Research On Tunnel Disease Detection Method Based On Multi-sensor Integration

The number and mileage of tunnels in China rank first in the world now,and tunnel safety is closely related to people's lives and social economic development.During the operation of the tunnel,various diseases which affect safety will appear.However,existing manual inspection has low efficiency and high security risks,and it is difficult to meet large-scale high-speed,efficient and high-precision tunnel inspection.Researching an automated tunnel detection technology has very practical application value.In this paper,we conducted systematic research on tunnel detection system,tunnel 3D point cloud processing and tunnel lining image processing.The main contents are as follows:(1)A rapid detection method for tunnel diseases based on multi-sensor integration is researched.The data characteristics of tunnel lining gray and infrared image acquired by CCD and infrared camera array under dynamic conditions are analyzed,and the problems existing in the tunnel disease assessment and the causes of these problems are analyzed.The characteristics of measuring 3D tunnel through LIDAR are researched,and the influencing factors of 3D point cloud modeling are analyzed.In order to collect tunnel lining images better,a lining image data acquisition strategy was designed.(2)Aiming at the problem that the auxiliary structure affects the 3D modeling of the tunnel and causes the calculation error of the array camera object distance,a new method for removing the noise of the auxiliary structure is designed,and a new method is used to realize the 3D reconstruction of the tunnel point cloud.Firstly,the tunnel point cloud is segmented.Secondly,the standard cross section of each part is calculated,and the standard cross section is used to reconstruct the part.Finally,each section was spliced.This method can effectively remove the noise of auxiliary structures.In order to reconstruct the 3D model of the tunnel,the 3D point cloud of the tunnel is expanded into planar point cloud.And triangulation algorithm is used to construct the triangle network between points in plane point cloud,then the triangle network is applied to 3D point cloud.This method can obtain a high-precision 3D model of the tunnel.The processing of point cloud data solves the problem of dynamic object distance calculation of array camera.(3)Aiming at the problem of adjacent image misalignment caused by the relative displacement of the sensor in dynamic measurement,a method for stitching tunnel lining images based on geometric relationship and data characteristics is designed.This method firstly roughly locates the position of the gray image of the tunnel lining in the panorama based on the geometric relationship during image acquisition.Then,for images with better quality,the SURF algorithm is used to perform high-precision image registration on adjacent images in order to update the position of the image in the panorama and improve the stitching accuracy.The prototype equipment has been applied in the inspection of the two and three lanes tunnels in China.It can realize the detection of tunnel crack damage,water damage and freezing damage.The point cloud processing method can eliminate the influence of the noise of the tunnel auxiliary structure and reconstruct the 3D model of the tunnel well.The sequence image stitching method can realize the lining cross-section image stitching.It meets the requirement of tunnel lining crack identification.