Vision-Based Concrete-Crack Detection on Railway Sleepers Using Dense U-Net Model DOI Creative Commons
Md. Al-Masrur Khan, Seong‐Hoon Kee, Abdullah-Al Nahid

и другие.

Algorithms, Год журнала: 2023, Номер 16(12), С. 568 - 568

Опубликована: Дек. 15, 2023

Crack inspection in railway sleepers is crucial for ensuring rail safety and avoiding deadly accidents. Traditional methods detecting cracks on are very time-consuming lack efficiency. Therefore, nowadays, researchers paying attention to vision-based algorithms, especially Deep Learning algorithms. In this work, we adopted the U-net first time a sleeper proposed modified architecture named Dense segmenting cracks. structure, established several short connections between encoder decoder blocks, which enabled obtain better pixel information flow. Thus, model extracted necessary more detail predict We collected images from sleepers, processed them dataset, finally trained with images. The achieved an overall F1-score, precision, Recall, IoU of 86.5%, 88.53%, 84.63%, 76.31%, respectively. compared our suggested original U-net, results demonstrate that performed than both quantitative qualitative results. Moreover, considered necessity crack severity analysis measured few parameters engineers must know have idea about most severe locations take steps repair badly affected sleepers.

Язык: Английский

Artificial intelligence-assisted visual inspection for cultural heritage: State-of-the-art review DOI Creative Commons
Mayank Mishra, Paulo B. Lourénço

Journal of Cultural Heritage, Год журнала: 2024, Номер 66, С. 536 - 550

Опубликована: Янв. 24, 2024

Applying computer science techniques such as artificial intelligence (AI), deep learning (DL), and vision (CV) on digital image data can help monitor preserve cultural heritage (CH) sites. Defects weathering, removal of mortar, joint damage, discoloration, erosion, surface cracks, vegetation, seepage, vandalism their propagation with time adversely affect the structural health CH Several studies have reported damage detection in concrete bridge structures using AI techniques. However, few quantified defects paradigm, limited case exist for applications. Hence, application AI-assisted visual inspections sites needs to be explored. assist inspection professionals increase confidence levels assessment buildings. This review summarizes processing techniques, focusing mainly DL applied conservation. study applications buildings are presented where traditional inspections.

Язык: Английский

Процитировано

43
Deep learning-based automated tile defect detection system for Portuguese cultural heritage buildings DOI Creative Commons
Narges Karimi, Mayank Mishra, Paulo B. Lourénço

и другие.

Journal of Cultural Heritage, Год журнала: 2024, Номер 68, С. 86 - 98

Опубликована: Май 31, 2024

A prominent feature in Portuguese historic architecture is Portugal's azulejos or tiles that cover cultural heritage buildings with colorful patterns. However, are prone to deterioration due the quality of masonry materials, exposure over time, and natural human factors. careful approach necessary detect assess tile damage time conserve heritage. Deep learning (DL) methods applied by automating vision-based monitoring. This study uses You Only Look Once (YOLO), method automatically. To obtain initial dataset, 5000 images were collected, including cracks, craters, glaze detachment, lacunae, as well no defects. Additionally, a MobileNet model was used for binary classification damaged intact compare detection approaches. Through fine-tuning hyperparameters updating an overall accuracy 72% YOLO (multiple classification) 97% achieved, demonstrating adequacy tool real-world applications.

Язык: Английский

Процитировано

20
Deep Learning and Computer Vision Techniques for Enhanced Quality Control in Manufacturing Processes DOI Creative Commons

Md Raisul Islam,

Md Zakir Hossain Zamil,

Md. Eshmam Rayed

и другие.

IEEE Access, Год журнала: 2024, Номер 12, С. 121449 - 121479

Опубликована: Янв. 1, 2024

Язык: Английский

Процитировано

13
Lightweight network for millimeter-level concrete crack detection with dense feature connection and dual attention DOI Creative Commons
Xiao Ma, Yang Li, Zijiang Yang

и другие.

Journal of Building Engineering, Год журнала: 2024, Номер 94, С. 109821 - 109821

Опубликована: Июнь 13, 2024

Development of lightweight deep learning crack detection method is essential for the future deployment mobile device-based structure inspection. The primary challenge involves analysis and extraction features from narrow cracks, typically 3–6 pixels wide, which are often obscured by noise such as water stains shadows. model should also maintain high accuracy while ensuring low computational complexity a minimal number parameters. To this end, paper proposes YOLO v5-DE (Dense Feature Enhancement Connection, Efficient Fast Convolution), network based on v5 architecture tailored to address these challenges, constructs datasets captured at different heights investigate impact shooting distances performance. utilizes efficient convolutions dense feature connections, with strategic reuse filtered shallow layers, significantly enhance model's fine-grained information gradient flow. experimental results demonstrate that achieves 96% cracks in concrete structures. Compared improved EfficientViT backbone network, 4.7% increase requiring fewer resources, only 1.4 million parameters 3.6 Giga Floating point Operations Per Second (GFLOPS). Additionally, reduces inference time 3.38 ms increases frame rate 295.8 FPS. Moreover, proposed exhibits better performance when facing complex backgrounds real-world environments.

