Study on the automated characterization of particle size and shape of stacked gravelly soils via deep learning

Acta Geotechnica, Год журнала: 2025, Номер unknown

Опубликована: Янв. 4, 2025

Язык: Английский

---

A structure‐oriented loss function for automated semantic segmentation of bridge point clouds

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

Опубликована: Янв. 12, 2025

Abstract Focusing on learning‐based semantic segmentation (SS) methods for bridge point cloud data (PCD), this study proposes a structure‐oriented concept (SOC) with training focused the spatial distribution patterns of components, including both horizontally absolute location each component and its vertically relative position compared other components. Then loss (SOL) function, which embodies core SOC, is defined accordingly, it to five cutting‐edge functions collected PCD dataset. In contrast limitations functions, SOL significantly improves overall evaluation metrics accuracy (6.53%) mean intersection over union (mean IoU: 8.67%). The IoU category “others” improved by 8.44%, very important automating time‐consuming denoising process. Furthermore, demonstrated robustness SOC reveal great potential improve performance SS models.

Язык: Английский

---

An interactive cross‐multi‐feature fusion approach for salient object detection in crack segmentation

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

Опубликована: Фев. 9, 2025

Abstract Salient object detection (SOD) is a crucial preprocessing technique in visual computing, which identifies the salient regions an image by simulating human perception system. It achieves remarkable results tasks such as quality assessment, editing, and recognition. However, due to particularity of pavement crack terms scale feature requirements, SOD model rarely applied surface at present. In order break existing dilemma, this paper proposes new (iU2Net) specialized for detection, based on encoder–decoder structure U2Net incorporates developed interactive cross‐multi‐feature fusion module (ICMFM). Compared with models, main contributions iU2Net are reflected two aspects. On one hand, current models difficult comprehensively extract complex features cracks while breakthrough extraction efficiently aggregating multiscale accurately reconstructing them through its unique architecture. other focuses infrastructure breaking limitation independent processing traditional channels facilitating information exchange. To validate model's effectiveness, comprehensive experiments conducted public benchmark dataset. compared eight (EGNet, PoolNet, MINet, F3Net, U2Net, SegNet, BASNet, DeepCrack). Training performance evaluated using average mean absolute error (AveMAE), maximum F1 score (MaxF1), (MeanF1), precision–recall curves, visualizations. Experimental indicate that exceeds behavior networks during both training testing phases, MaxF1 MeanF1 achieving values 0.912 0.730, respectively; AveMAE 0.048, only 0.005 higher than minimum value, demonstrates effectiveness indicating potential future applications involving fusion.

Язык: Английский

---

An optimized and precise road crack segmentation network in complex scenarios

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

Опубликована: Фев. 17, 2025

Abstract Road cracks pose a serious threat to the stability of road structures and traffic safety. Therefore, this paper proposes an optimized accurate crack segmentation network called MBGBNet, which can solve problems complex background, tiny cracks, irregular edges in segmentation. First, multi‐scale domain feature aggregation is proposed address interference background. Second, bidirectional embedding fusion adaptive attention capture features finally, Gaussian weighted edge algorithm ensure accuracy In addition, uses preheated bat optimization algorithm, quickly determine optimal learning rate converge equilibrium. validation experiments on self‐built dataset, mean intersection over union reaches 80.54% precision 86.38%. MBGBNet outperforms other seven state‐of‐the‐art networks three classical datasets, highlighting its advanced capabilities. effective auxiliary method for solving safety problems.

