Automatic Recognition Of Point Cloud Of Water Leakage In Subway Shield Tunnel

In recent years,the construction of urban rail transit in China has been developed rapidly,which has also brought tremendous pressure to the maintenance and repair work.Due to the complex underground geological conditions,water leakage has become a common problem in subway tunnels.Traditional tunnel leakage inspection methods are mainly based on manual inspection,and the accuracy and efficiency are difficult to meet the inspection requirements of the currently busy subway lines.3D laser scanning inspection technology can quickly obtain high-precision and high-density point cloud data of the tunnel,which has become a new trend for tunnel leakage inspection.In this paper,we propose a new method to accomplish the automatic identification of water leakage in the tunnel based on point cloud data,which is collected by the laser scanner mounted on the track inspection trolley.Our new method utilizes the point cloud data automatic processing to achieve the extraction accuracy of up to 92%.(1)We summarize the basic methods of tunnel leakage inspection,and find that 3D laser scanning technology has more advantages in tunnel leakage inspection.In addition,we study basic principles of static scanning and dynamic scanning in laser scanning,It shows that the point cloud data of dynamic scanning has good quality and higher efficiency in tunnel water leakage inspection.(2)Through the research on the distribution regularity of point cloud density,we decide to denoise the point cloud based on statistical filtering and then downsample them.We choose the tunnel point cloud segmentation algorithm based on geometric model fitting,which is determined by the spatial characteristics of the tunnel.The cross section of the tunnel lining point cloud is fitted based on the least square method.We can obtain the elliptic equation of the tunnel cross section from the fitting result.(3)The basic parameters of tunnel cross section ellipse are calculated and used to complete the plane expansion of tunnel point cloud.The tunnel point cloud plane is unrolled along the tunnel vault.We complete the point cloud intensity correction by fitting both the polynomial model of point cloud intensity and laser ranging.The final point cloud data intensity is only related to the reflectivity of the reflector.The planar point cloud are rasterized,then point cloud intensity value is converted to pixel gray value and gray image is generated.(4)In our method,we choose the appropriate threshold to binarize the tunnel gray image,and use median filtering algorithm to remove the noise in the binary image.After the binary image is denoised,the leakage area and normal area are marked on the image respectively.We use the edge inspection to identify the edge of the water leakage,and we find that the Sobel operator is more effective compared to other methods.According to the principle of connected domain,we calculate and count the location,area and number of water leakages in tunnel.Finally we can get the automatic identification results of tunnel leakage.Water leakage information is counted and analyzed per 3m as a section.Compared with the actual leakage in the tunnel,the inspection accuracy of our proposed method can reach 92%.Figure 64,table 16,reference 72.