A non-contact track wear and third rail geometric parameter real-time detection method and system
By using non-contact image acquisition and 3D point cloud processing, efficient and accurate detection of the relative position of the third track was achieved, solving the problems of low efficiency and large error in existing technologies and improving the safety of the third track.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- HANGZHOU SHENHAO TECH
- Filing Date
- 2024-12-25
- Publication Date
- 2026-06-26
AI Technical Summary
In existing technologies, the measurement efficiency of the third rail relative to the railway rail is low, and wear leads to large errors in the measurement results, affecting train power supply and safety.
Track images were acquired using a non-contact method, and then converted into 3D point cloud data through image preprocessing. Coordinate system normalization and registration fusion were performed to calculate the wear, pull-out value, and guide height value of the third track.
It improves the accuracy and speed of third-rail geometric inspection, eliminates measurement errors caused by wear, and enhances the safety of the third rail.
Smart Images
Figure CN119714066B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of track geometry detection technology, specifically to a non-contact method and system for real-time detection of track wear and third track geometric parameters. Background Technology
[0002] The third rail, also known as the power supply rail, is a rail installed beside the track line for power supply. It works in conjunction with the current receiving surface (collector shoe) to provide power support for all equipment on the rail transit train.
[0003] The train draws power from the third rail via a current collector shoe. If the position of the third rail relative to the railway rail changes, it will cause a power failure for the train; in severe cases, it may cause the train to intrude into the clearance gauge, potentially leading to a collision. Therefore, it is necessary to periodically measure the relative position and wear condition of the third rail relative to the railway rail.
[0004] The relative position of the third rail to the railway rail includes the guide height and pull-out value. Measurement methods generally include manual measurement, hand-pushed trolley measurement, and measurement using a train with an added measuring arm. Manual measurement requires the use of a third rail measuring ruler, and measurements are taken manually at monitoring points during the work period. This method is inefficient and has a long measurement cycle. Measurement using a hand-pushed trolley with an added measuring arm is generally a non-contact method, using third rail acquisition equipment to measure the geometric parameters of the third rail. This method is limited by the walking speed of the operator and the detection speed of the measuring tools; typically, the maximum speed for this method is 5 km / h. Using a train with an added measuring arm can affect the measurement results due to factors such as rail subsidence caused by the weight of the train and changes in the relative relationship between the train wheels and the rails in curved sections.
[0005] When designing and installing the third rail, the stress-bearing side and top surface of the rail closest to the third rail are taken into account. Long-term rail wear can also cause errors in the geometric measurement of the third rail. Therefore, the amount of rail wear on the side closest to the third rail should be taken into account during the geometric inspection of the third rail. Otherwise, the geometric inspection results of the third rail will have a large deviation in areas with greater rail wear. Summary of the Invention
[0006] Based on the above background, the purpose of this invention is to provide a non-contact method and system for real-time detection of track wear and third track geometric parameters.
[0007] To achieve the above objectives, the present invention adopts the following technical solution:
[0008] In a first aspect, the present invention provides a non-contact method for real-time detection of track wear and third track geometric parameters, including,
[0009] Images of the rail head on the stressed and unstressed sides of the first rail, images of the rail head on the stressed and unstressed sides of the second rail, and images of the contact surface of the third rail are collected and preprocessed.
[0010] The acquisition equipment is calibrated, and the preprocessed image data is converted into 3D point cloud data; and all point cloud data are normalized according to the coordinate system based on the rotation and translation matrix obtained from the calibration.
[0011] The point cloud data of the rail head on the unloaded side of the first and second rails are matched and registered with the standard rail profile in the template file to obtain the corresponding registration rotation matrix. Based on the obtained registration rotation matrix, the rail head data on the loaded side of the first and second rails are transformed by coordinate transformation. All the registered rail head point cloud data are then fused to obtain complete rail head point cloud data.
[0012] The spatial distance between the pre-set points of the standard rail profile in the template file and the corresponding points of the actual collected rail is calculated based on the fused rail head point cloud data, and the wear value of the corresponding points is obtained.
[0013] The actual wear plane is obtained by fitting a straight line to the contact surface of the third rail. The center point of the third rail is obtained by fitting the non-contact surface data of the edge of the third rail. The wear value of the third rail is calculated based on the distance between the fitted center point of the third rail and the actual wear plane, as well as the theoretical values of the non-contact surface and the contact surface of the third rail.
