A catenary line extraction method based on point cloud

Abstract

A kind of overhead line system line extraction method based on point cloud, mainly include: obtaining the point cloud scanned in railway field, pre-processing;According to the fluctuation of overhead line's elevation difference, intercept the point cloud of the area where overhead line is located;From the point cloud that is intercepted, using the relative geometric features of overhead line and the locator in overhead line, extract the locator in overhead line, and remove the locator from point cloud data;From the obtained point cloud without locator, the intersection of dropper and contact line is extracted using geometric features;From the previously obtained point cloud without locator, the intersection of extracted contact line and dropper is traversed in sequence along horizontal direction coordinate, detect whether the point cloud between two intersection points conforms to the line feature, if so, then consider that the two points are on the same contact line;Connect each intersection point belonging to the same contact line, determine the contact line.The method can realize the automatic line identification of contact line at turnout in overhead line point cloud, improve the efficiency of overhead line detection and maintenance.

Description

Technical Field

[0001] This invention relates to the field of intelligent detection technology for rail transit, and in particular to a method for extracting overhead contact lines based on point clouds. Background Technology

[0002] my country's railways are primarily electrified, powered by high-voltage generators. This entire system is known as the traction power supply system, and the overhead contact line is a crucial component of this system. The inspection and maintenance of the overhead contact line is a vital prerequisite for the safe operation of trains.

[0003] The earliest inspection method relied on inspectors climbing onto the overhead contact line maintenance ladder and using a plumb line (or measuring rod) to perform point-by-point contact measurements of the contact wire. This method was inefficient and prone to safety accidents. With the development and application of new sensors such as high-speed linear CCD cameras, lidar, and gratings in non-contact measurement methods, the advantages of non-contact methods—minimal impact on the contact wire and high measurement accuracy—have become more prominent, gradually replacing traditional contact measurement methods.

[0004] Currently, routine maintenance and measurement of railway overhead contact lines during maintenance windows in China mostly rely on ground surveyors using portable DJJ.8 contact line laser measuring instruments for manual inspection and calculation. This method requires significant manpower, and the accuracy resulting from manual inspection is relatively low. Inconsistent professional qualifications among inspectors can easily lead to data omissions and calculation errors. Furthermore, this measurement method requires long maintenance windows, which can place a burden on the busy railway network.

[0005] To improve the speed and accuracy of overhead contact line parameter measurement, both domestic and international approaches favor using scanners or lidar in conjunction with point cloud algorithms. Currently, the calculation of contact line parameters often involves dividing multiple track lines into individual track lines for identification and calculation. While it's not necessary to confirm the correspondence between rails and contact lines for a single track line, when multiple tracks appear in the point cloud, especially near railway switches where lines intersect, current algorithms cannot confirm the correspondence between rails and contact lines, thus making it impossible to calculate the geometric features of corresponding points. Summary of the Invention

[0006] This disclosure provides a point cloud-based method for extracting overhead contact lines. The method is designed based on point cloud data of multiple tracks in the field and existing overhead contact line identification algorithms for single-track railways. It can identify the contact lines at turnouts in the overhead contact line point cloud.

[0007] The point cloud-based contact network line extraction method disclosed herein mainly includes the following steps:

[0008] A method for extracting overhead contact lines based on point clouds includes the following steps:

[0009] S1, acquire the point cloud of the railway site scan, and preprocess the point cloud data;

[0010] S2, based on the elevation difference fluctuation of the overhead contact line, extract the point cloud of the area where the overhead contact line is located;

[0011] S3. From the point cloud extracted in step S2, the locators in the contact wire are extracted using the relative geometric features of the contact wire and the locators in the contact wire, and the locators are removed from the point cloud data.

[0012] S4. From the point cloud without a locator obtained in step S3, the intersection of the dropper and the contact line is extracted using geometric features.

