Railway tunnel construction clearance detection method based on laser point cloud

Abstract

The application discloses a railway tunnel construction clearance detection method based on a laser point cloud, and relates to the technical field of rail transit operation safety detection. The method comprises the following steps: acquiring three-dimensional laser point cloud data with absolute coordinates of a railway tunnel; extracting a plurality of first sections and a plurality of first section point cloud data based on a preset center line obtained from prior knowledge; segmenting first rail section point cloud data; constructing a standard rail section point cloud model, and adopting a track gauge constraint ICP algorithm to register the standard rail section point cloud model with the first rail section point cloud data; constructing a reference center line, extracting a plurality of second sections and a plurality of second section point cloud data, and determining tunnel boundaries of each second section; reducing the tunnel boundaries of the second sections; and respectively merging the tunnel boundaries of the second sections to obtain railway tunnel construction clearance results. The method has high detection precision and processing efficiency.

Description

Technical Field

[0001] This invention relates to the field of rail transit operation safety inspection technology, specifically to a method for detecting the structural clearance of railway tunnels based on laser point clouds. Background Technology

[0002] Railway construction clearance is a fundamental technical standard that all railway departments must adhere to, ensuring the safe operation of all locomotives and rolling stock on the tracks. The measurement and data management of actual railway construction clearance are prerequisites and foundations for ensuring safe railway operation, and are of paramount importance. Construction clearance refers to the clearance beyond the locomotive and rolling stock clearance, which prohibits buildings and equipment from exceeding during locomotive and rolling stock operation. It includes two types: straight construction clearance and curved construction clearance. Conducting construction clearance inspections on railway tunnels is one of the fundamental tasks for ensuring safe train operation.

[0003] Traditional clearance detection methods mainly rely on manual measurement or contact sensors, which have problems such as low detection efficiency, low measurement accuracy, and low level of intelligent data processing.

[0004] With the development of Mobile Laser Scanning (MLS) technology, non-contact clearance inspection using lidar has become an industry trend. Lidar obtains the distance to a target object through the principle of diffuse laser reflection. It emits laser pulses towards the target object during high-speed rotation; after reflection, the laser light scatters in all directions, and the sensor receives the reflected pulses in real time, enabling high-density, high-precision acquisition of structural cross-sectional information inside tunnels. However, existing technologies still suffer from low detection accuracy and low processing efficiency in practical applications. Summary of the Invention

[0005] The purpose of this invention is to provide a railway tunnel construction clearance detection method based on laser point cloud, which has high detection accuracy and high processing efficiency.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] A method for detecting the clearance of railway tunnels based on laser point clouds, comprising:

[0008] S1. Obtain 3D laser point cloud data with absolute coordinates for the railway tunnel;

[0009] S2. Based on the preset centerline obtained from prior knowledge, the three-dimensional laser point cloud data is initially segmented to extract multiple first cross-sections and multiple first cross-section point cloud data; the first rail cross-section point cloud data is segmented from each of the first cross-section point cloud data.

[0010] S3. Construct a standard rail cross-section point cloud model, and use the gauge-constrained ICP algorithm to register the standard rail cross-section point cloud model with the first rail cross-section point cloud data to reconstruct the second rail cross-section point cloud data.

[0011] S4. Based on the point cloud data of each of the second rail sections, construct the baseline centerline;

[0012] S5. Based on the reference centerline, the three-dimensional laser point cloud data is divided into multiple second sections and multiple second section point cloud data to extract multiple second sections, and an orbital coordinate system is established for each second section to determine the tunnel boundary of each second section.

[0013] S6. Based on the track alignment corresponding to the second cross section, reduce the tunnel boundary of the second cross section;

[0014] S7. For the second cross-section belonging to the same track alignment, the tunnel boundaries of the reduced second cross-section are merged separately to obtain the railway tunnel construction clearance results.

[0015] Optionally, in step S1, acquiring the three-dimensional laser point cloud data of the railway tunnel with absolute coordinates specifically includes:

[0016] A 2D cross-sectional laser scanner was used to perform a moving scan of the railway tunnel, and the scan data was fused and calculated to obtain three-dimensional laser point cloud data in the absolute coordinate system.

[0017] Alternatively, a stand-alone 3D laser scanner can be used to scan the railway tunnel, and the scanned data can be stitched and transformed based on the target data to obtain three-dimensional laser point cloud data in an absolute coordinate system.

