Railway catenary line recognition and parameterization method using three-dimensional laser point cloud

Abstract

The application discloses a railway catenary line identification and parameterization method using a three-dimensional laser point cloud, which comprises the following steps: S1, in the original point cloud of a target scene, based on the point cloud density and the height information of the catenary support, the position of the catenary support is determined; S2, based on the position of the catenary support, the point cloud data is classified; S3, taking the single-class railway point cloud data as a basic processing unit, the segmentation and coarse classification are carried out by considering the local geometric shape and topological knowledge, and the coarse-classification single-class catenary line point cloud is obtained; S4, the optimal three-dimensional mathematical parameter equation of the railway catenary line is obtained based on the RANSAC algorithm; S5, the distance of each point to the corresponding railway catenary three-dimensional mathematical parameter equation is calculated, a distance threshold is set, the calculation result is less than the threshold, and the result is classified as the catenary line point cloud, so that the fine classification of the single-class railway catenary line point cloud is completed. The method uses topological knowledge and a suspension line equation, and realizes the efficient and accurate identification of the catenary line.

Description

Technical Field

[0001] This invention relates to the fields of intelligent railway management and three-dimensional computer vision technology, specifically to a method for identifying and parameterizing railway catenary lines using three-dimensional laser point clouds. Background Technology

[0002] With the continuous advancement of railway electrification technology, the overhead contact system, as a core component of the railway electrification system, is responsible for providing electrical energy to electric locomotives. The stability and reliability of the overhead contact system directly affect the safety and efficiency of train operation. Therefore, accurate and rapid inspection, identification, and maintenance of the overhead contact system are particularly important. However, traditional manual inspection methods are limited by the constraints of manual operation, resulting in low inspection efficiency, high workload, and certain safety hazards, especially in complex or harsh environmental conditions where accuracy is even more difficult to guarantee. To solve this problem, utilizing modern technological means to achieve intelligent identification and precise parameterization of the overhead contact system has become a research hotspot.

[0003] In recent years, 3D laser scanning technology, as an advanced spatial data acquisition method, has achieved significant applications in fields such as urban modeling, bridge inspection, and tunnel surveying. 3D laser point cloud technology features high precision, high efficiency, and non-contact operation, enabling the rapid acquisition of 3D spatial information of a target area using laser scanning equipment. For the inspection of railway catenary systems, 3D laser point cloud technology can accurately reconstruct the 3D model of the catenary, automatically extracting the geometric parameters and spatial distribution of components such as the contact wire, catenary cables, and supports, avoiding the cumbersome manual operations and high error rates inherent in traditional methods.

[0004] The extraction of contact wires is one of the core issues in lidar point cloud applications. Since contact wires typically appear as long, thin lines in point cloud data, researchers have proposed various line extraction methods based on point cloud features. For example, curve fitting methods use the least squares method to fit the geometry of the contact wire, effectively extracting its spatial distribution and geometric features. Furthermore, based on clustering analysis and segmentation techniques, combined with the geometric constraints of the contact network, automatic identification and classification of elements such as contact wires, catenary wires, and supports can be achieved. In recent years, the application of deep learning in contact wire extraction has gradually increased, enabling the automatic identification and extraction of contact wires and other related components through model training, greatly improving the automation level and accuracy of extraction. In addition, parametric modeling technology for contact networks based on 3D lidar point cloud data has made significant progress. Through in-depth analysis of point cloud data, researchers can extract the geometric features of the contact wire, such as its height, tilt angle, and tension, while simultaneously identifying the location and structure of important components such as catenary wires, supports, and insulators, providing support for the design, analysis, and maintenance of the contact network. Despite the extensive work done by many scholars on the identification and parameterization of overhead contact lines, the automatic extraction and parameterization of these lines still faces numerous challenges due to the complex working conditions they present, such as bends, crossings, and overlapping layers.

