隧道衬砌内部空洞等病害检测已经成为隧道检修人员的主要工作之一.本文提出一种将探地雷达与深度学习相结合的隧道衬砌空洞检测方法,通过雷达探测和仿真模拟,得到大量衬砌雷达图像,并对图像进行标注和制作数据集.基于YOLOv5(You Only Look Once version 5)目标检测模型,结合数据集目标特征,提出一种检测衬砌空洞的算法,引入特征融合模块提高网络感受野,并采用K-means聚类算法提高检测准确率.通过现场检测,本文的检测方法准确率达到了97.7%,准确可靠,可在工程中进行应用.
Inspection Method of Tunnel Lining Cavity Based on Ground Penetrating Radar
The inspection of cavities and other diseases in tunnel lining had become one of the main tasks of tunnel maintenance personnel.This paper proposed a method to detect cavity in tunnel lining by combining ground penetrating radar and deep learning.Through radar detection and simulation,a large number of tunnel lining radar images were obtained,then the images were labeled to make a dataset.Based on YOLOv5 network model,combined with the target features of the dataset,an algorithm for detecting typical diseases of tunnel lining was proposed.The feature fusion module was introduced to improve the receptive field of the network,and the K-means clustering algorithm was used to improve the detection accuracy.Through on-site testing,the inspection accuracy of the method reaches 97.7%,which proves that the inspection method in this paper is accurate and reliable,and can be applied in engineering.
tunneltunnel lining diseasecavityintelligent inspectionground penetrating radardeep learningexperimental research
NETL
NSTL
万方数据
