Method and system for detecting distance from car coupler center line to rail surface and medium

Abstract

The invention discloses a method and system for detecting the distance from a car coupler center line to a rail surface and a medium. Relates to the technical field of distance measurement. The method comprises the following steps: improving on the basis of a traditional measurement technology, acquiring a coupler image and a track image at the same point by using a camera of an inspection robot, and respectively performing mask segmentation and projection on the coupler image and the track image to obtain a coupler center line point cloud and a track surface point cloud; after the coupler center line point cloud and the track plane point cloud are converted to a base coordinate system of the inspection robot based on the conversion matrix, a centroid point coordinate of the coupler center line point cloud and a track plane equation are obtained; the distance between the center line of the car coupler and the rail surface can be rapidly and accurately measured, a large amount of human resource investment is saved, and measurement errors caused by different manual business levels can be avoided.

Description

Technical Field

[0001] The present invention relates to the technical field of distance measurement, and in particular to a method, system and medium for detecting the distance from a coupler center line to a rail surface. Background Art

[0002] The height of the centerline of the coupler from the rail surface is the vertical distance between the centerline of the hook connecting the train carriage and the railway track. In railway transportation, the height of the centerline of the coupler from the rail surface is a key parameter, which plays an important role in ensuring the stability and safety of railway trains. The adjustment of the height of the centerline of the coupler from the rail surface directly affects the traction and braking force of the vehicle, and also affects the stability and safety of the vehicle. When the height of the centerline of the coupler from the rail surface is too large or too small, the train will generate lateral and longitudinal torsional forces when passing through curves or slopes, and may even cause derailment. Therefore, it is very important to accurately control the height of the centerline of the coupler from the rail surface. The traditional measurement method is to place the measuring ruler seat on both sides of the rail, so that the crossbeam spans the rail, adjust the position of the main ruler and lock it, and select the probe positioning method according to the EMU coupler. The value directly read from the reading device after the probe is aligned with the measurement position is the center height of the coupler at this time. The traditional measurement method has complicated steps and consumes a lot of manpower. Summary of the invention

[0003] The technical problem to be solved by the present invention is: the traditional measurement method is to put the measuring ruler seat on both sides of the rail, make the crossbeam span the rail, lock and fix the main ruler after adjusting the position, and select the probe positioning mode according to the EMU coupler, the steps are complicated and consume a lot of manpower; the purpose of the present invention is to provide a detection method, system and medium for the distance from the center line of the coupler to the rail surface, which is improved on the basis of the traditional measurement technology, uses the camera of the inspection robot to collect the coupler image and the track image at the same point, and performs corresponding computer calculation processing on the coupler image and the track image, so as to realize the rapid and accurate measurement of the distance from the center line of the coupler to the rail surface, save a lot of human resource investment, and avoid the measurement error caused by different manual business levels.

[0004] The present invention is achieved through the following technical solutions:

[0005] This solution provides a method for detecting the distance from the center line of a coupler to the rail surface, comprising:

[0006] Perform hand-eye calibration on the inspection robot camera to obtain the transformation matrix from the camera coordinate system to the end effector of the inspection robot's mechanical arm;

[0007] The inspection robot camera collects the coupler image and track image at the same point;

[0008] Mask segmentation is performed on the coupler image and the track image respectively, and then projected to obtain the coupler centerline point cloud and the track surface point cloud;

[0009] The coupler centerline point cloud and the track surface point cloud are converted to the inspection robot base coordinate system based on the conversion matrix, the center of mass point coordinates are extracted based on the coupler centerline point cloud, and the track plane equation is fitted based on the track surface point cloud;

[0010] Calculate the distance from the center of mass point coordinates to the track plane equation as the distance from the coupler center line to the rail surface.

[0011] Working principle of this scheme: the traditional measurement method is to put the measuring ruler seat on both sides of the rail, make the crossbeam span the rail, adjust the position of the main ruler and lock it, and select the probe positioning method according to the EMU coupler. The steps are complicated and consume a lot of manpower; the purpose of the present invention is to provide a detection method, system and medium for the distance from the center line of the coupler to the rail surface, which is improved on the basis of the traditional measurement technology, and uses the camera of the inspection robot to collect the coupler image and the track image at the same point, and performs corresponding computer calculation processing on the coupler image and the track image, so as to realize the rapid and accurate measurement of the distance from the center line of the coupler to the rail surface, save a lot of human resources investment, and avoid measurement errors caused by different manual business levels.