[0018] Optionally, in step S3, the gauge-constrained ICP algorithm is used to register the standard rail cross-section point cloud model with the first rail cross-section point cloud data, specifically including:

[0019] The standard rail cross-section point cloud model is registered with the first rail cross-section point cloud data based on the following formula:

[0020] Point set corresponding to the point cloud data of the first rail section: The point set corresponding to the standard rail cross-section point cloud model: A two-dimensional transformation makes: Where R is the rotation matrix and t is the translation matrix.

[0021] In the formula: Track gauge penalty item : Preset regularization coefficient; d: Template gauge; Standard track gauge, value is 1435mm;

[0022] ICP objective function: , .

[0023] Optionally, in step S5, establishing an orbital coordinate system for each of the second cross-sections specifically includes:

[0024] Extract the left and right rail vertices from the point cloud data of the second rail section;

[0025] Using the point on the reference centerline as the origin of the coordinate system, the point cloud data of the second section is rotated so that the vertices of the left and right rails are aligned, and the line connecting the vertices of the left and right rails is used as the X-axis to establish the track coordinate system.

[0026] Optionally, in step S5, determining the tunnel boundaries of each second section specifically includes:

[0027] Statistical filtering or radius filtering is used to remove discrete points from the second cross-section point cloud data;

[0028] Using the Y-axis of the track coordinate system as the vertical detection baseline, a horizontal detection line is laid out along the vertical detection baseline at preset intervals from the track surface elevation until the highest point of the arch.

[0029] For each horizontal detection line, all second section points at the corresponding height of the horizontal detection line are traversed, and the left and right wall position points are determined using the quantile method; each of the left and right wall position points constitutes the tunnel boundary of the second section.

[0030] Optionally, in step S6, the reduction of the tunnel boundary of the second cross-section based on the linear interval corresponding to the second cross-section specifically includes:

[0031] Based on the mileage information of the second cross section and the curve data of the railway tunnel, the linear section corresponding to the second cross section is determined;

[0032] If the linear interval is a circular curve interval, or if the linear interval is the interval from the midpoint of a transition curve to the endpoint of a circular curve, then the reduction is performed based on the following curve reduction formula: , ;in, : Measured limit value inside the curve; : Measured limit value inside the curve; : Limit value after reduction inside the curve; : Limit value after reduction on the outer side of the curve;

[0033] If the linear interval is the interval from 22m outside the straight-to-gradient point to 13m outside the midpoint of the transition curve, then it is reduced by 0.5 times the curve reduction value.

[0034] Optionally, in step S7, the step of merging the tunnel boundaries of the reduced second cross-sections belonging to the same track alignment to obtain the railway tunnel construction clearance results specifically includes:

[0035] The track alignment can be a straight line, a left curve, or a right curve;

[0036] For the second section tunnel boundary belonging to the same track alignment, at any detection height, the minimum left side wall value and the minimum right side wall value of each second section at that detection height are taken as the minimum clearance value; the minimum clearance value of each detection height constitutes the clearance result corresponding to the track alignment.

[0037] Draw a continuous outline of the clearance results corresponding to each track alignment and embed it into the standard building clearance diagram to obtain the railway tunnel building clearance results.

[0038] The present invention has the following beneficial effects: After acquiring three-dimensional laser point cloud data, the present invention first obtains a preset centerline based on prior knowledge. Based on the preset centerline, the three-dimensional laser point cloud data is segmented to obtain preliminary first cross-sectional point cloud data, and then the first rail cross-sectional point cloud data is segmented. Then, a standard rail cross-sectional point cloud model is established. By registering the established standard rail cross-sectional point cloud model with the first rail cross-sectional point cloud data, accurate second rail cross-sectional point cloud data is obtained. Based on the second rail cross-sectional point cloud data, a reference centerline is constructed. This reference centerline is closer to the actual centerline of the railway tunnel than the preset centerline. Using the reference centerline as a benchmark, the three-dimensional laser point cloud data is processed, resulting in more accurate clearance results. When reducing the construction clearance of curved sections of railway tunnels, the track alignment of the second cross-section can be determined based on the mileage information corresponding to the second cross-section and the curve data of the railway tunnel. The reduction calculation formula corresponding to the track alignment can then be called to complete the reduction calculation. Compared to existing technologies that often rely on manually consulting curve tables and performing manual reduction calculations section by section, this method is more efficient and accurate. Attached Figure Description

[0039] Figure 1 This is a flowchart of the railway tunnel construction clearance detection method based on laser point cloud according to the present invention;

[0040] Figure 2 This is a schematic diagram illustrating the segmentation of three-dimensional laser point cloud data and extraction of each point cloud cross-section according to the present invention.

