Real-time spatial contextual network based on deep learning for bridge exposed rebar segmentation
文献类型：期刊论文
| 作者 | Wang, Yong4; He, Zhenglong3,4; Zeng, Xiangqiang3,4; Cen, Zongxi3,4; Zeng, Juncheng2; Ren, Xiang4; Cheng, Xinyi1 |
| 刊名 | CONSTRUCTION AND BUILDING MATERIALS
 |
| 出版日期 | 2024-10-25 |
| 卷号 | 449页码:138379 |
| 关键词 | Deep learning Bridge exposed rebar Real-time semantic segmentation Lightweight Drone |
| DOI | 10.1016/j.conbuildmat.2024.138379 |
| 产权排序 | 1 |
| 英文摘要 | Bridge exposed rebar is a common and highly detrimental superficial defect that significantly affects the structural strength of bridges. Therefore, ensuring traffic safety requires timely inspections. However, bridge exposed rebar approaches are sporadic, and their accuracy, efficiency and robustness make meeting industrial requirements challenging. Consequently, to rapidly and accurately segment bridge exposed rebar, the first realtime segmentation model for bridge exposed rebar, Real-time Spatial Context Network (RSCNet), is proposed. The model constructs a spatial pathway to learn latent spatial relationships of exposed rebar in high-resolution images and a semantic pathway to capture the global contextual defect information, which also innovatively constructs a directional convolutional unit to capture the axial features of exposed rebar and designs a lightweight contextual unit to efficiently learn long-range dependencies. These strategies significantly improve the segmentation effect of RSCNet. An extensive series of experiments are conducted on the self-constructed dataset named BERD as well as publicly available concrete crack datasets DeepCrack and CFD. The results indicate that RSCNet is superior to mainstream models regarding accuracy, efficiency and robustness. Thus, this novel method supports the timely detection and maintenance of bridge surface defects, and has significant scientific and engineering value for promoting intelligent bridge inspection and digitized management. |
| WOS关键词 | NET |
| WOS研究方向 | Construction & Building Technology ; Engineering ; Materials Science |
| WOS记录号 | WOS:001324790800001 |
| 源URL | [http://ir.igsnrr.ac.cn/handle/311030/208229]  |
| 专题 | 资源与环境信息系统国家重点实验室_外文论文 |
| 通讯作者 | Wang, Yong |
| 作者单位 | 1.China Univ Geosci, Sch Geog & Informat Engn, Wuhan 430070, Peoples R China 2.Fujian Expressway Sci & Technol Innovat Res Inst C, Fuzhou 350001, Peoples R China 3.Univ Chinese Acad Sci, Coll Resources & Environm, Beijing 100049, Peoples R China 4.Chinese Acad Sci, Inst Geog Sci & Nat Resources Res, State Key Lab Resources & Environm Informat Syst, Beijing 100101, Peoples R China |
| 推荐引用方式 GB/T 7714 | Wang, Yong,He, Zhenglong,Zeng, Xiangqiang,et al. Real-time spatial contextual network based on deep learning for bridge exposed rebar segmentation[J]. CONSTRUCTION AND BUILDING MATERIALS,2024,449:138379. |
| APA | Wang, Yong.,He, Zhenglong.,Zeng, Xiangqiang.,Cen, Zongxi.,Zeng, Juncheng.,...&Cheng, Xinyi.(2024).Real-time spatial contextual network based on deep learning for bridge exposed rebar segmentation.CONSTRUCTION AND BUILDING MATERIALS,449,138379. |
| MLA | Wang, Yong,et al."Real-time spatial contextual network based on deep learning for bridge exposed rebar segmentation".CONSTRUCTION AND BUILDING MATERIALS 449(2024):138379. |
入库方式： OAI收割
来源：地理科学与资源研究所
浏览0
下载0
收藏0
其他版本
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。