基于YOLOv8的铁路隧道渗漏水智能检测

Intelligent Detection of Water Leakage in Railway Tunnel Based on YOLOv8

铁路隧道渗漏水会影响隧道的结构稳定性和运营安全,渗漏水病害的自动化检测亟待解决。传统的人工巡检方法自动化程度较低,检测效率低下,容易出现错检漏检,无法满足大规模隧道的快速检测需求。针对这一问题,提出了一种基于YOLOv8网络的铁路隧道渗漏水智能检测方法,并在自建的铁路隧道渗漏水数据集中进行模型训练和参数调优。实验结果表明,在不同版本的模型实验中,YOLOv8-n网络的综合性能最好。在不同模型对比实验中,YOLOv8模型的F_(1)分数、AP值分别为81.28%,81.38%,相比于YOLOv7、YOLOv5、SSD模型。分别提高了6.85%,6.89%,9.40%、8.19%,12.19%、10.57%。综合分析可得,YOLOv8模型综合性能最优秀,适用于铁路隧道工程的渗漏水检测任务。

The water leakage of railroad tunnel will affect the structural stability and operational safety of the tunnel,and the automated detection of water leakage defect needs to be solved urgently.The traditional manual inspection method is lower in automation and detection efficiency,and is prone to errors and omissions,which cannot meet the demand of rapid detection of large-scale tunnels.To solve this problem,a YOLOv8 network-based intelligent water leakage detection method for railroad tunnels is proposed,and the model training and parameter tuning are carried out in a self-constructed water leakage dataset of railroad tunnels.The experimental results show that the YOLOv8-n network has the best overall performance among the different versions of the model experiments.In the comparison experiments of different models,the F_(1) score and AP of the YOLOv8 model are 81.28%and 81.38%,respectively,which are 6.85%and 6.89%,9.40%and 8.19%,12.19%and 10.57%higher compared to the YOLOv7,YOLOv5,and SSD models,respectively.The comprehensive analysis shows that the YOLOv8 model has the best overall performance and is suitable for water leakage detection task in railroad tunnel projects.