3Z5+xJEzMJKC+9Z9aZx2g==, tableContent=null), ArticleFig(id=1327012213602234786, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, language=EN, label=Figure 3, caption=Demonstration of unreasonable label distribution due to the use of average pooling, figureFileSmall=+Rn4QZX5k3kwJ6B5VVitIg==, figureFileBig=1hQJIgVJgcATw6RPZtdfHg==, tableContent=null), ArticleFig(id=1327012213669343651, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, language=CN, label=图3, caption=使用平均池化导致的标签分布不合理现象的展示, figureFileSmall=+Rn4QZX5k3kwJ6B5VVitIg==, figureFileBig=1hQJIgVJgcATw6RPZtdfHg==, tableContent=null), ArticleFig(id=1327012213740646820, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, language=EN, label=Figure 4, caption=Using hard division leads to the display of unreasonable positive and negative sample judgments, figureFileSmall=0A8b3Uu9ESO26DJ7W04xxA==, figureFileBig=N0Qa0ZZUOMKidySqP1wMYA==, tableContent=null), ArticleFig(id=1327012213816144293, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, language=CN, label=图4, caption=使用硬划分导致正负样本判定不合理的展示, figureFileSmall=0A8b3Uu9ESO26DJ7W04xxA==, figureFileBig=N0Qa0ZZUOMKidySqP1wMYA==, tableContent=null), ArticleFig(id=1327012215510643110, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, language=EN, label=Figure 5, caption=Boundary performance comparison of different models on S3DIS dataset, figureFileSmall=LJ41XAmqhCvcC23gLb7LOw==, figureFileBig=SUgVTQw04UjK34tdIbw7Jw==, tableContent=null), ArticleFig(id=1327012215611306407, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, language=CN, label=图5, caption=不同模型在S3DIS数据集的边界效果展示, figureFileSmall=LJ41XAmqhCvcC23gLb7LOw==, figureFileBig=SUgVTQw04UjK34tdIbw7Jw==, tableContent=null), ArticleFig(id=1327012215690998184, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, language=EN, label=Figure 6, caption=Boundary performance comparison of different models on ShapeNet Part dataset, figureFileSmall=s7Its1dabzkUYemx83gZSg==, figureFileBig=2YB4xgdvRdEOGK7+6e974w==, tableContent=null), ArticleFig(id=1327012215758107049, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, language=CN, label=图6, caption=不同模型在ShapeNet Part数据集的边界效果展示, figureFileSmall=s7Its1dabzkUYemx83gZSg==, figureFileBig=2YB4xgdvRdEOGK7+6e974w==, tableContent=null), ArticleFig(id=1327012215850381738, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, language=EN, label=Figure 7, caption=Impact of different modules on the model, figureFileSmall=kRE2sMD3w+svMZe9JtR4/Q==, figureFileBig=FZpZEcaEtosOiAr0R4x7DA==, tableContent=null), ArticleFig(id=1327012215921684907, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, language=CN, label=图7, caption=不同模块对模型的影响, figureFileSmall=kRE2sMD3w+svMZe9JtR4/Q==, figureFileBig=FZpZEcaEtosOiAr0R4x7DA==, tableContent=null), ArticleFig(id=1327012215997182381, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, language=EN, label=Table 1, caption=Experimental Results of Each Model on S3DIS and ShapeNet Part Datasets %, figureFileSmall=null, figureFileBig=null, tableContent=
模型S3DISShapeNet Part
mIoUmIoU@boundarymIoU@innermIoUmIoU@boundarymIoU@inner
注:黑体数值表示最优值.
PointNet41.130.253.482.964.083.5
KPConv67.150.570.986.065.186.6
ConvNet67.450.171.285.965.285.3
Point Transformer70.051.573.886.465.387.3
CBL69.452.673.186.266.386.9
CL2M(本文)71.354.974.887.068.287.6
), ArticleFig(id=1327012216076874157, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, language=CN, label=表1, caption=

各模型在S3DIS和ShapeNet Part数据集上的实验结果

, figureFileSmall=null, figureFileBig=null, tableContent=
模型S3DISShapeNet Part
mIoUmIoU@boundarymIoU@innermIoUmIoU@boundarymIoU@inner
注:黑体数值表示最优值.
PointNet41.130.253.482.964.083.5
KPConv67.150.570.986.065.186.6
ConvNet67.450.171.285.965.285.3
Point Transformer70.051.573.886.465.387.3
CBL69.452.673.186.266.386.9
CL2M(本文)71.354.974.887.068.287.6
), ArticleFig(id=1327012216156565934, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, language=EN, label=Table 2, caption=Experimental Results of Each Model in Some Categories on S3DIS Dataset %, figureFileSmall=null, figureFileBig=null, tableContent=
模型mIoUceilingfloorwallwindowdoortablechairsofaboardclutter
注:黑体数值表示最优值.
PointNet41.188.897.369.846.310.859.052.65.926.433.2
KPConv67.192.897.382.458.069.081.591.075.466.758.9
ConvNet67.494.198.183.153.573.091.782.376.566.960.8
Point Transformer70.095.298.685.162.575.981.591.075.176.460.2
CBL69.493.998.484.257.771.991.781.877.869.162.9
CL2M(本文)71.394.798.687.664.178.583.290.576.681.563.0
), ArticleFig(id=1327012216236257711, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, language=CN, label=表2, caption=

各模型在S3DIS数据集上部分类别的实验结果

, figureFileSmall=null, figureFileBig=null, tableContent=
模型mIoUceilingfloorwallwindowdoortablechairsofaboardclutter
注:黑体数值表示最优值.
PointNet41.188.897.369.846.310.859.052.65.926.433.2
KPConv67.192.897.382.458.069.081.591.075.466.758.9
ConvNet67.494.198.183.153.573.091.782.376.566.960.8
Point Transformer70.095.298.685.162.575.981.591.075.176.460.2
CBL69.493.998.484.257.771.991.781.877.869.162.9
CL2M(本文)71.394.798.687.664.178.583.290.576.681.563.0
), ArticleFig(id=1327012216332726704, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, language=EN, label=Table 3, caption=Experimental Results of Each Model in Some Categories on ShapeNet Part Dataset %, figureFileSmall=null, figureFileBig=null, tableContent=
模型mIoUairplanecapcarchairguitarknifelamplaptopmugtable
注:黑体数值表示最优值.
PointNet82.982.980.474.187.890.985.179.795.291.380.6
KPConv86.086.186.579.091.892.588.584.397.194.083.3
ConvNet85.985.685.778.991.992.788.484.297.994.483.8
Point Transformer86.486.886.679.492.993.189.285.097.694.784.6
CBL86.286.086.779.593.093.288.984.898.095.084.7
CL2M(本文)87.087.487.279.793.193.389.886.297.994.985.2
), ArticleFig(id=1327012216433390001, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, language=CN, label=表3, caption=

各模型在ShapeNet Part数据集上部分类别的实验结果

, figureFileSmall=null, figureFileBig=null, tableContent=
模型mIoUairplanecapcarchairguitarknifelamplaptopmugtable
注:黑体数值表示最优值.
PointNet82.982.980.474.187.890.985.179.795.291.380.6
KPConv86.086.186.579.091.892.588.584.397.194.083.3
ConvNet85.985.685.778.991.992.788.484.297.994.483.8
Point Transformer86.486.886.679.492.993.189.285.097.694.784.6
CBL86.286.086.779.593.093.288.984.898.095.084.7
CL2M(本文)87.087.487.279.793.193.389.886.297.994.985.2
), ArticleFig(id=1327012216517276082, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, language=EN, label=Table 4, caption=Effect of Temperature Coefficient on Point Cloud Segmentation Performance, figureFileSmall=null, figureFileBig=null, tableContent=
温度系数$\tau $mIoU/%
S3DISShapeNet Part
0.370.6786.35
0.570.9886.65
171.3387.03
270.7386.40
570.0385.71
), ArticleFig(id=1327012216601162163, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, language=CN, label=表4, caption=

温度系数对点云分割性能的影响

, figureFileSmall=null, figureFileBig=null, tableContent=
温度系数$\tau $mIoU/%
S3DISShapeNet Part
0.370.6786.35
0.570.9886.65
171.3387.03
270.7386.40
570.0385.71
), ArticleFig(id=1327012216676659636, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, language=EN, label=Table 5, caption=Effect of Positive and Negative Sample Segmentation Thresholds on Point Cloud Segmentation Performance, figureFileSmall=null, figureFileBig=null, tableContent=
epsmIoU/%
S3DISShapeNet Part
0.170.0586.23
0.370.8786.69
0.571.3387.03
0.770.7486.65
0.970.0286.21
), ArticleFig(id=1327012216752157109, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, language=CN, label=表5, caption=