[0014] The nearest rail is determined based on the point cloud data of the third rail, and the third rail pull-out value and the third rail guide height value are obtained based on the relative positional relationship between the center point of the third rail and the surface of the nearest rail, the distance between the center point of the third rail and the actual wear plane, and the wear value of the corresponding point of the rail.
[0015] Secondly, the present invention provides a non-contact real-time detection system for track wear and third track geometric parameters, comprising,
[0016] The image data acquisition module is used to acquire images of the rail head on the stressed and unstressed sides of the first rail, images of the rail head on the stressed and unstressed sides of the second rail, and images of the contact surface of the third rail.
[0017] The image data preprocessing module is used to preprocess the acquired images;
[0018] The 3D point cloud conversion module is used to calibrate the image data acquisition module and convert the preprocessed image data into 3D point cloud data; and to normalize the coordinate system of all point cloud data according to the rotation and translation matrix obtained from the calibration.
[0019] The rail point cloud data registration and fusion module is used to match and register the rail head point cloud data of the unloaded side of the first and second rails with the standard rail contour in the template file to obtain the corresponding registration rotation matrix. Based on the obtained registration rotation matrix, the coordinate transformation is performed on the rail head data of the loaded side of the first and second rails, and all the registered rail head point cloud data are fused to obtain complete rail head point cloud data.
[0020] The rail wear value acquisition module is used to calculate the spatial distance between the preset points of the standard rail profile in the template file and the corresponding points of the actual collected rail based on the fused rail head point cloud data, and to obtain the wear value of the corresponding points.
[0021] The third rail wear value acquisition module is used to perform linear fitting on the contact surface of the third rail to obtain the actual wear plane, perform feature fitting on the non-contact surface data of the third rail edge to obtain the center point of the third rail, and calculate the wear value of the third rail based on the distance between the fitted center point of the third rail and the actual wear plane, as well as the theoretical value between the non-contact surface and the contact surface of the third rail.
[0022] The module for obtaining the third rail pull-out value and guide height value is used to determine the nearest rail based on the third rail point cloud data, and obtain the point wear value of the corresponding rail and the distance between the center point of the third rail and the actual wear plane. The module obtains the third rail pull-out value and the third rail guide height value based on the relative positional relationship between the center point of the third rail and the rail surface.
[0023] The beneficial effects of this invention are as follows:
[0024] 1) When performing rail wear detection, this invention uses the outer contour of the rail on the unloaded side to register the rail head area, which reduces the influence of the wear surface point cloud on the registration result and improves the accuracy of wear detection.
[0025] 2) When performing geometric parameter testing of the third rail, this invention divides the surface contour of the third rail into a contact area and a non-contact area. By detecting the arc of the non-contact area and fitting the straight line of the contact surface, the wear value of the third rail is measured, which solves the problem that the wear value of the third rail is not detected in the prior art and provides a reference for the service status of the third rail.
[0026] 3) This invention compensates for the geometric detection results of the third rail based on the wear detection results, realizes real-time detection of the third rail status under high-speed conditions, eliminates the geometric measurement error of the third rail caused by rail wear, improves the geometric detection accuracy of the third rail, and increases the safety of the third rail. Attached Figure Description
[0027] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.
[0028] Figure 1 This is a flowchart of a non-contact track wear and third track geometric parameter real-time detection method provided in an embodiment of the present invention;
[0029] Figure 2 A schematic diagram of a non-contact track wear and third track geometric parameter real-time detection system provided in an embodiment of the present invention. Detailed Implementation
[0030] To further understand the present invention, preferred embodiments of the present invention are described below in conjunction with examples. However, it should be understood that these descriptions are only for further illustrating the features and advantages of the present invention, and not for limiting the scope of the claims of the present invention.
[0031] Terminology Explanation:
[0032] The third rail, also known as the power supply rail, includes contact and non-contact surfaces. The contact surface refers to the part of the third rail that is in direct contact with the train's current collector (such as a current collector shoe); the non-contact surface refers to other surfaces besides the part that is in contact with the train's current collector. These parts do not directly participate in the transmission of current, but may be involved in structures such as protection, insulation, and fixing.
[0033] There are three types of current collection methods for the third rail: top contact, bottom contact, and side contact. In this implementation, the bottom contact type is used as an example for illustration. The bottom contact type means that the rail head of the third rail faces downward.