[0041] Figure 3This is a two-dimensional projection of a segment of three-dimensional laser point cloud data onto a first cross-section in this invention;

[0042] Figure 4 This is a schematic diagram of the parameterized standard rail cross-section point cloud model constructed in this invention;

[0043] Figure 5 This is a diagram showing the registration result between the standard rail cross-section point cloud model and the first rail cross-section point cloud data in this invention.

[0044] Figure 6 for Figure 5 Enlarged view of a portion of the image;

[0045] Figure 7 This is a two-dimensional projection of a segment of three-dimensional laser point cloud data onto a second section with an established orbital coordinate system, as described in this invention.

[0046] Figure 8 A schematic diagram of the tunnel boundary for a single second cross-section;

[0047] Figure 9 This is the final architectural clearance drawing for the entire railway tunnel. Detailed Implementation

[0048] Hereinafter, embodiments will be described with reference to the accompanying drawings.

[0049] See Figures 1-9 A method for detecting the clearance of railway tunnels based on laser point clouds, comprising:

[0050] S1. Obtain 3D laser point cloud data with absolute coordinates for the railway tunnel.

[0051] When performing laser scanning on railway tunnels, either a 2D cross-sectional laser scanner or a stationary 3D laser scanner can be used to obtain laser point cloud data. The absolute coordinates here can be geographic coordinates.

[0052] A 2D cross-sectional laser scanner is used: As the vehicle-mounted mobile laser scanning system moves along the railway track, the laser scanner performs a circumferential scan to obtain laser point cloud data of the railway tunnel scene; during the scanning process, it can be paired with an odometer to collect relative point cloud data with mileage information, and the 3D laser point cloud data under geographic coordinates can be obtained by fusing the mileage information and relative point cloud data.

[0053] It can also be used in conjunction with a POS system to collect point cloud data with absolute geographic information. It's important to note that when used with a POS system, the satellite lock-off time should not exceed 2 minutes; otherwise, the POS data will drift, affecting the accuracy of the point cloud data. The POS system, combined with a vehicle-mounted mobile laser scanning system, performs fusion and computation on the scanned data to obtain 3D laser point cloud data in a unified geographic coordinate system across the entire route.

[0054] A station-mounted 3D laser scanner is used: After the station is set up, the 3D laser scanner rotates 360° horizontally while the scanning lens rotates 360° vertically to acquire point cloud data for each station. Target data from different stations is then used to stitch together and convert the point cloud data to a geographic coordinate system. At least four targets should be deployed before and after the 3D laser scanner.

[0055] S2. Based on the preset centerline obtained from prior knowledge, the three-dimensional laser point cloud data is initially segmented to extract multiple first cross-sections and multiple first cross-section point cloud data; the first rail cross-section point cloud data is segmented from each of the first cross-section point cloud data.

[0056] The preset centerline is determined based on prior knowledge. When using a 2D cross-sectional laser scanner for scanning, the prior knowledge refers to the trajectory line of the 2D cross-sectional laser scanner. The 2D cross-sectional laser scanner travels along the railway track in the railway tunnel, and the centerline of the railway track is the trajectory line. The preset centerline can be obtained based on the track centerline data of the railway track during the design or the track centerline data obtained from subsequent measurements.

[0057] When using a gantry-mounted 3D laser scanner, prior knowledge refers to the design centerline when setting up the 3D laser scanner. This design centerline data can be obtained during the setup of the 3D laser scanner.

