Construction and application method of local descriptor for segmenting leakage in tunnel pipe

Abstract

The invention discloses a construction and application method of a local descriptor for segmenting leakage in a tunnel pipe, belongs to the technical field of intelligent monitoring of subway tunnels, and solves the technical problems of single strength characteristic and insufficient segmentation precision in a leakage segmentation method in the prior art. According to the technical scheme, the method comprises the following steps: S1, scanning a subway tunnel by adopting a mobile laser scanning system to obtain high-precision point cloud data; s2, preprocessing the tunnel point cloud, extracting a lining point cloud, and segmenting the lining point cloud into single segments; s3, constructing a neighborhood for points in a single segment and generating a local strength descriptor; and S4, the local strength descriptor and the CatBoost classifier are combined, and leakage in the tunnel pipe is segmented. The method has the beneficial effects that the local strength distribution of the leakage points and the non-leakage points can be accurately represented by effectively expanding the dimensions of the strength characteristics, the leakage segmentation precision is remarkably improved, and reliable technical support is provided for intelligent monitoring and safety maintenance of the subway tunnel.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent monitoring of subway tunnels, and particularly to a method for constructing and applying a local descriptor for segmenting internal leakage of tunnel pipes. Background Art

[0002] Leakage, as a typical disease of subway tunnels, seriously affects the durability of tunnels and the operation safety of subways. In recent years, due to its high precision and high efficiency, the mobile laser scanning system has been widely used in tunnel structure and disease monitoring. The laser reflection intensity can reflect the physical characteristics of the tunnel surface and provides a key basis for leakage segmentation in three-dimensional space. However, traditional segmentation methods only rely on the reflection intensity characteristics of a single dimension and are difficult to characterize the local characteristics of leakage points and non-leakage points, thus affecting the segmentation accuracy. Summary of the Invention

[0003] The purpose of the present invention is to overcome the problems in the background art and provide a method for constructing and applying a local descriptor for segmenting internal leakage of tunnel pipes. By effectively expanding the dimension of intensity features, it can accurately characterize the local intensity distribution of leakage points and non-leakage points, significantly improve the leakage segmentation accuracy, and provide a reliable technical support for the intelligent monitoring and safety maintenance of subway tunnels.

[0004] To achieve the above invention purpose, the technical solution adopted by the present invention is specifically as follows: A method for constructing and applying a local descriptor for segmenting internal leakage of tunnel pipes, including the following steps:

[0005] S1. Use a mobile laser scanning system to scan the subway tunnel to obtain high-precision point cloud data;

[0006] S2. Preprocess the tunnel point cloud, extract the lining point cloud and segment it into individual segments;

[0007] S3. Construct a neighborhood for the points within an individual segment and generate a local intensity descriptor;

[0008] S4. Combine the local intensity descriptor and the CatBoost classifier to segment the internal leakage of the tunnel pipe.

[0009] Further, in step S2, the preprocessing of the tunnel point cloud, extracting the lining point cloud and segmenting it into individual segments includes the following steps:

[0010] S2.1. The subway tunnel point cloud is P(x i , y i , z i ). Set the slice thickness thickness and divide the tunnel point cloud into multiple slices according to the following formula: i＝1,2,…,n

[0011] ​

[0012] Among them

[0013] nslice is the number of slices,

[0014] ceil(·) represents rounding up,

[0015] y max and y min are the maximum and minimum values of y i respectively, and the coordinates of the slice points are k = 1, 2, …, nslice;

[0016] S2.2. Project the slice points onto the XOZ plane to obtain the section points Perform ellipse fitting on the section points to obtain the standard equation of the ellipse as shown in the following formula:

[0017]

[0018] Among them,

[0019] (x center , z center ) are the coordinates of the center of the circle,

[0020] a is the major semi - axis,

[0021] b is the minor semi - axis;