[0034] Example 1:
[0035] This invention provides a non-contact method for real-time detection of track wear and third track geometric parameters. See [link to relevant documentation]. Figure 1 The method includes the following steps:
[0036] S1: Collect images of the rail head on the stressed and unstressed sides of the first rail, the rail head on the stressed and unstressed sides of the second rail, and the contact surface of the third rail;
[0037] S2: Preprocess the acquired image information;
[0038] S3: Calibrate the camera and convert the preprocessed image data into 3D point cloud data. Apply the calibration parameters to normalize the coordinate system of all point cloud data to obtain the 3D point cloud data of the first rail, the rail head of the second rail, and the rail surface of the third rail in the same coordinate system.
[0039] S4: Perform point cloud registration and fusion on the rail head point cloud data, and detect the wear value at preset points on the rail respectively;
[0040] S5: Detect the center point of the third track and fit the wear surface based on the third track point cloud data, and calculate the wear value of the third track;
[0041] S6: Determine the position area of the third rail relative to the steel rail based on the point cloud data of the third rail, and calculate the pull-out value and guide height value of the third rail based on the relative relationship between the center point of the third rail and the rail surface of its adjacent steel rail.
[0042] Specifically:
[0043] S1: Collect the contour data of the contact surfaces of the first, second, and third rails of the running track.
[0044] In this embodiment, the image acquisition device includes six line-scan camera acquisition modules and an odometer encoder. All line-scan cameras are simultaneously triggered by the odometer encoder. When the encoder's change value reaches a set threshold, the cameras are triggered to acquire profile images of the rail surface and third rail surface in the same elevation area under the current odometer condition, thus completing the data acquisition of the rail and third rail contours. The threshold setting needs to be adjusted according to the specific application scenario and environmental conditions, and the optimal threshold setting is determined experimentally based on the actual track environment and camera performance.
[0045] The specific implementation steps and principles of this process are as follows:
[0046] The four line-scan camera acquisition modules in the middle are divided into two groups, with two modules in each group, installed on the same plane. Because the camera on one side can only scan part of the rail head area, two line-scan camera modules are set up in each group to ensure that complete rail head data can be acquired. The two camera acquisition modules at the edge respectively acquire data on the surface contour of the third rail on both sides of the rail. It should be noted that, for the convenience of subsequent description, the six cameras in the middle are numbered from left to right as: Camera 1, Camera 2, Camera 3, Camera 4, Camera 5, and Camera 6.
[0047] The third rail is located beside the main rail, usually one in number. With the direction of train travel as the reference, the third rail is sometimes located on the right side of the main rail and sometimes on the left side. It provides power to all equipment on the train, but the specific number of third rails needs to be determined based on the railway system design and application scenario. Therefore, to improve the versatility of the detection equipment, two camera acquisition modules are symmetrically arranged on both sides to collect data on the surface contours of the third rail on both sides of the main rail. However, this embodiment only illustrates the third rail on one side of the guide rail and is not intended to limit the scope of protection of this invention.
[0048] S2: Preprocess the acquired image information.
[0049] The acquired images undergo preprocessing to provide high-quality image data for subsequent image analysis. Image preprocessing includes image filtering, edge detection, and contour extraction.
[0050] Image filtering includes methods such as Gaussian filtering, median filtering, and bilateral filtering. Gaussian filtering uses a Gaussian function as a convolution kernel to smooth the image, removing noise and preserving edge information. Median filtering removes noise by replacing each pixel value with the median of its neighborhood. Bilateral filtering reduces noise while maintaining edge sharpness and is suitable for smoothing processes that preserve details.
[0051] Edge detection typically employs the Canny multi-stage edge detection algorithm, which includes Gaussian filtering, gradient calculation, non-maximum suppression, and double thresholding, to detect edges in images. It also includes Sobel edge detection, which uses the Sobel operator to calculate the image gradient, and Laplacian edge detection, which uses the Laplacian operator to calculate the second derivative of the image.
[0052] Contour extraction uses OpenCV's findContours function to connect edges into a contour, accurately extracting the contour information of an object from an image, i.e., a contour map, providing a foundation for subsequent image analysis and understanding.
[0053] S3: Convert the preprocessed image data of the rails and third rail into 3D point cloud data, and perform coordinate normalization on the 3D point cloud data. This includes...