[0012] A further optimization scheme is that the inspection robot camera is calibrated by hand and eye to obtain a conversion matrix from the camera coordinate system to the end effector of the inspection robot mechanical arm, including the following method:

[0013] Install the camera on the end effector of the inspection robot's mechanical arm and obtain the transformation matrix from the camera coordinate system to the end effector coordinate system of the mechanical arm.

[0014] A further optimization scheme is that the coupler image and the track image are respectively subjected to mask segmentation and projection to obtain the coupler centerline point cloud and the track centerline point cloud; including the method:

[0015] The mask image of the coupler centerline is extracted from the coupler image based on the PIDNET deep learning semantic segmentation network. CG , mask the coupler centerline image CG Project to the point cloud to get the coupler centerline point cloud

[0016] Extract the track surface mask image of the track image based on the PIDNET deep learning semantic segmentation network gd ; Set the track mask image mask gd Projecting into the point cloud to get the track surface point cloud

[0017] A further optimization scheme is that the PIDNET deep learning semantic segmentation network includes a proportional real-time semantic segmentation branch, an integral semantic segmentation branch and a derivative semantic segmentation branch;

[0018] The scaled real-time semantic segmentation branch is used to parse and save detailed information in the high-resolution feature map;

[0019] The integrated real-time semantic segmentation branch is used to aggregate local and global context information to resolve long-range dependencies;

[0020] The derivative real-time semantic segmentation branch is used to extract high-frequency features to predict boundary areas;

[0021] The loss function of the PIDNET deep learning semantic segmentation network is:

[0022] Loss = λ 0 l 0 +λ 1 l 1 +λ 2 l 2 +λ 3 l 3

[0023] where λ 0 ,λ 1 ,λ 2 ,λ 3 are training hyperparameters, which are set to 0.4, 20, 1, and 1 respectively; l 0 represents the semantic loss, l 1 represents the weighted binary cross entropy loss, l 2 and l 3 They represent the cross entropy loss respectively.

[0024] A further optimization scheme is that the coupler centerline point cloud and the track surface point cloud are converted to the inspection robot base coordinate system based on the conversion matrix, including the following method:

[0025] Based on the transformation matrix, the coupler centerline point cloud is converted to the inspection robot base coordinate system:

[0026]

[0027] in, Represents the point cloud of the coupler centerline in the base coordinate system of the inspection robot; The intrinsic transformation matrix for transforming the coordinates of the end effector of the inspection robot to the base coordinate system;

[0028] The orbital surface point cloud is transformed into the base coordinate system based on the transformation matrix:

[0029]

[0030] in, Represents the track surface point cloud in the base coordinate system of the inspection robot.

[0031] A further optimization scheme is that the centroid coordinates are extracted based on the coupler centerline point cloud, including the following method:

[0032] Input the centerline point of the coupler into the following formula to extract the coordinates of the center of mass point:

[0033]

[0034] where r i =(x i ,y i ,z i )i=1,2,3,...,n represents the point cloud on the center line of the coupler, n represents the total number of point clouds; m i is the calculation weight of point cloud i; M is the total number of point clouds of the coupler centerline; Pc(x c ,y c ,z c ) are the calculated centroid coordinates.

[0035] A further optimization scheme is to fit the orbital plane equation based on the orbital surface point cloud; including the following method:

[0036] Get the track surface point cloud and construct the track plane equation model as follows:

[0037] Ax+By+Cz+D=0

[0038] Among them, A, B, and C represent the plane normal vector components, and D represents the offset from the origin;

[0039] Convert the orbital plane equation model to:

[0040] z=Ax+By+C

[0041] The following error function s is constructed for fitting. When the error reaches the minimum, the orbital plane equation is obtained. The error function s is expressed as:

[0042]

[0043] Where: (x i ,y i ,z i )i=0,1..n is the orbital plane point cloud data.

[0044] A further optimization scheme is that the distance from the calculated centroid coordinates to the track plane equation is used as the distance from the coupler centerline to the track surface, including the following method:

[0045] Calculate the distance d from the center line of the coupler to the rail surface according to the following formula:

[0046]

[0047] Among them, x c ,y c ,z c Represents the centroid coordinates of the coupler centerline.

