Construction and Application Methods of Global Descriptors for Detecting Tunnel Segment Leakage

Abstract

This invention discloses a method for constructing and applying a global descriptor for detecting leakage in tunnel segments, relating to the field of detection technology for leaking and non-leaking segments in subway shield tunnels. The method includes the following steps: S1, acquiring point cloud data of the subway tunnel using a mobile laser scanning system; S2, locating the circumferential joint based on the spatial relationship between bolt holes and the circumferential joint, thereby extracting individual segments; S3, generating a global strength descriptor for each individual segment; S4, fusing the generated global strength descriptor with a random forest classifier to detect leaking and non-leaking segments. The beneficial effects of this invention are: by constructing a global strength descriptor, the strength distribution of individual segments can be effectively characterized; by utilizing the difference in strength distribution between leaking and non-leaking segments, leaking and non-leaking segments can be accurately located, providing important technical support and assurance for the safe operation and maintenance of subway tunnels.

Description

Technical Field

[0001] This invention relates to the field of detection technology for leaking and non-leaking tunnel segments in subway shield tunnels, and particularly to the construction and application method of a global descriptor for detecting leakage in tunnel segments. Background Technology

[0002] As a crucial component of urban subway transportation, tunnel defect detection and maintenance are vital for ensuring subway safety. During long-term operation, tunnels frequently experience leakage due to factors such as construction errors, changes in geological conditions, and environmental influences. This not only threatens the durability of the tunnel structure but can also pose significant hidden dangers to subway operation safety. With technological advancements, lidar, capable of providing high-precision three-dimensional point cloud data, has been widely applied to tunnel leakage detection. However, existing leakage detection methods rely on single-dimensional reflection intensity characteristics, making it difficult to accurately locate the leaking segment. Summary of the Invention

[0003] The purpose of this invention is to overcome the problems in the prior art and provide a method for constructing and applying a global descriptor for detecting leakage in tunnel segments. Its core is to characterize the intensity distribution of a single segment and to detect leaking and non-leaking segments by utilizing the differences in intensity distribution among different segments.

[0004] To achieve the aforementioned objectives, the present invention employs the following technical solution: a method for constructing and applying a global descriptor for detecting leakage in tunnel segments, comprising the following steps:

[0005] S1, The mobile laser scanning system scans the subway tunnel to acquire point cloud data of the subway tunnel;

[0006] S2, locate the circumferential seam based on the spatial relationship between the bolt hole and the circumferential seam, and then extract individual segments;

[0007] S3 generates a global intensity descriptor for a single tunnel segment;

[0008] S4, which integrates global strength descriptors and random forest classifiers, detects leaking and non-leaking pipe segments.

[0009] Furthermore, in step S2, the step of locating the circumferential seam based on the spatial relationship between the bolt hole and the circumferential seam, and then extracting a single segment, includes the following steps:

[0010] S2.1, the collected tunnel point cloud is P(x i ,y i ,z i ) i＝1,2,…,n Projecting the point cloud onto the XOZ plane yields a two-dimensional point set (x i ,z i ) i＝1,2,…nPerform ellipse fitting on the two-dimensional point set to obtain the center coordinates (x0, z0); calculate the distance from each point in the two-dimensional point set to the center of the circle, set a distance threshold δ, and if the distance from a point to the center of the circle is greater than the distance threshold, then the point is identified as a bolt hole point and extracted.

[0011] S2.2, the DBSCAN algorithm is used to cluster the mileage of bolt holes;

[0012] S2.3, after clustering the bolt holes using DBSCAN, each segment will form two clusters, representing two groups of bolt holes with similar mileage values. Then, the average y-coordinate values ​​of corresponding clusters in adjacent segments are taken to determine the location of the circumferential joint (loc). v ;

[0013] S2.4, based on the circumferential seam position loc v Extract a single segment using the following formula:

[0014]

[0015] Where v is the index of the circumferential seam.

[0016] Further, in step S2.2, the clustering of bolt hole mileage using the DBSCAN algorithm includes the following steps:

[0017] S2.21, set two parameters: neighborhood radius ε and density threshold MinPts. Randomly select an unvisited point and calculate the number of its neighborhood points within a given radius Eps. If the number of neighborhood points exceeds the preset density threshold minPts, mark the point as a core point; otherwise, temporarily mark it as an edge point. At the same time, treat points outside the neighborhood as noise points.

[0018] S2.22 For a core point, traverse all points in its neighborhood and group these points into the same cluster. If the cluster contains other core points, then include these core points and their neighborhood points into the current cluster.

[0019] S2.23, Repeat steps S2.21 and S2.22 until all points have been visited.