[0005] There are often fixed knowledge structures and patterns between railway overhead contact lines and the different features in their surrounding environment, which can serve as prior knowledge to assist in the automatic identification and classification of overhead contact lines. However, current technologies have not scientifically and rationally utilized these knowledge structures to complete the identification and parameterization of overhead contact lines. Summary of the Invention

[0006] To address the technical shortcomings of existing methods, the purpose of this invention is to provide a method for railway catenary line identification and parameterization using three-dimensional laser point clouds.

[0007] Therefore, the present invention adopts the following technical solution:

[0008] A method for identifying and parameterizing railway catenary lines using 3D laser point clouds, the specific steps of which are as follows:

[0009] S1, In the original point cloud of the target scene, the position of the catenary support is determined based on the point cloud density and the height information of the catenary support.

[0010] S2, the original point cloud data is divided into grades based on the location of the overhead contact line support. The specific operation is as follows:

[0011] S21, based on the direction of the contact wire, determine the railway line to which the location of the contact wire support obtained in S1 belongs;

[0012] S22, establish the topological relationship between all catenary support point clouds belonging to the same railway line to obtain the set of support center points on both sides of the same railway.

[0013] S23, adjacent pillars along the same line are treated as pillar pairs, and the point cloud data is divided into grades according to the pillar pairs;

[0014] S3, taking into account local geometry and topology, uses single-span railway point cloud data as the basic processing unit to segment and coarsely classify the point cloud, resulting in a coarsely classified single-span contact wire point cloud.

[0015] S4. Based on the RANSAC algorithm, the optimal three-dimensional mathematical parameter equations of the railway catenary line corresponding to the point cloud of each coarse-classified single-level catenary line are obtained.

[0016] S5. Based on the three-dimensional mathematical parameter equation of the railway catenary line, calculate the distance from each point in the point cloud of each coarsely classified single-segment catenary line to its corresponding three-dimensional mathematical parameter equation, set a reasonable distance threshold, and if the calculation result is less than the specified threshold, classify it as a point cloud of the catenary line, thereby completing the fine classification of the point cloud of each single-segment railway catenary line.

[0017] The specific operation of step S1 above is as follows:

[0018] Project the original point cloud of the target scene onto the xoy plane along the z-axis, and rasterize the projection area. Count the number of points that fall into each grid cell and use it as the value of that grid cell.

[0019] By traversing the values ​​of each grid cell in the projection area and all grid cells within a certain range, if the value of the current grid cell is greater than the values ​​of all grid cells within the range, it is considered that its value is a local maximum. By obtaining the set of all grid cell locations with local maximum values, the potential catenary support area is obtained.

[0020] Based on the potential catenary support area, the final catenary support position is determined in the original point cloud using the height information of the catenary support.

[0021] The specific operation of step S21 above is as follows:

[0022] Based on the fact that the direction of the overhead contact line is highly consistent with the direction of the railway line, the supports obtained from S1 are grouped, with each group belonging to the same railway line. The specific operation is as follows:

[0023] Using a random sampling consensus algorithm, a straight line is fitted based on the center point of the support pillar. The expression for the fitted line is:

[0024] ,

[0025] in, and These represent the x and y coordinates of the current center point, respectively. , and These are the parameters for fitting the straight line;

[0026] Then, the distance from the center point of each support to the fitted line is calculated. The calculation formula is:

[0027] ,

[0028] in, and Represent any center point The x and y coordinates;

[0029] If the distance from the center point of a support pillar to the fitted line is less than a specified threshold, then the center point of that support pillar is considered to belong to the current fitted line. This process is repeated iteratively until each center point has a corresponding fitted line.

[0030] Based on the slope of the fitted line, if the difference in the angle between adjacent lines is less than a specified threshold, the two lines are considered to belong to the same railway line. Then, the center points of all pillars belonging to the two lines are merged. This process is repeated until the center points of all pillars are merged.

[0031] The specific operation of step S22 above is as follows: based on the x and y coordinates of the support center point, sort all the support center points belonging to the same railway line, first sorting them in ascending order of x coordinates, and then sorting them in ascending order of y coordinates when the x coordinates are equal, thereby establishing the topological relationship between them and obtaining the set of support center points on both sides of the same railway.

