A track plate fine-tuning method based on point cloud data

Abstract

The present invention provides a track plate fine-tuning method based on point cloud data, including: a laser scanner generates 3D point cloud data and transmits it to a control center; the control center performs denoising through an optimization algorithm; the control center identifies two optical prism control points before and after the track plate to be laid and the measuring equipment through an optimization algorithm, calculates the relative coordinates of the optical prism control points, and obtains a coordinate system conversion matrix in combination with the absolute coordinates of the optical prism control points in the world coordinate system; calculates the required fine-tuning amount of each point according to the design coordinates and actual coordinates of the four plane points on the track plate in the absolute coordinate system; and sends the displacement amount that needs to be fine-tuned to the corresponding intelligent fine-tuning device for fine-tuning. The present invention only requires one three-dimensional laser scanner to obtain real-time data on the track plate fine-tuning amount, reduces the initial investment in equipment and labor, uses an optimization algorithm to calculate the high-precision track plate fine-tuning amount, ensures the accuracy of fine-tuning, and improves the construction efficiency of track plate fine-tuning.

Description

Technical Field

[0001] The invention belongs to the technical field of track plate fine adjustment, and in particular relates to a track plate fine adjustment method based on point cloud data. Background Art

[0002] With the application of prefabricated construction methods in urban rail transit, the proportion of prefabricated track slabs in urban rail transit is increasing. The traditional fine-tuning method of prefabricated track slabs is that construction personnel use total stations and CPⅢ control networks to locate, and then use reflective prisms to determine the positions of the four corners of the track slabs one by one, calculate the distance that needs to be fine-tuned, and complete the fine-tuning of the entire track slab after several measurements and fine-tuning. The entire fine-tuning process consumes a lot of labor costs and time costs, and the accuracy of fine-tuning is greatly affected by the readings of construction personnel. At present, there are also some intelligent fine-tuning equipment on the market. The prior art (CN118065189A) discloses an intelligent prefabricated track slab fine-tuning system and fine-tuning method, but this method requires 4 optical transmitters and 8 receivers, and more equipment needs to be purchased and placed. The fine-tuning process is still cumbersome and time-consuming. In order to solve the above technical problems, the present invention provides a new track slab fine-tuning method based on point cloud data and combined with an optimization algorithm, which can efficiently and quickly complete the real-time data calculation of the track slab fine-tuning amount without setting up too much equipment. Summary of the invention

[0003] In view of the shortcomings in the prior art, the present invention provides a track plate fine-tuning method based on point cloud data, which utilizes the advantages of three-dimensional laser scanning such as non-contact, high data sampling rate and high precision. Through uninterrupted laser scanning, real-time three-dimensional point cloud data in the tunnel is obtained, and then an optimization algorithm is used to calculate to obtain high-precision track plate fine-tuning value, which solves the problems of cumbersome traditional fine-tuning process, long time consumption and low reading accuracy.

[0004] The present invention achieves the above technical objectives through the following technical means.

[0005] A track plate fine-tuning method based on point cloud data is implemented based on a measuring device and a CPⅢ control network positioning system. The measuring device includes a laser scanner, a control center, and a tripod. The laser scanner is used to scan a tunnel 360 degrees to generate 3D point cloud data. The control center is used to receive the 3D point cloud data generated by the laser scanner and complete data processing. The tripod is used to fix the measuring device. The fine-tuning method includes the following processes:

[0006] Step 1: Using the CPⅢ control network positioning system, when the track plate is roughly in place, set up a laser scanner in front of the track plate;

[0007] Step 2: The laser scanner generates 3D point cloud data of the current tunnel measuring device in 360 degrees and transmits it to the control center;

[0008] Step 3: The control center performs preliminary processing and denoising on the 3D point cloud data to obtain high-precision point cloud data P part ;

[0009] Step 4: The control center identifies and calculates the real-time relative coordinates of the four plane points on the track plate and the relative coordinates of the four optical prism control points through the optimization algorithm, and obtains the coordinate system conversion matrix by combining the absolute coordinates of the optical prism control points in the world coordinate system. After converting the coordinate system through the coordinate system conversion matrix, the actual coordinates of the four plane points on the track plate are obtained;

[0010] Step 5: According to the design coordinates and actual coordinates of the four points on the track plate in the absolute coordinate system, calculate the displacement required for fine adjustment of each point;

[0011] Step 6: The control center sends the real-time displacement that needs to be fine-tuned to the intelligent fine-tuning device to complete the fine-tuning operation.

