高级搜索

    桥梁高空构件螺栓病害小样本快筛识别与可视化定位方法

    Rapid Identification and Visual Localization Method for Few-Shot Inspection of Bolt Defects in Bridge Tall Components

    • 摘要: 为了实现桥梁高空构件的高效检测及病害的快速识别与精准定位,提出了一种基于无人机和深度学习的小样本桥梁螺栓病害识别算法.通过无人机在桥梁可达性较差的区域进行定制化航线采集,获取固定姿态、角度和距离的表观图像,保障数据的时序可分析性.针对桥梁螺栓的锈蚀问题,基于YOLOv8模型提出一种适用于小样本情况下的病害识别方法.多种基线模型对比实验结果表明:该算法在小样本条件下具有优异的性能,识别精度超过90%.结合图像采集的定位信息与桥梁三维数字模型,实现了螺栓病害的三维可视化定位.该方法为桥梁高空难以到达区域的检修养护提供了高效解决方案,有助于保障桥梁的长期服役与安全耐久.

       

      Abstract: To enhance the efficiency of bridge tall component inspections and achieve rapid detection and accurate localization of defects,this study proposed a few-shot inspection algorithm for bridge bolt corrosion using Unmanned Aerial Vehicle (UAV) imagery and deep learning techniques.UAVs were employed to capture images of hard-to-reach areas by flying along a customized low-altitude path.This path was specifically designed to collect visual data at fixed angles,distances,and postures,ensuring the temporal consistency necessary for analysis.The study developed a corrosion detection method for bridge bolts based on the You Only Look Once version 8 (YOLOv8),tailored for few-shot datasets.Comparative experiments with multiple baseline models demonstrate that the proposed approach achieves superior performance,with an accuracy exceeding 90%,even with limited annotated sample numbers.Combining the collected image data with a digital model of the bridge enabled the precise visualization and localization of bolt defects in three-dimensional space.This method offers an efficient solution for the maintenance and inspection of difficult-to-reach bridge tall components,contributing to the long-term durability and safety of bridge structures.

       

    /

    返回文章
    返回