[0048] The present invention also provides a system for detecting the distance from the center line of a coupler to the rail surface, which is used to implement the above-mentioned method for detecting the distance from the center line of a coupler to the rail surface. The system comprises:

[0049] The calibration module is used to perform hand-eye calibration on the inspection robot camera and obtain the transformation matrix from the camera coordinate system to the end effector of the inspection robot's mechanical arm;

[0050] An acquisition module is used to acquire images of the coupler and the track at the same point based on the inspection robot camera;

[0051] The segmentation and projection module is used to perform mask segmentation on the coupler image and the track image respectively and project them to obtain the coupler centerline point cloud and the track surface point cloud;

[0052] A fitting module is used to convert the coupler centerline point cloud and the track surface point cloud into the inspection robot base coordinate system based on the conversion matrix, extract the centroid coordinates based on the coupler centerline point cloud, and fit the track plane equation based on the track surface point cloud;

[0053] The calculation module is used to calculate the distance from the coordinates of the centroid point to the track plane equation as the distance from the center line of the coupler to the rail surface.

[0054] The present solution also provides a computer-readable medium on which a computer program is stored. The computer program is executed by a processor to implement a method for detecting the distance from the center line of a coupler to the rail surface as described above.

[0055] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0056] The present invention provides a method, system and medium for detecting the distance from the center line of a coupler to the rail surface. The method improves on the traditional measurement technology by using the camera of a patrol robot to collect a coupler image and a rail image at the same point, and performs corresponding computer calculation processing on the coupler image and the rail image, so as to realize rapid and accurate measurement of the distance from the center line of the coupler to the rail surface, save a lot of human resource investment, and avoid measurement errors caused by different manual business levels. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and should not be regarded as limiting the scope. For those of ordinary skill in the art, other relevant drawings can be obtained based on these drawings without creative work. In the drawings:

[0058] Figure 1 The figure is a flow chart of a method for detecting the distance from the center line of the coupler to the rail surface;

[0059] Figure 2 This is a schematic diagram of the detection principle of the distance from the center line of the coupler to the rail surface;

[0060] Figure 3 is the coupler centerline mask image;

[0061] Figure 4 This is a schematic diagram of the point cloud of the coupler centerline;

[0062] Figure 5 is the orbital plane mask image;

[0063] Figure 6 It is a schematic diagram of the track plane point cloud;

[0064] Figure 7 Schematic diagram of converting the coupler centerline point cloud and track plane point cloud to the robot basic coordinate system. DETAILED DESCRIPTION

[0065] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with embodiments and drawings. The exemplary embodiments of the present invention and their description are only used to explain the present invention and are not intended to limit the present invention.

[0066] The traditional measurement method is to place the measuring ruler base on both sides of the rail, so that the crossbeam spans the rail, adjust the position of the main ruler and lock it, and select the probe positioning method according to the EMU coupler. The steps are complicated and consume a lot of manpower. In view of this, the present solution provides the following embodiments to solve the above technical problems.

[0067] Example 1

[0068] This embodiment provides a method for detecting the distance from the center line of the coupler to the rail surface. Figure 1 and Figure 2 As shown, including:

[0069] Step 1: perform hand-eye calibration on the inspection robot camera to obtain the transformation matrix from the camera coordinate system to the end effector of the inspection robot's mechanical arm; this step specifically includes the following methods:

[0070] Install the camera on the end effector of the inspection robot's mechanical arm and obtain the transformation matrix from the camera coordinate system to the end effector coordinate system of the mechanical arm.

[0071] Step 2: The inspection robot camera collects the coupler image and the track image at the same point;

[0072] Step 3: Mask segmentation and projection of the coupler image and the track image are performed to obtain the coupler centerline point cloud and the track surface point cloud; this step specifically includes the following method:

[0073] The mask image of the coupler centerline is extracted from the coupler image based on the PIDNET deep learning semantic segmentation network. CG ,like Figure 3 As shown, Figure 3 The left side is the 2D image of the coupler, and the right side is the mask image of the coupler centerline. CG Project to the point cloud to get the coupler centerline point cloud like Figure 4 As shown, Figure 4 The black part in the middle of the left side is the point cloud of the coupler centerline, and the right side is an enlarged view of the extracted point cloud.