正负样本划分阈值对点云分割性能的影响

, figureFileSmall=null, figureFileBig=null, tableContent=
epsmIoU/%
S3DISShapeNet Part
0.170.0586.23
0.370.8786.69
0.571.3387.03
0.770.7486.65
0.970.0286.21
)], attaches=null, journal=Journal(id=1324759606947315908, delFlag=0, nameCn=计算机研究与发展, nameEn=Journal of Computer Research and Development, nameHistory1=null, nameHistory2=null, issn=1000-1239, eissn=null, cn=11-1777/TP, coden=null, periodic=0, 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=k2l8zgqwstuxxevFX2qxVA==, journalPrice=null, startedYear=null, abbrevIsoEn=Journal of Computer Research and Development, journalRemark=null, publicationField=null, createdTime=1793776903141, updatedTime=1793850698612, createdBy=18614031015, updatedBy=13701087609, firstLetterCn=J, firstLetterEn=J, subjectCode=Electronic Information, subjectName=null, subjectCodeEn=Electronic Information, subjectNameEn=null, picCn=k2l8zgqwstuxxevFX2qxVA==, picEn=r6k0Ry+RD+c/VXj5QarmDw==, jcr=null, cjcr=null, exts=[JournalExt(id=1325069127710286688, language=CN, name=计算机研究与发展, 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=1793850698640, updatedTime=1793850698640, createdBy=13701087609, updatedBy=13701087609, submissionGuidelinesUrl=, submissionAuthorUrl=https://crad.ict.ac.cn/journalx/authorLogOn.action, submissionEditorUrl=https://crad.ict.ac.cn/journalx/editorLogOn.action, submissionReviewUrl=https://crad.ict.ac.cn/journalx/expertLogOn.action, submissionCeEditorUrl=https://crad.ict.ac.cn/journalx/editorInChiefLogOn.action, submissionAeEditorUrl=, option={"copyright":""}), JournalExt(id=1325069127777395553, language=EN, name=Journal of Computer Research and Development, 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=1793850698656, updatedTime=1793850698656, createdBy=13701087609, updatedBy=13701087609, submissionGuidelinesUrl=, submissionAuthorUrl=https://crad.ict.ac.cn/journalx/authorLogOn.action, submissionEditorUrl=https://crad.ict.ac.cn/journalx/editorLogOn.action, submissionReviewUrl=https://crad.ict.ac.cn/journalx/expertLogOn.action, submissionCeEditorUrl=https://crad.ict.ac.cn/journalx/editorInChiefLogOn.action, submissionAeEditorUrl=, option={"copyright":""})], databaseList=null, tenantJournalId=1324759914146529478, websiteList=[Website(id=1325070300139892771, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1324759914146529478, 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/jsjyjyfz/CN, language=CN, createTime=1793850978169, createBy=18614031015, updateTime=1793851040259, updateBy=18614031015, name=计算机研究与发展-中文, tplId=1146099689490845704, title=计算机研究与发展, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1325070893751349360, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=articleTextType, value=kx, createTime=1793851119697, updateTime=1793851119697, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893730377837, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=banner, value=null, createTime=1793851119692, updateTime=1793851119692, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893772320883, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=grayFlag, value=0, createTime=1793851119702, updateTime=1793851119702, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893721989228, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=logo, value=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/file/pic?fileId=U27tDsy717YF7XSPA+Vwdw==, createTime=1793851119690, updateTime=1793851119690, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893784903797, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=minRunFlag, value=0, createTime=1793851119705, updateTime=1793851119705, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893742960751, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/file/pic, createTime=1793851119695, updateTime=1793851119695, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893780709492, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=silenceFlag, value=0, createTime=1793851119704, updateTime=1793851119704, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893738766446, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_cn_619/, createTime=1793851119694, updateTime=1793851119694, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893755543665, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=themeColor, value=null, createTime=1793851119698, updateTime=1793851119698, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893763932274, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=themeStyle, value=null, createTime=1793851119700, updateTime=1793851119700, creator=18614031015, updator=18614031015)]), Website(id=1325070300580294698, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1324759914146529478, 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/jsjyjyfz/EN, language=EN, createTime=1793850978274, createBy=18614031015, updateTime=1793851036638, updateBy=18614031015, name=计算机研究与发展-英文, tplId=1146101810881728533, title=Journal of Computer Research and Development, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1325070925288321148, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=articleTextType, value=kx, createTime=1793851127216, updateTime=1793851127216, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925263155321, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=banner, value=null, createTime=1793851127210, updateTime=1793851127210, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925309292671, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=grayFlag, value=0, createTime=1793851127221, updateTime=1793851127221, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925254766712, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=logo, value=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN/file/pic?fileId=U27tDsy717YF7XSPA+Vwdw==, createTime=1793851127208, updateTime=1793851127208, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925326069889, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=minRunFlag, value=0, createTime=1793851127225, updateTime=1793851127225, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925279932539, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN/file/pic, createTime=1793851127214, updateTime=1793851127214, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925317681280, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=silenceFlag, value=0, createTime=1793851127223, updateTime=1793851127223, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925271543930, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_en_623/, createTime=1793851127212, updateTime=1793851127212, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925292515453, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=themeColor, value=null, createTime=1793851127217, updateTime=1793851127217, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925300904062, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=themeStyle, value=null, createTime=1793851127219, updateTime=1793851127219, creator=18614031015, updator=18614031015)])], journalTitle=计算机研究与发展, weixinUrl=null, journalUrl=https://crad.ict.ac.cn/, iacademicId=null, status=1, seqNo=null, journalTitleEn=Journal of Computer Research and Development, journalPhotoCn=k2l8zgqwstuxxevFX2qxVA==, journalPhotoEn=r6k0Ry+RD+c/VXj5QarmDw==, journalFirstLetter=J, 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/jsjyjyfz/CN/10.7544/issn1000-1239.202330491, detailUrlEn=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN/10.7544/issn1000-1239.202330491, pdfUrlCn=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/PDF/10.7544/issn1000-1239.202330491, pdfUrlEn=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN/PDF/10.7544/issn1000-1239.202330491, aliStartDate=0, aliEndDate=0, collectionFlag=false, citedCount=null, citedUrl=null, previewStatus=0, delFlag=0, hasFullText=1, orderTime=1735660800000, fullTextJson=null, articleText=null, reference=null)
收藏切换
基于对比学习和标签挖掘的点云分割算法
黄华 1, 2, 4 , 卜一凡 1, 2 , 许宏丽 1, 3 , 王晓荣 1, 2
计算机研究与发展 | 人工智能 2025,62(1): 132-143
收起
收藏切换
计算机研究与发展 |人工智能 2025 , 62 (1) : 132 -143
基于对比学习和标签挖掘的点云分割算法
全屏
[Author(id=1327012209328238972, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, orderNo=0, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=hhua@bjtu.edu.cn, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1327012211068875136, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, authorId=1327012209328238972, language=EN, stringName=Hua Huang, firstName=Hua, middleName=null, lastName=Huang, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, 4, address=1School of Computer and Information Technology, Beijing Jiaotong University, Beijing 100044
2Beijing Key Lab of Traffic Data Analysis and Mining (Beijing Jiaotong University), Beijing 100044
4Frontiers Science Center for Smart High-speed Railway System (Beijing Jiaotong University), Beijing 100044, bio={"content":"

born in 1977. PhD, associate professor. Member of CCF. His main research interests include computer vision, machine learning, and the applications in the railway field

"}, bioImg=null, bioContent=

born in 1977. PhD, associate professor. Member of CCF. His main research interests include computer vision, machine learning, and the applications in the railway field

, aboutCorrespAuthor=null), CN=AuthorExt(id=1327012211135984001, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, authorId=1327012209328238972, language=CN, stringName=黄华, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, 4, address=1北京交通大学计算机与信息技术学院 北京 100044
2交通数据分析与挖掘北京市重点实验室(北京交通大学) 北京 100044
4智慧高铁系统前沿科学中心(北京交通大学) 北京 100044, bio={"img":"cTMd2Qm+tWqRhe0b6wF+FA==","content":"