[0012] Furthermore, the control center uses the PCL random sampling consistency algorithm to identify and calculate the real-time relative coordinates of the four points on the track plate. The specific process is as follows:

[0013] According to the plane formula, in the point cloud data P part Three points are randomly sampled, parameters are fitted, and the plane is determined. The PCL random sampling consistency algorithm is used for cyclic calculation to finally obtain multiple plane point cloud data. Among them, the plane with the largest number of point clouds is the required point cloud data P on the surface of the track plate. plane ; Since the size of the track plate is fixed, four straight line functions are constructed according to the geometric relationship of the outer contour of the track plate surface, and then the point cloud data P is calculated. plane The sum of the distances from all points to the four straight lines, where the function of the minimum sum is the outer contour of the track plate surface, and then the real-time relative coordinate information of the four points on the track plate is obtained: c1(x c1 ,y c1 ,z c1 ), c2(x c2 ,y c2 ,z c2 ), c3(x c3 ,y c3 ,z c3 ), c4(x c4 ,y c4 ,z c4 ).

[0014] Furthermore, the control center uses the PCL random sampling consistency algorithm to identify and calculate the relative coordinates of the four optical prism control points. The specific process is as follows:

[0015] Using point cloud data part Remove point cloud data P plane , get the point cloud data P part2 , which includes the optical prism point cloud and other interference point clouds; according to the spherical formula, random sampling, fitting parameters, determining the spherical surface, and cyclic calculation through the PCL random sampling consistency algorithm, finally obtaining 4 spherical point cloud data P sphere ; Since the size of the optical prism spherical surface is known, the four spherical surfaces are calculated separately to obtain the outer contour information of the four spherical surfaces, and then the relative coordinates of the four optical prism control points are obtained: s1(x s1 ,y s1 ,z c1 ), s2(x s2 ,y s2 ,z s2 ), s3(x s3 ,y s3 ,z s3 ), s4(x s4 ,y s4 ,z s4 ).

[0016] Furthermore, the specific process of obtaining the actual coordinates of the four corners of the plane on the track plate is as follows:

[0017] According to the CPⅢ control network positioning system, the absolute coordinates of the four optical prism control points in real space are: r1 (x sr1 ,y sr1 ,z cr1 ), s r2 (x sr2 ,y sr2 ,z sr2 ), s r3 (x sr3 ,y sr3 ,z sr3 ), s r4 (x sr4 ,y sr4 ,z sr4 ), the design coordinates of the four points on the track plate in the absolute coordinate system are: c r1 (x cr1 ,y cr1 ,z cr1 ), c r2 (x cr2 ,y cr2 ,z cr2 ), c r3(x cr3 ,y cr3 ,z cr3 ), c r4 (x cr4 ,y cr4 ,z cr4 ); According to the relative coordinates and absolute coordinates of the optical prism control point, the coordinate system conversion matrix is ​​obtained. After the coordinate system is converted by the coordinate system conversion matrix, the actual coordinates of the four points on the plane of the track plate are obtained as follows: c′1(x′ c1 ,y′ c1 ,z′ c1 ), c′2(x′ c2 ,y′ c2 ,z′ c2 ), c′3(x′ c3 ,y′ c3 ,z′ c3 ), c′4(x′ c4 ,y′ c4 ,z′ c4 ).

[0018] Furthermore, in step 5, among the four plane points on the track plate, the displacements that need to be fine-tuned on the x, y, and z axes for each point are: ck -x crk , y′ ck -y crk , z′ ck -z crk , where k = 1, 2, 3, 4.

