摘要
针对钢筋混凝土桥梁裂缝人工检测效率低、安全风险高、微小裂缝漏检及复杂背景误检等问题,提出无人机与改进YOLOv11-DeepLabv3+网络协同的钢筋混凝土桥裂缝检测方法。通过无人机实现对桥梁表面无接触高效成像,在YOLOv11中引入AKConv可变形卷积替换标准卷积,增强多尺度裂缝特征提取能力。结合DeepLabv3+语义分割网络,实现裂缝定位与形态量化分析。测试表明,该检测方法显著提升裂缝检测精度与召回率,且裂缝长度及宽度测量误差小于15%,可自动完成桥梁全表面裂缝检测。
To address the issues of low efficiency,high safety risks,missed detection of tiny cracks,and false detec⁃tion in complex backgrounds during manual detection of reinforced concrete bridge cracks,an detection method for re⁃inforced concrete bridge combining unmanned aerial vehicle and an improved YOLOv11⁃DeepLabv3+network is pro⁃posed.The method enables non⁃contact and efficient imaging of the bridge surface through unmanned aerial vehicle.AKConv deformable convolution is introduced to replace the standard convolution in YOLOv11 so as to enhance the ability to extract multi⁃scale crack features.After YOLOv11 network and the DeepLabv3+semantic segmentation net⁃work are combined,crack location and morphological quantitative analysis are achived.Tests show that the detection method significantly improves the accuracy and recall rate of crack detection,with measurement errors of crack length and width are less than 15%,and the detection of cracks on the entire surface of the bridge can be automatically com⁃pleted.
作者
张虎耀
吕丽
高田田
张丽
刘洪彬
张林
ZHANG Huyao;LYU Li;GAO Tiantian;ZHANG Li;LIU Hongbin;ZHANG Lin
出处
《铁道技术监督》
2025年第11期45-50,共6页
Railway Quality Control
关键词
钢筋混凝土桥
桥梁观测
裂缝检测
无人机
图像采集
语义分割
量化分析
Reinforced Concrete Bridge
Bridge Observation
Crack Detection
Unmanned Aerial Vehicle
Image Ac⁃quisition
Semantic Segmentation
Quantitative Analysis