[0032] The specific operation of step S23 above is as follows: based on the coordinate information of the center points of adjacent pillars, the point cloud data between each pair of adjacent pillars on the same side is taken as one file. By judging the size of the coordinates of each point in the point cloud data and the center points of all pillar pairs, it is determined which pillar pair it is located in, and then the point cloud data of the railway line is cut into multiple files.

[0033] The specific operation of step S3 above is as follows:

[0034] S31, determine the routing of the contact wire by using the position of the support, and define an identification buffer zone for the contact wire according to its relative position to the track;

[0035] S32, calculate the normal vector of the point cloud within the recognition buffer point by point;

[0036] S33 uses point coordinates and normal vectors as similarity criteria to segment the point cloud data in the recognition buffer, obtaining a series of homogeneous units. Based on the normal vectors of adjacent homogeneous units, Euclidean clustering is used to merge the homogeneous units.

[0037] S34. Utilizing the spatial topological relationship between the contact wire components, the results of merging homogeneous units obtained from clustering are labeled as different structures, thereby obtaining a coarse-classified single-level contact wire point cloud, wherein the structure includes contact wire, dropper, and catenary cable structures.

[0038] The specific operation of step S32 above is as follows: For any point in the point cloud, select its local K-neighborhood point set, then calculate the covariance matrix of the neighborhood points, and then perform eigenvalue decomposition on the covariance matrix. The eigenvector corresponding to the smallest eigenvalue is the normal vector direction of the best fitting plane.

[0039] The specific operation of step S4 above is as follows:

[0040] The three-dimensional mathematical equation of the contact wire can be expressed as follows: its projection onto the horizontal plane in the local coordinate system is a straight line, and its projection onto the vertical plane is a suspension line, where:

[0041] The equation of the line is:

[0042] ,

[0043] in, This represents the angle between the normal vector of the line projected onto the local coordinate system xoy plane and the X-axis. This represents the distance between the origin and the straight line obtained by projecting the contact wire onto the local coordinate system xoy plane;

[0044] The equation of the suspension line is:

[0045] ,

[0046] in, , and The parameters represent the current equation of the plumb line; and This represents the coordinates of a point on the current suspension line. Represents the hyperbolic cosine function;

[0047] The optimal three-dimensional mathematical parameter equations for railway catenary lines are obtained based on the RANSAC algorithm. Specifically, in each iteration, candidate parameter equations are generated by random sampling, the distance from the point cloud to the candidate parameter equations is calculated, local points with a distance less than a set threshold are selected, and the number of local points is recorded. The parameter equation with the most local points is obtained by iterative sampling and calculation.

[0048] The neighborhood mentioned in step S1 above is a region within a 1-meter radius.

[0049] In step S21 above:

[0050] If the distance from the center point of a support pillar to the fitted line is less than 0.1 meters, then the center point of that pillar is considered to belong to the current fitted line; based on the slope of the fitted line, if the angle difference between adjacent lines is less than... If the two straight lines are considered to belong to the same railway line, then they are considered to belong to the same railway line.

[0051] Compared with the prior art, the present invention has the following beneficial effects:

[0052] 1. This invention comprehensively utilizes the linear characteristics of the contact wire itself and the inherent knowledge structure of the surrounding ground features to achieve automatic extraction and classification of contact wire from three-dimensional laser point clouds.

[0053] 2. This invention utilizes the inherent linearity of the contact wire itself and the topological regularity between the components of the railway contact network to achieve automatic identification and classification of railway contact wires that takes into account structural topology knowledge;

[0054] 3. Based on the three-dimensional geometry of the contact wire, this invention designs an optimal estimation method for the three-dimensional mathematical parameter equation of the contact wire based on the suspension line equation. Through iterative optimization, the influence of factors such as point cloud noise on the convergence of the cost function is avoided.

[0055] 4. This invention provides a high-precision, robust, and automated solution for railway catenary wire identification and parameterization, reducing manual intervention and greatly improving the efficiency and accuracy of catenary detection and monitoring. It has significant application value for the maintenance and management of railway electrification systems. Attached Figure Description

[0056] Figure 1 This is a flowchart of the railway catenary wire identification and parameterization method of the present invention;

[0057] Figure 2 This is the contact wire identification result in one embodiment of the present invention. Detailed Implementation

[0058] The technical solution of the invention will be clearly and completely described below with reference to the accompanying drawings.

