融合视觉测量和Transformer的桥梁裂缝智能检测与三维映射OA
Bridge crack intelligent detection and three-dimensional mapping by integrating visual measurement and Transformer algorithm
无人机巡检方法已成为桥梁表观病害检测的重要手段.针对无人机(unmanned aerial vehicle,UVA)易受振动干扰、成像质量不稳定、卷积神经网络识别裂缝小目标效果差及裂缝定位困难等问题,提出了融合视觉测量与Transformer的桥梁裂缝智能检测与三维映射方法,建立基于Real ESRGan图像超分辨率技术的航拍图像采集方法,复原运动模糊图像,实现对高质量桥梁图像的高效采集;在桥梁复杂背景下,通过具有全局注意力机制的Detection Transformer算法和数字图像处理技术,实现对裂缝病害的识别提取与高亮标识;利用多视角影像匹配法处理裂缝高亮标识图片,生成密集点云数据,构建桥梁病害面的精细模型,实现对裂缝病害的三维映射定位.以长沙市靳江河大桥为实验桥梁,使用无人机拍摄2 316张桥梁表观图像用于生成三维实景模型,拍摄479张精细图像用于病害检测,使用Real ESRgan超分辨率重建算法对运动模糊图像进行复原,所建立的桥梁病害部位精细模型像素点分辨率达0.25 mm/pixel,裂缝宽度测量最小相对误差为1.37%,最大相对误差为9.90%.结果表明,融合视觉测量和Transformer的桥梁裂缝智能检测与三维映射方法,能够有效提升检测效率,保障人员安全,实现对桥梁裂缝的数字化、智能化、可视化检测,具备重要的研究价值与广阔的应用前景.
UAV inspection has been an important method for detecting the visible defects of bridges.To address the problems of unstable image quality caused by UAV vibration,poor performance of convolutional neural networks(CNNs)in small-scale crack detection and difficulty in crack positioning,a bridge crack intelligent detection and three-dimensional mapping method using integrated visual measurement and Transformer algorithm is proposed.This method establishes an image acquisition based on the Real-ESRGan image super-resolution algorithm to restore blurred images and to achieve efficient collection of high-quality bridge images.Subsequently,under the complex bridge background conditions,this method constructs a framework integrating the Detection Transformer algorithm and digital image processing method to identify,extract,and visually highlight the cracks.Then,this method processes crack-highlighted images through multi-view stereo matching to generate dense points cloud and construct a detailed bridge model,enabling the three-dimensional mapping and localization of cracks.Taking the Jinjiang River Bridge in Changsha City as the experimental subject,2 316 bridge appearance images were captured by UAV for generating a three-dimensional real-scene model,and 479 detailed images were captured for defect detection.The Real ESRGan algorithm was used to restore motion-blurred images.The detailed model of the bridge defect part established has a pixel resolution of 0.25 mm/pixel;the minimum relative error of crack width measurement is 1.37%,and the maximum relative error is 9.90%.The experimental results demonstrate that the intelligent detection and three-dimensional mapping method for bridge cracks,integrating visual measurement and Transformer algorithm,can effectively enhance detection efficiency,ensure safety,and achieve digital,intelligent,and visual bridge crack detection,holding significant research value and broad application prospects.
余加勇;杨睿韬;王昱东;彭志豪;周劲
湖南大学 土木工程学院,湖南 长沙 410082||湖南大学 桥梁工程安全与韧性全国重点实验室,湖南 长沙 410082||湖南大学 天-空-地一体化结构健康监测与运维中心,湖南 长沙 410082湖南大学 土木工程学院,湖南 长沙 410082||湖南大学 桥梁工程安全与韧性全国重点实验室,湖南 长沙 410082湖南大学 土木工程学院,湖南 长沙 410082||湖南大学 桥梁工程安全与韧性全国重点实验室,湖南 长沙 410082湖南大学 土木工程学院,湖南 长沙 410082||湖南大学 桥梁工程安全与韧性全国重点实验室,湖南 长沙 410082湖南大学 土木工程学院,湖南 长沙 410082||湖南大学 桥梁工程安全与韧性全国重点实验室,湖南 长沙 410082
交通工程
桥梁裂缝检测无人机三维建模Transformer
bridgescrack detectionUAVthree-dimensional modelingTransformer
《湖南大学学报（自然科学版）》 2026 (3)
108-118,11
湖南省水利科技项目(XSKJ2021000-46),Hunan Water Resources Science and Technology Project(XSKJ2021000-46)湖南省自然资源厅科技计划项目(20230120DZ),Science and Technology Plan Project of Department of Natural Resources of Hunan Province(20230120DZ)
评论