Article
Remote Sensing
Zhengsen Xu, Haiyan Guan, Jian Kang, Xiangda Lei, Lingfei Ma, Yongtao Yu, Yiping Chen, Jonathan Li
Summary: This paper proposes a locally enhanced Transformer network (LETNet) for efficient detection of pavement cracks. By utilizing Transformer to model long-range dependencies, designing convolution stem and local enhancement modules to compensate both low-level and high-level local features, and using skip connection strategy and efficient upsampling module to restore detailed information, LETNet outperforms four advanced deep learning-based models in terms of both efficiency and effectiveness.
INTERNATIONAL JOURNAL OF APPLIED EARTH OBSERVATION AND GEOINFORMATION
(2022)
Article
Automation & Control Systems
Jian Zhang, Songrong Qian, Can Tan
Summary: This research proposes an automatic detection and segmentation method for bridge surface cracks based on computer vision deep learning models, which is able to effectively identify and segment bridge cracks. Experimental results demonstrate that our method outperforms other baseline methods, with smaller model size and higher frame per second (FPS) performance.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2022)
Article
Construction & Building Technology
Jinchao Guan, Xu Yang, Ling Ding, Xiaoyun Cheng, Vincent C. S. Lee, Can Jin
Summary: An automated pixel-level pavement distress detection framework integrating stereo vision and deep learning is developed in this study, which establishes multi-feature pavement image datasets based on a multi-view stereo imaging system and proposes a modified U-net deep learning architecture introducing depthwise separable convolution for efficient crack and pothole segmentation. The results show that the 3D pavement image achieves millimeter-level accuracy, and the enhanced 3D crack segmentation model outperforms other models in terms of accuracy and speed, enabling high-precision automated pothole volume measurement.
AUTOMATION IN CONSTRUCTION
(2021)
Article
Engineering, Multidisciplinary
Pengfei Shi, Fengting Zhu, Yuanxue Xin, Shen Shao
Summary: In this paper, an automatic pavement crack detection method called U(2)CrackNet is proposed. The method uses an encoding and decoding architecture with a two-level nested U-structure. The experimental results show that U(2)CrackNet can obtain clearer and more continuous cracks.
STRUCTURAL HEALTH MONITORING-AN INTERNATIONAL JOURNAL
(2023)
Article
Construction & Building Technology
Lei Yang, Hanyun Huang, Shuyi Kong, Yanhong Liu
Summary: Accurate detection of crack defects in infrastructure is crucial but challenging. Traditional image processing techniques are not effective due to complex backgrounds, various shapes and sizes of crack defects, and class imbalance. Deep learning-based segmentation networks have been proposed, but they still face challenges in high-precision crack segmentation. To address these issues, an end-to-end deep crack segmentation network called PHCF-Net is proposed, incorporating progressive and hierarchical context fusion. Experimental results show that PHCF-Net achieves better crack detection results than other advanced segmentation models.
JOURNAL OF BUILDING ENGINEERING
(2023)
Article
Automation & Control Systems
Rong Gui, Qian Sun, Wenqing Wu, Dejin Zhang, Qingquan Li
Summary: Inspired by the heterogeneity of 3D pavement data and the generalization ability of transfer learning, this paper proposes a robust and generalized framework called PCDM-HED for cross-scene 3D pavement-crack detection. By constructing enhanced deep edge features, PCDM-HED highlights the essential properties of cracks in heterogeneous 3D data. Experimental results show that PCDM-HED has strong transfer learning capability and provides an effective solution for heterogeneous 3D pavement-crack detection tasks in engineering.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2023)
Article
Engineering, Civil
Jiang Chen, Ye Yuan, Hong Lang, Shuo Ding, Jian John Lu
Summary: This paper proposes an automatic concrete pavement crack segmentation framework with an enhanced graph network branch. The framework divides the image into nodes and generates their attributes, and uses Gaussian distribution to determine the edges of the graph. The graph is input into the graph branch, and its feature map is fused with the image feature map of the encoder, then passed to the decoder to recover the image resolution and obtain the crack segmentation result. The proposed method achieves the highest F1 and IoU scores in the comparison experiments.
