146 / 2023-08-24 10:31:48

Research on Tunnel Water Leakage Detection Method Based on Mobile Laser Scanning

Mine tunnel,shield tunnel,Water Leakage,tunnle position,target detection

Subway is the main form of urban transportation construction, and currently China's tunnel operating mileage is at a global leading position. Tunnels are susceptible to diseases such as water leakage, cracks, and lining peeling due to geological conditions, stress conditions, and changes in groundwater. Water leakage, as the most common apparent disease in subway tunnels, can cause damage to infrastructure such as steel rails and lines inside the tunnel. Water leakage can affect the service life of the tunnel and threaten train operation and passenger safety. Therefore, it is necessary to accurately locate and detect water leakage diseases quickly. This article uses laser point cloud intensity images as the data source to study precise positioning and rapid detection methods for tunnel leakage water.



A point cloud localization method for tunnels was studied based on the types of tunnel construction and structural characteristics. (1) A tunnel ring positioning method combining bolt hole recognition is proposed for shield tunneling. Firstly, the tunnel wall point cloud is unfolded, and the adaptive parameter adjustment CSF algorithm combined with DBSCAN algorithm is used to extract bolt holes. The Mean Shift algorithm is used to obtain the center point of the bolt hole, and the position relationship of the bolt hole is designed based on the pipe segment to identify the center point of the splicing block. The standard straight line method is used to fit the straight line to locate the joint, and then the shield tunnel point cloud is divided into rings and blocks. (2) For mining method tunnels, a method for locating mining method tunnels based on automatic target recognition was proposed. A mobile tunnel laser detection scheme combining target layout was designed. Combined with the YOLOv7 model, the confidence threshold, target spatial position, and target grayscale pattern were comprehensively predicted to achieve automatic target recognition. Based on the point cloud index relationship, the positioning of mining method tunnels was achieved.



Secondly, an improved YOLOv7 model was proposed to achieve rapid detection of water leakage diseases. Adding ConvNeXt Block to the backbone network to enhance the network's ability to extract leakage water features; Introducing the Convolutional Block Attention Module (CBAM) attention mechanism in the neck network to enhance the weighted feature information of water leakage in both channel and spatial dimensions, capturing local correlations of feature information, making the model more focused on target feature information and improving detection accuracy; Using GIOU-Loss as the positioning loss function to better calculate the distance between the predicted value and the true value, providing accurate data for the backpropagation of the model, and enabling the model to better update parameters; At the same time, the soft-NMS method is used to reduce the number of missed and false detections of water leakage, and improve the overall performance of the model.



The research method was validated based on the measured point cloud data of the subway. The accuracy of extracting the inter ring joint of the shield tunneling method can reach 3.4mm, and the success rate of extracting the target of the mining tunneling method can reach 100%. At the same time, for the leakage water detection algorithm, through comparative experiments and ablation experiments, it has been verified that the proposed model has higher detection accuracy and efficiency.