Язык: Английский

Процитировано

11
Automated Surface Crack Detection in Historical Constructions with Various Materials Using Deep Learning-Based YOLO Network DOI
Narges Karimi, Mayank Mishra, Paulo B. Lourénço

и другие.

International Journal of Architectural Heritage, Год журнала: 2024, Номер unknown, С. 1 - 17

Опубликована: Июль 17, 2024

Язык: Английский

Процитировано

10
Generative adversarial network based on domain adaptation for crack segmentation in shadow environments DOI Creative Commons
Yingchao Zhang, Cheng Liu

Computer-Aided Civil and Infrastructure Engineering, Год журнала: 2025, Номер unknown

Опубликована: Март 2, 2025

Abstract Precision segmentation of cracks is important in industrial non‐destructive testing, but the presence shadows actual environment can interfere with results cracks. To solve this problem, study proposes a two‐stage domain adaptation framework called GAN‐DANet for crack shadowed environments. In first stage, CrackGAN uses adversarial learning to merge features from shadow‐free and datasets, creating new dataset more domain‐invariant features. second CrackSeg network innovatively integrates enhanced Laplacian filtering (ELF) into high‐resolution net enhance edges texture while out shadow information. model, addresses shift by generating features, avoiding direct feature alignment between source target domains. The ELF module effectively enhances suppresses interference, improving model's robustness Experiments show that improves accuracy, mean intersection over union value increasing 57.87 75.03, which surpasses performance existing state‐of‐the‐art algorithms.

Язык: Английский

Процитировано

1
Unsupervised domain adaptation-based crack segmentation using transformer network DOI
Daniel Asefa Beyene, Dai Quoc Tran, Michael Bekele Maru

и другие.

Journal of Building Engineering, Год журнала: 2023, Номер 80, С. 107889 - 107889

Опубликована: Окт. 11, 2023

Язык: Английский

Процитировано

15
Deep Learning-Based Automated Detection of Cracks in Historical Masonry Structures DOI Creative Commons
Kemal Hacıefendioğlu, Ahmet Can Altunışık,

Tuğba Abdioğlu

и другие.

Buildings, Год журнала: 2023, Номер 13(12), С. 3113 - 3113

Опубликована: Дек. 15, 2023

The efficient and precise identification of cracks in masonry stone structures caused by natural or human-induced factors within a specific region holds significant importance detecting damage subsequent secondary harm. In recent times, remote sensing technologies have been actively employed to promptly identify crack regions during repair reinforcement activities. Enhanced image resolution has enabled more accurate sensitive detection these areas. This research presents novel approach utilizing deep learning techniques for area cellphone images, achieved through segmentation object methods. developed model, named the CAM-K-SEG combines Grad-CAM visualization K-Mean clustering approaches with pre-trained convolutional neural network models. A comprehensive dataset comprising photographs numerous historical buildings was utilized training model. To establish comparative analysis, widely used U-Net model employed. testing datasets technique were meticulously annotated masked. evaluation results based on Intersection-over-Union (IoU) metric values. Consequently, it concluded that exhibits suitability recognition localization, whereas is well-suited segmentation.

Язык: Английский

Процитировано

15
Building Surface Defect Detection Using Machine Learning and 3D Scanning Techniques in the Construction Domain DOI Creative Commons
Alexandru Marin Mariniuc, Dorian Cojocaru, Marian Abagiu

и другие.

Buildings, Год журнала: 2024, Номер 14(3), С. 669 - 669

Опубликована: Март 2, 2024

The rapid growth of the real estate market has led to appearance more and residential areas large apartment buildings that need be managed maintained by a single developer or company. This scientific article details development novel method for inspecting in semi-automated manner, thereby reducing time needed assess requirements maintenance building. paper focuses on an application which purpose detecting imperfections range building sections using combination machine learning techniques 3D scanning methodologies. research design learning-based utilizes Python programming language PyTorch library; it builds team′s previous study, they investigated possibility applying their expertise creating construction-related applications real-life situations. Using Zed camera system, pictures various components were used, along with stock images when needed, train artificial intelligence model could identify surface damage defects such as cracks differentiate between naturally occurring elements shadows stains. One goals is develop can while readily available tools order ensure practical affordable solution. findings this study have potential greatly enhance availability defect detection procedures construction sector, will result better structural integrity.

Язык: Английский

Процитировано

6
Crack pattern–based machine learning prediction of residual drift capacity in damaged masonry walls DOI Creative Commons
Mauricio Pereira, Antonio Maria D’Altri, Stefano de Miranda

и другие.

Computer-Aided Civil and Infrastructure Engineering, Год журнала: 2024, Номер unknown

Опубликована: Май 2, 2024

Abstract In this paper, we present a method based on an ensemble of convolutional neural networks (CNNs) for the prediction residual drift capacity in unreinforced damaged masonry walls using as only input crack pattern. We use accurate block‐based numerical model to generate mechanically consistent patterns induced by external actions (earthquake‐like loads and differential settlements). For wall, extract width cumulative distribution, derive exceedance curve (CWEC), evaluate loss (DL) with respect undamaged wall. Numerous pairs CWEC DL are thus generated used training (and validating) CNNs via repeated ‐folding cross validation shuffling. As result, damage prognosis (Level IV SHM) is provided. Such appears general, inexpensive, able adequately predict CWEC, providing real‐time support decision making structures.

Язык: Английский

Процитировано

6