[0054] S301: Calibrate the camera to obtain the camera's rotation and translation matrix Rt;
[0055] Camera calibration involves determining the camera's intrinsic and extrinsic parameters. Intrinsic parameters include focal length, principal point, and distortion coefficients, forming the camera's intrinsic parameter matrix. Extrinsic parameters include rotation and translation matrix parameters. Using the camera's intrinsic and extrinsic parameters, combined with the depth information of each pixel in the 2D image, the 2D image coordinates can be inversely projected into 3D space to obtain the corresponding 3D point cloud data. The depth information of each pixel in the 2D image represents the depth in 3D space (i.e., the distance from the camera), which can typically be obtained through methods such as structured light, depth sensors, and stereo vision. The rotation and translation matrix describes the camera's position and orientation relative to the world coordinate system, transforming the image's 3D coordinates from the camera coordinate system to the world coordinate system to obtain the final 3D point cloud data.
[0056] In this embodiment, the four line-scan camera acquisition modules that collect data on the rail contour area at the middle position are calibrated to obtain the rotation and translation matrices Rt for the four cameras; the rotation and translation matrices Rt are saved in a configuration file; the third rail data acquisition camera is calibrated to obtain the corresponding rotation and translation matrix, and the rotation and translation matrix parameters are saved in the configuration file. After calibration, no further calibration is required without changing the camera installation position.
[0057] It should be noted that camera calibration and the corresponding conversion of image data into 3D point cloud data are existing technologies, which can be implemented by those skilled in the art using existing knowledge. Therefore, the specific process will not be described in detail here.
[0058] S302: Convert rail image data into 3D point cloud data and normalize the coordinate system;
[0059] After converting the rail image data acquired by the line scan camera acquisition module into 3D point cloud data, the rotation and translation matrix Rt of the corresponding camera is read according to the camera number to perform coordinate system normalization. The coordinate systems of the two cameras in each group are transformed to the same coordinate system and fused to obtain complete rail head point cloud data.
[0060] S303: Convert the third track image data into 3D point cloud data and perform coordinate system normalization;
[0061] After converting the third track image data acquired by the line scan camera acquisition module into 3D point cloud data, the rotation and translation matrix Rt of the corresponding camera is read according to the camera number for coordinate system normalization processing. That is, the third track point cloud data is transformed into the same coordinate system as the rail point cloud data through the rotation and translation matrix.
[0062] S4: Based on the rail head point cloud data obtained in step S3, perform point cloud registration and fusion, and detect the wear value of the rail at the preset points.
[0063] In this embodiment, the preset points on the rail are the top horizontal points, that is, the points where the horizontal tangent is tangent to the rail head, denoted as the 0° point; the points where the vertical tangent is tangent to the stress-bearing side of the rail head are denoted as the 90° point. For simplicity, the 0° and 90° points will be used interchangeably in the following description. The specific process is as follows:
[0064] S401: Based on the rail head point cloud data Pts2 and Pts5 collected by camera 2 and camera 5, the collected rail profile is registered with the standard rail profile in the template file using the ICP (Iterative Closest Point) registration algorithm to obtain the registration rotation matrix Rt2' and Rt5' and the registered point cloud set Pts2' and Pts5'.
[0065] S402: Apply the registration rotation matrices Rt2' and Rt5' to the point cloud data Pts3 and Pts4 of camera 3 and camera 4 respectively to perform coordinate transformation, and obtain new point cloud sets Pts3' and Pts4'.
[0066] S403: After obtaining the new point cloud set, Pts2' and Pts3', Pts4' and Pts5' are fused respectively to obtain new point cloud data Pts23 and Pts45. The new point cloud data are the point cloud data of the first rail head and the second rail head.
[0067] S404: Based on the fused rail head point cloud data, calculate the spatial distance between the 0° and 90° points of the standard rail profile in the template file and their corresponding positions on the actual collected rail. This distance is the wear value of the 0° and 90° points. Specifically,
[0068] Based on the spatial coordinates of the 0° and 90° points measured in the template file, the 10 point cloud data closest to the 0°(x,y,z) and 90°(x,y,z) points in the template file are extracted from the point cloud sets Pts23 and Pts45 respectively, resulting in two new point cloud sets Pts_0 and Pts_90. Based on the two point cloud sets Pts_0 and Pts_90, two straight lines L_0 and L_90 are fitted respectively using the least squares method. The wear values w_0 and w_90 are obtained by taking the vertical distance from the 0°(x,y,z) and 90°(x,y,z) points in the point template file to the straight lines L_0 and L_90.