[0020] Further, in step S2.2, the mileage of the bolt is the y-coordinate value.

[0021] Further, in step S3, generating a global intensity descriptor for a single segment includes the following steps:

[0022] S3.1, normalize the reflection intensity of a single segment using the following formula:

[0023]

[0024] in,

[0025] I i The reflection intensity value at each point,

[0026] I i 'Normalized reflection intensity value for each point,

[0027] N is the total number of points in a single segment;

[0028] S3.2, the normalized reflection intensity value range is [0,1], and the intensity interval range is [b k ,b k+1 ), k=1,2,...K, based on the strength interval width b k+1 -b k Divide [0,1] into K intensity intervals;

[0029] S3.3, calculate the number n of points falling into each intensity interval using an indicator function. K As shown in the following formula:

[0030]

[0031] in,

[0032] δ(I i '∈[b k ,b k+1 )) represents the indicator function, indicating when a point falls within the intensity interval [b k ,b k+1 When the condition is met, the function takes the value 1; otherwise, the function takes the value 0.

[0033] S3.4, Calculate the intensity probability density value p for each intensity interval. K Each intensity probability density value is combined into a matrix [p1, p2, ..., p K The K-dimensional global intensity descriptor for a single tunnel segment is obtained, and the intensity probability density is calculated as follows:

[0034]

[0035] in,

[0036] ΔI represents the length of the intensity interval, ΔI = b k+1 -b k .

[0037] Further, in step S4, the fusion of the global strength descriptor and the random forest classifier to detect leaking and non-leaking pipe segments includes the following steps:

[0038] S4.1, Set the random forest parameters: number of decision trees n_estimators, criterion for measuring the quality of feature selection, and maximum depth of decision trees max_depth;

[0039] S4.2, construct n_estimators decision trees, divide the data into training set and test set, and generate M subsets from the training set by sampling with replacement. Each subset is used for training a single decision tree. The subset has M1 samples and contains duplicate samples.

[0040] S4.3, take the global descriptor as the input feature and select k sub-features from the input feature;

[0041] S4.4 For each decision tree, by maximizing the criterion, select the optimal splitting feature and splitting node, and recursively split the nodes of the decision tree. If the maximum depth of the tree, max_depth, is reached, then the splitting stops.

[0042] S4.5 uses the trained decision trees to classify the test set and outputs the prediction results according to the majority voting mechanism, as shown in the following formula:

[0043]

[0044] in,

[0045] argmax is the maximum value function.

[0046] This is the final classification result for the tunnel segments.

[0047] T m (x) represents the classification result of the m-th tree.

[0048] c represents the segment label; a leaking segment label is 0, and a non-leaking segment label is 1.

[0049] Beneficial effects: Compared with the prior art, the significant advantages of this invention are:

[0050] 1. This invention breaks through the reliance of traditional tunnel leakage detection methods on single-dimensional strength features. By constructing a global strength descriptor, it effectively expands the dimensions of strength features and achieves accurate characterization of the strength distribution of tunnel segments.

[0051] 2. This invention uses the intensity distribution of a single pipe segment as the input data for the classifier, which greatly reduces the computational complexity and significantly improves the detection efficiency.

[0052] 3. This invention provides key technical support for the efficient detection and accurate positioning of leakage segments in subway tunnels, laying an important foundation for ensuring the safe operation of subway tunnels.

[0053] 4. This invention constructs a global strength descriptor to effectively characterize the strength distribution of a single tunnel segment. By utilizing the difference in strength distribution between leaking and non-leaking tunnel segments, it accurately locates leaking and non-leaking tunnel segments, providing important technical support and guarantee for the safe operation and maintenance of subway tunnels. Attached Figure Description

[0054] Figure 1 This is a schematic diagram of the process of the present invention.

[0055] Figure 2 This is a scatter plot of point cloud data of a subway tunnel obtained for the implementation of this invention.

[0056] Figure 3 This is a scatter plot of tunnel point cloud extracted for the implementation of a single tunnel segment in this invention.

[0057] Figure 4 This is a schematic diagram illustrating the generation of the global intensity descriptor in this invention.

[0058] Figure 5 This is a schematic diagram illustrating the fusion of the global strength descriptor and the random forest classifier in this invention for detecting leaking and non-leaking pipe segments. Detailed Implementation

[0059] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. Of course, the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0060] The invention will now be further described with reference to the accompanying drawings.