[0019] Furthermore, all the point cloud data P obtained by scanning with the laser scanner all Including the ground, track slabs, tunnel walls, optical prisms and noise points, before denoising, the control center first removes the point cloud data with a height less than 200 mm on the z-axis according to the spatial coordinate information in the spatial coordinate system with the tunnel width as the x-axis, the tunnel length as the y-axis and the tunnel height as the z-axis, that is, removes the ground point cloud data and the point cloud data below half the height of the track slab; then cuts the remaining point cloud data into partitions along the x-axis direction, with each partition being 10 mm wide, and then deletes the point cloud data with the largest z-axis value in each partition and all point cloud data within 10 mm below the maximum z-axis value, that is, removes the point cloud data of the tunnel wall.

[0020] Furthermore, the control center processes the noise points in all the point cloud data obtained by the laser scanner by using the bilateral filtering method, as follows:

[0021] p′ i =p i +α*n i

[0022] Among them, p i is any point in the point cloud data before denoising, p′ i is the point after denoising, n i For p i The normal vector of point p i The offset is calculated as follows:

[0023]

[0024] Where j = 1, 2, ... N, representing the N neighboring points around point i, point p j For point p i The neighborhood point set of <n i , p i -p j > is a vector n i and p i -p j The inner product of s (x) is the feature domain weight, W c (x) is the spatial domain weight. The specific calculation formulas of the two weights are as follows:

[0025]

[0026] Among them, δ s For point p i Gaussian weight for the Euclidean distance to the neighbors, δ c For point p i Gaussian weight for the distance to the neighboring tangent plane.

[0027] The present invention has the following beneficial effects:

[0028] The present invention can realize automatic fine adjustment of the track plate. Only one three-dimensional laser scanner is needed, and no receiver needs to be installed separately, so that real-time data of the track plate fine adjustment amount can be obtained, reducing the initial investment in equipment and labor. The present invention also uses an optimization algorithm to calculate the high-precision track plate fine adjustment amount, ensuring the accuracy of the fine adjustment and improving the construction efficiency of the track plate fine adjustment. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 Fine-tune the site top view for the precast track slabs in the tunnel;

[0030] Figure 2 Calculate the flow chart for point cloud data;

[0031] Figure 3 Schematic diagram of point cloud data cutting and partitioning in step 3.

[0032] In the figure: 1-track plate; 2-laser scanner; 3-optical prism; 4-tunnel wall. DETAILED DESCRIPTION

[0033] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments, but the protection scope of the present invention is not limited thereto.

[0034] The track plate fine-tuning method based on point cloud data of the present invention is implemented based on measuring equipment and a CPⅢ control network positioning system used in traditional track laying projects; wherein the measuring equipment includes a laser scanner 2, a control center, and a tripod, the laser scanner 2 is used to scan the tunnel 360 degrees and generate 3D point cloud data; the control center is responsible for receiving the 3D point cloud data generated by the laser scanner 2 and completing data processing; the tripod is responsible for fixing the measuring equipment firmly.

[0035] Reference Figure 1 , 2 The track plate fine-tuning method based on point cloud data of the present invention specifically includes the following process:

[0036] Step 1: Using the CPⅢ control network positioning system in the traditional track laying project, when the track plate 1 is roughly in place, set up the measuring equipment in front of the track plate 1;

[0037] Step 2: The laser scanner 2 on the measuring device starts working, generates 3D point cloud data of the measuring device in the current tunnel in a 360-degree direction, and transmits it to the control center;

[0038] Step 3: The control center uses an optimization algorithm to denoise the 3D point cloud data and generate a high-precision point cloud data. The specific process is as follows:

[0039] All point cloud data P obtained by laser scanner 2 all It mainly includes the ground, the track plate 1, the tunnel wall 4, the optical prism 3 and the noise point; the control center first removes the point cloud data with a height less than 200 mm on the z-axis according to the spatial coordinate information (where the tunnel width direction is the x-axis, the tunnel length direction is the y-axis, and the tunnel height direction is the z-axis), that is, removes the ground point cloud data and the point cloud data below half the height of the track plate 1; Figure 3 As shown, the remaining point cloud data is then divided into partitions along the x-axis direction, with each partition being 10 mm wide, and then the point cloud data with the largest z-axis value in each partition and all point cloud data within 10 mm below the maximum z-axis value are deleted, that is, the point cloud data of the tunnel wall 4 is removed; for the processing of noise points, a high-precision bilateral filtering method is selected, and the calculation formula is as follows:

[0040] p′ i =p i +α*n i

[0041] Among them, p i is any point in the point cloud data before denoising, p′ i is the point after denoising, n i For p i The normal vector of point p i The offset is calculated as follows:

[0042]

[0043] Where j = 1, 2, ... N, representing the N neighboring points around point i, point p j For point p i The neighborhood point set of <n i , p i -p j > is a vector n i and p i -p j The inner product of s (x) is the feature domain weight, W c (x) is the spatial domain weight. The specific calculation formulas of the two weights are as follows:

[0044]

[0045] Among them, δ s For point p i Gaussian weight for the Euclidean distance to the neighbors, δ c For point p i Gaussian weight for the distance to the neighboring tangent plane.

[0046] At this point, the final processed point cloud data P part Only the upper half orbital plate data and optical prism control point data remain.

[0047] Step 4: The control center then uses the optimization algorithm to identify the track plate 1 to be laid and the two CPⅢ optical prism control points before and after the measuring equipment, calculates the relative coordinates of the optical prism control points, and obtains the coordinate system conversion matrix in combination with the absolute coordinates of the optical prism control points in the world coordinate system. The specific process is as follows:

[0048] Step 4.1: Use the PCL random sampling consensus (RANSAC) algorithm to identify the plane data of track plate 1: According to the plane formula ax+by+cz+d=0, in the point cloud data P part Three points are randomly sampled, parameters are fitted, and the plane is determined. The RANSAC algorithm is used for cyclic calculation to finally obtain several plane point cloud data. Among them, the plane with the largest number of point clouds is the required point cloud data P on the upper surface of the track plate 1. plane; Since the size of the track plate 1 is fixed, four straight line functions are constructed according to the geometric relationship of the outer contour of the upper surface of the track plate 1, and then the point cloud data P is calculated. plane The sum of the distances from all points in the track to the four straight lines, where the function of the minimum sum is the outer contour of the upper surface of the track plate 1; after obtaining the outer contour of the track plate 1, the real-time relative coordinate information of the four corners of the plane on the track plate 1 (i.e., the four points on the plane of the track plate 1) can be obtained: c1(x c1 ,y c1 ,z c1 ), c2(x c2 ,y c2 ,z c2 ), c3(x c3 ,y c3 ,z c3 ), c4(x c4 ,y c4 ,z c4 ).

[0049] Step 4.2: Use the RANSAC algorithm to identify the spherical surface data of the optical prism 3: Use the point cloud data P part Remove point cloud data P plane , get the point cloud data P part2 , which includes the optical prism point cloud and other interference point clouds; according to the spherical formula x 2 +y 2 +z 2 +ax+by+cz+d=0, random sampling, fitting parameters, determining the sphere, and calculating through the RANSAC algorithm cycle, finally obtaining 4 spherical point cloud data P sphere Similarly, since the size of the spherical surface of the optical prism 3 is known, the outer contour information of the four spherical surfaces can be obtained by calculating the four spherical surfaces separately, thereby obtaining the relative coordinates of the four optical prism control points: s1(x s1 ,y s1 ,z c1 ), s2(x s2 ,y s2 ,z s2 ), s3(x s3 ,y s3 ,z s3 ), s4(x s4 ,y s4 ,z s4 ).