[0069] S5: Detect the center point of the third track and calculate the wear value of the third track based on the third track point cloud data. This includes...
[0070] S501: Detect the non-contact surface of the third track to obtain the coordinates of the center point of the third track;
[0071] The non-contact surface of the third track has different characteristics in different models, meaning different feature detection methods can be performed, such as arc feature detection and straight line feature detection. In this embodiment, the arc feature presented in the edge region of the third track is fitted with a circle to obtain the coordinates of the center point of the third track.
[0072] Based on the arc features presented by the edge region of the third track, a circle is fitted. The methods for circle fitting include the least squares method and the RANSAC (Random Sample Consensus) algorithm. In this embodiment, the RANSAC algorithm is specifically used to fit the point cloud of the third track contour, and two fitted circle equations C_l and C_r and the center point coordinates of the two fitted circles (c_l_cent(x, y, z) and c_r_cent(x, y, z)).
[0073] Based on the fitted circle center coordinates, the coordinates of the center point of the third track, t_cent(x', y', z'), are obtained, where x' = t_cent((c_l_cent.x + c_r_cent.x) * 0.5, y' = (c_l_cent.y + c_r_cent.y) * 0.5, z' = (c_l_cent.z + c_r_cent.z) * 0.5).
[0074] S502: Perform a straight-line test on the contact surface of the third rail, locate the actual wear plane of the third rail, and calculate the wear value of the third rail based on the theoretical value and the actual test difference between the center point of the arc on both sides of the third rail surface and the third rail surface.
[0075] Based on the point cloud data of the third rail profile, the RANSAC algorithm is used to detect straight lines on the contact surface of the third rail, and the corresponding rail contact surface point cloud set Pts_plane and straight line equation L are obtained to locate the actual wear plane of the third rail.
[0076] The height difference between the center point of the arc detection and the wear surface is calculated by subtracting the theoretical value between the non-contact and contact surfaces of the third rail to obtain the wear value of the third rail: railcost = Ht - Hw, where Hw is the distance between the fitted circle center and the straight line equation; and Ht is the theoretical design value between the center point of the arc on both sides of the third rail and the contact surface of the third rail for the corresponding model.
[0077] S6: Determine the position of the third rail relative to the main rail based on the third rail point cloud data, and calculate the third rail pull-out value and the third rail guide height value based on the relative relationship between the center point of the third rail and the rail surface; including,
[0078] S601: Determine the position area of the third rail relative to the steel rail and extract the wear value of the nearest steel rail;
[0079] After the point cloud of the third track profile is normalized, the size of the y-coordinate value of the point cloud data collected by the third track is used to determine which side of the rail the area of the third track is closer to, that is, to determine whether the location of the third track is closer to the first rail or the second rail; and based on the judgment result, the wear values of the corresponding rail 0° and 90° points detected in step S4 are extracted.
[0080] S602: Calculate the pull-out value and guide height value of the third rail;
[0081] The third rail pull-out value refers to the lateral distance between the center point C of the third rail and the wear measurement point at a 90° angle to the adjacent rail within the same cross section; that is, the difference in the Y-axis coordinates of the two points in the coordinate system of this embodiment.
[0082] The third rail guide height value refers to the vertical distance between the center point C of the third rail and the 0° point of the adjacent rail wear measurement point within the same cross section, which is the difference in Z-axis coordinates between the two points in the coordinate system of this embodiment.
[0083] The specific calculation process is as follows:
[0084] The coordinate difference between the center point coordinates and the wear point of the rail on the nearest side is calculated to obtain the initial pull value and leadhight of the third rail.
[0085] Based on the rail wear values w_0 and w_90 obtained in step S4, and the distance HW between the center point of the third rail and the actual wear plane obtained in step S5, the initial pull-out value and leadhight of the third rail are compensated for to obtain the specific geometric detection results of the third rail:
[0086] Pullvalue' = pullvalue + w_90
[0087] The leadhigh value is defined as leadhight' = leadhight + w_0 + HW.
