Detecting pavement cracks is critical for road safety and infrastructure management.Traditional methods,relying on manual inspection and basic image processing,are time-consuming and prone to errors.Recent deep-learni...Detecting pavement cracks is critical for road safety and infrastructure management.Traditional methods,relying on manual inspection and basic image processing,are time-consuming and prone to errors.Recent deep-learning(DL)methods automate crack detection,but many still struggle with variable crack patterns and environmental conditions.This study aims to address these limitations by introducing the Masker Transformer,a novel hybrid deep learning model that integrates the precise localization capabilities of Mask Region-based Convolutional Neural Network(Mask R-CNN)with the global contextual awareness of Vision Transformer(ViT).The research focuses on leveraging the strengths of both architectures to enhance segmentation accuracy and adaptability across different pavement conditions.We evaluated the performance of theMaskerTransformer against other state-of-theartmodels such asU-Net,TransformerU-Net(TransUNet),U-NetTransformer(UNETr),SwinU-NetTransformer(Swin-UNETr),You Only Look Once version 8(YoloV8),and Mask R-CNN using two benchmark datasets:Crack500 and DeepCrack.The findings reveal that the MaskerTransformer significantly outperforms the existing models,achieving the highest Dice SimilarityCoefficient(DSC),precision,recall,and F1-Score across both datasets.Specifically,the model attained a DSC of 80.04%on Crack500 and 91.37%on DeepCrack,demonstrating superior segmentation accuracy and reliability.The high precision and recall rates further substantiate its effectiveness in real-world applications,suggesting that the Masker Transformer can serve as a robust tool for automated pavement crack detection,potentially replacing more traditional methods.展开更多
M50轴承钢中主要的碳化物类型为MC、M_(2)C和M_(23)C_(6)。扫描电子显微镜(Scanning Electron Microscopy,SEM)下,3种碳化物的形状、尺寸和在材料中的分布存在明显的区别。有些碳化物的尺寸较大且分布不均匀。轴承受载过程中,这些碳化...M50轴承钢中主要的碳化物类型为MC、M_(2)C和M_(23)C_(6)。扫描电子显微镜(Scanning Electron Microscopy,SEM)下,3种碳化物的形状、尺寸和在材料中的分布存在明显的区别。有些碳化物的尺寸较大且分布不均匀。轴承受载过程中,这些碳化物会成为应力集中的区域,对轴承疲劳性能产生负面影响。为了高效地获得材料中的碳化物信息,提出一种改进的掩膜基于区域的卷积神经网络(Mask Region-based Convolutional Neural Network,Mask R-CNN)模型,可批量鉴别SEM图像中3种碳化物的种类,确定其尺寸大小及分布。网络模型输出的图像和数值结果显示,M50轴承钢中M_(2)C型碳化物尺寸大且分布不均匀,但总体尺寸最大的MC型碳化物和尺寸最小的M_(23)C_(6)型碳化物分布相对均匀。展开更多
船舶长期暴露在高盐高湿的海洋环境中,腐蚀问题严重影响其结构安全与运营效率。传统人工检测存在效率低、精度不足的局限性,而传统目标检测算法并未针对船舶腐蚀问题进行有针对性的优化。因此,本文提出了一种基于Mask R-CNN的智能船舶...船舶长期暴露在高盐高湿的海洋环境中,腐蚀问题严重影响其结构安全与运营效率。传统人工检测存在效率低、精度不足的局限性,而传统目标检测算法并未针对船舶腐蚀问题进行有针对性的优化。因此,本文提出了一种基于Mask R-CNN的智能船舶腐蚀检测方法。本文通过多源数据整合与模型优化,实现腐蚀区域的精准识别,研究汇集各开源数据,构建包含135张高分辨率图像的标准数据集,并在预处理后利用Labelme进行像素级精细的多边形标注。在模型上,本研究选择结合特征金字塔网络(FPN)与改进型ROIAlign模块,并采用动态学习率策略优化训练,通过迁移学习策略解决了船舶腐蚀检测数据集合数量不足的问题。最后的实验结果表明,Mask R-CNN结合FPN结构与128 × ROIAlign模块的模型在多尺度检测方面表现优异,平均精度(AP)达到0.414,较基准配置(Faster R-CNN结合C4结构与64 × ROIAlign模块)提升14.0%;小目标检测精度(APS)提高11.4%;边界定位精度(AP75)提升15.6%。本研究验证了深度学习在船舶腐蚀检测中的有效性,并提出轻量化部署方案,为工程应用提供技术支撑。Ships are chronically exposed to high-salinity and high-humidity marine environments, where corrosion severely compromises structural safety and operational efficiency. Traditional manual inspection suffers from inefficiency and limited accuracy, while conventional object detection algorithms lack targeted optimization for ship corrosion issues. To address these challenges, this paper proposes an intelligent ship corrosion detection method based on Mask R-CNN. Through multi-source data integration and model optimization, our approach achieves precise identification of corrosion areas. We aggregated open-source datasets to construct a standardized dataset comprising 135 high-resolution images, followed by pixel-level polygonal annotations using Labelme after preprocessing. For the model architecture, we integrated a Feature Pyramid Network (FPN) with an enhanced ROIAlign module and adopted a dynamic learning rate strategy to optimize training. A transfer learning strategy was employed to mitigate the limited dataset size for ship corrosion detection. Experimental results demonstrate that the Mask R-CNN model combined with FPN and a 128 × ROIAlign module excels in multi-scale detection, achieving a mean Average Precision (AP) of 0.414—a 14.0% improvement over the baseline configuration (Faster R-CNN with C4 backbone and 64×ROIAlign). Notably, small-target detection precision (APS) increased by 11.4%, and boundary localization accuracy (AP75) improved by 15.6%. This study validates the effectiveness of deep learning in ship corrosion detection and proposes a lightweight deployment solution, offering technical support for engineering applications.展开更多