[0059] See Figure 1 The present invention provides a method for railway catenary line identification and parameterization using three-dimensional laser point clouds, comprising the following technical steps:

[0060] S1, based on point cloud density and height information, locate and identify the overhead contact line support pillars:

[0061] The original point cloud of the target scene is projected onto the xoy plane along the z-axis, and the projection area is rasterized. The number of points falling into each grid cell is counted and used as the value of that grid cell. The original point cloud is a 3D laser point cloud, belonging to a local coordinate system. The x-axis is the direction of the track, the y-axis is horizontal (perpendicular to the track direction), the z-axis is vertical, and the origin is on the center line of the track.

[0062] By traversing the values ​​of each grid cell in the projection area and all grid cells within a certain range (usually set to 1 meter), if the value of the current grid cell is greater than the values ​​of all grid cells within its 1-meter range, it is considered to have a local maximum. By obtaining the set of all grid cell locations with local maximum values, the potential contact wire support area is obtained.

[0063] To ensure the stability and safety of the overhead contact system, the height of the supports is generally required to be high (the supports are the supporting structures of the overhead contact system, supporting the suspension devices of the catenary and contact wire in the air. In point cloud data, supports are generally vertically upward-pointing rod-like structures); based on the potential overhead contact support areas, the final positions of the overhead contact supports are determined in the original point cloud using the height information of the supports.

[0064] S2, the point cloud data is divided into grades based on the location of the overhead contact line support. The specific operation is as follows:

[0065] S21. Based on the direction of the overhead contact line, determine the railway line to which the location of the overhead contact support obtained in S1 belongs.

[0066] The railway overhead contact system is the power source for powering railway trains, and its direction is closely related to the direction and layout of the railway line, the location of stations, and the design of the power supply system. The contact system is generally designed according to the layout and direction of the railway line, and the direction of the contact wires is highly consistent with the direction of the railway line. To ensure train safety, the direction of the railway line generally does not change abruptly. Based on this assumption, the supports obtained from S1 are grouped, with each group belonging to the same railway line. The specific operation is as follows:

[0067] First, the Random Sampling Consensus (RANSAC) algorithm is used to fit a straight line based on the center point of the support pillar. The expression for the fitted line is:

[0068] ,

[0069] in, and These represent the x and y coordinates of the current center point, respectively. , and These are the parameters for fitting the straight line;

[0070] Then, the distance from the center point of each support to the fitted line is calculated. The calculation formula is:

[0071] ,

[0072] in, and Represent any center point The x and y coordinates.

[0073] If the distance from the center point of a support pillar to the fitted line is less than a specified threshold (usually set to 0.1 meters), then the center point of that support pillar is considered to belong to the current fitted line. This fitting process is repeated iteratively until each center point has a corresponding fitted line.

[0074] Based on the slope of the fitted line, if the difference in the angle between adjacent lines is less than a specified threshold (generally...), then... If two straight lines are considered to belong to the same railway line, then the center points of all supports belonging to those two lines are merged. This process is repeated until the center points of all supports are merged.

[0075] S22 establishes the topological relationships between all catenary supports belonging to the same railway line.

[0076] Based on the x and y coordinates of the support center point, all support center points belonging to the same railway line are sorted, first by ascending x coordinate, and then by ascending y coordinate if x coordinates are equal. This establishes the topological relationship between them, resulting in the set of support center points on both sides of the same railway line.

[0077] S23, using adjacent pillars along the same line to classify point cloud data.

[0078] Based on the coordinate information of the center points of adjacent supports, the point cloud data between each pair of adjacent supports on the same side (hereinafter referred to as "support pairs") is divided into one file. By judging the magnitude of the coordinates of each point in the point cloud data and the center points of all support pairs, it is determined which support pair it belongs to, and then the point cloud data of the railway line is divided into multiple files.