[0061] like Figure 1 As shown, this embodiment provides a method for constructing and applying a global descriptor for detecting leakage in tunnel segments, including the following steps:

[0062] Step 1: The mobile laser scanning system scans the subway tunnel to obtain the subway tunnel point cloud P(x i ,y i ,z i ) i＝1,2,…,n ,like Figure 2 As shown.

[0063] Step Two: Locate the circumferential seam based on the spatial relationship between the bolt holes and the circumferential seam, and then extract individual segments, such as... Figure 3 As shown, the details are as follows:

[0064] (1) The collected tunnel point cloud is P(x i ,y i ,z i ) i＝1,2,…,nProjecting the point cloud onto the XOZ plane yields a two-dimensional point set (x i ,z i ) i＝1,2,…n Perform ellipse fitting on the two-dimensional point set to obtain the center coordinates (x0, z0); calculate the distance from each point in the two-dimensional point set to the center of the circle, set the distance threshold δ to 2.752m, and if the distance from a point to the center of the circle is greater than the distance threshold, then the point is identified as a bolt hole point and extracted.

[0065] (2) The DBSCAN algorithm is used to cluster the bolt holes by mileage, where the mileage of the bolt is the y-coordinate value. The algorithm steps are as follows:

[0066] ① Set the neighborhood radius ε to 0.15m and the density threshold MinPts to 10. Randomly select an unvisited point and calculate the number of its neighborhood points within the given radius Eps. If the number of neighborhood points exceeds the preset density threshold minPts, mark the point as a core point; otherwise, temporarily mark it as an edge point. At the same time, treat points outside the neighborhood as noise points.

[0067] ② For a core point, traverse all points in its neighborhood and group these points into the same cluster. If the cluster contains other core points, then include these core points and their neighborhood points into the current cluster.

[0068] ③ Repeat steps ① and ② until all points have been visited;

[0069] (3) After clustering the bolt holes using DBSCAN, each segment will form two clusters, representing two groups of bolt holes with similar mileage values. Then, the average y-coordinate values ​​of the corresponding clusters in adjacent segments are taken to determine the location of the circumferential joint. v ;

[0070] (4) Based on the circumferential seam position loc v Extract a single segment using the following formula:

[0071]

[0072] Where v is the index of the circumferential seam.

[0073] Step 3: As Figure 4 As shown, a global intensity descriptor is generated for a single tunnel segment, as detailed below:

[0074] (1) Normalize the reflection intensity of a single tube segment using the following formula:

[0075]

[0076] Among them, I i The reflection intensity value at each point, I i'Normalized reflection intensity value for each point, where N is the total number of points in a single segment;

[0077] (2) The normalized reflection intensity values ​​range from [0,1], and the intensity interval ranges from [b...]. k ,b k+1 ), k=1,2,...K, set the intensity interval width to 0.05, and divide [0,1] into 20 intensity intervals;

[0078] (3) Calculate the number of points n falling into each intensity range using the indicator function. K As shown in the following formula:

[0079]

[0080] Wherein, δ(I) i '∈[b k ,b k+1 )) represents the indicator function, indicating when a point falls within the intensity interval [b k ,b k+1 When the condition is met, the function takes the value 1; otherwise, the function takes the value 0.

[0081] (4) Calculate the intensity probability density value p for each intensity interval. K Each intensity probability density value is combined into a matrix [p1, p2, ..., p K The 20-dimensional global intensity descriptor for a single segment is obtained, and the intensity probability density is calculated as follows:

[0082]

[0083] Where ΔI represents the length of the intensity interval, ΔI = b k+1 -b k .

[0084] Step 4: Combine the global strength descriptor and random forest classifier to detect leaking and non-leaking pipe segments, as detailed below:

[0085] (1) Set the random forest parameters: the number of decision trees n_estimators is 100, the criterion for measuring the quality of feature selection is information entropy, and the maximum depth of decision trees max_depth is 6;

[0086] (2) Construct 100 decision trees, divide the data into training set and test set, and generate 100 subsets from the training set with replacement. Each subset is used for training a single decision tree. The number of samples in the subset M1 is the total number of points and includes duplicate samples.

[0087] (3) Using the global descriptor as the input feature, randomly select 5 features from the input features;

[0088] (4) For each decision tree, the optimal splitting feature and splitting node are selected by maximizing the criterion, and the decision tree is recursively split. If the maximum depth of the tree is reached, the growth is stopped.

[0089] (5) Figure 5 As shown, the test set is classified using multiple decision trees after growth, and the prediction results are output according to the majority voting mechanism, as shown in the following formula:

[0090]

[0091] Where argmax is the maximum value function. For the final classification result of the tunnel segments, T m (x) represents the classification result of the m-th tree, and c represents the segment label, with 0 for leaking segments and 1 for non-leaking segments.