[0058] S21. The 3D laser point cloud data is segmented into several segments according to the preset centerline. The thickness of each segment is called the section thickness d. When setting the section thickness d, the section thickness can be adaptively calculated according to the preset centerline to keep the elevation difference in the section mileage direction within 2mm, while also ensuring the positive versine distance. Less than 1mm. This is to avoid the effects of slope and planar bending deformation. It can be determined according to the formula. The range of selection for calculating the cross-sectional thickness includes: : sine distance; : String length; Radius. Since the minimum curve radius for railways is 300m, the calculated cross-sectional thickness should be less than or equal to 1.5m. After setting the cross-sectional thickness d, the cross-sectional width w also needs to be set. The cross-sectional width w should be sufficient to include all laser point cloud data within the cross-section. During segmentation, the unit normal direction of the preset centerline at each cross-section cutting mileage is T(A, B, C), and a point on the preset centerline is... Then the equation of the cutting plane at that mileage position is: For ease of description, this plane will be referred to as the first section. See also Figure 2 , Figure 2 Five first cross-sections are shown in the figure.

[0059] S22, for any point in space Calculate the signed distance t along the unit normal direction: If point P satisfies the formula If so, then this point belongs to this first cross section.

[0060] For the first cross-section mentioned above, construct a local Cartesian coordinate system Oxy, where the origin is O= When the Z-axis is vertically upward, take the global upward unit vector. Construct e x Strictly perpendicular to the unit normal T, and located in the horizontal plane; e y It is also perpendicular to the unit normal T and parallel to e. x Orthogonal, oriented upwards but strictly within the first cross-section: , The coordinates of any point P in the cross-sectional point cloud projected onto the cross-section. for: , The set of projected coordinates of point P constitutes the point cloud data for the first cross-section. (See also...) Figure 3 , Figure 3 The image shows a two-dimensional projection of a segment of three-dimensional laser point cloud data onto a first cross-section.

[0061] Extracting the first rail section point cloud data from the first section point cloud data: Projecting the data points on the preset centerline onto the corresponding first section to form a projection point. By setting an elevation or distance threshold, comparing all the first section point clouds on the first section with the projection point, thereby extracting the first rail section point cloud data.

[0062] S3. Construct a standard rail cross-section point cloud model, and use the gauge-constrained ICP algorithm to register the standard rail cross-section point cloud model with the first rail cross-section point cloud data to reconstruct the second rail cross-section point cloud data.

[0063] Based on the rail cross-sectional dimensions specified in railway industry standards, key geometric parameters of the rail cross-section are obtained, including rail head width, rail head height, rail web thickness, rail base width, rail base height, and the radius of each circular arc segment. These parameters collectively define the complete geometric shape of the rail cross-section. The cross-sectional profile is discretized and sampled at a certain sampling density to generate a standard rail cross-section point cloud model. This standard rail cross-section point cloud model serves as a template point cloud for registration and should have sufficient point density to ensure registration accuracy, while maintaining a reasonable number of points to control computational efficiency. (See also...) Figure 4 , Figure 4 The constructed parametric standard rail cross-section point cloud model is shown.

[0064] By using gauge-constrained ICP registration, the optimal transformation matrix (including rotation and translation parameters) is obtained from the standard rail cross-section point cloud model to the first cross-section point cloud data (measured rail point cloud). Applying this transformation matrix to the entire standard rail cross-section point cloud model yields a reconstructed rail model (the second rail cross-section point cloud data) that accurately fits the measured rail point cloud (here, the first cross-section rail point cloud data). Since the standard rail cross-section point cloud model is parametric, the reconstructed rail model (the second rail cross-section point cloud data) also possesses parametric characteristics; all cross-sectional geometric parameters can be directly extracted from the transformed model, achieving a parametric expression of the rail's three-dimensional morphology. The second rail cross-section point cloud data is used to represent the left and right rail cross-sections within the first cross-section.

[0065] S4. Based on the point cloud data of each of the second rail sections, construct the baseline centerline.

[0066] Based on the definition of the reference track, the reference rail is determined from the left and right rail sections represented by the point cloud data of the second rail section. Then, based on the reference rail, the reference rail is horizontally offset by 1 / 2 standard gauge (the standard gauge defined in China is 1435mm) to another rail to obtain the reference midpoint of the first section. The reference midpoints of each first section are mapped back to three-dimensional space. The reference centerline is obtained by fitting the reference midpoints of each first section mapped back to three-dimensional space using the least squares method.

[0067] S5. Based on the reference centerline, the three-dimensional laser point cloud data is segmented a second time to extract multiple second cross sections and multiple second cross section point cloud data, and an orbital coordinate system is established for each second cross section to determine the tunnel boundary of each second cross section.

