基于激光雷达的隧道轮廓三维点云重构与形变检测研究  

Research on three-dimensional point cloud reconstruction and deformation detection of tunnel contours based on LiDAR

作　　者：王耀东[1] 苏广思 方恩权 周伟[3] WANG Yaodong;SU Guangsi;FANG Enquan;ZHOU Wei(State Key Laboratory of Advanced Rail Autonomous Operation,Beijing Jiaotong University,Beijing 100044,China;Guangzhou Metro Group Co.Ltd.,Guangzhou 510335,China;School of Traffic&Transportation Engineering,Central South University,Changsha 410075,China)

机构地区：[1]先进轨道交通自主运行全国重点实验室(北京交通大学),北京100044 [2]广州地铁集团有限公司,广东广州510335 [3]中南大学交通运输工程学院,湖南长沙410075

出　　处：《中南大学学报（自然科学版）》2024年第6期2393-2403,共11页Journal of Central South University:Science and Technology

基　　金：中央高校基本科研业务费专项资金资助项目(2022JBXT005);先进轨道交通自主运行全国重点实验室(北京交通大学)项目(RAO2023ZZ003);国能朔黄铁路技术开发项目(SHTL-22-28)。

摘　　要：为实现地铁隧道轮廓全面、高效、精准的数字化检测,提出一种基于三维点云的隧道形变检测方法。该方法通过相对定位算法将激光雷达获取的多周期隧道轮廓点云数据进行数据融合,利用地铁隧道建模算法对融合数据进行处理建立标准隧道轮廓模型,根据测量值与模型输出值对比结果完成形变检测。相对定位算法利用转速传感器与激光位移传感器获取的公里标与轨道特征数据实现对地铁隧道轮廓特征数据粗、细校准定位,以解决转速传感器定位误差较大导致的相同位置不同检测周期激光点云数据无法对齐融合的问题。地铁隧道建模算法基于径向基神经网络(RBFNN)对融合后点云数据进行多重训练并不断剔除大误差数据建立隧道普通内壁模型,结合聚类算法对被剔除数据训练建立隧道管线区域模型。研究结果表明:相对定位算法可实现多周期数据融合,相对定位误差小于10 cm;隧道建模算法利用点云数据可建立标准隧道模型实现隧道形变分析,形变分析误差小于10 mm,达到预期效果。In order to realize comprehensive,efficient and accurate digital detection of subway tunnel contour,a tunnel deformation detection method based on three-dimensional point cloud was proposed in this paper.In this method,the multi-period tunnel contour point cloud data obtained by LiDAR wad fused by positioning algorithm,and a standard tunnel contour model was established by using subway tunnel modeling algorithm to process the fused data.The deformation detection was completed by comparing the measured value with the output value of the model.The positioning algorithm used the mileage mark and track feature data obtained by the speed sensor and the laser displacement sensor to realize coarse and fine calibration positioning of the subway tunnel contour feature data,so as to solve the problem that the laser point cloud data of the same position and different detection periods could not be aligned and fused due to the large positioning error of the speed sensor.The subway tunnel modeling algorithm was based on radial basis neural network(RBFNN),which trained the fused point cloud data multiple times and continuously removed the large error data to establish the common tunnel inner wall model,and combined the clustering algorithm to train the eliminated data to establish the tunnel pipeline region model.The results show that the relative positioning algorithm can realize multi-period data fusion,and the relative positioning error is less than 10 cm.The tunnel modeling algorithm can use the point cloud data to establish the standard tunnel inner wall model to analyze the tunnel deformation and achieve the expected effect.

关 键 词：地铁隧道 定位算法 RBF神经网络 隧道模型 形变分析 

分 类 号：U456.3[建筑科学—桥梁与隧道工程]