[0022] S2.3. Calculate the distance from the section points to the fitted ellipse Set the distance thresholds δ1 and δ2. If the distance from a certain point to the fitted ellipse satisfies δ1 < d < δ2, then this point is a tunnel lining point; otherwise, it is a non - lining point. The obtained lining point cloud is P'(x j , y j , z j ) j＝1,2,…,m ;

[0023] S2.4. According to the designed width of the segment, divide the lining point cloud into nSegment segments according to the following formula:

[0024]

[0025] Among them,

[0026] nSegment is the number of segments,

[0027] ceil(·) represents rounding up,

[0028] and are the maximum and minimum values of y j respectively,

[0029] The width is the designed width of the segment.

[0030] Further, in step S3, constructing a neighborhood for points within a single segment and generating local intensity descriptors includes the following steps:

[0031] S3.1, project a single segment onto the XOZ plane to obtain cross-section points (x j , z j ), perform ellipse fitting on the cross-section points to obtain the center (x o ', z o '), translate the cross-section so that the center is moved to the origin. The translation process is shown in the following formula:

[0032]

[0033] Where,

[0034] (x j ', z j ) represents the translated cross-section points;

[0035] S3.2, in a single segment, calculate the central angle θ of the query point p o (x o , y o , z o ). The central angle calculation formula is as follows:

[0036]

[0037] Where,

[0038] Q1 represents the first quadrant,

[0039] Q2 represents the second quadrant,

[0040] Q3 represents the third quadrant,

[0041] Q4 represents the fourth quadrant,

[0042] arctan() is the arctangent function;

[0043] S3.3, for the current query point, set the step sizes Δy o and Δθ, and construct a neighborhood using the mileage range [y o - Δy o , y o + Δy o and the angle range [θ - Δθ, θ + Δθ]. If the mileage value y1 and the central angle θ1 of a certain point respectively satisfy y o - Δy o ≤ y1 ≤ y o + Δy oIf the angle range is θ - Δθ ≤ θ1 ≤ θ + Δθ, then this point is a neighborhood point of the query point; when Δy o is small and Δθ is large, the neighborhood is vertically distributed; when Δy o is large and Δθ is small, the neighborhood is horizontally distributed; by adjusting the two step parameters Δy o and Δθ, construct a vertically distributed neighborhood N1 and a horizontally distributed neighborhood N2, and extract the corresponding two neighborhood point sets N1(p o ) and N2(p o ); calculate the mean I mean and standard deviation I std of the reflection intensity of the neighborhood point set. If the reflection intensity I o of the current query point satisfies I o <I mean - 3I std , then this point is preliminarily determined as a leakage point; otherwise, this point is preliminarily determined as a non - leakage point; calculate the proportion of the number of leakage points prob1 and prob2 in the two neighborhood point sets respectively, and determine the neighborhood point set of the current query point according to the following formula:

[0044]

[0045] S3.4. After determining the neighborhood point set, normalize the reflection intensity of the neighborhood point set according to [I i - min(I i )] / [max(I i ) - min(I i )]. The range of the normalized reflection intensity value is [0, 1]. With ω as the intensity interval width, divide [0, 1] into K intensity intervals, and count the number of points n K falling into each intensity interval in the neighborhood point set. Calculate the probability density value p K of each intensity interval according to the following formula to obtain the local intensity descriptor;

[0046]

[0047] where N is the total number of points in the neighborhood point set.

[0048] Furthermore, in step S4, the segmentation of the leakage in the tunnel pipe by combining the local intensity descriptor and the CatBoost classifier includes the following steps:

[0049] S4.1, construct num decision trees, set the maximum number of iterations iterations, learning rate η, decision tree depth depth, loss function loss_function parameters, and initialize the CatBoost model;

[0050] S4.2. Use the local descriptors as training features, fit the features using the constructed decision tree, and calculate the value of the loss function according to the following formula:

[0051]