1977年生. 博士,副教授. CCF会员. 主要研究方向为计算机视觉、机器学习及其在铁路领域中的应用

"}, bioImg=cTMd2Qm+tWqRhe0b6wF+FA==, bioContent=

1977年生. 博士,副教授. CCF会员. 主要研究方向为计算机视觉、机器学习及其在铁路领域中的应用

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1327012208984306031, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, xref=1, ext=[AuthorCompanyExt(id=1327012208992694640, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, companyId=1327012208984306031, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1School of Computer and Information Technology, Beijing Jiaotong University, Beijing 100044), AuthorCompanyExt(id=1327012209001083249, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, companyId=1327012208984306031, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1北京交通大学计算机与信息技术学院 北京 100044)]), AuthorCompany(id=1327012209059803506, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, xref=2, ext=[AuthorCompanyExt(id=1327012209068192115, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, companyId=1327012209059803506, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2Beijing Key Lab of Traffic Data Analysis and Mining (Beijing Jiaotong University), Beijing 100044), AuthorCompanyExt(id=1327012209072386420, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, companyId=1327012209059803506, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2交通数据分析与挖掘北京市重点实验室(北京交通大学) 北京 100044)]), AuthorCompany(id=1327012209256935800, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, xref=4, ext=[AuthorCompanyExt(id=1327012209265324409, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, companyId=1327012209256935800, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=4Frontiers Science Center for Smart High-speed Railway System (Beijing Jiaotong University), Beijing 100044), AuthorCompanyExt(id=1327012209273713018, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, companyId=1327012209256935800, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=4智慧高铁系统前沿科学中心(北京交通大学) 北京 100044)])]), Author(id=1327012211236647299, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, 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=1327012211349893510, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, authorId=1327012211236647299, language=EN, stringName=Yifan Bu, firstName=Yifan, middleName=null, lastName=Bu, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, address=1School of Computer and Information Technology, Beijing Jiaotong University, Beijing 100044
2Beijing Key Lab of Traffic Data Analysis and Mining (Beijing Jiaotong University), Beijing 100044, bio={"content":"

born in 1998. Master. Her main research interests include 3D vision and point cloud segmentation

"}, bioImg=null, bioContent=

born in 1998. Master. Her main research interests include 3D vision and point cloud segmentation

, aboutCorrespAuthor=null), CN=AuthorExt(id=1327012211429585287, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, authorId=1327012211236647299, language=CN, stringName=卜一凡, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, address=1北京交通大学计算机与信息技术学院 北京 100044
2交通数据分析与挖掘北京市重点实验室(北京交通大学) 北京 100044, bio={"img":"QJULnevITJkz6Qm5SVSfiQ==","content":"

1998年生. 硕士. 主要研究方向为3D视觉、点云分割

"}, bioImg=QJULnevITJkz6Qm5SVSfiQ==, bioContent=

1998年生. 硕士. 主要研究方向为3D视觉、点云分割

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1327012208984306031, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, xref=1, ext=[AuthorCompanyExt(id=1327012208992694640, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, companyId=1327012208984306031, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1School of Computer and Information Technology, Beijing Jiaotong University, Beijing 100044), AuthorCompanyExt(id=1327012209001083249, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, companyId=1327012208984306031, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1北京交通大学计算机与信息技术学院 北京 100044)]), AuthorCompany(id=1327012209059803506, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, xref=2, ext=[AuthorCompanyExt(id=1327012209068192115, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, companyId=1327012209059803506, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2Beijing Key Lab of Traffic Data Analysis and Mining (Beijing Jiaotong University), Beijing 100044), AuthorCompanyExt(id=1327012209072386420, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, companyId=1327012209059803506, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2交通数据分析与挖掘北京市重点实验室(北京交通大学) 北京 100044)])]), Author(id=1327012211484111241, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, 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=1327012211559608716, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, authorId=1327012211484111241, language=EN, stringName=Hongli Xu, firstName=Hongli, middleName=null, lastName=Xu, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 3, address=1School of Computer and Information Technology, Beijing Jiaotong University, Beijing 100044
3Key Laboratory of Beijing for Railway Engineering (Beijing Jiaotong University), Beijing 100044, bio={"content":"

born in 1963. PhD, professor. Her main research interests include machine learning and machine vision

"}, bioImg=null, bioContent=

born in 1963. PhD, professor. Her main research interests include machine learning and machine vision

, aboutCorrespAuthor=null), CN=AuthorExt(id=1327012211668660621, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, authorId=1327012211484111241, language=CN, stringName=许宏丽, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 3, address=1北京交通大学计算机与信息技术学院 北京 100044
3轨道工程北京市重点实验室(北京交通大学) 北京 100044, bio={"img":"zPg9nj+uzosLm1GKyMwxLg==","content":"

1963年生. 博士,教授. 主要研究方向为机器学习、机器视觉

"}, bioImg=zPg9nj+uzosLm1GKyMwxLg==, bioContent=

1963年生. 博士,教授. 主要研究方向为机器学习、机器视觉

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1327012208984306031, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, xref=1, ext=[AuthorCompanyExt(id=1327012208992694640, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, companyId=1327012208984306031, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1School of Computer and Information Technology, Beijing Jiaotong University, Beijing 100044), AuthorCompanyExt(id=1327012209001083249, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, companyId=1327012208984306031, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1北京交通大学计算机与信息技术学院 北京 100044)]), AuthorCompany(id=1327012209147883893, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, xref=3, ext=[AuthorCompanyExt(id=1327012209173049718, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, companyId=1327012209147883893, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3Key Laboratory of Beijing for Railway Engineering (Beijing Jiaotong University), Beijing 100044), AuthorCompanyExt(id=1327012209177244023, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, companyId=1327012209147883893, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3轨道工程北京市重点实验室(北京交通大学) 北京 100044)])]), Author(id=1327012211748352399, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, 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=1327012211823849874, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, authorId=1327012211748352399, language=EN, stringName=Xiaorong Wang, firstName=Xiaorong, middleName=null, lastName=Wang, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, address=1School of Computer and Information Technology, Beijing Jiaotong University, Beijing 100044
2Beijing Key Lab of Traffic Data Analysis and Mining (Beijing Jiaotong University), Beijing 100044, bio={"content":"

born in 1999. Master candidate. Student member of CCF. His main research interests include 3D vision and point cloud segmentation

"}, bioImg=null, bioContent=

born in 1999. Master candidate. Student member of CCF. His main research interests include 3D vision and point cloud segmentation

, aboutCorrespAuthor=null), CN=AuthorExt(id=1327012211890958739, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, authorId=1327012211748352399, language=CN, stringName=王晓荣, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, address=1北京交通大学计算机与信息技术学院 北京 100044
2交通数据分析与挖掘北京市重点实验室(北京交通大学) 北京 100044, bio={"img":"xRyHRB9THrxS/RsrNh9xjQ==","content":"

1999年生. 硕士研究生. CCF学生会员. 主要研究方向为3D视觉、点云分割

"}, bioImg=xRyHRB9THrxS/RsrNh9xjQ==, bioContent=

1999年生. 硕士研究生. CCF学生会员. 主要研究方向为3D视觉、点云分割

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1327012208984306031, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, xref=1, ext=[AuthorCompanyExt(id=1327012208992694640, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, companyId=1327012208984306031, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1School of Computer and Information Technology, Beijing Jiaotong University, Beijing 100044), AuthorCompanyExt(id=1327012209001083249, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, companyId=1327012208984306031, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1北京交通大学计算机与信息技术学院 北京 100044)]), AuthorCompany(id=1327012209059803506, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, xref=2, ext=[AuthorCompanyExt(id=1327012209068192115, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, companyId=1327012209059803506, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2Beijing Key Lab of Traffic Data Analysis and Mining (Beijing Jiaotong University), Beijing 100044), AuthorCompanyExt(id=1327012209072386420, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011715012735599, companyId=1327012209059803506, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2交通数据分析与挖掘北京市重点实验室(北京交通大学) 北京 100044)])])]
黄华1, 2, 4 , 卜一凡1, 2, 许宏丽1, 3, 王晓荣1, 2
作者信息
  • 1北京交通大学计算机与信息技术学院 北京 100044
  • 2交通数据分析与挖掘北京市重点实验室(北京交通大学) 北京 100044
  • 3轨道工程北京市重点实验室(北京交通大学) 北京 100044
  • 4智慧高铁系统前沿科学中心(北京交通大学) 北京 100044
作者简介:

1977年生. 博士,副教授. CCF会员. 主要研究方向为计算机视觉、机器学习及其在铁路领域中的应用

1998年生. 硕士. 主要研究方向为3D视觉、点云分割

1963年生. 博士,教授. 主要研究方向为机器学习、机器视觉

1999年生. 硕士研究生. CCF学生会员. 主要研究方向为3D视觉、点云分割

Point Cloud Segmentation Algorithm Based on Contrastive Learning and Label Mining
Hua Huang1, 2, 4 , Yifan Bu1, 2, Hongli Xu1, 3, Xiaorong Wang1, 2
Affiliations
  • 1School of Computer and Information Technology, Beijing Jiaotong University, Beijing 100044
  • 2Beijing Key Lab of Traffic Data Analysis and Mining (Beijing Jiaotong University), Beijing 100044
  • 3Key Laboratory of Beijing for Railway Engineering (Beijing Jiaotong University), Beijing 100044
  • 4Frontiers Science Center for Smart High-speed Railway System (Beijing Jiaotong University), Beijing 100044
  • born in 1977. PhD, associate professor. Member of CCF. His main research interests include computer vision, machine learning, and the applications in the railway field

    born in 1998. Master. Her main research interests include 3D vision and point cloud segmentation

    born in 1963. PhD, professor. Her main research interests include machine learning and machine vision

    born in 1999. Master candidate. Student member of CCF. His main research interests include 3D vision and point cloud segmentation

出版时间: 2025-01-01 doi: 10.7544/issn1000-1239.202330491
文章导航
收藏切换

基于深度学习的点云分割算法通过设计复杂的特征提取模块,可以对高维空间点云进行有效的分割. 但由于缺乏对边界点集的特征挖掘,使得其对边界分割的精度欠佳. 已有将对比学习思想用于点云分割以解决边界区域分割性能不足问题的研究中,忽略了点云无序和稀疏特性,特征提取不够准确. 对此,设计了基于对比学习和标签挖掘的点云分割模型CL2M (contrastive learning label mining),通过自注意力机制学习不同位置处点云更为精准的特征,并引入对比学习方法,提高了点云边界处的分割精度. 在对比边界学习过程中通过深入挖掘语义空间中的标签并设计了基于标签分布的对比边界学习模块,使得高维空间点云标签分布包含更多的语义信息. CL2M充分利用标签的分布规律计算分布间的距离,可准确划分正负样本,减少了常规硬划分带来的累计错误. 在2个公开数据集上进行的实验结果表明,CL2M在多个评价指标上优于既有的点云分割模型,验证了模型的有效性.

