摘要
随着我国交通建设的蓬勃发展,桥梁数量的激增,桥梁定期检测业务量日益繁重。桥梁外观缺陷检测主要靠人工目测和简单工具的检测方式工作效率低、劳动强度大、环境和人为主观因素影响大。为了提高桥梁检测质量和效率,降低人工成本,采用YOLOv5算法的目标检测方法对混凝土桥梁表观缺陷(剥落、露筋、空洞等)智能识别进行分析研究,并通过定量的桥梁工程检测数据进行验证。结果表明:1)相较其他方法,采用YOLOv5算法的目标识别检测方法,其识别速率提升了4倍,能更快地识别出常见的混凝土表观缺陷,当IoU阈值为0.5时,其检测精度mAP值达到了72.5%,识别精度明显高于其他方法;2)YOLOv5算法应用于混凝土桥梁表观缺陷能很好地满足缺陷检测实时、快速和准确的需求,并可用于桥梁、隧道、路面及建筑物的外观检测。
With the rapid development of transportation construction in China,the surge in the number of bridges,the regular inspection business has become increasingly heavy.The detection of bridge surface defects mainly relies on manual visual inspection and simple tool detection,which is inefficient,laborintensive,and greatly affected by environmental and human subjective factors.In order to improve the detection quality and efficiency,reduce the cost of manual labor,the YOLOv5 algorithm′s object detection method is adopted to analyze the intelligent recognition of concrete bridge surface defects(peeling,exposed steel bars,hollows,etc.)and verify through quantitative real detection data.The results show that:1)Compared with other methods,the object recognition detection method using the YOLOv5 algorithm has a recognition rate increased by 4 times,which can identify common concrete surface defects faster.When the IoU threshold is 0.5,the mAP detection accuracy value reaches to 72.5%,which is significantly higher than other methods;2)The YOLOv5 algorithm applied to concrete bridge surface defect detection can meet the needs of fast and accurate defect detection and defect detection in real time,and can be used for the detection of the appearance of bridges,tunnels,roads and buildings.
作者
粟寒
斯新华
黄倩文
SU Han;SI Xinhua;HUANG Qianwen(China Merchants Chongqing Highway Engineering Testing Center Co.,Ltd.,Chongqing 400067 China;TYLIN International Engineering Consulting(China)Co.,Ltd.,Chongqing 401121 China)
出处
《公路交通技术》
2024年第6期90-94,共5页
Technology of Highway and Transport
关键词
YOLOv5
混凝土桥梁
表观缺陷
智能识别
实时检测
YOLOv5
concrete bridge
apparent defects
intelligent recognition
real-time detection