[0052] where,

[0053] n is the number of samples,

[0054] x i is the feature of the i-th sample,

[0055] y i is the true label of the i-th sample,

[0056] f(x i ) is the predicted label of the model,

[0057] l(y i ,f(x i )) represents the loss of the i-th sample;

[0058] S4.3. Calculate the gradient of the prediction function f (t) (x i ) in each iteration t according to the following formula:

[0059]

[0060] where,

[0061] g i (t) is the gradient of the i-th sample in the t-th iteration, representing the prediction error of the sample,

[0062] represents taking the partial derivative of the function;

[0063] S4.4. Based on the gradient calculated in the t-th iteration, refit the training features to obtain a new predicted label h t (x i ). Minimize the loss function step by step according to the following formula to optimize the model:

[0064] f (t+1) (x i ) = f (t) (x i ) + η·h t (x i ) (10)

[0065] where,

[0066] η represents the learning rate, which is a hyperparameter used to control the step size of model update in each iteration;

[0067] S4.5. Repeat steps S4.2, S4.3, and S4.4 until the maximum number of iterations is reached. Terminate the training and obtain the optimal model. Subsequently, use this model to perform inference on the test data to obtain the predicted labels for each point within a single segment.

[0068] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0069] 1. The present invention breaks through the limitation of the traditional neighborhood lacking directionality and being inapplicable to the construction of the local neighborhood of leakage points. By constructing a new type of neighborhood, it can effectively fit the distribution direction of leakage, improving the ability to accurately describe the distribution direction of leakage.

[0070] 2. The present invention breaks through the limitation of the single - dimensional intensity feature in the traditional tunnel leakage segmentation method. By constructing a local intensity descriptor, the dimension of the intensity feature is effectively expanded, thereby effectively characterizing the local intensity distributions of leakage points and non - leakage points.

[0071] 3. The present invention realizes the accurate segmentation of leakage within a single segment, providing an important technical guarantee for the intelligent monitoring and safety maintenance of subway tunnels.

[0072] 4. The present invention focuses on the innovative construction and application of local intensity descriptors, breaks through the limitation of the traditional leakage segmentation technology in the dimension of intensity features, can effectively expand the dimension of intensity features, and realizes the high - precision segmentation of leakage within a single segment by accurately characterizing the local characteristics of leakage points and non - leakage points, significantly improving the leakage segmentation accuracy, and providing reliable technical support and guarantee for the intelligent monitoring and safety maintenance of subway tunnels. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] Figure 1 is a schematic flow chart of the present invention.

[0074] Figure 2 is the test data of the present invention.

[0075] Figure 3 is a scatter plot of tunnel point clouds after the implementation of point cloud pre - processing of the present invention.

[0076] Figure 4 is a schematic diagram of the implementation of neighborhood construction of the present invention.

[0077] Figure 5 is a schematic diagram of the implementation of local intensity descriptor extraction of the present invention.

[0078] Figure 6 is a schematic diagram of the implementation of the combination of the global intensity descriptor and the CatBoost classifier to achieve leakage segmentation of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0079] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. Of course, the specific embodiments described here are only used to explain the present invention and are not used to limit the present invention.

[0080] The present invention will be further described below in conjunction with the accompanying drawings.

[0081] Example

[0082] like Figure 1 As shown, this embodiment provides a method for constructing and applying a local descriptor for segmenting leakage in a tunnel pipe, comprising the following steps:

[0083] Step 1: If Figure 2 As shown in the figure, a mobile laser scanning system is used to scan the subway tunnel to obtain a high-precision point cloud P(x i ,y i ,z i ) i＝1,2,…,n , where n is the number of points in the collected point cloud, and the observation value is the three-dimensional coordinate of the surface point inside the tunnel.

