摘要
桥梁病害(如混凝土剥落、桥梁裂纹、铆钉锈蚀等)大多发生在局部区域,而目前对桥梁病害的检测通常没有实现病害在全桥的定位。结合无人机的高清摄像功能和YOLOv5的实时目标检测能力,可以实现对桥面病害的全场快速定位和检测。为此,提出一种基于YOLOv5的无人机桥面病害检测算法。采用无人机对桥梁路面进行数据采集,将轻量化模型YOLOv5s作为基础检测模型,并对YOLOv5s模型做如下改进:在模型现有的3个不同尺度特征图检测的基础上额外增加2个尺度,提高较大目标和较小目标的检测准确性;采用Soft-NMS算法替代NMS算法,同时为保证密集病害检测的全场快速定位和检测精度,将采集到的桥梁路面数据输入到改进YOLOv5s模型中,该模型输出即为桥面病害检测结果。实验结果表明,经过优化的YOLOv5s模型的mAP@0.5值达到了92.0%,mAP@0.5:0.95值也达到了73.2%。此外,该模型处理速度达到134 f/s,高效准确地识别了桥梁路面病害,显著提升了检测的精准性和效率。
Bridge diseases,such as concrete peeling,bridge cracks,rivet corrosion,etc.,mostly occur in local areas,but most of the bridge diseases are not located in the whole bridge at present.The full-field rapid location and detection of bridge deck diseases can be realized by combining the high-definition camera function of unmanned aerial vehicle(UAV)with the real-time target detection ability of YOLOv5.Therefore,an UAV bridge deck disease detection algorithm based on YOLOv5 is proposed.The UAV is used to collect data on the bridge pavement,and the lightweight model YOLOv5s is used as the basic detection model.The YOLOv5s model is improved as follows:two scales are added on the basis of the existing three characteristic maps detection with different scales to improve the detection accuracy of larger targets and smaller targets;Soft-NMS algorithm is used to instead of NMS algorithm.In order to ensure the full-field rapid positioning and detection accuracy of dense diseases,the collected bridge pavement data is input into the improved YOLOv5s model,and the output of the model is the detection result of bridge deck diseases.The experimental results show that the value of mAP@0.5 of the optimized YOLOv5s model can reach 92.0%,and the value of mAP@0.5:0.95 also can reach 73.2%.The processing speed of the model can reach 134 f/s,which effectively and accurately identifies bridge pavement diseases,and significantly improves the accuracy and efficiency of detection.
作者
戴鹏飞
邹京汕
杨柳
刘恒
阴慧颖
DAI Pengfei;ZOU Jingshan;YANG Liu;LIU Heng;YIN Huiying(Nanjing Tech University,Nanjing 211816,China;China Railway Bridge and Tunnel Technology Co.,Ltd.,Nanjing 210000,China;School of Information Science and Technology,Southwest JiaoTong University,Chengdu 611756,China;National Engineering Laboratory of Comprehensive Transportation Big Data Application Technology,Southwest JiaoTong University,Chengdu 611756,China;Tangshan Institute,Southwest Jiaotong University,Tangshan 063000,China)
出处
《现代电子技术》
北大核心
2025年第16期7-12,共6页
Modern Electronics Technique
关键词
桥面病害检测
YOLOv5
无人机
图像采集
多尺度检测
特征融合
bridge deck disease detection
YOLOv5
unmanned aerial vehicle
image collection
multi-scale detection
feature fusion
