摘要
[目的]地铁隧道内壁缺陷主要以裂纹和渗漏水为主,目前以人工和半自动化设备为主的检测方式存在强度大、效率低、可靠度不高等问题。应研究基于智能的检测识别算法及检测系统,以实现地铁隧道衬砌缺陷检测的信息化及智能化。[方法]分析了地铁隧道缺陷巡检技术现状,提出了一套适用于地铁隧道衬砌缺陷的识别算法,主要包括图像处理算法、缺陷分类检测算法及缺陷分级检测算法等,并选用了4个指标,用以评估该识别算法的检测效果。进一步从软件和硬件2个方面,建立了基于深度学习法的地铁隧道衬砌缺陷智能检测系统。最后将该系统应用于北京地铁3条线路上,分析其现场应用的可靠性。[结果及结论]应用该智能检测系统后,地铁隧道衬砌裂纹缺陷的检测率为91.95%,误检率为0.89%;渗漏水缺陷的检测率为93.83%,误检率为0.65%。该系统可作为地铁隧道智能化检测的核心平台,对地铁隧道各种缺陷进行有效检测。
[Objective]The main defects on metro tunnel inner wall are cracks and water seepage.Current detection methods based on manual and semi-automatic equipment have problems such as high intensity,low efficiency,and low reliability.The intelligence-based detection/identification algorithm and detection system should be studied to realize informative and intelligent detection for metro tunnel lining defects.[Method]The current status of metro tunnel defect inspection technology is analyzed,and a set of algorithms applicable to the identification of metro tunnel lining defects are put forward,mainly including image processing algorithm,defect category detection algorithm,and defect grading detection algorithm,etc.Four indexes are chosen to evaluate the detection effect of the identification algorithm.Furthermore,an intelligent detection system for metro tunnel lining defects based on deep learning method is established from both software and hardware aspects.Finally,the system is implemented on Beingjing Metro Line 3 to analyze the reliability of its on-site application.[Result&Conclusion]After the application of the intelligent detection system,the detection rate of metro tunnel lining crack defects reaches 91.95%with a false detection rate of 0.89%.The detection rate of seepage defect is 93.83%and the false detection rate 0.65%.This system can be used as the core intelligent detection platform of metro tunnel,effectively detecting various defects in metro tunnel.
作者
张悦
韩静
关祈峰
侯珏
陆婷婷
张亚芹
ZHANG Yue;HAN Jing;GUAN Qifeng;HOU Jue;LU Tingting;ZHANG Yaqin(Beijing MTR Co.,Ltd.,100068,Beijing,China;Shanghai Oriental Maritime Engineering Technology Co.,Ltd.,200011,Shanghai,China)
出处
《城市轨道交通研究》
北大核心
2024年第9期311-316,共6页
Urban Mass Transit
关键词
地铁
隧道衬砌缺陷
智能检测
裂纹识别
渗漏水识别
图像算法
深度学习
metro
tunnel lining defect
intelligent detection
crack identification
water seepage identification
image algorithm
deep learning
