Subway Tunnel Structure Detection Based On Improved 3D Point Cloud Instance Segmentatio

With the continuous development of urban rail transit,more and more attention has been paid to the safety operation of subway.Subway because its saving energy,less pollution,large volume,high efficiency advantages widely loved by the masses,but because of the subway tunnel conditions and repeated vibration cause the tunnel structure is easy to deformation,cause great losses,at the same time,the subway tunnel with high construction cost,long construction period,the late maintenance costs such as big shortcoming.As a new technology,3D laser scanning technology has more exploration and application in the protection of cultural relics,architecture,traffic accident handling,large container loading and so on.In this paper,3d laser scanning technology is used to carry out key technology research on deformation detection of subway tunnel structure based on 3d point cloud data,and the existing 3D point cloud instance segmentation network model is improved based on segmentation accuracy and calculation efficiency.Thesis first introduces the research status and application of 3d laser scanner in tunnel structure detection at home and abroad,intelligent processing of point cloud data acquisition,neural network model,point cloud instance segmentation technology,transfer learning and other basic knowledge.Then,the improved 3d point cloud instance segmentation network model is introduced in detail: In the existing 3 d-Bonet three-dimensional point cloud instance in multi-scale segmentation model group(MSG)module and innovative to apply migration study in three-dimensional point cloud data,to test and verify the effectiveness of the improved model,the improved model on the same public data sets and the 3 d point cloud instance segmentation effect is good,comparing the two models Experimental results show that the improved model has improved both computational efficiency and segmentation accuracy.Secondly,taking a subway tunnel in Qingdao as an example,a 3D laser scanner is used to collect relatively complete point cloud data of subway tunnel segment,and the point cloud data is preprocessed.The improved 3D point cloud instance segmentation network model and migration learning are directly applied to 3D point cloud tunnel data.The experiment proves that the improved case segmentation model can effectively detect the height guide distance of catenary,height difference of left and right track,parallelism of left and right track,center line of left and right track plane in tunnel.At the same time,in order to further verify the application of this method in the tube ring and segment of subway tunnel,the deformation detection is carried out from the roundness of the tube ring,the dislocation between the rings and the misalignment between segments,and the practicability of this method is verified by experiments.In terms of 3D data fitting,for instance the tunnel after the 3D data segmentation consistency by using random sampling combined with least square method for data fitting,through construction requirements index to compare the results with the actual measurement result,that put forward a new method of fitting error range can meet the needs of practical production,further prove the effectiveness of the proposed method.Research results show that the 3D laser scanner subway tunnel point cloud data,using the improved 3D point cloud instance network segmentation model can implement the subway tunnel structure deformation test,the experimental results of the analysis of metro tunnel in the late research provides the theoretical basis,at the same time,the method for the metro tunnel structure detection provides more innovation means.This method has a broad application prospect in the future maintenance and monitoring of tunnel safety and stability.