Automatic crack segmentation model based on multi-branch aggregation transformer

IF 2.1 4区 工程技术 Q2 CONSTRUCTION & BUILDING TECHNOLOGY
Jin Wang, Zhigao Zeng, Jianxin Wang, Jianming Zhang, Siyuan Zhou
{"title":"Automatic crack segmentation model based on multi-branch aggregation transformer","authors":"Jin Wang, Zhigao Zeng, Jianxin Wang, Jianming Zhang, Siyuan Zhou","doi":"10.1177/13694332241266538","DOIUrl":null,"url":null,"abstract":"Crack detection plays a crucial role in evaluating the safety and durability of civil infrastructure. However, detecting cracks of uneven intensity in complex backgrounds is challenging. To overcome this problem, we propose a dual decoder network (CSMT) based on a multi-branch aggregation Transformer, which uses residual atrous spatial pyramid pooling (RASPP) and Transformer dual decoding branches to extract local and global features of different structures. To enhance global feature extraction, we designed a multi-branch aggregation Transformer (MAT) that adaptively weights the features of two attention heads from spatial and channel dimensions to achieve intra block feature aggregation between dimensions. Meanwhile, to obtain multi-scale semantic information, we constructed a new decoding branch, RASPP, which embeds a squeeze-and-excitation (SE) module and residual structures into standard ASPP. Finally, we propose a feature adaptive fusion module (FAM) to enhance feature fusion between adjacent layers and codec layers. Many experiments on three benchmark datasets have shown that the proposed CSMT segmentation network provides excellent performance in a variety of complex scenarios.","PeriodicalId":50849,"journal":{"name":"Advances in Structural Engineering","volume":"28 1","pages":""},"PeriodicalIF":2.1000,"publicationDate":"2024-07-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Advances in Structural Engineering","FirstCategoryId":"5","ListUrlMain":"https://doi.org/10.1177/13694332241266538","RegionNum":4,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"CONSTRUCTION & BUILDING TECHNOLOGY","Score":null,"Total":0}
引用次数: 0

Abstract

Crack detection plays a crucial role in evaluating the safety and durability of civil infrastructure. However, detecting cracks of uneven intensity in complex backgrounds is challenging. To overcome this problem, we propose a dual decoder network (CSMT) based on a multi-branch aggregation Transformer, which uses residual atrous spatial pyramid pooling (RASPP) and Transformer dual decoding branches to extract local and global features of different structures. To enhance global feature extraction, we designed a multi-branch aggregation Transformer (MAT) that adaptively weights the features of two attention heads from spatial and channel dimensions to achieve intra block feature aggregation between dimensions. Meanwhile, to obtain multi-scale semantic information, we constructed a new decoding branch, RASPP, which embeds a squeeze-and-excitation (SE) module and residual structures into standard ASPP. Finally, we propose a feature adaptive fusion module (FAM) to enhance feature fusion between adjacent layers and codec layers. Many experiments on three benchmark datasets have shown that the proposed CSMT segmentation network provides excellent performance in a variety of complex scenarios.
基于多分支聚集变换器的自动裂缝分割模型
裂缝检测在评估民用基础设施的安全性和耐久性方面发挥着至关重要的作用。然而,在复杂背景中检测强度不均匀的裂缝具有挑战性。为了克服这一问题,我们提出了一种基于多分支聚合变换器的双解码器网络(CSMT),它使用残差无规空间金字塔池化(RASPP)和变换器双解码分支来提取不同结构的局部和全局特征。为了加强全局特征提取,我们设计了一个多分支聚合变换器(MAT),从空间和通道维度对两个注意头的特征进行自适应加权,以实现维度间的块内特征聚合。同时,为了获取多尺度语义信息,我们构建了一个新的解码分支 RASPP,它在标准 ASPP 中嵌入了挤压激励(SE)模块和残差结构。最后,我们提出了特征自适应融合模块(FAM),以加强相邻层和编解码层之间的特征融合。在三个基准数据集上进行的大量实验表明,所提出的 CSMT 细分网络在各种复杂情况下都能提供出色的性能。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
求助全文
约1分钟内获得全文 求助全文
来源期刊
Advances in Structural Engineering
Advances in Structural Engineering 工程技术-工程:土木
CiteScore
5.00
自引率
11.50%
发文量
230
审稿时长
2.3 months
期刊介绍: Advances in Structural Engineering was established in 1997 and has become one of the major peer-reviewed journals in the field of structural engineering. To better fulfil the mission of the journal, we have recently decided to launch two new features for the journal: (a) invited review papers providing an in-depth exposition of a topic of significant current interest; (b) short papers reporting truly new technologies in structural engineering.
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
copy
已复制链接
快去分享给好友吧!
我知道了
右上角分享
点击右上角分享
0
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:481959085
Book学术官方微信