文摘Detecting pavement cracks is critical for road safety and infrastructure management.Traditional methods,relying on manual inspection and basic image processing,are time-consuming and prone to errors.Recent deep-learning(DL)methods automate crack detection,but many still struggle with variable crack patterns and environmental conditions.This study aims to address these limitations by introducing the Masker Transformer,a novel hybrid deep learning model that integrates the precise localization capabilities of Mask Region-based Convolutional Neural Network(Mask R-CNN)with the global contextual awareness of Vision Transformer(ViT).The research focuses on leveraging the strengths of both architectures to enhance segmentation accuracy and adaptability across different pavement conditions.We evaluated the performance of theMaskerTransformer against other state-of-theartmodels such asU-Net,TransformerU-Net(TransUNet),U-NetTransformer(UNETr),SwinU-NetTransformer(Swin-UNETr),You Only Look Once version 8(YoloV8),and Mask R-CNN using two benchmark datasets:Crack500 and DeepCrack.The findings reveal that the MaskerTransformer significantly outperforms the existing models,achieving the highest Dice SimilarityCoefficient(DSC),precision,recall,and F1-Score across both datasets.Specifically,the model attained a DSC of 80.04%on Crack500 and 91.37%on DeepCrack,demonstrating superior segmentation accuracy and reliability.The high precision and recall rates further substantiate its effectiveness in real-world applications,suggesting that the Masker Transformer can serve as a robust tool for automated pavement crack detection,potentially replacing more traditional methods.
文摘船舶长期暴露在高盐高湿的海洋环境中,腐蚀问题严重影响其结构安全与运营效率。传统人工检测存在效率低、精度不足的局限性,而传统目标检测算法并未针对船舶腐蚀问题进行有针对性的优化。因此,本文提出了一种基于Mask R-CNN的智能船舶腐蚀检测方法。本文通过多源数据整合与模型优化,实现腐蚀区域的精准识别,研究汇集各开源数据,构建包含135张高分辨率图像的标准数据集,并在预处理后利用Labelme进行像素级精细的多边形标注。在模型上,本研究选择结合特征金字塔网络(FPN)与改进型ROIAlign模块,并采用动态学习率策略优化训练,通过迁移学习策略解决了船舶腐蚀检测数据集合数量不足的问题。最后的实验结果表明,Mask R-CNN结合FPN结构与128 × ROIAlign模块的模型在多尺度检测方面表现优异,平均精度(AP)达到0.414,较基准配置(Faster R-CNN结合C4结构与64 × ROIAlign模块)提升14.0%;小目标检测精度(APS)提高11.4%;边界定位精度(AP75)提升15.6%。本研究验证了深度学习在船舶腐蚀检测中的有效性,并提出轻量化部署方案,为工程应用提供技术支撑。Ships are chronically exposed to high-salinity and high-humidity marine environments, where corrosion severely compromises structural safety and operational efficiency. Traditional manual inspection suffers from inefficiency and limited accuracy, while conventional object detection algorithms lack targeted optimization for ship corrosion issues. To address these challenges, this paper proposes an intelligent ship corrosion detection method based on Mask R-CNN. Through multi-source data integration and model optimization, our approach achieves precise identification of corrosion areas. We aggregated open-source datasets to construct a standardized dataset comprising 135 high-resolution images, followed by pixel-level polygonal annotations using Labelme after preprocessing. For the model architecture, we integrated a Feature Pyramid Network (FPN) with an enhanced ROIAlign module and adopted a dynamic learning rate strategy to optimize training. A transfer learning strategy was employed to mitigate the limited dataset size for ship corrosion detection. Experimental results demonstrate that the Mask R-CNN model combined with FPN and a 128 × ROIAlign module excels in multi-scale detection, achieving a mean Average Precision (AP) of 0.414—a 14.0% improvement over the baseline configuration (Faster R-CNN with C4 backbone and 64×ROIAlign). Notably, small-target detection precision (APS) increased by 11.4%, and boundary localization accuracy (AP75) improved by 15.6%. This study validates the effectiveness of deep learning in ship corrosion detection and proposes a lightweight deployment solution, offering technical support for engineering applications.