JOURNAL OF ADVANCED TRANSPORTATION
(2022)
Article
Computer Science, Hardware & Architecture
Jongwoo Ha, Dongsoo Kim, Minsoo Kim
Summary: This study proposed a system that can automatically detect, classify, and assess the severity of road cracks by expanding the number of crack types and including crack severity assessment, achieving an accuracy of 91.2% for both crack type and severity assessment.
JOURNAL OF SUPERCOMPUTING
(2022)
Article
Construction & Building Technology
Feng Guo, Jian Liu, Chengshun Lv, Huayang Yu
Summary: Currently, there is a need for automatic approaches to detect pavement cracks for maintenance. This study proposes a Transformer-based network for accurate pixel-level pavement crack detection. By utilizing the hierarchical architecture of Swin Transformer and attention module in the UperNet, the proposed model achieves the best performance compared to other models on three public pavement crack datasets. This paves the way for future applications of automatic pavement crack detection using Transformer-based networks.
CONSTRUCTION AND BUILDING MATERIALS
(2023)
Article
Engineering, Civil
Tianjie Zhang, Donglei Wang, Yang Lu
Summary: This paper presents ECSNet, a customized deep learning model for accelerated real-time pavement crack detection and segmentation. The model achieves high accuracy and efficiency by incorporating novel components such as small kernel convolutional layers and parallel max pooling and convolutional operations. Experimental results on the DeepCrack Dataset demonstrate that ECSNet performs well in terms of accuracy and efficiency compared to other state-of-the-art models.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2023)
Article
Engineering, Civil
Bin Yu, Xiangcheng Meng, Qiannan Yu
Summary: A two-step convolutional neural network method is proposed for pavement crack detection on pixel-levels, reducing time consumption while maintaining accuracy. Experimental results demonstrate that this method significantly decreases processing time with minimal loss of accuracy compared to pure-segmentation networks.
JOURNAL OF TRANSPORTATION ENGINEERING PART B-PAVEMENTS
(2021)
Article
Environmental Sciences
Md. Al-Masrur Khan, Regidestyoko Wasistha Harseno, Seong-Hoon Kee, Abdullah-Al Abdullah-Al Nahid
Summary: Crack inspection is crucial for monitoring the health of pavement structures and simplifying the repair process. Existing manual inspection methods are inefficient and expensive. This study developed a robotic system for real-time automated data collection and analysis. A deep-learning-based semantic segmentation framework named RCDNet was proposed to identify cracks from visual images collected by the robot. Simulation results demonstrated 96.29% accuracy of the deep learning model in predicting images. The proposed robotic system significantly reduced the inspection time and generated a severity map to prioritize repair areas.
Article
Engineering, Multidisciplinary
Chun Zhang, Le Wan, Ruo-Qing Wan, Jian Yu, Rui Li
Summary: This paper proposes a novel ensemble deep neural network that can effectively improve the identification accuracy of bridge cracks in complex backgrounds by simulating human multi-scale observation, reasoning, and decision-making processes.
Article
Computer Science, Artificial Intelligence
Qiang Wu, Xunpen Qin, Kang Dong, Aixian Shi, Zeqi Hu
Summary: This paper presents a two-stage convolutional neural network (CNN) method for crack defect detection and segmentation of metal parts. The first stage detects potential cracks and crops them to a small area, while the second stage learns the context of cracks in the detected patches. A window-based stereo matching method is then used to map crack pixels to 3D world points. Experimental results show that the proposed method achieves high accuracy and efficiency in target detection and pixel-level segmentation.
EXPERT SYSTEMS WITH APPLICATIONS
(2023)
Article
Construction & Building Technology
Weixing Wang, Lei Li, Ya Han
Summary: A crack detection method based on gray level standard deviation and distance standard deviation was proposed in this study, which is different from traditional image processing algorithms. Extensive testing on different shadowed images showed that the new method performed well in crack detection, outperforming traditional algorithms.
CONSTRUCTION AND BUILDING MATERIALS
(2021)