[0092] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for construction and application of global descriptors for detecting leakage of tunnel segments, characterized in that, Includes the following steps: S1, The mobile laser scanning system scans the subway tunnel to acquire point cloud data of the subway tunnel; S2, locate the circumferential seam based on the spatial relationship between the bolt hole and the circumferential seam, and then extract individual segments; S3 generates a global intensity descriptor for a single tunnel segment; S4, which integrates global strength descriptors and random forest classifiers to detect leaking and non-leaking pipe segments; In step S3, generating a global intensity descriptor for a single segment includes the following steps: S3.1, normalize the reflection intensity of a single segment using the following formula: (2) in, S3.2, the normalized reflection intensity value range is [0,1], and the intensity interval range is... According to the width of the intensity range Divide [0,1] into One intensity range; S3.3, calculate the number of points falling into each intensity range using an indicator function. As shown in the following formula: (3) in, This represents the indicator function, indicating when a point falls within the intensity range. When the condition is met, the function takes the value 1; otherwise, the function takes the value 0. S3.4, Calculate the intensity probability density value for each intensity interval. Each intensity probability density value is combined into a matrix. To obtain a single segment The global intensity descriptor is calculated using the following formula: (4) in, Indicates the length of the intensity interval. .

2. The method for constructing and applying a global descriptor for detecting tunnel segment leakage according to claim 1, characterized in that, In step S2, the process of locating the circumferential seam based on the spatial relationship between the bolt hole and the circumferential seam, and then extracting a single segment, includes the following steps: S2.1, the collected tunnel point cloud is Project the point cloud onto the XOZ plane to obtain a two-dimensional point set. ; Ellipse fitting is performed on a two-dimensional point set to obtain the coordinates of the circle's center. ; Calculate the distance from each point in the 2D point set to the center of the circle, and set a distance threshold. If the distance from a point to the center of the circle is greater than the distance threshold, then the point is identified as a bolt hole point and extracted. S2.2, the DBSCAN algorithm is used to cluster the mileage of bolt holes; S2.3 After clustering the bolt holes using DBSCAN, each segment will form two clusters, representing two groups of bolt holes with similar mileage values. Then, the y-coordinate values ​​of corresponding clusters in adjacent segments are averaged to determine the location of the circumferential joint. ; S2.4, based on the position of the circumferential seam Extract a single segment using the following formula: (1) in, This is the index for the circumferential seam.

3. The method for constructing and applying a global descriptor for detecting tunnel segment leakage according to claim 2, characterized in that, In step S2.2, the clustering of bolt hole mileage using the DBSCAN algorithm includes the following steps: S2.21, Set neighborhood radius and density threshold Two parameters are used to randomly select an unvisited point and calculate its radius. The number of neighboring points within a given area; if the number of neighboring points exceeds a preset density threshold. If the condition is met, mark the point as a core point; otherwise, temporarily mark it as an edge point, and treat points outside the neighborhood as noise points. S2.22 For a core point, traverse all points in its neighborhood and group these points into the same cluster. If the cluster contains other core points, then include these core points and their neighborhood points into the current cluster. S2.23, Repeat steps S2.21 and S2.22 until all points have been visited.

4. The method for constructing and applying a global descriptor for detecting tunnel segment leakage according to claim 2, characterized in that, In step S2.2, the mileage of the bolt is the y-coordinate value.

5. The method for constructing and applying a global descriptor for detecting tunnel segment leakage according to claim 1, characterized in that, In step S4, the fusion of the global strength descriptor and the random forest classifier to detect leaking and non-leaking pipe segments includes the following steps: S4.1, Set Random Forest Parameters: Number of Decision Trees Standards for measuring the quality of feature selection Maximum depth of decision tree ; S4.2, Construction A decision tree is generated by dividing the data into training and test sets, and generating a test set by sampling with replacement from the training set. There are 3 subsets of data, each used for training a single decision tree, with a total number of samples in each subset. And it contains duplicate samples; S4.3, using global descriptors as input features, select from the input features. Individual characteristics; S4.4, for each decision tree, by maximizing Select the optimal splitting feature and splitting node, and recursively split the nodes of the decision tree until the maximum depth of the tree is reached. Then the splitting will stop; S4.5 uses the trained decision trees to classify the test set and outputs the prediction results according to the majority voting mechanism, as shown in the following formula: (5) in, It is a function with maximum value. This is the final classification result for the tunnel segments. The classification result for the m-th tree. The label is for the segment; 0 indicates a leaking segment and 1 indicates a non-leaking segment.

Images