计算机视觉  /  点云分割  /  对比学习  /  自注意力机制  /  边界挖掘

Point cloud segmentation algorithm based on deep learning can effectively segment point clouds in high-dimensional space by designing complex feature extraction modules. However, the lack of feature mining for boundary point set results in suboptimal accuracy in boundary segmentation. Some studies have applied the idea of contrastive learning to point cloud segmentation to solve the problem of insufficient boundary region segmentation performance, but the disorder and sparse characteristics of point cloud have not been fully utilized, and the feature extraction is not accurate enough. To solve these problems, we propose CL2M to learn more accurate features of point clouds at different locations through the self-attention mechanism, and the contrastive learning method is introduced to improve the segmentation accuracy of point cloud boundaries. In the process of contrastive boundary learning, labels in semantic space are deeply mined and a contrastive boundary learning module based on label distribution is designed to make the label distribution of point cloud in high-dimensional space contain more semantic information. The model makes full use of the label distribution law to calculate the distance between distributions, and can accurately divide positive and negative samples, reducing the cumulative errors caused by conventional hard partition. The results on two public data sets show that CL2M is superior to the existing point cloud segmentation model on several evaluation indexes, which verifies the effectiveness of the model.

computer vision  /  point cloud segmentation  /  contrastive learning  /  self-attention mechanism  /  boundary mining
黄华, 卜一凡, 许宏丽, 王晓荣. 基于对比学习和标签挖掘的点云分割算法. 计算机研究与发展, 2025 , 62 (1) : 132 -143 . DOI: 10.7544/issn1000-1239.202330491
Hua Huang, Yifan Bu, Hongli Xu, Xiaorong Wang. Point Cloud Segmentation Algorithm Based on Contrastive Learning and Label Mining[J]. Journal of Computer Research and Development, 2025 , 62 (1) : 132 -143 . DOI: 10.7544/issn1000-1239.202330491
点云分割[1]是将点云数据集中的点按照一定的规则或标准划分为互不相交的子集,以实现对点云数据的进一步分析、处理或应用的过程. 在点云分割算法中,如何学习得到合理的特征或标准对于精准分割至关重要.
点云分割的研究难点主要在于点云数据的自身特性:无序性、稀疏性和不规则性. 由于点云数据中有很多个点,这些点无论以什么顺序出现,它们指代的信息并不改变,这就是无序性. 稀疏性指的是采集环境、物体形状等因素可能导致点云数据的不均匀分布和缺失. 点云的不规则性指的是这些点可以任意分布,不受任何规则的限制.
早期的点云分割技术主要利用传统的方法[2],如基于边缘[3-4]、基于区域增长[5-6]、基于属性[7-9]、基于模型[10-12]等,它们通常基于手工提取的特征来进行分类或聚类. 此外,基于图像处理技术的点云分割方法也比较流行,如基于2维图像的投影方法[13-15],将点云数据投影到2维图像上进行分割. 但这些方法存在一定的局限性,难以处理复杂的3维几何形状和表面纹理. 近年来,深度学习算法和模型得到快速发展,并在各个领域[1,16]中得到广泛应用,在性能和评价指标方面有了较大的提升. 深度学习模型可以自动从数据中学习到最具表征能力的特征,无需手动提取特征,具备大规模数据和并行计算的能力. 因此在点云分割领域,从PointNet[17],PointNet++[18]到各种卷积模型如KPConv[19],PAConv[20]等,都设计了深度学习网络模型,能够很好地自适应提取点云特征并进行点云分割.
Transformer是一种用于自然语言处理和其他序列任务[21-23]的神经网络架构,由于其网络核心自注意力操作原本就是一种集合操作,对于输入元素的顺序和数量都是不敏感的,因此特别适合点云数据处理. Point Transformer[24]将Transformer应用到点云分割中,通过结合自注意力机制来捕捉点云数据中的全局和局部信息. 相比于传统的基于卷积操作的点云处理方法,Point Transformer具有更强的表达能力和适应性,参数量少,计算效率更高.
尽管深度学习的引入在一定程度上提升了点云的整体分割精度,但在场景边界上的分割表现欠佳,从而导致整体分割性能下降. 既有的模型鲜少提及边界的概念,且边界区域中相对非边界区域点的数量较少,因而难以提取其特征. 对比学习(contrastive learning)[25-27]可以通过学习对象之间的相似性和差异性来提取更具判别性的特征,且只需要少量的标记数据就可以有效地学习,对于数据集较小或标记数据难以获取的任务非常适用. 对比边界学习(contrastive boundary learning, CBL)方法[28]将对比学习应用到点云分割任务中,将场景中不同位置的信息结合起来,提供更全面、更准确的特征表示. 同时通过跨尺度的特征比较,能更好地捕捉场景的结构信息和语义关系. 但该方法在处理点云数据时忽略了点云数据的2个特性:无序性和稀疏性,因此对于点云特征提取的精度不够准确.
本文据此提出基于对比学习和标签挖掘的点云分割模型CL2M(contrastive learning label mining),借鉴了CBL模型中利用对比学习来提高点云边界精度的思想,设计了正负样本的定义方式,使得模型更加关注边界处的分割效果;通过引入自注意力机制进行特征提取,计算每个位置与其他所有位置的特征相似度并加权聚合,可以有效应对点云数据的无序性. CL2M引入基于语义空间的标签挖掘模块,缓解平均池化进行标签聚合导致的语义信息缺失问题;并引入基于标签分布的对比边界学习模块,利用标签的分布规律缓解硬划分带来的错误累计问题.
本文的主要贡献有3个方面:
1)设计了基于对比学习和标签挖掘的点云分割模型CL2M,加入自注意力机制以更好地挖掘点云特征. 同时,通过对比学习可以强化边界处的分割以提升整体分割性能.
2)提出了基于语义空间的标签分布挖掘算法,其包含了2个模块,分别是基于语义空间的标签挖掘和基于标签分布的对比边界学习模块. 利用深度学习模型从输入的数据中提取出高层次的语义信息,并用于生成高维空间中的点云标签;利用KL散度计算标签向量之间的距离,可以更好地反映标签分布的差异来更精准地划分正负样本.
