An Method For 3-D As-Built Shield Tunnel Model Reconstruction From Point Cloud

The laser scanners can quickly obtain the point cloud of shield tunnels.The point cloud can show the state of each segment in the tunnel lining.The method of obtaining segment information from the point cloud requires further research.This dissertation studies the reconstruction method of the as-built circular shield tunnel lining model based on the point cloud.And a calculation method of segment dislocation and ovality based on the as-built model is proposed.The methods of segment point cloud classification,parameter fitting,and evaluation of assembly quality are studied in the process of automatic processing of point cloud.The proposed algorithms are utilized to analyze two shield tunnels.The current analysis level of the processing point cloud with the tunnel section as the basic cell has been improved to the analysis level with the segment as the basic cell.The research work and results obtained in this dissertation are as follows:(1)According to the relevant specifications of the shield tunnel and the structural characteristics of the segment,the calculation method of the tunnel alignment and the control parameters in the segment model are studied,and the tunnel alignment and segment models are parametrically designed.Parametric control alignment and segment models have been implemented.(2)To solve the misclassification problem caused by the distortion of segments in the unwrapped map,a method is proposed to classify the point cloud according to the segment joints in the unwrapped depth map.This method can extract point clouds for cylinder fitting for each segment.A method for obtaining point clouds for determining segment boundaries based on the above classification results is proposed.Finally,two kinds of point clouds are extracted for each segment,which solves the problem of fitting accuracy under one type of point cloud of a segment.(3)For the classified segment point cloud,a new cylindrical fitting algorithm is proposed based on the characteristics of incomplete cylinder surfaces.The algorithm innovatively proposes a method for finding the cylinder axis direction.The sheaf of planes is used to reduce the two variables for finding the direction to one variable,and the calculation efficiency is improved by the dichotomy.The calculation time of the algorithm is about 20% of the nonlinear least-squares method,and high-precision results can still be obtained in noise data.(4)A method for determining segment boundaries using template matching in the unwrapped map of the segment point cloud is proposed.To ensure that the unwrapped map of the segment point cloud is consistent with the design,the unwrapped map is generated according to the cylindrical surface parameters of the segment.The structural features in the unwrapped map include bolt holes,grouting well,etc.Using multiple features to jointly determine the segment boundary can accurately determine the boundary when the boundary is blocked or missing.(5)The reconstruction of the as-built model of the shield tunnel lining is completed based on the segment parameters obtained from the point cloud.The as-built model is continuously updated to solve the problem that the model is different from the actual situation.Based on the as-built model,an algorithm for evaluating segment dislocation and ovality is proposed.This method takes a single segment as the research object and upgrades the random inspection items to a full inspection.This method reduces the workload and provides more refined and accurate results.(6)The segment reconstruction and assembly quality detection algorithm has been verified in two shield tunnels,and the expected results have been achieved.The results show that the classification algorithm can correctly extract various segments in the two tunnels.The fitting algorithm also achieves high-precision calculation results in the actual collected data.The standard deviation of the point-to-model distance is only 1 mm.Calculating the dislocation can obtain the dislocation at any position,and the difference is within 1mm compared with the manual measurement results.