[0068] The 3D laser point cloud data is segmented a second time using the baseline centerline. The specific process can be referred to in step S2 to extract multiple second sections and multiple second section point cloud data. During the second segmentation, the mileage selected for the second section can be the same as or different from the mileage selected for the first section. This second segmentation is a fine-grained segmentation of the 3D laser point cloud data.

[0069] Extract the left and right rail vertices from the point cloud data of the second rail section;

[0070] Using the point on the reference centerline as the origin, the second cross-sectional point cloud data is rotated so that the vertices of the left and right rails are aligned. A track coordinate system is established using the line connecting the vertices of the left and right rails as the X-axis and a line passing through the origin and perpendicular to the X-axis as the Y-axis. Specifically, the rotation angle is calculated based on the track gauge and superelevation, and the second cross-sectional point cloud data is rotated around the origin until both the left and right rail vertices are horizontal and aligned in height. (See also...) Figure 5 and Figure 7 , Figure 5 This is a projection of the 3D laser point cloud data in the original coordinate system onto the second section. It shows the point cloud data of the second section before rotation. Figure 7 This is a projection of the 3D laser point cloud data onto the second section in the orbital coordinate system. It shows the rotated point cloud data of the second section. Figure 7 The two rail apexes on the left and right are at the same height.

[0071] For a given second cross-section, if it is located on a curved section (where the railway is curved), there is an elevation difference between the left and right rails at that cross-section. When establishing the track coordinates, the line connecting the vertices of the two rails should be used as the X-axis. A Cartesian coordinate system should be established using the straight line perpendicular to the X-axis within the second cross-section as the Y-axis. The initial coordinates of the point cloud of the second cross-section are represented as follows: Rotate the coordinates around the origin by an angle α to construct a rotation matrix, and then obtain the projection equation of the orbital coordinate system: , In the formula: H: superelevation; S: track gauge; u: abscissa in the track coordinate system; v: ordinate in the track coordinate system. This yields the coordinate point set of the second cross-section point cloud data in the track coordinate system.

[0072] Discrete points in the second cross-section point cloud data are removed using statistical filtering or radius filtering. Using the Y-axis of the track coordinate system as the vertical detection baseline, a horizontal detection line is laid out along the vertical detection baseline at preset intervals from the track surface elevation until the highest point of the arch.

[0073] For each horizontal detection line, all second section points at the corresponding height of the horizontal detection line are traversed, and the left and right wall position points are determined using the quantile method; each of the left and right wall position points constitutes the tunnel boundary of the second section.

[0074] The left and right clearance results are detected by setting horizontal detection lines at different heights. This process automatically and accurately finds the true boundary positions of the left and right tunnel walls from the two-dimensional point cloud data of the second section, thereby calculating the left and right boundary values.

[0075] Due to environmental and instrumental influences during scanning, discrete points inevitably exist in the scanned point cloud, such as suspended dust particles and scanning error points. Directly using the original point cloud to find the outermost point will result in significant noise interference. Statistical filtering or radius filtering is necessary to remove obviously isolated discrete points. Furthermore, auxiliary equipment and cables exist on the tunnel walls. The quantile method is used to eliminate the influence of these facilities and obtain the true tunnel boundary. Preferably, the 2% quantile of the u-value within this layer is used as the location point of the left wall, and the 98th quantile as the location point of the right wall. Figure 8 The tunnel boundary of a single second section is shown.

[0076] S6. Based on the linear interval corresponding to the second cross section, reduce the tunnel boundary of the second cross section.

[0077] Based on the mileage information of the second cross section and the curve data of the railway tunnel, the linear section corresponding to the second cross section is determined;

[0078] If the linear interval is a circular curve interval, or if the linear interval is the interval from the midpoint of a transition curve to the endpoint of a circular curve, then the reduction is performed based on the following curve reduction formula: , ;in, : Measured limit value inside the curve; : Measured limit value inside the curve; : Limit value after reduction inside the curve; : Limit value after reduction on the outer side of the curve;

[0079] If the linear interval is the interval from 22m outside the straight-to-gradient point to 13m outside the midpoint of the transition curve, then it is reduced by 0.5 times the curve reduction value.