[0088] This embodiment uses the rail closest to the third rail as the testing benchmark. The wear test result at the 90° point of the rail is used as the compensation value for the third rail pull-out test result, and the wear test result at the 0° point of the rail is used as the compensation value for the third rail guide height test result. This method solves the problem of measurement errors in the third rail guide height and pull-out values caused by rail wear. Simultaneously, the difference between the distance between the arc area of the third rail edge and the wear surface of the third rail and the design value of the third rail model is used as the third rail wear test result, providing a reference for the service status of the third rail.
[0089] Example 2:
[0090] See Figure 2This embodiment provides a non-contact track wear and third track geometric parameter real-time detection system, including,
[0091] The image data acquisition module is used to acquire images of the rail head on the stressed side of the first rail, the rail head on the unstressed side of the first rail, the rail head on the stressed side of the second rail, the rail head on the unstressed side of the second rail, and the contact surface of the third rail.
[0092] The image data preprocessing module is used to preprocess the acquired images;
[0093] The 3D point cloud conversion module is used to calibrate the image data acquisition module and convert the preprocessed image data into 3D point cloud data; and to normalize the coordinate system of all point cloud data according to the rotation and translation matrix obtained from the calibration.
[0094] The rail point cloud data registration and fusion module is used to match and register the rail head point cloud data of the unloaded side of the first and second rails with the standard rail contour in the template file to obtain the corresponding registration rotation matrix. Based on the obtained registration rotation matrix, the coordinate transformation is performed on the rail head data of the loaded side of the first and second rails, and all the registered rail head point cloud data are fused to obtain complete rail head point cloud data.
[0095] The rail wear value acquisition module is used to calculate the spatial distance between the preset points of the standard rail profile in the template file and the corresponding positions of the actual collected rails based on the fused rail head point cloud data, and to obtain the wear value of the corresponding points.
[0096] The third rail wear value acquisition module is used to perform linear fitting on the contact surface of the third rail to obtain the actual wear plane, perform feature fitting on the non-contact surface data of the third rail edge to obtain the center point of the third rail, and calculate the wear value of the third rail based on the distance between the fitted center point of the third rail and the actual wear plane, as well as the theoretical value between the non-contact surface and the contact surface of the third rail.
[0097] The module for obtaining the third rail pull-out value and guide height value is used to determine the nearest rail based on the third rail point cloud data, and to obtain the point wear value of the corresponding rail and the distance between the center point of the third rail and the actual wear plane. The module calculates the third rail pull-out value and the third rail guide height value based on the relative positional relationship between the center point of the third rail and the rail surface.
[0098] The above description of the embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. It should be noted that those skilled in the art can make several improvements and modifications to the present invention without departing from the principles of the present invention, and these improvements and modifications also fall within the protection scope of the claims of the present invention.
[0099] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of this invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.
[0100] It should be noted that some exemplary embodiments are described as processes or methods depicted as flowcharts. Although the flowcharts describe the steps as sequential processes, many of these steps can be performed in parallel, concurrently, or simultaneously. Furthermore, the order of the steps can be rearranged. A process can be terminated when its operation is complete, but it may also have additional steps not included in the figures. A process can correspond to a method, function, procedure, subroutine, subroutine, etc.