3)在S3DIS和ShapeNet Part数据集上进行了实验,结果表明CL2M在整体分割精度和边界分割精度这2个评价指标上优于对比方法,验证了模型的有效性.
本节主要介绍对比学习和点云分割的国内外研究现状.
对比学习[29]是一种无监督学习的方法,旨在学习如何比较输入中的不同部分,从而提高模型对于数据的表示能力. 与有监督学习不同,对比学习不需要标注数据,因此可以应用于各种领域. 对比学习的核心思想是将相同类别的样本映射到空间中的相近位置,而将不同类别的样本映射到空间中的相对较远位置,来学习更加鲁棒和具有区分性的表示.
近年来,对比学习成为了机器学习、计算机视觉和自然语言处理等领域的研究热点之一[30],在分类、分割、预测等下游任务中均有重要应用. 对比学习最常见的下游应用就是分类任务. 针对图像分类任务,Hou等人[31]把对比学习融入半监督高光谱图像分类问题中,分阶段对模型分别进行基于对比学习方法的预训练,并利用有标注样本进行监督学习. 除了分类任务,对比学习在分割任务中的作用也十分显著. Wang等人[32]提出密集对比学习自监督视觉预训练方法(DenseCL),通过将对比学习的思想应用于自监督学习中,训练一个神经网络来学习图像的特征表示,以此提高自监督学习的性能. 在医学图像分割领域,Chaitanya等人[33]提出基于全局和局部特征之间的对比学习,通过将不同部位的图像划分为不同的子区域来提取局部特征,并使用全局池化来获取全局特征,该方法旨在解决医学图像分割中标注数据不足的问题. 康健等人[34]为了解决高分辨率SAR图像的分割问题,通过对监督对比学习进行正则化,即使用生成的建筑物足迹与真实建筑物足迹之间的差异来约束训练过程,提高了建筑物足迹的生成质量. 这些论文[29-34]都在对比学习领域做出了重要的贡献,推动了该领域的发展和应用.
近年来,基于深度学习的点云分割算法吸引了越来越多的学者关注[35-36],他们在公开发布的数据集上进行研究,提出了许多效果显著的点云分割模型. 基于逐点多层感知器的算法都是通过点云变换网络来消除点云无序性的影响,其优点在于可以直接处理点云数据,减少引入辅助手段带来的性能开销. 2017年,Qi等人[17]提出了PointNet模型,直接操作点云数据,利用深度学习通过对每个点进行处理来学习全局特征. PointNet++[18]在PointNet的基础上提出改进,通过采样、分组,并使用PointNet模块提取特征,以此提高模型精度,但是此类方法缺乏局部特征的表述. 由于卷积神经网络在2维图像上进行特征提取效果显著,而在3维点云数据上受其不规则性限制,传统卷积核无法进行卷积,因此,研究者们致力于解决并突破这一局限. 2019年,Thomas等人[19]提出的KPConv模型具有可变形卷积核的特性,可以在不同的位置对卷积核进行形变,从而适应不同形状的点云. Wu等人[37]提出了PointConv,利用一组自适应权重来计算点云中每个点的特征表示,从而在点云数据上执行卷积操作,使得神经网络能够直接处理点云数据而无需转换为其他形式.
2021年,Zhao等人[24]提出了Point Transformer,将Transformer应用到点云处理上,通过使用自注意力机制在点云中计算特征表示,使其能够捕获不同点之间的复杂关系,其整体结构类似于Transformer,由编码器和解码器组成. 编码器包括多层点自注意力模块,其中每个点自注意力模块处理点云中所有点之间的关系. 解码器由多层传统的全连接神经网络组成,用于预测每个点的类别. 该模型可以对任意数量的点进行端到端的处理,不需要任何先验网格或采样,在多个点云数据集上取得了不错的效果,并在速度上有显著优势. 尽管上述模型在不同的模块中取得了这些进展,但点云分割中的边界问题却很少被探讨. 2022年,Tang等人[28]设计了CBL方法,提出边界的概念,将对比学习的思想融入到点云分割领域中,通过设计正负样本定义方式、计算距离、选择合适的损失函数,使正样本对相互靠近、负样本对逐渐远离,进一步提高了分割的准确性. 尽管该方法是一种有效的点云分割方法,但是它存在一个问题:确定高维空间点云标签的过程比较简单,因此难以精确地判定正负样本,这会导致边界分割效果不够理想.
本节主要介绍基于对比学习和标签挖掘的点云分割模型CL2M的2个部分,分别是融合自注意力机制和对比学习的点云分割算法、基于语义空间的标签分布挖掘算法,并对边界问题给出了定义.
本节阐述了边界的定义[28],并对CL2M模型进行了形式化定义,给出了相关的符号含义说明. 若在一个点的邻域内存在不同标注标签的点,则定义其为边界点. 用$ X = \{ {x_i}\} $表示点云数据,点云中第i个点记为${x_i}$,将${x_i}$的局部邻域记为$N({x_i})$,对应的标签记为${{\boldsymbol{l}}_i}$,边界点集${B_l}$的定义如式(1)所示:
$ {B_l} = \{ {x_i} \in X|\exists {x_j} \in N({x_i}),{{\boldsymbol{l}}_j} \ne {{\boldsymbol{l}}_i}\} . $
CL2M模型网络整体采用类似U-Net[38]的编码器-解码器的结构完成点云分割任务. 在编码器阶段,通过设计基于语义空间的标签挖掘模块,挖掘高维空间中语义信息的点云标签分布,以此指导点云的特征学习;在解码器阶段,通过设计基于标签分布的对比边界学习模块,精确划分正负样本来提升点云边界的分割精度. 编码器和解码器之间通过跳跃连接,可以实现编码器和解码器的特征融合,进一步充分利用编码器中一些有用的信息. 模型结构示意图如图1所示,骨干网络搭建如图2所示. 其中,基于语义空间的标签分布挖掘包含基于语义空间的标签挖掘、基于标签分布的对比边界学习2个模块,在图2中对应为各阶段损失的计算. 此外上下采样模块依旧沿用传统方法PointNet++中的操作.
本节主要介绍融合自注意力机制和对比学习的点云分割,并对其基于位置编码的特征提取模块和基于对比学习的点云分割模块进行详细阐述.
受Point Transformer启发[24],使用Transformer层能够适应点云的数据特性以此提取更精准的点云特征. 对于进行了n次下采样阶段的点$x_i^n$,将经过Transformer层后输出的特征$f_i^{n + 1}$定义为:
$ \begin{split}{f}_{i}^{n+1}=\;&{\displaystyle \sum _{{x}_{j}^{n}\in N({x}_{i}^{n})}\rho (\gamma (\phi ({f}_{i}^{n})-\psi ({f}_{j}^{n})+\delta ))}\;\odot\\ & (\alpha ({f}_{j}^{n})+\delta )\text{,}\end{split} $
其中$f_i^n$和$f_j^n$分别表示目标点$x_i^n$和其邻域范围内节点$x_j^n$的特征,二者均为输入特征. 通过使用减法运算可以求得目标点输入特征$f_i^n$与其邻域内某点输入特征$f_j^n$的差异. 映射函数$\gamma ,\varphi ,\psi ,\alpha ,\rho $和位置编码$\delta $[24]均沿用Point Transformer中的定义. 映射函数$\gamma $是一个包含2个线性层和非线性变换ReLU的多层感知器,其功能在于形成可学习的映射,便于模型训练;函数$\varphi ,\psi ,\alpha $对特征进行线性变换;函数$\rho $采用softmax旨在对权重进行归一化. CL2M通过Transformer层促进了特征向量之间的信息交流,以此输出具有更准确表征的特征向量.
提升边界分割的准确度,不仅有利于整体分割性能,而且还能很大程度上减少稀疏数据的影响. 尽管现在已经提出了各种各样的点云分割方法,但对3维点云边界的关注却很少. 受CBL启发[28],对比学习可以通过将点云数据分成不同的局部区域,并将每个局部区域中的点作为正样本,将其他区域中的点作为负样本. 通过比较正样本和负样本之间的相似性,可以学习到区分边界处的点和非边界处的点的特征表示. 这样就可以强化边界处的分割,从而提升整体分割性能. CL2M通过加入对比学习来增强跨边界的特征识别能力,并定义正负样本对来约束损失函数. 具体来说,对于边界点${x_i} \in {B_l}$,使用InfoNCE损失[29]定义边界点上的对比优化目标,如式(3)所示:
$ {L_{{\text{con}}}} = \dfrac{{ - 1}}{{\left| {{B_l}} \right|}}\displaystyle\sum\limits_{{x_i} \in {B_l}} {\ln \dfrac{{\displaystyle\sum\limits_{{x_j} \in N({x_i}) \wedge {{\boldsymbol{l}}_j} = {{\boldsymbol{l}}_i}} {\exp ( - d({f_i},{f_j})/\tau )} }}{{\displaystyle\sum\limits_{{x_k} \in N({x_i})} {\exp ( - d({f_i},{f_k})/\tau )} }}} , $