[0079] S3, taking into account local geometry and topological knowledge, uses single-track railway point cloud data as the basic processing unit to segment and coarsely classify the point cloud.

[0080] The overhead contact line is typically suspended at a certain height above the track surface, generally requiring a safe distance. Different types of overhead contact lines, such as the catenary, contact wire, and return wire, also exhibit spatial layering. Furthermore, the overhead contact line has a relative positional relationship with the track in the horizontal direction. Generally, the overhead contact line is installed above or slightly offset from the track centerline.

[0081] Generally, the contact wire, droppers, and catenary wire have fixed topological relationships with each other. The operation of the overhead contact system is accomplished through the cooperation of the contact wire, droppers, and catenary wire. The catenary wire supports the droppers, and the droppers support the contact wire; each part plays an interdependent role in the overhead contact system.

[0082] The supports for the overhead contact system are typically installed on one side of the track, maintaining a stable lateral distance from the track centerline. Assuming the track centerline is a spatial curve denoted by L, each support is located on one side of the track and extends parallel to the track centerline. The lateral offset between the support and the track centerline is generally 2.5–3.0 meters.

[0083] Therefore, using single-track railway point cloud data as the basic processing unit, the steps for segmenting and coarsely classifying the point cloud are as follows:

[0084] S31, the routing of the contact wire is determined by the position of the support, and a potential identification buffer is defined for the contact wire based on its relative position to the track.

[0085] Using the set of support center points obtained from S22, the centerline direction of the track is obtained by fitting spline curves. To identify the catenary line in the point cloud data, a potential buffer zone based on the track-support position relationship can be defined for each track. This buffer zone is a three-dimensional strip-shaped space extending along the track direction with a circular or elliptical longitudinal section (z-axis direction), used to indicate the possible area where the catenary line may exist.

[0086] S32, calculate the normal vector of the point cloud within the recognition buffer point by point.

[0087] For any point in the point cloud, select its local K-neighborhood point set, then calculate the covariance matrix of the neighborhood points, and then perform eigenvalue decomposition on the covariance matrix. The eigenvector corresponding to the smallest eigenvalue is the normal vector direction of the best-fit plane.

[0088] S33 uses point coordinates and normal vectors as similarity criteria to segment the point cloud data within the recognition buffer, obtaining a series of homogeneous units. Based on the normal vectors of adjacent homogeneous units, Euclidean clustering is used to merge the homogeneous units.

[0089] S34 utilizes the spatial topological relationships between catenary components to label the merging results (point cloud data) of homogeneous units obtained from clustering as structures such as contact wires, droppers, and catenary cables, thereby obtaining a coarsely classified single-segment catenary line point cloud.

[0090] S4, the optimal estimation of the three-dimensional mathematical parameter equations of railway catenary lines based on the RANSAC algorithm, is as follows:

[0091] Generally, railway contact wires, under tension, form a suspended shape. The three-dimensional mathematical equation of the contact wire can be expressed as follows: its projection onto the horizontal plane in a local coordinate system is a straight line, and its projection onto the vertical plane (xoz) is a suspended line. Wherein:

[0092] The equation of the line is:

[0093] ,

[0094] in, This represents the angle between the normal vector of the line projected onto the local coordinate system xoy plane and the X-axis. This represents the distance between the origin and the straight line obtained by projecting the contact wire line onto the local coordinate system xoy plane.

[0095] The equation of the suspension line is:

[0096] ,

[0097] in, , and The parameters represent the current equation of the plumb line; and This represents the coordinates of a point on the current suspension line. This represents the hyperbolic cosine function.

[0098] By solving for the tangents at any two points on the randomly sampled suspension line, the parameters of the suspension line equation can be calculated. , and However, because the coarse classification point cloud of a single-span contact wire obtained by S34 may contain other ground features, these noise points do not satisfy the suspension line equation of a single-span contact wire.