[0013] S5. From the point cloud without a locator obtained in step S3, traverse the intersection points of the extracted contact line and the dropper along the horizontal coordinate, and detect whether the point cloud between the two intersection points conforms to the connection feature. If so, it is considered that the two points are on the same contact line; otherwise, it is considered that they are on different contact lines.

[0014] S6. Connect the intersections that belong to the same contact line to determine the contact line.

[0015] Furthermore, step S1 includes:

[0016] S11: Acquire point cloud data from the railway site scan, and segment the point cloud data according to the point cloud density, retaining the higher density portion;

[0017] S12 filters the segmented point cloud to remove noise that could affect subsequent calculations.

[0018] Furthermore, step S2 specifically includes:

[0019] Based on the ground location and elevation coordinates, the height range of the overhead contact line can be calculated.

[0020] The area where the overhead contact line is located is extracted using a direct-pass filtering method.

[0021] Furthermore, in step S3, the method for extracting the locator from the overhead contact line specifically includes:

[0022] For any point X in the point cloud extracted from step S2 i :

[0023] Find point X i For all points within a radius R, calculate their mean and sample variance.

[0024] Calculate the covariance matrix using the sample variance;

[0025] Eigenvalue decomposition is performed on the covariance matrix to obtain eigenvalues ​​and eigenvectors;

[0026] The planar features of the point cloud are calculated using the following formula:

[0027]

[0028] Where: P λ Planar features representing points;

[0029] λ1 represents the eigenvalue in the X-axis direction;

[0030] λ2 represents the eigenvalue in the Y-axis direction;

[0031] λ3 represents the eigenvalue in the Z-axis direction;

[0032] According to P λ If a certain threshold is met, a set of point clouds with planar characteristics is found, and a locator is obtained.

[0033] Furthermore, the specific method of step S4 includes:

[0034] S41, In the point cloud without a locator obtained in step S3, all points in the point cloud are traversed sequentially, and the linear feature value of the point cloud within a radius R1 around each point is calculated. When the feature value meets a certain threshold, the point is considered to have linear characteristics. In this way, points with linear characteristics in the point cloud are found and removed.

[0035] S42, in the point cloud obtained in step S41, all points in the point cloud are traversed in turn, and the spherical feature value of the point cloud within the radius R2 around each point is calculated. When the feature value meets a certain threshold, the point is considered to conform to the spherical characteristics. In this way, the points in the point cloud that conform to the spherical characteristics are found, which are the intersections of the contact line and the suspension wire.

[0036] Furthermore, step S41 specifically includes:

[0037] For any point X in the point cloud obtained in step S3 i :

[0038] Find point X i For all points within a radius R1, calculate their mean and sample variance;

[0039] And calculate the covariance matrix using the sample variance;

[0040] Eigenvalue decomposition is performed on the covariance matrix to obtain eigenvalues ​​and eigenvectors;

[0041] Calculate point X i Linear eigenvalues:

[0042]

[0043] Where: L λ λ1 represents the linear characteristics of a point; λ2 represents the eigenvalue in the X-axis direction; λ3 represents the eigenvalue in the Y-axis direction.

[0044] According to L λ Whether a certain threshold is met determines whether the point has linear characteristics;

[0045] Therefore, all points with linear properties in the point cloud are identified and removed.

[0046] Furthermore, step S42 specifically includes:

[0047] For any point X in the point cloud obtained in step S41 i :

[0048] Find point X i For all points within a radius of R2, calculate their mean and sample variance;

[0049] And calculate the covariance matrix using the sample variance;

[0050] Eigenvalue decomposition is performed on the covariance matrix to obtain eigenvalues ​​and eigenvectors;

[0051] Calculate point X i Sphere eigenvalues:

[0052]

[0053] Wherein: S λ λ1 represents the spherical characteristics of a point; λ3 represents the eigenvalue in the X-axis direction; λ4 represents the eigenvalue in the Z-axis direction.

[0054] According to S λ Whether a certain threshold is met determines whether the point has linear characteristics. By doing so, we can find the point with spherical characteristics, which is the intersection of the contact line and the dropper.

