基于YOLOv5的钢桥面多类别病害检测方法

A Method for Detecting Multiple Category Diseases of Steel Bridge Deck Based on YOLOv5

针对钢桥面检测中出现的病害种类繁多、形状多样等问题,对照路面病害检测评定标准,需找到并训练一种适合实际钢桥面养护作业的钢桥面病害识别算法,满足全面自动化检测钢桥面病害的应用要求。文中通过多功能检测车采集钢桥面图像,并利用labelimg标注9种病害建立自制数据集,采用YOLOv5算法完成检测模型训练。结果表明,YOLOv5模型的病害检测精确率可达到80%,其中,对于裂缝类病害的检测精度相对较低,块状病害的检测精度较高。

Due to the wide variety of types and shapes of defects in steel bridge deck detection,as well as the higher requirements compared to pavement disease detection and evaluation standards,it is necessary to find and train a steel bridge deck disease recognition algorithm suitable for practical steel bridge deck maintenance operations to meet the application requirements for fully automated detection of steel bridge deck defects.The multi-function detection vehicle is used to collect images of the steel bridge deck,and nine types of defects are labeled using labelimg to establish a self-made dataset.The YOLOv5 algorithm is used to complete the training of the detection model.The experimental results show that the disease detection accuracy of the YOLOv5 model can reach 80%,with relatively low detection accuracy for crack-type defects and high detection accuracy for block-type defects.