[0074] like Figure 5 As shown in the figure, the track surface mask image of the track image is extracted based on the PIDNET deep learning semantic segmentation network. gd ; Figure 5 The left side shows the captured orbital image, and the right side shows the extracted orbital plane mask; Figure 6 As shown, the track mask image mask gd Projecting into the point cloud to get the track surface point cloud

[0075] The PIDNET deep learning semantic segmentation network includes a proportional real-time semantic segmentation branch, an integral semantic segmentation branch and a derivative semantic segmentation branch;

[0076] The scaled real-time semantic segmentation branch is used to parse and save detailed information in the high-resolution feature map;

[0077] The integrated real-time semantic segmentation branch is used to aggregate local and global context information to resolve long-range dependencies;

[0078] The derivative real-time semantic segmentation branch is used to extract high-frequency features to predict boundary areas;

[0079] The loss function of the PIDNET deep learning semantic segmentation network is:

[0080] Loss = λ 0 l0 +λ 1 l 1 +λ 2 l 2 +λ 3 l 3

[0081] where λ 0 ,λ 1 ,λ 2 ,λ 3 are training hyperparameters, which are set to 0.4, 20, 1, and 1 respectively; l 0 represents the semantic loss, l 1 represents the weighted binary cross entropy loss, l 2 and l 3 They represent the cross entropy loss respectively.

[0082] Step 4: Figure 7 As shown, based on the transformation matrix, the coupler centerline point cloud and the track surface point cloud are transformed into the inspection robot base coordinate system, the centroid coordinates are extracted based on the coupler centerline point cloud, and the track plane equation is fitted based on the track surface point cloud; Figure 7 The upper middle picture is the point cloud of the coupler centerline, and the lower picture is the point cloud of the track plane;

[0083] The method of converting the coupler centerline point cloud and the track surface point cloud to the inspection robot base coordinate system based on the conversion matrix includes:

[0084] Based on the transformation matrix, the coupler centerline point cloud is converted to the inspection robot base coordinate system:

[0085]

[0086] in, Represents the point cloud of the coupler centerline in the base coordinate system of the inspection robot; The intrinsic transformation matrix for transforming the coordinates of the end effector of the inspection robot to the base coordinate system;

[0087] The orbital surface point cloud is transformed into the base coordinate system based on the transformation matrix:

[0088]

[0089] in, Represents the track surface point cloud in the base coordinate system of the inspection robot.

[0090] The method of extracting the centroid coordinates based on the coupler centerline point cloud includes:

[0091] Input the centerline point of the coupler into the following formula to extract the coordinates of the center of mass point:

[0092]

[0093] where r i =(x i ,y i ,z i )i=1,2,3,...,n represents the point cloud on the center line of the coupler, n represents the total number of point clouds; m i is the calculation weight of point cloud i; M is the total number of point clouds of the coupler centerline; Pc(x c ,y c ,z c ) are the calculated centroid coordinates.

[0094] The method of fitting the orbital plane equation based on the orbital surface point cloud includes:

[0095] Get the track surface point cloud and construct the track plane equation model as follows:

[0096] Ax+By+Cz+D=0

[0097] Among them, A, B, and C represent the plane normal vector components, and D represents the offset from the origin;

[0098] Convert the orbital plane equation model to:

[0099] z=Ax+By+C

[0100] The following error function s is constructed for fitting. When the error reaches the minimum, the orbital plane equation is obtained. The error function s is expressed as:

[0101]

[0102] Where: (x i ,y i ,z i )i=0,1..n is the orbital plane point cloud data.

[0103] Step 5, calculate the distance from the coordinates of the centroid to the track plane equation as the distance from the center line of the coupler to the rail surface.

[0104] This step specifically includes the following methods:

[0105] Calculate the distance d from the center line of the coupler to the rail surface according to the following formula:

[0106]

[0107] Among them, x c ,y c ,z c Represents the centroid coordinates of the coupler centerline.

[0108] Since the field of view of the coupler and the track plane is very large, the current 3D camera cannot simultaneously capture the coupler and the track plane in one image. This solution innovatively proposes to capture the coupler and the track surface separately at one point, and use hand-eye calibration to convert them into the robot's basic coordinate system, so as to accurately measure the coupler to the track surface in the same coordinate system. The specific description is as follows: a color structured light 3D camera is built at the end of the robot's mechanical arm, and the camera collects RGB-D images, which include 3D point clouds and 2D images. After the robot arrives at the shooting point, it uses the mechanical arm to teach the posture, and captures the coupler horizontally and the upper surface of the track vertically, respectively, to obtain RGB-D images of the coupler and the track. The 2D image of the coupler is subjected to deep learning semantic segmentation to segment the centerline of the coupler, and then mapped to the 3D point cloud to extract the centerline point cloud. The 2D color image of the track is subjected to deep learning semantic segmentation to segment the 2D image of the track, and then mapped to the 3D point cloud to extract the 3D point cloud of the track. Finally, the extracted 3D point cloud of the coupler centerline and the extracted 3D point cloud of the track are converted to the robot reference coordinate system, so that the coupler centerline point cloud and the track point cloud are in the same coordinate system. After conversion to the same coordinate system, fit the track point cloud to the plane, calculate the distance from the coupler centerline to the track plane, and then calculate the height of the coupler centerline from the track surface.

