An advanced method for surface damage detection of concrete structures in low-light environments based on image enhancement and object detection networks DOI Creative Commons
Tianyong Jiang, Lin Liu,

Chunjun Hu

и другие.

Advances in Bridge Engineering, Год журнала: 2024, Номер 5(1)

Опубликована: Дек. 13, 2024

Abstract Surface damage detection in concrete structures is critical for maintaining structural integrity, yet current object algorithms often struggle low-light environments. To address this challenge, study proposed a methodology that integrates image enhancement and networks to improve identification such conditions. Specifically, we employ the self-calibrated illumination (SCI) model reconstruct images, which are then processed by an improved YOLOv5-based network, YOLOv5-GAM-ASFF, incorporating global attention mechanism (GAM) adaptive spatial feature fusion (ASFF). The performance of YOLOv5-GAM-ASFF evaluated on dataset structure demonstrating its superiority over YOLOv5s, YOLOv6s, YOLOv7-tiny. results show achieves [email protected] 79.1%, surpassing other models 1.3%, 3.3%, 5.8%, respectively. This approach provides reliable solution surface environments, advancing field health monitoring improving accuracy under challenging

Язык: Английский

Review on computer vision-based inspection and monitoring for bridge cables DOI
Wei Ji, Ke Luo, Kui Luo

и другие.

Measurement, Год журнала: 2025, Номер unknown, С. 116892 - 116892

Опубликована: Фев. 1, 2025

Язык: Английский

Процитировано

0

An Advanced Computer Vision Method for Noncontact Vibration Measurement of Cables in Cable‐Stayed Bridges DOI Creative Commons
Naiwei Lu,

W. Y. Zeng,

Jian Cui

и другие.

Structural Control and Health Monitoring, Год журнала: 2025, Номер 2025(1)

Опубликована: Янв. 1, 2025

With the development of computer and image processing technologies, vision (CV) has been attracting increasing attention in field civil engineering measurement monitoring. Cables slender structures have unique challenges for CV‐based vibration methods, such as low pixel proportion sensitivity to environmental conditions. This study proposes a noncontact method based on line tracking algorithm (LTA). The robustness applicability proposed under varying resolutions, signal‐to‐noise ratios, cable inclination angles were systematically evaluated through experimental test specimen. To validate effectiveness practical detection applications, scaled cable‐stayed bridge model was carried out. numerical result indicates that LTA provides high reliability accuracy values force. maximum errors first‐order self‐vibration frequency force is 0.99% 2%, respectively. maintains strong stability across various conditions, which reference long‐term structural health monitoring bridges.

Язык: Английский

Процитировано

0

Bayesian continuous wavelet transform for time-varying damping identification of cables using full-field measurement DOI
Junying Wang, Qiankun Zhu,

Qiong Zhang

и другие.

Automation in Construction, Год журнала: 2024, Номер 168, С. 105791 - 105791

Опубликована: Сен. 21, 2024

Язык: Английский

Процитировано

3

Vision-based identification of cable tensions and finite element model verification of a cable-stayed bridge DOI
Cevdet Enes Cukaci, Serdar Soyöz

Journal of Civil Structural Health Monitoring, Год журнала: 2024, Номер unknown

Опубликована: Ноя. 20, 2024

Язык: Английский

Процитировано

1

Bridge component segmentation for health monitoring an enhanced DeepLabV3+ model with lightweight network and multi-scale channel attention mechanism DOI
Tianyong Jiang, Yali Huang,

Chunjun Hu

и другие.

Advances in Structural Engineering, Год журнала: 2024, Номер unknown

Опубликована: Ноя. 5, 2024

Due to the influence of various factors, such as complex environments and sustained load effects, long-term service life bridge structures will lead a gradual deterioration in performance. Therefore, health monitoring is utmost importance, component identification crucial step evaluating overall structural integrity bridges. With advancement deep learning algorithms, semantic segmentation methods can effectively classify identify components environments, thereby facilitating assessment their state. Nevertheless, conventional for segmenting suffer from drawbacks intensive computation, inadequate feature extraction, low accuracy, failing meet requirements current monitoring. Consequently, this paper proposes method based on an improved DeepLabV3 + model, named DeepLabV3-MS, which enhanced model. This utilizes MobileNetV2 backbone network reduce parameter count improve computational speed The Strip Pooling (SP) also integrated into ASPP, known SP_ASPP, enhance capture more comprehensive contextual information. Additionally, Multi-scale Channel Attention Mechanism (MS-CAM) incorporated integration efficiency multi-semantic multi-scale features. results indicate that compared with original Mean Intersection over Union Pixel Accuracy DeeplabV3-MS model increased by 5.90%, 4.92%, respectively. Furthermore, comparison classic models PSPNet U-Net, demonstrated increase 19.50% 8.88% MIoU MPA, respectively, well 13.50% 5.34%, proposed has superior performance across evaluation metrics, exerting significant impact safety components. it offers valuable technical support research applications related fields.

Язык: Английский

Процитировано

0

An advanced method for surface damage detection of concrete structures in low-light environments based on image enhancement and object detection networks DOI Creative Commons
Tianyong Jiang, Lin Liu,

Chunjun Hu

и другие.

Advances in Bridge Engineering, Год журнала: 2024, Номер 5(1)

Опубликована: Дек. 13, 2024

Abstract Surface damage detection in concrete structures is critical for maintaining structural integrity, yet current object algorithms often struggle low-light environments. To address this challenge, study proposed a methodology that integrates image enhancement and networks to improve identification such conditions. Specifically, we employ the self-calibrated illumination (SCI) model reconstruct images, which are then processed by an improved YOLOv5-based network, YOLOv5-GAM-ASFF, incorporating global attention mechanism (GAM) adaptive spatial feature fusion (ASFF). The performance of YOLOv5-GAM-ASFF evaluated on dataset structure demonstrating its superiority over YOLOv5s, YOLOv6s, YOLOv7-tiny. results show achieves [email protected] 79.1%, surpassing other models 1.3%, 3.3%, 5.8%, respectively. This approach provides reliable solution surface environments, advancing field health monitoring improving accuracy under challenging

Язык: Английский

Процитировано

0