[0055] Furthermore, let R1 = 0.2 and R2 = 0.06.

[0056] Furthermore, in step S5, the RANSAC algorithm is used to detect whether the point cloud between two intersection points has a connecting feature, specifically including:

[0057] A cylindrical region with the line connecting the two points as its axis is cut out using a pass-through filter;

[0058] RANSAC is used to detect whether the point cloud in this cylindrical region has linear characteristics.

[0059] Compared with the prior art, the beneficial effects of this disclosure are: (1) it realizes automatic extraction and branch matching of contact wires in the contact network point cloud; (2) it realizes automatic extraction of locators in the contact network point cloud; and (3) it improves the efficiency of automatic detection and maintenance of contact network. Attached Figure Description

[0060] The above and other objects, features and advantages of this disclosure will become more apparent from the more detailed description of exemplary embodiments of this disclosure taken in conjunction with the accompanying drawings, in which the same reference numerals generally represent the same components.

[0061] Figure 1 Example of a contact network point cloud;

[0062] Figure 2 Example of locator location;

[0063] Figure 3 The section where the contact wire and the suspension wire intersect;

[0064] Figure 4 This is a flowchart illustrating an exemplary embodiment of the present disclosure. Detailed Implementation

[0065] Preferred embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While preferred embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that the present disclosure will be thorough and complete, and will fully convey the scope of the present disclosure to those skilled in the art.

[0066] This disclosure provides a point cloud-based method for extracting overhead contact lines, used for extracting locators in the overhead contact system and automatically extracting and matching contact lines from point clouds.

[0067] The flowchart of the exemplary embodiment is attached. Figure 4 As shown, the main steps include:

[0068] 1. Read the point cloud data scanned from the railway site, and preprocess the point cloud to obtain a point cloud set of the rails and overhead contact line. This includes:

[0069] Acquire point cloud data from railway site scanning, segment the point cloud data according to point cloud density, and retain the higher density portion;

[0070] 2. Filter the segmented point cloud to remove noise that could affect subsequent calculations;

[0071] Filtering methods can be chosen from radius filtering and statistical filtering, among which:

[0072] The basic idea of ​​radius filtering is to assume that each laser point in the original point cloud contains at least a certain number of nearest neighbors in a specified radius neighborhood. Laser points in the original point cloud that meet the assumption are considered normal points and retained; otherwise, they are considered noise points and removed.

[0073] Statistical filtering is based on calculating the distribution of distances from a point to its neighbors in the input data. For each point, the average distance to all its neighbors is calculated. If the result conforms to a Gaussian distribution, the shape of which is determined by the mean and standard deviation, points whose average distance is outside the standard range are considered outliers. The standard range is defined by the global distance mean and variance.

[0074] 3. The ground (track) and contact network point cloud are separated by elevation difference fluctuations;

[0075] The point cloud data collected for track inspection includes trees, traffic lights, mountains, etc., with the roadbed area containing the track making up the vast majority of the point cloud. The height of the overhead contact line is usually a fixed value; therefore, the height range of the contact line can be calculated based on the ground location's height coordinates, and the area containing the contact line can be extracted using a direct-pass filtering method. (Appendix) Figure 1 This represents an exemplary contact network point cloud segmented using the elevation difference fluctuation method.

[0076] 4. From the area of ​​the overhead contact line extracted in step 3, utilizing the relative geometric characteristic of the planarity of the locator points, extract the locators from the overhead contact line. After marking the locators, separate them from the overhead contact line point cloud. The separated locators are shown below. Figure 2 The white portion is shown.

[0077] The specific steps include:

[0078] (1) Find point X i For all points X within a radius R, calculate the mean and the sample variance.

[0079] (2) Calculate the covariance matrix using the sample variance;

[0080] (3) Perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and eigenvectors;

[0081] (4) Calculate the set of point clouds with planarity using the following formula.