[0084] Step 2: Preprocess the tunnel point cloud, extract the lining point cloud and segment it into individual segments, as follows:

[0085] (1) Set the slice thickness to 0.2 m and divide the tunnel point cloud into multiple slices according to the following formula:

[0086]

[0087] Among them, nslice is the number of slices, ceil(·) means rounding up, and y max and min y i The maximum and minimum values ​​of the slice points are k=1,2,…,nslice;

[0088] (2) Project the slice point to the XOZ plane to obtain the cross-section point Perform ellipse fitting on the cross-section points to obtain the standard ellipse equation, as shown in the following formula:

[0089]

[0090] Among them, (x center ,z center ) are the coordinates of the center of the circle, a is the major semi-axis, and b is the minor semi-axis;

[0091] (3) Calculate the distance from the cross-section point to the fitted ellipse Set the distance thresholds δ1 to -0.01 m and δ2 to 0.01 m. If the distance from a point to the fitted ellipse satisfies δ1 < d < δ2, then the point is a tunnel lining point; otherwise, the point is a non-lining point. The obtained lining point cloud is P'(x j , y j , z j ); j＝1,2,…,m ;

[0092] (4) The design width width of the segment is 1.2 m, and the lining point cloud is divided into 20 segments, as shown in Figure 3 . The single-segment segmentation process is shown in the following formula:

[0093]

[0094] where nSegment is the number of segments, ceil(·) represents rounding up, and are the maximum and minimum values of y j respectively, and width is the design width of the segment.

[0095] Step 3: Construct a neighborhood for the points within a single segment and generate a local intensity descriptor, specifically as follows:

[0096] (1) Project a single segment onto the XOZ plane to obtain the section points (x j , z j ). Perform ellipse fitting on the section points to obtain the center of the circle (x o ', z o '). Translate the section so that the center of the circle is moved to the origin. The translation process is shown in the following formula:

[0097]

[0098] where (x j ', z j ') represents the translated section points;

[0099] (2) In a single segment, calculate the central angle θ of the query point p o (x o , y o , z o ). The central angle calculation formula is as follows:

[0100]

[0101] where Q1 represents the first quadrant, Q2 represents the second quadrant, Q3 represents the third quadrant, Q4 represents the fourth quadrant, and arctan() is the arctangent function;

[0102] (3) As shown in Figure 4As shown, for the current query point (white point), set the step sizes Δy o and Δθ. Using the mileage range S = [y o -Δy o , y o +Δy o and the angle range θ s = [θ - Δθ, θ + Δθ], construct the neighborhood: If the mileage value y1 and the central angle θ1 of a certain point respectively satisfy y o -Δy o ≤y1≤y o +Δy o and the angle range θ - Δθ ≤ θ1 ≤ θ + Δθ, then this point is a neighborhood point of the query point; when Δy o is small and Δθ is large, the neighborhood is vertically distributed; when Δy o is large and Δθ is small, the neighborhood is horizontally distributed; by adjusting the two step size parameters Δy o and Δθ, construct the vertically distributed neighborhood N1 and the horizontally distributed neighborhood N2, and extract the corresponding two neighborhood point sets N1(p o ) and N2(p o ); calculate the mean I mean and the standard deviation I std of the reflection intensities of the neighborhood point sets. If the reflection intensity I o of the current query point satisfies I o <I mean - 3I std , then this point is preliminarily determined to be a leakage point, otherwise, this point is preliminarily determined to be a non - leakage point; calculate the proportion prob1 and prob2 of the number of leakage points in the two neighborhood point sets respectively, and determine the neighborhood point set of the current query point according to the following formula:

[0103]

[0104] (4) After determining the neighborhood point set, normalize the reflection intensities of the neighborhood point set according to [I i - min(I i )] / [max(I i ) - min(I i )]. The range of the normalized reflection intensity value is [0, 1]. Set the intensity interval width ω to 0.05, then [0, 1] is divided into 20 intensity intervals, and count the number of points n K falling into each intensity interval in the neighborhood point set. Calculate the probability density value p K of each intensity interval according to the following formula, and obtain a 20 - dimensional local intensity descriptor, as Figure 5 shown,

[0105]

[0106] Among them, N is the total number of points in the neighborhood point set.