其中${f_i}$是目标点${x_i}$的特征,d是距离度量函数,$\tau $是温度系数. 对于每个边界点${x_i} \in {B_l}$,在其局部邻域$N({x_i})$内采样时,选择与${x_i}$标签相同的点作为其正样本,即$\{ {x_j} \in N({x_i}) \wedge {{\boldsymbol{l}}_j} = {{\boldsymbol{l}}_i}\} $,选择与其标签不同的点作为${x_i}$的负样本,即$\{ {x_j} \in N({x_i}) \wedge {{\boldsymbol{l}}_j} \ne {{\boldsymbol{l}}_i}\} $. CL2M引入对比学习对提高边界区域的分割效果具有重要意义.
由于模型在下采样后缺失高维空间点的标签,难以得到边界点集${B_l}$的实际定义,进而无法充分将对比学习的思想应用到点云分割算法中. 因此,模型需要对每个高维空间中的点进行标签标注[28]. 具体来说,定义下采样第n阶段的点云为${X^n}$,输入点云即为${X^0}$. 由于下采样过程中每个点$x_i^n$都是由上一阶段点云${X^{n - 1}}$中的一组点经过最远点采样(farthest point sampling, FPS) [18]产生的,因此可以在下采样时通过迭代的方式来确定某点在高维空间中的标签. 对于$x_i^0 \in {X^0}$,其标签${\boldsymbol{l}}_i^0$为one-hot类型的向量,与GroundTruth中的类别相对应. 因此,定义目标点$x_i^n$的标签向量${{\boldsymbol{L}}}_i^n$如式(4)所示:
$ {{\boldsymbol{L}}}_i^n = AVG(\{ {{\boldsymbol{L}}}_j^{n - 1}|x_j^{n - 1} \in N(x_i^{n - 1})\} ), $
其中AVG表示平均池化,$N(x_i^{n - 1})$表示目标点$x_i^n$在第n-1阶段邻域范围内的点的集合. 对于$x_i^n \in {X^n}$,其标签向量可以表示为${{\boldsymbol{L}}}_i^n = (l_{i,0}^n,l_{i,1}^n,…,l_{i,k}^n,…)$,${{\boldsymbol{L}}}_i^n$向量中第k个位置的$l_{i,k}^n$表示输入点云${X^n}$中的点$x_i^n$属于第k个类别的概率. 式(4)表示了通过计算$x_i^n$在第$ n - 1 $阶段邻域范围内属于不同类别的点所占的比例可以确定标签向量${{\boldsymbol{L}}}_i^n$. 接着,CBL[28]通过arg max ${{\boldsymbol{L}}}_i^n$确定第n阶段节点$x_i^n$的类别标签,再根据式(1)边界点集的定义可以得到当前下采样阶段的边界点. 如图2所示,在此基础上结合下采样点的特征,根据infoNCE损失函数将对比学习应用到各个下采样阶段就可以达到优化边界的目的,最终的损失函数定义为:
$ L = {L_{{\text{cross entropy}}}} + \lambda \sum\limits_n {L_{{\text{con}}}^n} , $
其中$L_{{\text{con}}}^n$表示第n阶段利用函数infoNCE求得的对比损失,$\lambda $为损失权重,${L_{{\text{cross entropy}}}}$表示交叉熵损失.
本节主要介绍基于语义空间的标签分布挖掘算法,并对其中的基于语义空间的标签挖掘模块和基于标签分布的对比边界学习模块进行详细阐述.
由于CBL[28]使用平均池化来确定高维空间点云标签的方法比较简单,忽略了空间中的语义信息,这会为精确判定正负样本造成困难,因而导致边界分割效果不够理想. 图3展示了使用平均池化造成标签分布不合理现象. 图3中用不同颜色区分标签类别,即蓝色和黄色表示不同的标签;用不同形状区分特征,即三角形和圆形表示点与点之间具有不同的特征,具有相同形状的2个点表示它们的特征更相似. 对于目标点$x_i^{n - 1}$,在其邻域范围内只有3个点$x_j^{n - 1}$,$x_k^{n - 1}$,$x_q^{n - 1}$的标签与$x_i^{n - 1}$的标签相同,其余6个点与$x_i^{n - 1}$的标签不同,但是实际上$x_i^{n-1} $邻域范围内某一关键点$x_k^{n - 1}$与$x_i^{n - 1}$的特征相似度更高,因此$ {{\boldsymbol{L}}}_k^{n - 1} $对于$x_i^n$的标签$ {{\boldsymbol{L}}}_i^n $贡献程度更大,也就是说目标点$x_i^{n - 1}$在经过下采样后在第n阶段的标签$ {{\boldsymbol{L}}}_i^n $应该更趋向于目标点在第$ n - 1 $阶段邻域范围$N(x_i^{n - 1})$内与其特征更相似的点的标签,即点$x_j^{n - 1}$,$x_k^{n - 1}$,$x_q^{n - 1}$的标签,而非简单依靠计算目标点第$ n - 1 $阶段邻域范围内属于不同类别的点所占的比例来确定其高维空间中的标签. 因此,如果使用平均池化来确定高维空间点的标签会导致边界分割效果不够理想.
基于此,本文提出基于语义空间的标签挖掘方法,利用自注意力机制的思想将目标点在第$ n - 1 $阶段的邻域内每一个点$x_j^{n - 1} \in N(x_i^{n - 1})$的标签${{\boldsymbol{L}}}_j^{n - 1}$聚合生成目标点${x_i}$在第n阶段的标签${{\boldsymbol{L}}}_i^n$,核心公式如式(6)所示:
$ {{\boldsymbol{L}}}_i^n = \sum\limits_{x_j^{n - 1} \in N(x_i^{n - 1})} {\rho (\gamma (\beta (\varphi (f_i^{n - 1}),\psi (f_j^{n - 1}))))} \odot {\boldsymbol{l}}_j^{n - 1}, $
其中$f_i^{n - 1}$表示目标点${x_i}$在第$ n - 1 $阶段的特征,${{\boldsymbol{L}}}_j^{n - 1}$表示节点${x_j}$在第$ n - 1 $阶段的标签. $\varphi $和$\psi $是可学习的变换,负责将$f_i^{n - 1}$和$f_j^{n - 1}$进行线性映射,并采用点乘的方法作为函数$\beta $进行相似度计算. 映射函数$\gamma $是包含2个线性层和非线性变换ReLU的多层感知器. 函数$\rho $采用softmax函数旨在对权重进行归一化. 总的来说,CL2M模型根据第$ n - 1 $阶段目标点${x_i}$的特征$f_i^{n - 1}$与邻域内节点${x_j}$的特征$f_j^{n - 1}$计算注意力系数,为与目标点特征相似度更高的点$x_k^{n - 1}$的标签赋予更高的权值,更精确地获取目标点第n阶段的标签分布${{\boldsymbol{L}}}_i^n$. 通过上述方法,能够有效解决使用平均池化导致的标签分布不合理问题.
基于语义空间的标签挖掘方法充分利用了点云的语义特征与位置关系,获得了高维空间中点的标签分布. 然而,CBL[28]通过arg max ${{\boldsymbol{L}}}_i^n$确定目标点在高维空间中的标签,可能会导致正负样本判定错误,进而影响边界的分割效果. 如图4所示用不同颜色来区分点的标签,展示了使用硬划分导致正负样本判定不合理的现象. 例如对于点${x_P}$与点${x_Q}$,其标签分布分别为${{{\boldsymbol{L}}}_P}$= (0.4, 0.3, 0.3)和${{{\boldsymbol{L}}}_Q}$= (0.7, 0.2, 0.1),如果使用arg max ${{\boldsymbol{L}}}_i^n$的方法确定该点的标签类别,则会将${x_Q}$判定为${x_P}$的正样本,然而实际上${x_Q}$与${x_P}$构成负样本对,这就造成了正负样本判定错误. 因此,简单通过arg max ${{\boldsymbol{L}}}_i^n$方法会导致正负样本采样不精准,模型无法学习样本之间的差异,这会影响对比学习的效果.
为了更好地衡量正负样本之间的距离,本文提出基于标签分布的对比边界学习方法. KL散度(Kullback-Leibler divergence)也称为相对熵,是一种用于衡量2个概率分布之间差异的指标. KL散度是非对称的,它衡量的是从一个概率分布到另一个概率分布的信息损失. 在这里,KL散度计算标签向量分布间的距离,并据此进行负样本采样,其核心公式为:
$ {F_{{\text{KL}}}}(P\parallel Q) = \sum_x{P(x)\ln \frac{{P(x)}}{{Q(x)}}} , $
其中P(x)表示目标点${x_P}$的标签分布,Q(x)表示目标点邻域范围内节点${x_Q}$的标签分布,${F_{{\text{KL}}}}(P\parallel Q)$表示${x_P}$与${x_Q}$标签分布之间的距离. 若$ {F_{{\text{KL}}}}(P\parallel Q) $$ \gt eps $,则定义${x_Q}$为${x_P}$的负样本,反之则称${x_Q}$为${x_P}$的正样本,CL2M使用这种方法能更好地划分正负样本,有效提高边界点云的挖掘精度.
本节主要介绍CL2M的实验结果,并对其进行定量和定性分析,最后通过消融实验验证各模块的有效性.
实验使用2个公开的数据集,包括室内场景数据集S3DIS和部件分割ShapeNet Part数据集.
S3DIS(Stanford large-scale 3D indoor spaces)数据集[39]是一个用于室内3维场景分割的数据集. 该数据集包含6个不同建筑物内的11种场景,包括办公室、会议室、走廊、楼梯等不同类型的房间共272个场景,并且标记了13个语义类别的标签,如墙壁、地板、天花板、家具等. 该数据集通过激光扫描设备采集点云数据以高精度捕捉室内环境,被广泛用于3维场景分割的研究中,并已成为评估算法性能和比较不同算法的标准数据集之一.
ShapeNet Part数据集[40]是一个用于部件分割的大型3维数据集. 该数据集包含16个形状类别的16880个点云模型,其中14006个用于训练,2874个用于测试,每个类别标注2~6个部件,共标注了50个不同的部件. 该数据集可用于训练深度学习模型自动识别或分割3维模型中的不同部分,以实现更高级别的3维场景理解.
本节主要介绍实验的相关设置,包含对比模型的简要介绍、点云分割算法的评价指标以及训练模型的相关设置.
为了验证CL2M的性能,将其与5种方法进行对比:
1)PointNet[17]通过将点云中的每个点作为网络的输入,并对每个点进行特征提取和聚合,最终得到整个点云的特征表示,这种处理方式具有可扩展性和不变性.
2)KPConv[19]通过对点云进行采样和聚类,得到一组核点. 然后,对于每个点,KPConv会根据它周围的核点计算出1组权重,再根据这些权重将周围点的特征进行加权平均. 这个过程类似于传统的卷积操作,但是针对的是点云数据.
3)ConvNet是CBL中[29]提出的一种基础网络,通过在3维空间中构建的卷积神经网络来实现点云分割.
4)Point Transformer[24]通过引入Transformer的自注意力机制来对点云进行特征提取,这些特征表示可以用于各种点云处理任务,如点云分割、点云分类、点云重建等.
5)CBL[28]通过对点云边界区域进行对比学习、划分正负样本,来提高点云边界分割的准确性.
为了评估CL2M的分割性能,本文使用mIoU,mIoU@boundary,mIoU@inner[29]作为模型的性能评价指标.
IoU是计算2个集合的交集和并集之比,均交并比mIoU是IoU的平均值,可以表现出每类点云的分割精度,计算如式(8)所示:
$ IoU(X) = \frac{{Gt \cap Pre}}{{Gt \cup Pre}} \times 100\% , $
其中Gt表示整片点云X中实际属于某一类标签的点的集合,Pre表示整片点云X中预测属于这一类标签的点的集合,分子求交集表示二者标签均相同的个数,分母求并集表示实际是此标签或者预测是此标签的个数,最后再换化成百分比的形式. 通过改变定义域的范围,可以定义mIoU@boundary和mIoU@inner,如式(9)(10)所示:
$ mIoU@boundary = mIoU({B_l}), $
$ mIoU@inner = mIoU(X - {B_l}), $
其中${B_l}$为边界区域点的集合,$X - {B_l}$为非边界区域点的集合.
CL2M模型通过PyTorch实现. 对于S3DIS数据集,为了公平比较,训练CL2M模型时将batchsize设为16,总的点云数量均为2048. 使用SGD优化器训练600个epoch,权重衰减为0.0001,学习率设置为0.5. 设置优化器动量为0.9,丢失率dropout为0.5. CL2M中对比学习相关参数设置温度系数$\tau = 1$,正负样本阈值$eps = 0.5$,损失权重$\lambda = 0.1$,下采样过程中使用k近邻的k=16. 对于ShapeNet Part数据集,除学习率设置为0.05外,其余参数一致. 模型使用GeForce RTX 2080 Ti的GPU完成训练.
表1展示了CL2M与对比模型在S3DIS和ShapeNet Part数据集上的mIoU,mIoU@boundary,mIoU@inner,表2和表3分别展示了CL2M与对比模型在S3DIS和ShapeNet Part数据集上的实验结果. 其中,在S3DIS数据集上进行的对比实验均引用CBL[29]中展示的实验结果. 通过表1~3得出3个结论:
1)如表1~3所示,CL2M对比其他方法在S3DIS数据集上mIoU指标最高提升了1.3个百分点,在ShapeNet Part数据集上mIoU指标最高提升了0.6个百分点. CL2M在2个数据集上对比其他方法mIoU指标均有所提升,因此,实验证明了CL2M模型能有效提升点云的总体分割精度.
2)如表1所示,无论是在S3DIS数据集还是在ShapeNet Part数据集上,CL2M对比CBL模型mIoU的提升幅度均高于对比Point Transformer模型的提升幅度. 由于加入对比学习,略优于Point Transformer,但提升性能不高,可能重在提升边界性能;由于加入自注意力机制模型分割效果远优于先前的其他模型,证明了Transformer结构对模型分割的优越性.
3)如表1所示,CL2M在S3DIS和ShapeNet Part数据集上对比其他方法,mIoU@boundary评价指标的提升幅度高于mIoU@inner. 因此,实验证明了CL2M能够有效提升边界区域分割精度.
本节展示不同模型对边界点云分割的效果. 如图5、图6所示,分别为各模型在S3DIS数据集和ShapeNet Part数据集上的边界效果可视化,其中:由图5(c)(e)和图6(c)(e)可知,CL2M可以有效识别出边界,而Point Transformer模型对于边界点云的识别没有明显作用. 由于Point Transformer重在特征提取,没有专注于挖掘边界特征,因此对边界识别效果有限. 由图5(d)(e)和图6(d)(e)可知,CBL和CL2M都能够识别出边界,说明对比学习方法对于识别边界有效. 由于对比学习通过挖掘子场景的标签可以更好地区分正负样本,因此能提升边界处的点云分割性能. 并且图5(e)比图5(d)识别的边界更加精细,说明CL2M模型提出的标签分布学习对于识别更加精细的边界有益. 综上所述,CL2M模型对于边界点云分割效果有较大提升.
本节主要验证CL2M仅使用单一模块对模型性能的影响,图7展示了不同模块对模型分割精度mIoU指标的影响. 本节将这些模型与CBL基准模型进行比较,并在其基础上引入自注意力机制. 通过分别应用标签语义挖掘模块和标签分布对比模块,我们证明了CL2M模型在性能上优于仅使用标签分布学习结构和仅使用Transformer结构的模型. 实验结果验证了CL2M 的2个结构对性能的提升效果,Transformer结构对提升整体分割准确度的效果明显;标签分布学习结构对模型整体分割性能提升不大,其主要提升的是边界分割性能.
因此,CL2M模型在语义理解和标签分布学习方面具有较高的优势. 通过引入自注意力机制,CL2M模型进一步增强了模型的表达能力和泛化能力. 而标签语义挖掘模块和标签分布对比模块能够更好地捕捉点云边界的关键信息,从而提高了模型的性能.
为验证不同参数对模型性能的影响,本节对温度系数和正负样本划分阈值进行参数灵敏度分析. 表4展示了设置不同的温度系数对模型的影响,实验结果表明,当温度系数$\tau = 1$时,CL2M效果较好. 表5展示了不同阈值eps对模型的影响,实验结果表明,当$eps = 0.5$时CL2M模型效果较好.
本文提出了基于对比学习和标签挖掘的点云分割模型CL2M,加入自注意力机制构建Transformer层和对比学习分别用于学习每个位置点云更为精准的特征和提高点云边界处的分割精度. 通过引入基于语义空间的标签挖掘和基于标签分布的对比边界学习,使高维空间中的点云标签分布包含更多的语义信息,同时充分利用标签的分布规律计算标签向量分布间的距离,能够更加准确地划分正负样本,减少了通过硬划分带来的累计错误. 实验结果表明,CL2M对比其他点云分割算法在各个数据集上性能均有所提升,可以有效提升边界和整体的分割效果.
作者贡献声明:黄华提出研究思路、工作方案并修改论文;卜一凡开展研究实验并撰写论文;许宏丽审阅并提出修改意见;王晓荣参与实验.