[0050] Step 4.3: According to the existing CPⅢ control network positioning system, the absolute coordinates of the four optical prism control points in real space are obtained as s r1 (x sr1 ,y sr1 ,z cr1 ), s r2(x sr2 ,y sr2 ,z sr2 ), s r3 (x sr3 ,y sr3 ,z sr3 ), s r4 (x sr4 ,y sr4 ,z sr4 ), the design coordinates of the four points on the plane of the track plate 1 in the absolute coordinate system are c r1 (x cr1 ,y cr1 ,z cr1 ), c r2 (x cr2 ,y cr2 ,z cr2 ), c r3 (x cr3 ,y cr3 ,z cr3 ), c r4 (x cr4 ,y cr4 ,z cr4 ); According to the relative coordinates and absolute coordinates of the optical prism control point, the coordinate system conversion matrix is ​​obtained. After the coordinate system is converted by the coordinate system conversion matrix, the actual coordinates of the four corners of the plane on the track plate 1 are obtained as c′1(x′ c1 ,y′ c1 ,z′ c1 ), c′2(x′ c2 ,y′ c2 ,z′ c2 ), c′3(x′ c3 ,y′ c3 ,z′ c3 ), c′4(x′ c4 ,y′ c4 ,z′ c4 ).

[0051] Step 5: According to the design coordinates and actual coordinates of the four points on the track plate 1 in the absolute coordinate system, calculate the required fine adjustment of each point on the x, y, and z axes, which are x′ and ck -x crk , y′ ck -y crk , z′ ck -z crk , where k = 1, 2, 3, 4.

[0052] Step 6: The control center sends the real-time displacement that needs to be fine-tuned to the corresponding intelligent fine-tuning device, which compares the absolute values ​​of all fine-tuning amounts and starts fine-tuning in order from large to small until all fine-tuning amounts are less than 0.1mm, completing the fine-tuning operation.

[0053] The embodiments are preferred implementations of the present invention, but the present invention is not limited to the above-mentioned implementations. Any obvious improvements, substitutions or modifications that can be made by those skilled in the art without departing from the essential content of the present invention belong to the protection scope of the present invention.

Claims

1. A track plate fine-tuning method based on point cloud data, the fine-tuning method is implemented based on a measuring device and a CPⅢ control network positioning system, the measuring device comprises a laser scanner (2), a control center, and a tripod, the laser scanner (2) is used to scan the tunnel 360 degrees to generate 3D point cloud data, the control center is used to receive the 3D point cloud data generated by the laser scanner (2) and complete data processing; the tripod is used to fix the measuring device, characterized in that, The fine-tuning method includes the following steps: Step 1: Use the CPⅢ control network positioning system until the track plate (1) is roughly placed in place and set up the laser scanner (2); Step 2: The laser scanner (2) scans to generate 3D point cloud data and transmits it to the control center; Step 3: The control center performs preliminary processing and denoising on the 3D point cloud data to obtain high-precision point cloud data P part ; Step 4: The control center identifies and calculates the real-time relative coordinates of the four plane points on the track plate (1) and the relative coordinates of the four optical prism control points through an optimization algorithm, and obtains a coordinate system conversion matrix by combining the absolute coordinates of the optical prism control points in the world coordinate system. After converting the coordinate system through the coordinate system conversion matrix, the actual coordinates of the four plane points on the track plate (1) are obtained; Step 5: Calculate the displacement required for fine adjustment of each point based on the design coordinates and actual coordinates of the four points on the plane of the track plate (1) in the absolute coordinate system; Step 6: The control center sends the real-time displacement that needs to be fine-tuned to the intelligent fine-tuning device to complete the fine-tuning operation; The control center uses a random sampling consistency algorithm to identify and calculate the real-time relative coordinates of four points on the plane of the track plate (1). The specific process is as follows: According to the plane formula, in the point cloud data P part Three points are randomly sampled, parameters are fitted, and the plane is determined. The random sampling consistency algorithm is used for cyclic calculation to finally obtain multiple plane point cloud data. Among them, the plane with the largest number of point clouds is the required point cloud data P of the upper surface of the track plate (1). plane Since the size of the track plate (1) is fixed, four straight line functions are constructed according to the geometric relationship of the outer contour of the upper surface of the track plate (1), and then the point cloud data P is calculated. plane The sum of the distances from all points to the four straight lines, where the function of the minimum sum is the outer contour of the upper surface of the track plate (1), and then the real-time relative coordinates of the four points on the plane of the track plate (1) are obtained as follows: c1(x c1 ,y c1 ,z c1 ), c2(x c2 ,y c2 ,z c2 ), c3(x c3 ,y c3 ,z c3 ), c4(x c4 ,y c4 ,z c4 ).