[0082]

[0083] Where: P λ Representing point X i Planar features;

[0084] λ1 represents the eigenvalue in the X-axis direction;

[0085] λ2 represents the eigenvalue in the Y-axis direction;

[0086] λ3 represents the eigenvalue in the Z-axis direction;

[0087] Based on a certain threshold, a set of point clouds that conforms to planar characteristics is extracted to obtain a locator.

[0088] 5. From the point cloud data without a positioner extracted in step 4, extract the intersection points of the dropper and the contact line using geometric features:

[0089] Taking advantage of the clear linear characteristics of the dropper and contact line regions in the point cloud, all points in the point cloud without a locator are traversed, and the linear characteristics of the point cloud within a certain radius R1 around each point are calculated. The radius R1 is preferably set to 0.2m. At this point, the regions belonging to the dropper and contact line (excluding the intersection of the two) exhibit linear characteristics. Therefore, by setting an appropriate threshold, points with linear characteristics are identified, which are the points belonging to the contact line and dropper respectively. These points are then marked and separated from the point cloud.

[0090] From the point cloud obtained in the previous step, taking advantage of the strong spherical characteristics of the intersection of the suspension wire and the contact wire, we traverse all the points in the point cloud and calculate the spherical characteristics within a certain radius R2 around each point. The preferred value of R2 is 0.06m. By setting an appropriate threshold, we find the points with spherical characteristics, which are the intersections of the suspension wire and the contact wire.

[0091] Appendix Figure 3 The gray area represents the intersection of the contact wire and the suspension wire.

[0092] The specific algorithms include:

[0093] Find point X i For all points X within a radius of R1 or R2, calculate their mean and sample variance;

[0094] Calculate the covariance matrix using the sample variance;

[0095] Eigenvalue decomposition is performed on the covariance matrix to obtain eigenvalues ​​and eigenvectors;

[0096] Calculate X i Linear characteristics:

[0097]

[0098] Where: L λ Planar features representing points;

[0099] When L λ If a point meets a certain threshold, it is considered to have linear characteristics.

[0100] Calculate the sphericity characteristic:

[0101]

[0102] Wherein: S λ This indicates the spherical nature of a point;

[0103] When S λ If a point meets a certain threshold, it is considered to have spherical characteristics.

[0104] 6. Based on the point cloud data without a locator extracted in step 4, for the intersection points of the extracted contact line and suspension line, traverse each intersection point in order of increasing X coordinate, and check whether there is a point cloud with linear characteristics between two adjacent intersection points. If so, the two intersection points are considered to be on the same contact line; otherwise, they are considered to belong to different contact lines.

[0105] One method for detecting whether a linear point cloud exists between two points is the RANSAC algorithm:

[0106] A cylindrical region with the line connecting the two points as its axis is cut out using a pass-through filter;

[0107] RANSAC was used to extract and record point clouds with linear features from the point cloud in this region.

[0108] 7. Connect the intersections that belong to the same contact line to determine the contact line.

[0109] Output the labeled point cloud, as well as the position of each contact line and the point where it intersects with the dropper and the contact line to which it belongs.

[0110] The above technical solutions are merely exemplary embodiments of the present invention. For those skilled in the art, based on the application methods and principles disclosed in the present invention, it is easy to make various types of improvements or modifications, and not limited to the methods described in the specific embodiments of the present invention. Therefore, the methods described above are merely preferred and not restrictive.