参考文献 引证文献
排序方式:
1
Gong Zheng, Sun Junhua, Zhang Guangjun. Dynamic structured-light measurement for wheel diameter based on the cycloid constraint[J]. Applied Optics, 2016, 55(1): 198−207
2
秦彩杰,管强. 三维点云数据分割研究现状[J]. 宜宾学院学报,2017,17(6):30−35
Qin Caijie, Guan Qiang. Research status of 3D point cloud data segmentation[J]. Journal of Yibin University, 2017, 17(6): 30−35 (in Chinese)
3
Bhanu B, Lee S, Ho C C, et al. Range data processing: Representation of surfaces by edges [C] //Proc of the 8th Int Conf on Pattern Recognition. Los Alamitos, CA: IEEE Computer Society, 1986: 236−238
4
Jiang Xiaoyi, Bunke H. Edge detection in range images based on scan line approximation[J]. Computer Vision and Image Understanding, 1999, 73(2): 183−99
5
Wang Zhe, Hong Liu, Yueliang Qian, et al. Real-time plane segmentation and obstacle detection of 3D point clouds for indoor scenes [C] //Proc of the Computer Vision–ECCV. Berlin: Springer, 2012: 22−31
6
Papon J, Alexey A, Markus S, et al. Voxel cloud connectivity segmentation-supervoxels for point clouds [C] //Proc of the 31st IEEE Conf on Computer Vision and Pattern Recognition. Piscataway, NJ: IEEE, 2013: 2027−2034
7
Zhan Qingming, Liang Yu, Liang Yubing. A point cloud segmentation method based on vector estimation and color clustering [C] //Proc of the 2nd Int Conf on Information Science and Engineering. Piscataway, NJ: IEEE, 2010: 3463−3466
8
Zhan Qingming, Liang Yu. Segmentation of LiDAR point cloud based on similarity measures in multi-dimension Euclidean space [C] //Proc of the 2nd Advances in Computer Science and Engineering. Berlin: Springer, 2012: 349−357
9
Holz D, Holzer S, Rusu R B, et al. Real-Time plane segmentation using RGB-D cameras [C] //Proc of the 15th RoboCup. Berlin: Springer, 2011: 306−317
10
Li Yangyan, Wu Xiaokun, Chrysathou Y, et al. Globfit: Consistently fitting primitives by discovering global relations [C/OL] //Proc of the 11th ACM SIGGRAPH. New York: ACM, 2011[2023-01-20]. https://dl.acm.org/doi/abs/10.1145/1964921.1964947
11
Awadallah M, Lynn A, Sherin G. Segmentation of sparse noisy point clouds using active contour models [C]//Proc of the 21st IEEE Int Conf on Image Processing (ICIP). Piscataway, NJ: IEEE, 2014: 6061−6065
12
Wang Yanmin, Shi Hongbin. A segmentation method for point cloud based on local sample and statistic inference [C] //Proc of the 2nd Int Conf on Geo-Informatics in Resource Management and Sustainable Ecosystem(GRMSE). Berlin: Springer, 2015: 274−282
13
Boulch A, Guerry J, Le Saux B, et al. SnapNet: 3D point cloud semantic labeling with 2D deep segmentation networks[J]. Computers & Graphics, 2018, 71: 189−198
14
Wang Wei, Xu Yuan, Ren Yingchao, et al. Parsing of urban facades from 3D point clouds based on a novel multi-view domain[J]. Photogrammetric Engineering & Remote Sensing, 2021, 87(4): 283−293
15
Kundu A, Yin X, Fathi A, et al. Virtual multi-view fusion for 3D semantic segmentation [C] //Proc of the 16th European Conf on Computer Vision (ECCV). Berlin: Springer, 2020: 518−535
16
Chen Xiaozhi, Ma Huimin, Wan Ji, et al. Multi-view 3D object detection network for autonomous driving [C] //Proc of the 35th IEEE Conf on Computer Vision and Pattern Recognition. Piscataway, NJ: IEEE, 2017: 1907−1915
17
Qi C R, Su Hao, Mo K, et al. PointNet: Deep learning on point sets for 3D classification and segmentation [C] //Proc of the 35th IEEE Conf on Computer Vision and Pattern Recognition. Piscataway, NJ: IEEE, 2017: 652−660
18
Qi C R, Yi L, Su Hao, et al. PointNet++: Deep hierarchical feature learning on point sets in a metric space[J/OL]Advances in Neural Information Processing Systems, 2017, 30. [2024-01-20]. https://proceedings.neurips.cc/paper_files/paper/2017/hash/d8bf84be3800d12f74d8b05e9b89836f-Abstract.html
19
Thomas H, Qi C R, Deschaud J E, et al. KPConv: Flexible and deformable convolution for point clouds [C] //Proc of the 37th IEEE/CVF Int Conf on Computer Vision. Piscataway, NJ: IEEE, 2019: 6411−6420
20
Xu Mutian, Ding Runyu, Zhao Hengshuang, et al. PAConv: Position adaptive convolution with dynamic kernel assembling on point clouds [C] //Proc of the 39th IEEE/CVF Conf on Computer Vision and Pattern Recognition. Piscataway, NJ: IEEE, 2021: 3173−3182
21
Wu Felix, Fan A, Baevski A, et al. Pay less attention with lightweight and dynamic convolutions [J]. arXiv preprint, arXiv: 1901.10430, 2019
22
Yang Zhilin, Dai Zihang, Yang Yiming, et al. XLNet: Generalized autoregressive pretraining for language understanding[J]. Neural Information Processing Systems, 2019, 33: 5753−5763
23
Devlin J, Chang Ming-Wei, Lee K, et al. BERT: Pre-training of deep bidirectional transformers for language understanding [J]. arXiv preprint, arXiv: 1810.04805, 2018
24
Zhao Hengshuang, Li Jiang, Jia Jiaya, et al. Point Transformer [C] //Proc of the 39th IEEE/CVF Int Conf on Computer Vision. Piscataway, NJ: IEEE, 2021: 16259−16268
25
Chen Ting, Kornblith S, Norouzi M, et al. A simple framework for contrastive learning of visual representations [C] //Proc of the 37th Int Conf on Machine Learning. Washington, DC: ICLR, 2020: 1597−1607
26
Khosla P, Teterwak P, Wang Chen, et al. Supervised contrastive learning[J]. Neural Information Processing Systems, 2020, 33: 18661−18673
27
Xu Yufei, Zhang Qiming, Zhang Jing, et al. RegionCL: Can simple region swapping contribute to contrastive learning[J]. arXiv preprint, arXiv: 2111.12309, 2021
28
Tang Liyao, Zhan Yibing, Chen Zhe, et al. Contrastive boundary learning for point cloud segmentation[C]//Proc of the 40th IEEE/CVF Conf on Computer Vision and Pattern Recognition. Piscataway, NJ: IEEE, 2022: 8489−8499
29
Oord A, Li Yazhe, Vinyals O. Representation learning with contrastive predictive coding[J]. arXiv preprint, arXiv: 1807.03748, 2018
30
张重生,陈杰,李岐龙,等. 深度对比学习综述[J]. 自动化学报,2023,49(1):15−39
Zhang Chongsheng, Chen Jie, Li Qilong, et al. A review of deep contrast learning[J]. Acta Automatica Sinica, 2023, 49(1): 15−39 (in Chinese)
31
Hou Sikang, Shi Hongye, Cao Xianghai, et al. Hyperspectral imagery classification based on contrastive learning[J]. IEEE Transactions on Geoscience and Remote Sensing, 2021, 60: 1−13
32
Wang Xinlong, Zhang Rufeng, Shen Chunhua, et al. Dense contrastive learning for self-supervised visual pre-training[C] //Proc of the 39th IEEE/CVF Conf on Computer Vision and Pattern Recognition. Piscataway, NJ: IEEE, 2021: 3024−3033
33
Chaitanya K, Erdil E, Karani N, et al. Contrastive learning of global and local features for medical image segmentation with limited annotations[J]. Advances in Neural Information Processing Systems, 2020, 33: 12546−12558
34
康健,王智睿,祝若鑫,等. 基于监督对比学习正则化的高分辨率 SAR 图像建筑物提取方法[J]. 雷达学报,2022,11(1):157−167
Kang Jian, Wang Zhirui, Zhu Ruoxin, et al. High-resolution SAR image building extraction method based on supervised contrast learning regularization[J]. Journal of Radar, 2022, 11(1): 157−167 (in Chinese)
35
侯琳琳. 基于几何特征的三维点云分割算法研究[D]. 北京:北京交通大学,2019
Hou Linlin. Research on 3D point cloud segmentation algorithm based on geometric features [D]. Beijing: Beijing Jiaotong University, 2019 (in Chinese)
36
李娇娇,孙红岩,董雨,等. 基于深度学习的3维点云处理综述[J]. 计算机研究与发展,2022,59(5):1160−1179
Li Jiaojiao, Sun Hongyan, Dong Yu, et al. Survey of 3-dimensional point cloud processing based on deep learning[J]. Journal of Computer Research and Development, 2022, 59(5): 1160−1179 (in Chinese)
37
Wu Wenxuan, Qi Zhongang, Li Fuxi. PointConv: Deep convolutional networks on 3D point clouds[C]//Proc of the 39th IEEE/CVF Conf on Computer Vision and Pattern Recognition. Piscataway, NJ: IEEE, 2019: 9621−9630
38
Ronneberger O, Fischer P, Brox T. U-Net: Convolutional networks for biomedical image segmentation[C] //Proc of the 18th Medical Image Computing and Computer-Assisted Intervention (MICCAI). Berlin: Springer, 2015: 234−241
39
Armeni I, Sax S, Zamir A R, et al. Joint 2D-3D-semantic data for indoor scene understanding[J]. arXiv preprint, arXiv: 1702.01105, 2017
40
Yi Li, Shao Lin, Manolis S, et al. Large-scale 3D shape reconstruction and segmentation from shapenet core55[J]. arXiv preprint, arXiv: 1710.06104, 2017
2025年第62卷第1期
100
42
0
引用本文
BibTeX
文章信息
doi: 10.7544/issn1000-1239.202330491
  • 接收时间:2023-06-12
  • 首发时间:2026-11-10
  • 出版时间:2025-01-01
补充材料
相关文章
文章信息
作者
出版历史
  • 收稿日期:2023-06-12
  • 修回日期:2024-02-06
基金
作者信息
    1北京交通大学计算机与信息技术学院 北京 100044
    2交通数据分析与挖掘北京市重点实验室(北京交通大学) 北京 100044
    3轨道工程北京市重点实验室(北京交通大学) 北京 100044
    4智慧高铁系统前沿科学中心(北京交通大学) 北京 100044
参考文献
分享链接
https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/10.7544/issn1000-1239.202330491
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