Contact line extraction method for electrified railway based on geometric features

Abstract

The present invention proposes a method for extracting contact lines of electrified railways based on geometric features, comprising: obtaining contact network point cloud data and selecting two seed points on the target contact line; determining a starting point and a search direction, constructing a cylindrical search area, extracting contact line points within the search area, performing straight line fitting using the contact line points, updating the starting point and the search direction, and iterating the search until all points on the current single contact line are obtained, thereby completing the first extraction of the single contact line; performing principal component analysis and clustering, and performing straight line fitting according to the clustering results; updating the starting point and the search direction using the straight line direction obtained by fitting and the projection point of the current single contact line in the straight line direction; merging the single contact line extraction results based on the number of intersection points between different contact lines to obtain the final contact line extraction result. The present application obtains all points on the current contact line through cyclic iterative search and clustering, thereby solving the difficult problems of contact line extraction and individualization in the point cloud.

Description

Technical Field

[0001] The present invention relates to the technical field of surveying and mapping existing railway lines, and in particular to a method for extracting contact lines of electrified railways based on geometric features. Background Art

[0002] As the core infrastructure of electrified railways, the catenary system is crucial for providing stable power to electric locomotives. Its performance directly impacts the safety and reliability of train operations. As a high-voltage transmission network installed along the tracks, the catenary system consists of contact suspensions, support devices, and positioning systems. It transmits power through dynamic contact with the pantograph. During railway operations, the catenary system is subjected to various external forces and environmental corrosion, which can lead to unreasonable deviations in key geometric parameters (such as conductor height and pullout) or damage. Therefore, regular catenary inspection is crucial to ensuring safe train operation, and this inspection requires a large amount of accurate catenary data.

[0003] Currently, catenary inspection data primarily consists of three categories: traditional contact measurement data, two-dimensional image data, and point cloud data. Contact measurement data is acquired through direct contact with the catenary, but is inefficient and requires manual on-site operation, posing a high safety risk. Two-dimensional images are significantly affected by the imaging environment, making data quality difficult to guarantee. Three-dimensional laser point cloud technology, through active scanning, can capture rich railway information, but the complex topological structure and dense spatial arrangement of the catenary system make it difficult to extract the catenary from the point cloud.

[0004] To address the problems of difficulty in contact network extraction and low extraction accuracy, the present invention proposes a method for extracting electrified railway contact networks based on geometric features, which can quickly extract contact lines from railway point cloud data and provide basic data support for subsequent contactless existing line mapping and railway contact network detection. Summary of the Invention

[0005] The object of the present invention is to solve at least one of the technical drawbacks.

[0006] To this end, one purpose of the present invention is to propose a method for extracting contact lines of electrified railways based on geometric features. In order to solve the problems of difficult contact network extraction and low extraction accuracy, the contact lines are directly extracted from point cloud data through cluster analysis, principal component analysis and other methods. Only a small amount of manual interaction is required to achieve the extraction of contact lines, which can provide data support for contactless surveying and mapping of existing lines and operation and maintenance of railway electrification.

[0007] To achieve the above-mentioned object, an embodiment of one aspect of the present invention provides a method for extracting contact lines of electrified railways based on geometric features, comprising the following steps:

[0008] S1, obtain the contact network point cloud data and select two seed points on the target contact line;

[0009] S2, using the selected seed point, determine the starting point and initial search direction, build a cylindrical search area based on the starting point and the initial search direction, extract the contact line points in the search area, use the contact line points to perform straight line fitting, update the starting point and search direction based on the obtained straight line direction and the projection point of the contact line on the straight line, iterate the search until all points on the current single contact line are obtained, and complete the first extraction of the single contact line;

[0010] S3, based on the first obtained single contact line, perform principal component analysis and clustering, and perform straight line fitting according to the process in S2 based on the clustering results; update the starting point and search direction in S2 with the fitted straight line direction and the projection point of the current single contact line in the straight line direction; use the steps of S2 to extract the next single contact line;

[0011] S4, based on the number of intersection points between different contact lines, the single contact line extraction results are merged to obtain the final contact line extraction result.

[0012] Further, preferably, in S2, using the selected seed point to determine the starting point and the initial search direction includes:

[0013] S201. Select two initial points of the target contact line and ;

[0014] S202, calculate the initial search direction using the following formula ;

[0015]

[0016] S203, select one of the seed points As a starting point.

[0017] Further, preferably, in S2, a cylindrical search area is constructed based on the starting point and the search direction, contact line points in the search area are extracted, a straight line is fitted using the contact line points, and the starting point and the search direction are updated according to the obtained straight line direction and the projection point of the contact line on the straight line, including:

[0018] S211, define parameterized cylindrical model;

[0019] S212, extracting the point set of all points in the cylindrical search domain from the contact network point cloud data ;

[0020] S213, Point Set Perform RANSAC straight line fitting on the points in the equation to get the optimal straight line ;

[0021] S214, based on the optimal straight line and the initial search direction Update the search direction, calculated as follows:

[0022]

[0023] in, is the current search direction, is the next search direction, The best straight line obtained by fitting The corresponding direction vector;

[0024] S215, for any point , its projection parameters on the line for:

[0025]

[0026] in, The optimal straight line Previous point, determine the starting point of the next stage ;

[0027]

[0028] Where i is any point Serial number.