2. The track plate fine-tuning method based on point cloud data according to claim 1, characterized in that: The control center uses a random sampling consistency algorithm to identify and calculate the relative coordinates of the four optical prism control points. The specific process is as follows: Using point cloud data part Remove point cloud data P plane , get the point cloud data P part2 , which includes the optical prism point cloud and other interference point clouds; according to the spherical formula, random sampling, fitting parameters, determining the spherical surface, and cyclically calculating through the random sampling consistency algorithm, finally obtaining 4 spherical point cloud data P sphere Since the size of the spherical surface of the optical prism (3) is known, the four spherical surfaces are calculated separately to obtain the outer contour information of the four spherical surfaces, and then the relative coordinates of the four control points of the optical prism are obtained.

3. The track plate fine-tuning method based on point cloud data according to claim 2, characterized in that: The specific process of obtaining the actual coordinates of the four points on the plane of the track plate (1) is as follows: According to the CPⅢ control network positioning system, the absolute coordinates of the four optical prism control points in the real space are obtained. The design coordinates of the four plane points on the track plate (1) in the absolute coordinate system are: r1 (x cr1 ,y cr1 ,z cr1 ), c r2 (x cr2 ,y cr2 ,z cr2 ), c r3 (x cr3 ,y cr3 ,z cr3 ), c r4 (x cr4 ,y cr4 ,z cr4 ); The coordinate system conversion matrix is ​​obtained according to the relative coordinates and absolute coordinates of the optical prism control point. After the coordinate system is converted by the coordinate system conversion matrix, the actual coordinates of the four points on the plane of the track plate (1) are obtained as follows: c1′(x c ′1,y c ′1,z c ′1), c2′(x c ′2,y c ′2,z c ′2), c3′(x c ′3,y c ′3,z c ′3), c4′(x c ′4,y c ′4,z c ′4).

4. The track plate fine-tuning method based on point cloud data according to claim 3 is characterized in that: In step 5, among the four points on the plane of the track plate (1), the displacements that need to be fine-tuned on the x, y, and z axes of each point are: c ' k -x crk ,y c ' k -y crk , z c ' k -z crk , where k = 1, 2, 3, 4.

5. The track plate fine-tuning method based on point cloud data according to claim 1, characterized in that: All point cloud data P obtained by scanning by the laser scanner (2) all The invention comprises the ground, the track plate (1), the tunnel wall (4), the optical prism (3) and the noise points. Before the denoising process, the control center first removes the point cloud data with a height less than 200 mm on the z-axis according to the spatial coordinate information on the spatial coordinate system with the tunnel width direction as the x-axis, the tunnel length direction as the y-axis and the tunnel height direction as the z-axis, that is, removes the ground point cloud data and the point cloud data with a height less than half of the track plate (1); then cuts the remaining point cloud data into partitions along the x-axis direction, with each partition being 10 mm wide, and then deletes the point cloud data with the largest z-axis value in each partition and all the point cloud data within a range of 10 mm below the maximum z-axis value, that is, removes the point cloud data of the tunnel wall (4).

6. The track plate fine-tuning method based on point cloud data according to claim 1, characterized in that: The control center processes the noise points in all the point cloud data obtained by scanning with the laser scanner (2) by using a bilateral filtering method, specifically as follows: p′ i =p i +a*n i Among them, p i is any point in the point cloud data before denoising, p i ′ is the point after denoising, n i For p i The normal vector of point p i The offset is calculated as follows: Where j = 1, 2, ... N, representing the N neighboring points around point i, point p j For point p i The neighborhood point set of <n i , p i -p j > is a vector n i and p i -p j The inner product of s (x) is the feature domain weight, W c (x) is the spatial domain weight. The specific calculation formulas of the two weights are as follows: Among them, δ s For point p i Gaussian weight for the Euclidean distance to the neighbors, δ c For point p i Gaussian weight for the distance to the neighboring tangent plane.

Images