[0080] Because the point cloud data of the second cross-section is associated with mileage information, railway tunnel lines are equipped with curve data during design and construction. This curve data records the track alignment at each mileage point. Based on the mileage information corresponding to the second cross-section, it can be determined which alignment section the second cross-section is located in. Railways have straight sections and circular curve sections, connected by transition curve sections. The intersection of a straight section and a transition curve section is called the straight-to-transition point (also known as the transition-to-straight point). If the second cross-section is located on a straight section more than 22m outside the straight-to-transition point, no reduction is required. If the second cross-section is located between the midpoint of the transition curve section and the endpoint of the circular curve (the endpoint where the circular curve section intersects the transition curve section), or if the second cross-section is located within the circular curve section, reduction is performed according to the curve reduction formula. If the second cross-section is located between 22m outside the straight-to-transition point and the midpoint of the transition curve section, reduction is performed using 0.5 times the curve reduction value.

[0081] The corresponding calculation formula can be retrieved based on the track alignment, thereby reducing the tunnel boundary of the second section.

[0082] S7. For the second cross-section belonging to the same track alignment, the tunnel boundaries of the reduced second cross-section are merged separately to obtain the railway tunnel construction clearance results.

[0083] Similarly, based on the mileage information of the second section, the track alignment is retrieved from the curve data, wherein the track alignment is a straight line, a left curve, or a right curve;

[0084] The tunnel boundaries of the second cross-sections belonging to the same track alignment are merged. Specifically, the tunnel boundaries of all second cross-sections belonging to straight track alignments are merged, the tunnel boundaries of all second cross-sections belonging to left-curved track alignments are merged, and the tunnel boundaries of all second cross-sections belonging to right-curved track alignments are merged. For a second cross-section of a track alignment, for any detection height (i.e., the height corresponding to each horizontal detection line), the minimum left and right wall values ​​of each second cross-section at that detection height are taken as the minimum clearance values. The minimum clearance values ​​at each detection height constitute the clearance results corresponding to that track alignment.

[0085] When there are multiple values ​​at the same elevation for multiple second sections of the same tunnel, the minimum value is taken. The half-width of the construction clearance at different heights from the rail surface is detected from top to bottom (from the highest point to the rail plane), and described separately for straight sections, left curved sections and right curved sections.

[0086] Draw a continuous outline of the clearance results corresponding to each track alignment and embed it into the standard building clearance diagram to obtain the railway tunnel building clearance results.

[0087] See Figure 9The figure shows the minimum comprehensive clearance profiles for the left curve segment and the straight segment. Since the railway tunnel does not have a right curve segment, therefore... Figure 9 The minimum comprehensive clearance outline of the right curve segment is not displayed. The minimum comprehensive clearance outline of the left curve segment (represented by dashed lines) and the minimum comprehensive clearance outline of the straight line segment (represented by solid lines) are the results of railway tunnel construction clearance. Figure 9 The standard building clearance line is also shown, through Figure 9 As can be seen, the minimum comprehensive clearance outline of both the left curve segment and the straight segment is outside the standard building clearance outline, and there is no encroachment.

[0088] This invention constructs a high-precision benchmark through a dual-drive approach of prior knowledge guidance and rail model registration. Combined with key technologies such as track coordinate system projection, quantile-based anti-interference boundary extraction, and automatic alignment recognition and reduction, it achieves fully automated, high-precision, and robust detection of tunnel structural clearances. Compared to traditional manual or contact-based methods, it significantly improves detection efficiency, accuracy, and intelligence, making it particularly suitable for clearance detection tasks in complex environments (such as tunnels with numerous ancillary facilities and curved sections).

[0089] Optionally, in step S1, acquiring the three-dimensional laser point cloud data of the railway tunnel with absolute coordinates specifically includes:

[0090] A 2D cross-sectional laser scanner was used to perform a moving scan of the railway tunnel, and the scan data was fused and calculated to obtain three-dimensional laser point cloud data in the absolute coordinate system.

[0091] Alternatively, a stand-alone 3D laser scanner can be used to scan the railway tunnel, and the scanned data can be stitched and transformed based on the target data to obtain three-dimensional laser point cloud data in an absolute coordinate system.

[0092] Optionally, in step S3, the gauge-constrained ICP algorithm is used to register the standard rail cross-section point cloud model with the first rail cross-section point cloud data, specifically including:

[0093] The standard rail cross-section point cloud model is registered with the first rail cross-section point cloud data based on the following formula:

[0094] Point set corresponding to the point cloud data of the first rail section: The point set corresponding to the standard rail cross-section point cloud model: A two-dimensional transformation makes: Where R is the rotation matrix and t is the translation matrix. In the formula: Track gauge penalty item : Preset regularization coefficient, preferably 0.5; d: Template gauge, used to represent the gauge obtained during the registration process of the left and right rails; Standard track gauge, value is 1435mm;

[0095] ICP objective function: , .