Язык: Английский

---

Bridge damage description using adaptive attention-based image captioning

Automation in Construction, Год журнала: 2024, Номер 165, С. 105525 - 105525

Опубликована: Июнь 8, 2024

Язык: Английский

---

A machine vision‐based intelligent segmentation method for dam underwater cracks using swarm optimization algorithm and deep learning

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

Опубликована: Окт. 2, 2024

Abstract Ensuring the safety of water networks is a research hotspot in current conservancy industry, and dams are an important part. However, over time, dam prone to varying degrees aging disease, most which structural cracks. If they cannot be discovered repaired normal operation will affected, even catastrophic accidents such as failure occur. complex backgrounds blurred images can easily lead misjudgments by machine vision detection models, high‐efficiency accurate evaluation technology urgently needed. This paper combines deep semantic segmentation network model hyperparameters optimization algorithm propose data‐intelligent perception method underwater cracks driven knowledge coupling. Taking concrete face rockfill example, effectiveness verified using vehicle carrier. Experimental results indicate that developed achieves intersection‐union ratio 0.9301, precision rate 0.9678, 0.9472, recall 0.9577 test set. shows constructed has high crack fine performance. In addition, better performance different scenes, further illustrates method.

Язык: Английский

---

Attention‐optimized 3D segmentation and reconstruction system for sewer pipelines employing multi‐view images

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

Опубликована: Июнь 3, 2024

Abstract Existing deep learning‐based defect inspection results on images lack depth information to fully demonstrate the sewer, despite their high accuracy. To address this limitation, a novel attention‐optimized three‐dimensional (3D) segmentation and reconstruction system for sewer pipelines is presented. First, real‐time method called AM‐Pipe‐SegNet developed inspect defects (i.e., misalignment, obstacle, fracture) efficiently. Attention mechanisms (AMs) are introduced improve performance of segmentation. Second, an sparse‐initialized estimation network AM‐Pipe‐DepNet presented generate maps from multi‐view images. Third, 2D‐to‐3D mapping algorithm proposed remove noise transform into 3D spaces. Comparison experiments reveal that incorporating AMs significantly enhances pipe performance. Finally, two digital replicas real pipes built based photos taken by probes, providing valuable insights maintenance.

Язык: Английский

---

Automatic tiny crack positioning and width measurement with parallel laser line‐camera system

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

Опубликована: Янв. 16, 2025

Abstract Quantifying tiny cracks is crucial for assessing structural conditions. Traditional non‐contact measurement technologies often struggle to accurately measure crack widths, especially in hard‐to‐access areas. To address these challenges, this study introduces an image‐based, handheld parallel laser line‐camera (PLLC) system designed automated localization and width from multiple angles safe distances. Established by processing strips, the camera coordinate addresses positioning pixel scale distortion challenges typical non‐perpendicular photography. The determined enables accurate measurement. An improved U‐Net model automatically identifies pixels, enhancing detection accuracy. Additionally, newly developed Equal Area algorithm sub‐pixel of cracks. Comprehensive laboratory field testing demonstrates system's accuracy feasibility across various This PLLC achieves quantitative one shot, significantly efficiency utility on‐site inspections.

Язык: Английский

---

Automated corrosion surface quantification in steel transmission towers using UAV photogrammetry and deep convolutional neural networks

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

Опубликована: Фев. 4, 2025

Abstract Corrosion in steel transmission towers poses a challenge to structural integrity and safety, requiring efficient detection methods. Traditional visual inspections are unsustainable due the complexity volume of structures. Their manual, qualitative, subjective nature often leads inconsistencies maintenance planning. This study proposes deep learning‐based approach for semantic segmentation corroded areas on towers. Using DeepLabv3+ model, network was trained validated 999 field photographs. MobileNetV2, serving as feature extractor, chosen its optimal balance between accuracy computational efficiency, achieving validation 90.8% loss 0.23. The applied real‐world using orthomosaics derived from photogrammetric reconstructions South‐East tower at Torino Eremo broadcasting center. These products not only enabled precise but also provided foundation corrosion quantification with metrical accuracy, critical advantage Unlike traditional image methods, which lack spatial reference scaling, ensures that extent distribution quantified exact physical dimensions, enhancing reliability analysis. results show can automate detection, providing reliable data reducing reliance manual inspections, accuracy.

Язык: Английский

---

CASF-MNet: multi-scale network with cross attention mechanism and spatial dimension feature fusion for maize leaf disease detection

Crop Protection, Год журнала: 2024, Номер 180, С. 106667 - 106667

Опубликована: Март 19, 2024

Язык: Английский

---