[0109] The present invention utilizes the 3D camera of the inspection robot to quickly and accurately measure the coupler, saving a large amount of human resource investment and avoiding measurement errors caused by different manual skills.

[0110] Example 2

[0111] This embodiment provides a system for detecting the distance from the center line of a coupler to the rail surface, which is used to implement a method for detecting the distance from the center line of a coupler to the rail surface as described in Embodiment 1. The system includes:

[0112] The calibration module is used to perform hand-eye calibration on the inspection robot camera and obtain the transformation matrix from the camera coordinate system to the end effector of the inspection robot's mechanical arm;

[0113] An acquisition module is used to acquire images of the coupler and the track at the same point based on the inspection robot camera;

[0114] The segmentation and projection module is used to perform mask segmentation on the coupler image and the track image respectively and project them to obtain the coupler centerline point cloud and the track surface point cloud;

[0115] A fitting module is used to convert the coupler centerline point cloud and the track surface point cloud into the inspection robot base coordinate system based on the conversion matrix, extract the centroid coordinates based on the coupler centerline point cloud, and fit the track plane equation based on the track surface point cloud;

[0116] The calculation module is used to calculate the distance from the coordinates of the centroid point to the track plane equation as the distance from the center line of the coupler to the rail surface.

[0117] Example 3

[0118] This embodiment provides a computer-readable medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement a method for detecting the distance from the center line of a coupler to a rail surface as described in Embodiment 1; specifically, the following steps are performed:

[0119] Step 1: Perform hand-eye calibration on the inspection robot camera to obtain the transformation matrix from the camera coordinate system to the end effector of the inspection robot's mechanical arm;

[0120] Step 2: The inspection robot camera collects the coupler image and the track image at the same point;

[0121] Step 3, mask segmentation is performed on the coupler image and the track image respectively, and then projected to obtain the coupler centerline point cloud and the track surface point cloud;

[0122] Step 4: based on the transformation matrix, the coupler centerline point cloud and the track surface point cloud are transformed into the inspection robot base coordinate system, the centroid coordinates are extracted based on the coupler centerline point cloud, and the track plane equation is fitted based on the track surface point cloud;

[0123] Step 5, calculate the distance from the coordinates of the centroid to the track plane equation as the distance from the center line of the coupler to the rail surface.

[0124] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for detecting the distance from the center line of a coupler to the rail surface, characterized in that: include: Perform hand-eye calibration on the inspection robot camera to obtain the transformation matrix from the camera coordinate system to the end effector of the inspection robot's mechanical arm; The inspection robot camera collects the coupler image and track image at the same point; Mask segmentation is performed on the coupler image and the track image respectively, and then projected to obtain the coupler centerline point cloud and the track surface point cloud; The coupler centerline point cloud and the track surface point cloud are converted to the inspection robot base coordinate system based on the conversion matrix, the center of mass point coordinates are extracted based on the coupler centerline point cloud, and the track plane equation is fitted based on the track surface point cloud; Calculate the distance from the center of mass point coordinates to the track plane equation as the distance from the coupler center line to the rail surface.

2. A method for detecting the distance from the center line of a coupler to the rail surface according to claim 1, characterized in that: The hand-eye calibration of the inspection robot camera is performed to obtain a conversion matrix from the camera coordinate system to the end effector of the inspection robot mechanical arm, including the following method: Install the camera on the end effector of the inspection robot's mechanical arm and obtain the transformation matrix from the camera coordinate system to the end effector coordinate system of the mechanical arm.