[0099] Considering the existence of noise points, this invention obtains the optimal three-dimensional mathematical parameter equations for railway catenary lines based on the RANSAC algorithm. Specifically, in each iteration, candidate parameter equations are generated by random sampling, the distance from the point cloud to the candidate parameter equations is calculated, local points with a distance less than a set threshold are selected, and the number of local points is recorded. The parameter equation with the most local points is continuously iterated and sampled as the optimal three-dimensional mathematical parameter equation for railway catenary lines.

[0100] The above operations yield the three-dimensional mathematical parameter equations of the railway catenary line corresponding to the point cloud of each coarse-classified single-level catenary line.

[0101] S5 utilizes the three-dimensional mathematical parametric equations of railway catenary lines to perform fine classification of the point cloud of a single-span railway catenary line, including:

[0102] Based on the three-dimensional mathematical parameter equations of the railway catenary lines corresponding to the point clouds of each coarse-classified single-segment catenary line obtained by S4, the distance from each point to its corresponding three-dimensional mathematical parameter equation is calculated. A reasonable distance threshold is set. If the calculation result is less than the specified threshold, it is classified as a point cloud of the catenary line, thus completing the fine classification of the point clouds of each single-segment railway catenary line.

[0103] Figure 2 The figure shows a visualization result of an embodiment of contact wire point cloud recognition using the method of the present invention. In the figure, the red dots are the identified contact wires, and the white dots are other ground features in the scene.

Claims

1. A method for railway catenary line identification and parameterization using three-dimensional laser point clouds, characterized in that, The specific steps are as follows: S1, In the original point cloud of the target scene, the position of the catenary support is determined based on the point cloud density and the height information of the catenary support. S2, the original point cloud data is divided into grades based on the location of the overhead contact line support. The specific operation is as follows: S21, based on the direction of the contact wire, determine the railway line to which the location of the contact wire support obtained in S1 belongs; S22, establish the topological relationship between all catenary support point clouds belonging to the same railway line to obtain the set of support center points on both sides of the same railway. S23, adjacent pillars along the same line are treated as pillar pairs, and the point cloud data is divided into grades according to the pillar pairs; S3, taking into account local geometry and topology, uses single-span railway point cloud data as the basic processing unit to segment and coarsely classify the point cloud, resulting in a coarsely classified single-span contact wire point cloud. S4. Based on the RANSAC algorithm, the optimal three-dimensional mathematical parameter equations of the railway catenary line corresponding to the point cloud of each coarse-classified single-level catenary line are obtained. S5. Based on the three-dimensional mathematical parameter equation of the railway catenary line, calculate the distance from each point in the point cloud of each coarsely classified single-segment catenary line to its corresponding three-dimensional mathematical parameter equation, set a reasonable distance threshold, and if the calculation result is less than the specified threshold, classify it as a point cloud of the catenary line, thereby completing the fine classification of the point cloud of each single-segment railway catenary line.

2. The method for identifying and parameterizing railway catenary wires according to claim 1, characterized in that, The specific operation of S1 is as follows: Project the original point cloud of the target scene onto the xoy plane along the z-axis, and rasterize the projection area. Count the number of points that fall into each grid cell and use it as the value of that grid cell. By traversing the values ​​of each grid cell in the projection area and all grid cells within a certain range, if the value of the current grid cell is greater than the values ​​of all grid cells within the range, it is considered that its value is a local maximum. By obtaining the set of all grid cell locations with local maximum values, the potential catenary support area is obtained. Based on the potential catenary support area, the final catenary support position is determined in the original point cloud using the height information of the catenary support.

3. The railway catenary wire identification and parameterization method according to claim 2, characterized in that, The specific operation of S21 is as follows: Based on the fact that the direction of the overhead contact line is highly consistent with the direction of the railway line, the supports obtained from S1 are grouped, with each group belonging to the same railway line. The specific operation is as follows: Using a random sampling consensus algorithm, a straight line is fitted based on the center point of the support pillar. The expression for the fitted line is: ， in, and These represent the x and y coordinates of the current center point, respectively. , and These are the parameters for fitting the straight line; Then, the distance from the center point of each support to the fitted line is calculated. The calculation formula is: ， in, and Represent any center point The x and y coordinates; If the distance from the center point of a support pillar to the fitted line is less than a specified threshold, then the center point of that support pillar is considered to belong to the current fitted line. This process is repeated iteratively until each center point has a corresponding fitted line. Based on the slope of the fitted line, if the difference in the angle between adjacent lines is less than a specified threshold, the two lines are considered to belong to the same railway line. Then, the center points of all pillars belonging to the two lines are merged. This process is repeated until the center points of all pillars are merged.