[0096] This method employs an improved ICP algorithm that integrates railway physical constraints to achieve precise registration of the rail model. Unlike traditional unconstrained ICP, it introduces dual regularization terms of gauge constraints and rail top coplanar constraints, embedding prior knowledge of railway engineering into the registration optimization process, which significantly improves the robustness and accuracy of rail reconstruction in complex tunnel environments.

[0097] Its core principle is to construct a joint optimization objective function based on minimizing the point cloud matching residuals in traditional ICP. The track gauge constraint uses the standard track gauge of 1435mm as a benchmark. A secondary penalty term constrains the deviation of the center distance between the top surfaces of the left and right rails during the registration process, forcing the model registration result to conform to railway track gauge specifications. This effectively suppresses track gauge distortion caused by noise points, enabling the rail model to achieve millimeter-level precise registration even in tunnel scenarios with noise interference, geometric distortion, and curve superelevation. This provides a high-precision benchmark for subsequent line centerline fitting and tunnel construction clearance detection.

[0098] Optionally, in step S5, establishing an orbital coordinate system for each of the second cross-sections specifically includes:

[0099] Extract the left and right rail vertices from the point cloud data of the second rail section;

[0100] Using the point on the reference centerline as the origin of the coordinate system, the point cloud data of the second section is rotated so that the vertices of the left and right rails are aligned, and the line connecting the vertices of the left and right rails is used as the X-axis to establish the track coordinate system.

[0101] Optionally, in step S5, determining the tunnel boundaries of each second section specifically includes:

[0102] Statistical filtering or radius filtering is used to remove discrete points from the second cross-section point cloud data;

[0103] Using the Y-axis of the track coordinate system as the vertical detection baseline, a horizontal detection line is laid out along the vertical detection baseline at preset intervals from the track surface elevation until the highest point of the arch.

[0104] For each horizontal detection line, all second section points at the corresponding height of the horizontal detection line are traversed, and the left and right wall position points are determined using the quantile method; each of the left and right wall position points constitutes the tunnel boundary of the second section.

[0105] Optionally, in step S6, the reduction of the tunnel boundary of the second cross-section based on the linear interval corresponding to the second cross-section specifically includes:

[0106] Based on the mileage information of the second cross section and the curve data of the railway tunnel, the linear section corresponding to the second cross section is determined;

[0107] If the linear interval is a circular curve interval, or if the linear interval is the interval from the midpoint of a transition curve to the endpoint of a circular curve, then the reduction is performed based on the following curve reduction formula: , ;in, : Measured limit value inside the curve; : Measured limit value inside the curve; : Limit value after reduction inside the curve; : Limit value after reduction on the outer side of the curve;

[0108] If the linear interval is the interval from 22m outside the straight-to-gradient point to 13m outside the midpoint of the transition curve, then it is reduced by 0.5 times the curve reduction value.

[0109] Optionally, in step S7, the step of merging the tunnel boundaries of the second cross-sections belonging to the same track alignment to obtain the railway tunnel construction clearance results specifically includes:

[0110] The track alignment can be a straight line, a left curve, or a right curve;

[0111] For the second section tunnel boundary belonging to the same track alignment, at any detection height, the minimum left side wall value and the minimum right side wall value of each second section at that detection height are taken as the minimum clearance value; the minimum clearance value of each detection height constitutes the clearance result corresponding to the track alignment.

[0112] Draw a continuous outline of the clearance results corresponding to each track alignment and embed it into the standard building clearance diagram to obtain the railway tunnel building clearance results.