3. A method for detecting the distance from the center line of a coupler to the rail surface according to claim 2, characterized in that: The method includes performing mask segmentation on the coupler image and the track image respectively and projecting them to obtain the coupler centerline point cloud and the track centerline point cloud; The mask image of the coupler centerline is extracted from the coupler image based on the PIDNET deep learning semantic segmentation network. CG , mask the coupler centerline image CG Project to the point cloud to get the coupler centerline point cloud Extract the track surface mask image of the track image based on the PIDNET deep learning semantic segmentation network gd ; Set the track mask image mask gd Projecting into the point cloud to get the track surface point cloud 4. A method for detecting the distance from the center line of a coupler to the rail surface according to claim 3, characterized in that: The PIDNET deep learning semantic segmentation network includes a proportional real-time semantic segmentation branch, an integral semantic segmentation branch and a derivative semantic segmentation branch; The scaled real-time semantic segmentation branch is used to parse and save detailed information in the high-resolution feature map; The integrated real-time semantic segmentation branch is used to aggregate local and global context information to resolve long-range dependencies; The derivative real-time semantic segmentation branch is used to extract high-frequency features to predict boundary areas; The loss function of the PIDNET deep learning semantic segmentation network is: Loss=λ0l0+λ1l1+λ2l2+λ3l3 Among them, λ0, λ1, λ2, and λ3 are training hyperparameters, which are set to 0.4, 20, 1, and 1 respectively; l0 represents semantic loss, l1 represents weighted binary cross entropy loss, and l2 and l3 represent cross entropy losses.

5. A method for detecting the distance from the center line of a coupler to the rail surface according to claim 3, characterized in that: The method of converting the coupler centerline point cloud and the track surface point cloud to the inspection robot base coordinate system based on the conversion matrix includes: Based on the transformation matrix, the coupler centerline point cloud is converted to the inspection robot base coordinate system: in, Represents the point cloud of the coupler centerline in the base coordinate system of the inspection robot; The intrinsic transformation matrix for transforming the coordinates of the end effector of the inspection robot to the base coordinate system; The orbital surface point cloud is transformed into the base coordinate system based on the transformation matrix: in, Represents the track surface point cloud in the base coordinate system of the inspection robot.

6. A method for detecting the distance from the center line of a coupler to the rail surface according to claim 5, characterized in that: The method of extracting the centroid coordinates based on the coupler centerline point cloud includes: Input the centerline point of the coupler into the following formula to extract the coordinates of the center of mass point: where r i =(x i ,y i ,z i )i=1, 2, 3, ..., n represents the point cloud on the center line of the coupler, and n represents the total number of point clouds; m i is the calculation weight of point cloud i; M is the total number of point clouds of the coupler centerline; Pc(x c ,y c ,z c ) are the calculated centroid coordinates.

7. A method for detecting the distance from the center line of a coupler to the rail surface according to claim 5, characterized in that: The track plane equation is fitted based on the track surface point cloud; Included methods: Get the track surface point cloud and construct the track plane equation model as follows: Ax+By+Cz+D=0 Among them, A, B, and C represent the plane normal vector components, and D represents the offset from the origin; Convert the orbital plane equation model to: z=Ax+By+C The following error function s is constructed for fitting. When the error reaches the minimum, the orbital plane equation is obtained. The error function s is expressed as: Where: (x i ,y i ,z i )i=0,1..n is the orbital plane point cloud data.

8. A method for detecting the distance from the center line of a coupler to the rail surface according to claim 7, characterized in that: The method of calculating the distance from the coordinates of the centroid to the track plane equation as the distance from the center line of the coupler to the track surface includes: Calculate the distance d from the center line of the coupler to the rail surface according to the following formula: Among them, x c ,y c ,z c Represents the centroid coordinates of the coupler centerline.

9. A system for detecting the distance from the center line of a coupler to the rail surface, characterized in that: A method for detecting the distance from a coupler centerline to a rail surface as claimed in any one of claims 1 to 8, the system comprising: The calibration module is used to perform hand-eye calibration on the inspection robot camera and obtain the transformation matrix from the camera coordinate system to the end effector of the inspection robot's mechanical arm; An acquisition module is used to acquire images of the coupler and the track at the same point based on the inspection robot camera; The segmentation and projection module is used to perform mask segmentation on the coupler image and the track image respectively and project them to obtain the coupler centerline point cloud and the track surface point cloud; A fitting module is used to convert the coupler centerline point cloud and the track surface point cloud into the inspection robot base coordinate system based on the conversion matrix, extract the centroid coordinates based on the coupler centerline point cloud, and fit the track plane equation based on the track surface point cloud; The calculation module is used to calculate the distance from the coordinates of the centroid point to the track plane equation as the distance from the center line of the coupler to the rail surface.

10. A computer readable medium having a computer program stored thereon, characterized in that: The computer program is executed by a processor to implement a method for detecting the distance from the center line of a coupler to the rail surface as described in any one of claims 1 to 8.

Images