朔黄铁路隧道衬砌表观病害检测技术  

Inspection Technology of Tunnel Lining Apparent Defects in Shuozhou-Huanghua Railway

作　　者：王敬 王宁 李健超 段培勇 WANG Jing;WANG Ning;LI Jianchao;DUAN Peiyong(Guoneng Shuohuang Railway Development Co.Ltd.,Suning Hebei 062350,China;Railway Engineering Research Institute,China Academy of Railway Sciences Corporation Limited,Beijing 100081,China)

机构地区：[1]国能朔黄铁路发展有限责任公司,河北肃宁062350 [2]中国铁道科学研究院集团有限公司铁道建筑研究所,北京100081

出　　处：《铁道建筑》2022年第8期122-125,共4页Railway Engineering

基　　金：国家能源投资集团有限责任公司科技创新项目(GJNY-20-231)。

摘　　要：为实现朔黄铁路隧道衬砌表观病害远距离非接触快速检测,提出一种基于多源数据深度融合的隧道病害检测方法。首先利用高清线阵相机、激光扫描传感器等检测设备获取隧道衬砌表观高清图像和激光点云数据,然后利用特征提取网络提取图像和点云特征图,并采用空间变换方法将图像特征图投影到点云特征俯视图上得到融合特征图,最后利用候选区域网络和金字塔场景分析网络对融合特征图进行检测识别,输出病害的类型与位置信息。在朔黄铁路重点隧道开展的现场试验表明,该方法能检测隧道裂缝、掉块、渗水等表观病害状态,有效提升重载铁路隧道运维的智能化程度及综合检测水平。In order to realize the long-distance non-contact rapid inspection of tunnel lining apparent defects in Shuozhou-Huanghua Railway,a tunnel defect inspection method based on multi-source data deep fusion was proposed.Firstly,the high-definition linear camera,laser scanning sensor and other detection equipment were used to obtain the high-definition tunnel lining apparent image and laser point cloud data. Then,the feature extraction network was used to extract the image and the point cloud feature map,and the image feature map was projected onto the point cloud feature vertical view by the spatial transformation method to obtain the fusion feature map. Finally,the candidate area network and pyramid scene parsing network were used to inspect and identify the fusion feature map,and the type and location information of the defect were output. The field test carried out in the key tunnel of Shuozhou-Huanghua Railway shows that this method can inspect the tunnel crack,block falling,water seepage and other apparent defects,and effectively improve the intelligent level and comprehensive inspection level of the operation and maintenance of the heavy haul railway tunnel.

关 键 词：铁道隧道 病害智能检测算法 试验研究 隧道病害 深度学习 激光点云 融合网络模型 

分 类 号：U456[建筑科学—桥梁与隧道工程]