Claims

1. A method for extracting overhead contact lines based on point clouds, comprising the following steps: S1, acquire the point cloud of the railway site scan, and preprocess the point cloud data; S2, based on the elevation difference fluctuation of the overhead contact line, extract the point cloud of the area where the overhead contact line is located; S3. From the point cloud extracted in step S2, the locators in the contact wire are extracted using the relative geometric features of the contact wire and the locators in the contact wire, and the locators are removed from the point cloud data. S4. From the point cloud without a locator obtained in step S3, the intersection of the dropper and the contact line is extracted using geometric features. S5. From the point cloud without a locator obtained in step S3, traverse the intersection points of the extracted contact line and the dropper along the horizontal coordinate, and detect whether the point cloud between the two intersection points conforms to the connection feature. If so, it is considered that the two points are on the same contact line; otherwise, it is considered that they are on different contact lines. S6. Connect the intersections that belong to the same contact line to determine the contact line. In step S3, the method for extracting the locator in the overhead contact line specifically includes: For any point in the point cloud extracted from step S2 : Find the point For all points within a radius R, calculate their mean and sample variance. Calculate the covariance matrix using the sample variance; Eigenvalue decomposition is performed on the covariance matrix to obtain eigenvalues ​​and eigenvectors; The planar features of the point cloud are calculated using the following formula: ； in: Planar features representing points; The eigenvalues ​​represent the features along the X-axis. Characteristic values ​​representing the Y-axis direction; The eigenvalues ​​represent the features along the Z-axis. according to If a certain threshold is met, find the set of point clouds with planar characteristics to obtain the locator; The specific method of step S4 includes: S41, In the point cloud without a locator obtained in step S3, all points in the point cloud are traversed sequentially, and the linear feature value of the point cloud within a radius R1 around each point is calculated. When the feature value meets a certain threshold, the point is considered to have linear characteristics. In this way, points with linear characteristics in the point cloud are found and removed. S42, In the point cloud obtained in step S41, all points in the point cloud are traversed in turn, and the spherical feature value of the point cloud within the radius R2 around each point is calculated. When the feature value meets a certain threshold, the point is considered to conform to the spherical characteristics. In this way, the points in the point cloud that conform to the spherical characteristics are found, which are the intersection points of the contact line and the suspension wire. In step S5, the RANSAC algorithm is used to detect whether the point cloud between two intersection points has a connecting line feature, specifically including: A cylindrical region with the line connecting the two points as its axis is cut out using a pass-through filter; RANSAC is used to detect whether the point cloud in this cylindrical region has linear characteristics.

2. The method according to claim 1, characterized in that, Step S1 includes: S11: Acquire point cloud data from the railway site scan, and segment the point cloud data according to the point cloud density, retaining the higher density portion; S12 filters the segmented point cloud to remove noise that could affect subsequent calculations.

3. The method according to claim 1, characterized in that, Step S2 specifically includes: Based on the ground location and elevation coordinates, the height range of the overhead contact line can be calculated. The area where the overhead contact line is located is extracted using a direct-pass filtering method.

4. The method according to claim 1, characterized in that, Step S41 specifically includes: For any point in the point cloud obtained in step S3 : Find the point For all points within a radius R1, calculate their mean and sample variance; And calculate the covariance matrix using the sample variance; Eigenvalue decomposition is performed on the covariance matrix to obtain eigenvalues ​​and eigenvectors; Calculate this point Linear eigenvalues: ； in: Represents the linear characteristics of points; The eigenvalues ​​represent the features along the X-axis. Characteristic values ​​representing the Y-axis direction; according to Whether a certain threshold is met determines whether the point has linear characteristics; Therefore, all points with linear properties in the point cloud are identified and removed.

5. The method according to claim 1 or 4, characterized in that, Step S42 specifically includes: For any point in the point cloud obtained from step S41 : Find the point For all points within a radius of R2, calculate their mean and sample variance; And calculate the covariance matrix using the sample variance; Eigenvalue decomposition is performed on the covariance matrix to obtain eigenvalues ​​and eigenvectors; Calculate this point Sphere eigenvalues: ； in: The sphere's characteristics represent points; The eigenvalues ​​represent the features along the X-axis. The eigenvalues ​​represent the features along the Z-axis. according to If a certain threshold is met, it is determined whether the point has spherical characteristics. Thus, the point with spherical characteristics is found, which is the intersection of the contact line and the dropper.

6. The method according to claim 1, characterized in that, Let R1 = 0.2 and R2 = 0.06.

Images