[0107] Step Four: As Figure 6 shown, combine the local intensity descriptor and the CatBoost classifier to achieve leakage segmentation inside the tunnel pipe, specifically as follows:

[0108] (1) Construct 1000 decision trees, set the maximum number of iterations iterations to 1000, the learning rate η to 0.01, the decision tree depth depth to 10, the loss function loss_function to the Logloss function, and initialize the CatBoost model;

[0109] (2) Use the local descriptor as the training feature, fit the feature with the constructed decision tree, and calculate the value of the loss function according to the following formula:

[0110]

[0111] where n is the number of samples, x i is the feature of the i-th sample, y i is the true label of the i-th sample, f(x i ) is the predicted label of the model, and l(y i , f(x i ) represents the loss of the i-th sample;

[0112] (3) Calculate the gradient of the prediction function f (t) (x i ) in each round of iteration t according to the following formula:

[0113]

[0114] where,

[0115] g i (t) is the gradient of the i-th sample in the t-th round of iteration, representing the prediction error of the sample;

[0116] represents taking the partial derivative of the function;

[0117] (4) Based on the gradient calculated in the t-th round of iteration, refit the training feature to obtain a new predicted label h t (x i ), and gradually minimize the loss function according to the following formula to optimize the model:

[0118] f (t+1) (x i ) = f (t) (x i ) + η · h t (xi ) (10)

[0119] Among them,

[0120] η represents the learning rate, which is a hyperparameter used to control the step size of model update in each iteration;

[0121] (5) Repeat steps (2), (3), and (4) in step four until the maximum number of iterations is reached, terminate the training and obtain the optimal model; subsequently, use this model to infer the test data to obtain the predicted labels of each point within a single segment, such as Figure 6 shown, according to the prediction results, extract the points marked as leakage.

[0122] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for constructing and applying a local descriptor for segmenting leakage in a tunnel pipe, characterized in that: The following steps are involved: S1, uses a mobile laser scanning system to scan subway tunnels and obtain high-precision point cloud data; S2, preprocessing the tunnel point cloud, extracting the lining point cloud and segmenting it into individual segments; S3, constructs a neighborhood for points within a single segment and generates a local intensity descriptor; S4, combining local intensity descriptor and CatBoost classifier to segment tunnel leakage.

2. The method for constructing and applying a local descriptor for segmenting leakage in a tunnel pipe according to claim 1 is characterized in that: In step S2, the preprocessing of the tunnel point cloud, extracting the lining point cloud and segmenting it into individual segments comprises the following steps: S2.1, subway tunnel point cloud is P(x i ,y i ,z i ) i＝1,2,…,n , set the slice thickness, and divide the tunnel point cloud into multiple slices according to the following formula: in, nslice is the number of slices, ceil(·) means rounding up. y max and min y i The maximum and minimum values ​​of the slice points are k=1,2,…,nslice; S2.2, project the slice point to the XOZ plane to obtain the cross-section point Perform ellipse fitting on the cross-section points to obtain the standard ellipse equation, as shown in the following formula: in, (x center ,z center ) are the coordinates of the circle center, a is the major semi-axis, b is the minor semi-axis; S2.3, calculate the distance from the cross-section point to the fitted ellipse Set the distance thresholds δ1 and δ2. If the distance from a point to the fitted ellipse satisfies δ1<d<δ2, then the point is a tunnel lining point. Otherwise, the point is a non-lining point. The obtained lining point cloud is P'(x j ,y j ,z j ) j＝1,2,…,m ; S2.4, according to the designed width of the segment, divide the lining point cloud into nSegment segments according to the following formula: in, nSegment is the number of segments, ceil(·) means rounding up. y jmax and jmin y j The maximum and minimum values ​​of width is the design width of the segment.