4. The railway catenary wire identification and parameterization method according to claim 3, characterized in that, The specific operation of S22 is as follows: Based on the x and y coordinates of the support center point, sort all the support center points belonging to the same railway line. Prioritize sorting by the x coordinate of the support center point from smallest to largest. If the x coordinates are equal, then sort by the y coordinate from smallest to largest. In this way, establish the topological relationship between them and obtain the set of support center points on both sides of the same railway.

5. The railway catenary wire identification and parameterization method according to claim 4, characterized in that, The specific operation of S23 is as follows: Based on the coordinate information of the center points of adjacent pillars, the point cloud data between each pair of adjacent pillars on the same side is taken as one file. By judging the size of the coordinates of each point in the point cloud data and the center points of all pillar pairs, it is determined which pillar pair it is located in, and then the point cloud data of the railway line is cut into multiple files.

6. The railway catenary wire identification and parameterization method according to claim 5, characterized in that, The specific operation of S3 is as follows: S31, determine the routing of the contact wire by using the position of the support, and define an identification buffer zone for the contact wire according to its relative position to the track; S32, calculate the normal vector of the point cloud within the recognition buffer point by point; S33 uses point coordinates and normal vectors as similarity criteria to segment the point cloud data in the recognition buffer, obtaining a series of homogeneous units. Based on the normal vectors of adjacent homogeneous units, Euclidean clustering is used to merge the homogeneous units. S34. Utilizing the spatial topological relationship between the contact wire components, the results of merging homogeneous units obtained from clustering are labeled as different structures, thereby obtaining a coarse-classified single-level contact wire point cloud, wherein the structure includes contact wire, dropper, and catenary cable structures.

7. The railway catenary wire identification and parameterization method according to claim 6, characterized in that, The specific operation of S32 is as follows: For any point in the point cloud, select its local K-neighborhood point set, then calculate the covariance matrix of the neighborhood points, and then perform eigenvalue decomposition on the covariance matrix. The eigenvector corresponding to the smallest eigenvalue is the normal vector direction of the best-fit plane.

8. The method for identifying and parameterizing railway catenary wires according to claim 7, characterized in that, The specific operation of S4 is as follows: The three-dimensional mathematical equation of the contact wire can be expressed as follows: its projection onto the horizontal plane in the local coordinate system is a straight line, and its projection onto the vertical plane is a suspension line, where: The equation of the line is: ， in, This represents the angle between the normal vector of the line projected onto the local coordinate system xoy plane and the X-axis. This represents the distance between the origin and the straight line obtained by projecting the contact wire onto the local coordinate system xoy plane; The equation of the suspension line is: ， in, , and The parameters represent the current equation of the plumb line; and This represents the coordinates of a point on the current suspension line. Represents the hyperbolic cosine function; The optimal three-dimensional mathematical parameter equations for railway catenary lines are obtained based on the RANSAC algorithm. Specifically, in each iteration, candidate parameter equations are generated by random sampling, the distance from the point cloud to the candidate parameter equations is calculated, local points with a distance less than a set threshold are selected, and the number of local points is recorded. The parameter equation with the most local points is obtained by iterative sampling and calculation.

9. The method for identifying and parameterizing railway catenary wires according to claim 2, characterized in that: The neighborhood mentioned in S1 is a region with a range of 1 meter.

10. The method for identifying and parameterizing railway catenary wires according to claim 3, characterized in that, In S21: If the distance from the center point of a support pillar to the fitted line is less than 0.1 meters, then the center point of that pillar is considered to belong to the current fitted line; based on the slope of the fitted line, if the angle difference between adjacent lines is less than... If the two straight lines are considered to belong to the same railway line, then they are considered to belong to the same railway line.

Images