Claims
1. A non-contact method for real-time detection of track wear and third track geometric parameters, characterized in that, include, Images of the rail head on the stressed and unstressed sides of the first rail, images of the rail head on the stressed and unstressed sides of the second rail, and images of the contact surface of the third rail are collected and preprocessed. The acquisition equipment is calibrated, and the preprocessed image data is converted into 3D point cloud data; and all point cloud data are normalized according to the coordinate system based on the rotation and translation matrix obtained from the calibration. The point cloud data of the rail head on the unloaded side of the first and second rails are matched and registered with the standard rail profile in the template file to obtain the corresponding registration rotation matrix. Based on the obtained registration rotation matrix, the rail head data on the loaded side of the first and second rails are transformed by coordinate transformation. All the registered rail head point cloud data are then fused to obtain complete rail head point cloud data. The spatial distance between the pre-set points of the standard rail profile in the template file and the corresponding points of the actual collected rail is calculated based on the fused rail head point cloud data to obtain the wear value of the corresponding points. This includes: extracting the point cloud data closest to the pre-set points in the template file from the rail point cloud data of the pre-set measurement points according to the spatial coordinates of the pre-set measurement points in the template file, and obtaining the point cloud set corresponding to each pre-set point; wherein, the pre-set measurement points are the points where the horizontal tangent is tangent to the rail head, denoted as 0° points; and the points where the vertical tangent is tangent to the force-bearing side of the rail head, denoted as 90° points; performing linear fitting based on the point cloud set corresponding to each pre-set point; and using the vertical distance from the pre-set measurement point in the template file to the corresponding fitted line as the wear value of that point, obtaining the wear value w_0 for the 0° point and the wear value w_90 for the 90° point; The actual wear plane is obtained by fitting a straight line to the contact surface of the third rail. Feature fitting is performed on the non-contact surface data of the third rail edge to obtain the center point of the third rail. The wear value of the third rail is calculated based on the distance between the fitted center point and the actual wear plane, as well as the theoretical values of the non-contact and contact surfaces. This includes: performing straight line detection on the contact surface of the third rail to obtain the corresponding straight line equation L, and locating the actual wear plane of the third rail; performing circle fitting based on the arc characteristics of the third rail edge region to obtain two fitted circle equations C_l and C_r, and the center coordinates of the two fitted circles (c_l_cent(x, y, z), c_r_cent(x, y, z)); and obtaining the coordinates of the center point t_cent(x', y', z') of the third rail based on the center coordinates of the fitted circles, where x' = t_cent((c_l_cent.x + c_r_cent.x) * 0.5, y' = (c_l_cent.y + c_r_cent.y) * 0.
5. z'=(c_l_cent.z+c_r_cent.z)*0.5); The wear value of the third rail is railcost = Ht- Hw, where Hw is the distance between the fitted circle center and the straight line equation; Ht is the theoretical design value between the center point of the arc on both sides of the third rail and the contact surface of the third rail under the corresponding model. The nearest rail is determined based on the point cloud data of the third rail, and the third rail pull-out value and the third rail guide height value are obtained based on the relative positional relationship between the center point of the third rail and the surface of the nearest rail, the distance between the center point of the third rail and the actual wear plane, and the wear value of the corresponding point of the rail.
2. The non-contact method for real-time detection of track wear and third track geometric parameters according to claim 1, characterized in that, The device for acquiring images of the railheads of the first rail (both the stressed and unstressed sides), the second rail (both the stressed and unstressed sides), and the contact surface of the third rail includes six line-scan cameras located in the same vertical area. These line-scan cameras are simultaneously triggered by a mileage encoder. The six line-scan cameras in the same vertical area are sequentially numbered as Camera 1, Camera 2, Camera 3, Camera 4, Camera 5, and Camera 6. Camera 1 and / or Camera 6 are used to acquire contact surface data of the third rail; Camera 2 is used to acquire the railhead image of the unstressed side of the first rail; Camera 3 is used to acquire the railhead image of the stressed side of the first rail; Camera 4 is used to acquire the railhead image of the stressed side of the second rail; and Camera 5 is used to acquire the railhead image of the unstressed side of the second rail.
3. The non-contact method for real-time detection of track wear and third track geometric parameters according to claim 1, characterized in that, The image preprocessing includes image filtering, edge detection, and / or contour extraction.
4. The non-contact method for real-time detection of track wear and third track geometric parameters according to claim 2, characterized in that, The process includes calibrating the acquisition device and converting the preprocessed image data into 3D point cloud data; and normalizing all point cloud data according to the rotation and translation matrix obtained from the calibration. Each camera is calibrated to obtain and store the corresponding rotation and translation matrix for each camera. The rail image data is converted into 3D point cloud data. The rotation and translation matrices of the corresponding cameras are read according to the camera numbers. The rail head data of the stressed and unstressed sides of the rail are transformed into the same coordinate system according to the corresponding rotation and translation matrices and then fused to obtain complete rail head point cloud data. The third track image data is converted into 3D point cloud data. The corresponding rotation and translation matrix is read according to the camera number. The third track point cloud data is then converted into the same coordinate system as the rail point cloud data through the corresponding rotation and translation matrix for normalization.
5. The non-contact method for real-time detection of track wear and third track geometric parameters according to claim 1, characterized in that, The method for matching and registering the rail head point cloud data of the unloaded side of the first and second rails with the standard rail profile in the template file is the Iterative Closest Point (ICP) Registration Algorithm.