3. The method for constructing and applying a local descriptor for segmenting leakage in a tunnel pipe according to claim 1 is characterized in that: In step S3, constructing a neighborhood for points in a single segment and generating a local intensity descriptor comprises the following steps: S3.1, project a single segment onto the XOZ plane and obtain the cross-sectional point (x j ,z j ), perform ellipse fitting on the cross-section points and obtain the center of the circle (x o ',z o ') Translate the cross section and move the center of the circle to the origin. The translation process is shown in the following formula: in, (x j ',z j ') represents the cross-section point after translation; S3.2, in a single segment, calculate the query point p o (x o ,y o ,z o ) is the center angle θ, and the center angle calculation formula is as follows: in, Q1 represents the first quadrant, Q2 represents the second quadrant, Q3 represents the third quadrant, Q4 means the fourth quadrant; arctan() is the inverse tangent function; S3.3, for the current query point, set the step size Δy o and Δθ, using the mileage range [y o -Δy o ,y o +Δy o ] and the angle range [θ-Δθ,θ+Δθ] to construct a neighborhood. If the mileage value y1 and the center angle θ1 of a point satisfy y o -Δy o ≤y1≤y o +Δy o and the angle range θ-Δθ≤θ1≤θ+Δθ, then the point is the neighborhood point of the query point; when Δy o When Δθ is large, the neighborhood is vertically distributed. o When Δθ is small, the neighborhood is horizontally distributed. By adjusting the two step parameters Δy o and Δθ, construct the vertical distribution neighborhood N1 and the horizontal distribution neighborhood N2, and extract the corresponding two neighborhood point sets N1(p o ) and N2(p o );Calculate the mean reflection intensity of the neighborhood point set I mean and standard deviation I std , if the reflection intensity I of the current query point o Satisfy I o <I mean -3I std , then the point is preliminarily determined to be a leakage point, otherwise, the point is preliminarily determined to be a non-leakage point; the proportion of leakage points in the two neighborhood point sets prob1 and prob2 are calculated respectively to determine the neighborhood point set of the current query point according to the following formula: S3.4, after determining the neighborhood point set, press [I i -min(I i )] / [max(I i )-min(I i )] The reflection intensity of the neighborhood point set is normalized. The normalized reflection intensity value range is [0,1]. Taking ω as the intensity interval width, [0,1] is divided into K intensity intervals. The number of points n in the neighborhood point set that fall into each intensity interval is counted. K The probability density value p of each intensity interval is calculated as follows: K , get the local intensity descriptor; in, N is the total number of points in the neighborhood point set.

4. The method for constructing and applying a local descriptor for segmenting leakage in a tunnel pipe according to claim 1, characterized in that: In step S4, the method of combining the local intensity descriptor and the CatBoost classifier to segment the leakage in the tunnel comprises the following steps: S4.1, build num decision trees, set the maximum number of iterations, learning rate η, decision tree depth depth, loss function loss_function parameters, and initialize the CatBoost model; S4.2, using the local descriptor as the training feature, using the constructed decision tree fitting feature, the value of the loss function is calculated as follows: in, n is the number of samples, x i is the feature of the i-th sample, y i is the true label of the i-th sample, f(x i ) is the model’s predicted label, l(y i ,f(x i )) represents the loss of the i-th sample; S4.3, calculate the prediction function f in each iteration t as follows (t) (x i )’s gradient: in, g i (t) is the gradient of the i-th sample in the t-th iteration, indicating the prediction error of the sample, It means to find partial derivative of the function; S4.4, based on the gradient calculated in the tth round of iteration, refit the training features to obtain a new prediction label h t (x i ), and gradually minimize the loss function as follows to optimize the model: f (t+1) (x i )=f (t) (x i )+η·h t (x i ) (10) Among them, η represents the learning rate, which is a hyperparameter used to control the step size of the model update in each iteration; S4.5, repeat steps S4.2, S4.3 and S4.4 until the maximum number of iterations is reached, terminate the training and obtain the optimal model, then use the model to infer the test data to obtain the predicted labels of each point in a single segment.

Images