Claims

1. A method for detecting the structural clearance of railway tunnels based on laser point clouds, characterized in that, include: S1. Obtain 3D laser point cloud data with absolute coordinates for the railway tunnel; S2. Based on the preset centerline obtained from prior knowledge, the three-dimensional laser point cloud data is initially segmented to extract multiple first cross-sections and multiple first cross-section point cloud data; the first rail cross-section point cloud data is segmented from each of the first cross-section point cloud data. S3. Construct a standard rail cross-section point cloud model, and use the gauge-constrained ICP algorithm to register the standard rail cross-section point cloud model with the first rail cross-section point cloud data to reconstruct the second rail cross-section point cloud data. S4. Based on the point cloud data of each of the second rail sections, construct the baseline centerline; S5. Based on the reference centerline, the three-dimensional laser point cloud data is divided into multiple second sections and multiple second section point cloud data to extract multiple second sections, and an orbital coordinate system is established for each second section to determine the tunnel boundary of each second section. S6. Based on the linear interval corresponding to the second section, reduce the tunnel boundary of the second section; S7. For the second section belonging to the same track alignment, merge the tunnel boundaries of the reduced second section separately to obtain the railway tunnel construction clearance results. In step S3, the gauge-constrained ICP algorithm is used to register the standard rail cross-section point cloud model with the first rail cross-section point cloud data, specifically including: The standard rail cross-section point cloud model is registered with the first rail cross-section point cloud data based on the following formula: Point set corresponding to the point cloud data of the first rail section: The point set corresponding to the standard rail cross-section point cloud model: A two-dimensional transformation makes: Where R is the rotation matrix and t is the translation matrix. , In the formula: Track gauge penalty item : Preset regularization coefficient; d: Template gauge; Standard track gauge, value is 1435mm; ICP objective function: , ; In step S5, determining the tunnel boundaries of each second section specifically includes: Statistical filtering or radius filtering is used to remove discrete points from the second cross-section point cloud data; Using the Y-axis of the track coordinate system as the vertical detection baseline, a horizontal detection line is laid out along the vertical detection baseline at preset intervals from the track surface elevation until the highest point of the arch. For each horizontal detection line, all second section points at the corresponding height of the horizontal detection line are traversed, and the left and right wall position points are determined using the quantile method; each of the left and right wall position points constitutes the tunnel boundary of the second section. In step S7, the tunnel boundaries of the reduced second cross-sections belonging to the same track alignment are merged separately to obtain the railway tunnel construction clearance results, specifically including: The track alignment can be a straight line, a left curve, or a right curve; For the second section tunnel boundary belonging to the same track alignment, at any detection height, the minimum left side wall value and the minimum right side wall value of each second section at that detection height are taken as the minimum clearance value; the minimum clearance value of each detection height constitutes the clearance result corresponding to the track alignment. Draw a continuous outline of the clearance results corresponding to each track alignment and embed it into the standard building clearance diagram to obtain the railway tunnel building clearance results.

2. The railway tunnel construction clearance detection method based on laser point cloud as described in claim 1, characterized in that, In step S1, acquiring the three-dimensional laser point cloud data of the railway tunnel with absolute coordinates specifically includes: A 2D cross-sectional laser scanner was used to perform a moving scan of the railway tunnel, and the scan data was fused and calculated to obtain three-dimensional laser point cloud data in the absolute coordinate system. Alternatively, a stand-alone 3D laser scanner can be used to scan the railway tunnel, and the scanned data can be stitched and transformed based on the target data to obtain three-dimensional laser point cloud data in an absolute coordinate system.

3. The railway tunnel construction clearance detection method based on laser point cloud as described in claim 1, characterized in that, In step S5, an orbital coordinate system is established for each of the second cross sections, specifically including: Extract the left and right rail vertices from the point cloud data of the second rail section; Using the point on the reference centerline as the origin of the coordinate system, the point cloud data of the second section is rotated so that the vertices of the left and right rails are aligned, and the line connecting the vertices of the left and right rails is used as the X-axis to establish the track coordinate system.

4. The railway tunnel construction clearance detection method based on laser point cloud as described in claim 3, characterized in that, In step S6, the reduction of the tunnel boundary of the second cross-section based on the linear interval corresponding to the second cross-section specifically includes: Based on the mileage information of the second cross section and the curve data of the railway tunnel, the linear section corresponding to the second cross section is determined; If the linear interval is a circular curve interval, or if the linear interval is the interval from the midpoint of a transition curve to the endpoint of a circular curve, then the reduction is performed based on the following curve reduction formula: , ;in, : Measured limit value inside the curve; Measured limit value outside the curve; : Limit value after reduction inside the curve; : Limit value after reduction on the outer side of the curve; If the linear interval is the interval from 22m outside the straight-to-gradient point to 13m outside the midpoint of the transition curve, then it is reduced by 0.5 times the curve reduction value.

Images