6. The non-contact method for real-time detection of track wear and third track geometric parameters according to claim 1, characterized in that, The step of determining the nearest rail based on the third rail point cloud data, and obtaining the third rail pull-out value and third rail guide height value based on the relative positional relationship between the center point of the third rail and the rail surface of the nearest rail, as well as the distance between the center point of the third rail and the actual wear plane and the wear value of the corresponding point on the rail, includes: The initial pull-out value and lead-up value of the third rail are obtained by calculating the coordinate difference between the center point coordinates of the third rail and the 0° wear point and 90° wear point in the template file of the rail closest to it. Based on the wear values at the 0° and 90° points of the rail, and the distance HW between the center point of the third rail and the actual wear plane, compensation calculations are performed on the initial pull-out value and leadhight of the third rail to obtain the specific geometric inspection results of the third rail: Pullvalue' = pullvalue + w_90, The leadhigh value is defined as leadhight' = leadhight + w_0 + HW.
7. A non-contact real-time detection system for track wear and third track geometric parameters, characterized in that, include, The image data acquisition module is used to acquire images of the rail head on the stressed and unstressed sides of the first rail, images of the rail head on the stressed and unstressed sides of the second rail, and images of the contact surface of the third rail. The image data preprocessing module is used to preprocess the acquired images; The 3D point cloud conversion module is used to calibrate the image data acquisition module and convert the preprocessed image data into 3D point cloud data; and to normalize the coordinate system of all point cloud data according to the rotation and translation matrix obtained from the calibration. The rail point cloud data registration and fusion module is used to match and register the rail head point cloud data of the unloaded side of the first and second rails with the standard rail contour in the template file to obtain the corresponding registration rotation matrix. Based on the obtained registration rotation matrix, the coordinate transformation is performed on the rail head data of the loaded side of the first and second rails, and all the registered rail head point cloud data are fused to obtain complete rail head point cloud data. The rail wear value acquisition module is used to calculate the spatial distance between the pre-set points of the standard rail profile in the template file and the corresponding points of the actual collected rail, based on the fused rail head point cloud data, and obtain the wear value of the corresponding point. Specifically, it includes extracting several point cloud data points that are closest to the pre-set points in the rail point cloud data of the pre-set measurement points in the template file according to the spatial coordinates of the pre-set measurement points in the template file, and obtaining the point cloud set corresponding to each pre-set point. Among them, the pre-set measurement points are the points where the horizontal tangent is tangent to the rail head, denoted as 0° points; and the points where the vertical tangent is tangent to the stress side of the rail head, denoted as 90° points. A straight line is fitted according to the point cloud set corresponding to each pre-set point. The vertical distance from the pre-set measurement point in the template file to the corresponding fitted straight line is used as the wear value of that point, and the wear value of the 0° point is w_0, and the wear value of the 90° point is w_90. The third rail wear value acquisition module is used to obtain the actual wear plane by performing straight line fitting on the contact surface of the third rail, and to obtain the center point of the third rail by performing feature fitting on the non-contact surface data of the third rail edge. Based on the distance between the fitted center point of the third rail and the actual wear plane, as well as the theoretical values of the non-contact and contact surfaces of the third rail, the third rail wear value is calculated. Specifically, this includes performing straight line detection on the contact surface of the third rail to obtain the corresponding straight line equation L, and locating the actual wear plane of the third rail; performing circle fitting based on the arc features presented by the edge region of the third rail to obtain two fitted circle equations C_l and C_r, and the center coordinates of the two fitted circles (c_l_cent(x, y, z), c_r_cent(x, y, z)); and obtaining the coordinates of the center point of the third rail t_cent(x', y', z') based on the center coordinates of the fitted circles, where x'=t_cent((c_l_cent.x+c_r_cent.x)*0.5, y'=(c_l_cent.y+c_r_cent.y)*0.
5. z'=(c_l_cent.z+c_r_cent.z)*0.5); The wear value of the third rail is railcost = Ht- Hw, where Hw is the distance between the fitted circle center and the straight line equation; Ht is the theoretical design value between the center point of the arc on both sides of the third rail and the contact surface of the third rail under the corresponding model. The module for obtaining the third rail pull-out value and guide height value is used to determine the nearest rail based on the third rail point cloud data, and obtain the point wear value of the corresponding rail and the distance between the center point of the third rail and the actual wear plane. The module obtains the third rail pull-out value and the third rail guide height value based on the relative positional relationship between the center point of the third rail and the rail surface.