[0029] Furthermore, preferably, in S211, the parameterized cylindrical model is represented by the following formula:

[0030] ,

[0031] Among them, R 3 Represents a real number in three-dimensional space; is the center point of the cylinder, is the direction of the cylinder axis, is the radius of the cylinder, is the search step size.

[0032] Further, preferably, in S3, performing principal component analysis based on the first obtained single contact line includes:

[0033] S301, segmenting the contact network point cloud data based on the contact lines extracted in S2 and the preset upper and lower buffer zones of the point cloud to obtain a region ROI containing all contact lines;

[0034] S302, based on principal component analysis, the contact lines are roughly classified. , calculate the neighborhood point set The covariance matrix of :

[0035]

[0036] Among them, q represents the neighborhood point, is the neighborhood point mean, ; Calculate the eigenvalues ​​of the covariance matrix , according to the characteristic value , select Line feature metrics for:

[0037] Classify the points according to the preset line feature threshold to obtain the rough classification results:

[0038]

[0039] Furthermore, preferably, in S3, clustering is also performed according to the principal component analysis results.

[0040] Clustering classification results based on Euclidean distance Perform clustering and remove categories with too few points to remove noise.

[0041] Further, preferably, in S4, based on the number of intersection points between different contact lines, the single contact line extraction results are merged to obtain the final contact line extraction result, including: when the number of intersection points between different contact lines exceeds a preset threshold, the single contact line extraction results are merged to obtain the final contact line extraction result.

[0042] Compared with the prior art methods, the present invention has the following advantages and positive effects:

[0043] 1. Based on the geometric structure characteristics of the contact network, the present invention uses principal component analysis, clustering and other methods to achieve rapid extraction of the contact line, which can provide data support for subsequent contact network geometric parameter detection;

[0044] 2. This invention designs a single contact line based on seed points. Based on the manually selected initial seed points, a search area is constructed to extract contact line points. Then, the search point and search direction are updated through linear fitting. All points on the current contact line are obtained through cyclic iteration, solving the problem of contact line extraction and individualization in point clouds.

[0045] 3. The present invention designs a contact line extraction method based on principal component analysis and clustering. Based on the extraction results of a single contact line, it automatically completes the extraction and individualization of all contact lines in the point cloud, improves the efficiency of contact line extraction, and has strong practical application and promotion value.

[0046] Additional aspects and advantages of the present invention will be set forth in part in the description which follows and, in part, will be obvious from the description which follows, or may be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments with reference to the accompanying drawings, in which:

[0048] Figure 1 A flow chart of the method for extracting contact lines of electrified railways based on geometric features provided by the present invention;

[0049] Figure 2 Schematic diagram of contact line extraction of the present invention. DETAILED DESCRIPTION

[0050] The following describes embodiments of the present invention in detail, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and are not to be construed as limiting the present invention.

[0051] like Figure 1 As shown, an embodiment of the present invention provides a method for extracting contact lines of electrified railways based on geometric features, comprising the following steps:

[0052] S1, obtain the contact network point cloud data and select two seed points on the target contact line;

[0053] S2, using the selected seed point, determine the starting point and search direction, build a cylindrical search area based on the starting point and search direction, extract the contact line points in the search area, use the contact line points to perform straight line fitting, update the starting point and search direction based on the obtained straight line direction and the projection point of the contact line on the straight line, iterate the search until all points on the current single contact line are obtained, and complete the first extraction of the single contact line;

[0054] like Figure 2 As shown in the figure, the original point cloud and seed points are used to first construct the search area, i.e., the cylindrical search area. It is judged whether the search is completed. If so, the result is output. If not, RANSAC linear fitting is performed and the starting point and search direction are recalculated, and the search area is constructed again. Specifically:

[0055] In S2, the starting point and search direction are determined using the selected seed point, including:

[0056] S201. Select two initial points of the target contact line and ;

[0057] S202, calculate the initial search direction using the following formula ;

[0058]

[0059] S203, obtain initial search direction And select one of the seed points As a starting point.

[0060] Furthermore, a cylindrical search area is constructed based on the starting point and the search direction, contact line points within the search area are extracted, a straight line is fitted using the contact line points, and the starting point and the search direction are updated according to the obtained straight line direction and the projection point of the contact line on the straight line, including:

[0061] S211, define a parameterized cylindrical model; the parameterized cylindrical model is represented by the following formula:

[0062] ,

[0063] Among them, R 3 Represents a real number in three-dimensional space; is the center point of the cylinder, is the direction of the cylinder axis, is the radius of the cylinder, is the search step size.

[0064] S212, extracting the point set of all points in the cylindrical search domain from the contact network point cloud data ;

[0065] S213, Point Set Perform RANSAC straight line fitting on the points in the equation to get the optimal straight line ;

[0066] S214, based on the optimal straight line and the initial search direction Update the search direction, calculated as follows:

[0067]

[0068] in, is the current search direction, is the next search direction, The best straight line obtained by fitting The corresponding direction vector;

[0069] S215, for any point , and its projection parameters on the line are:

[0070]

[0071] in, The optimal straight line Previous point, determine the starting point of the next stage ;

[0072]

[0073] Indicates taking projection parameters The maximum value in .

[0074] S3, based on the first obtained single contact line, perform principal component analysis and clustering, and perform straight line fitting according to the process in S2 based on the clustering results; update the starting point and search direction in S2 with the fitted straight line direction and the projection point of the current single contact line in the straight line direction; use the steps of S2 to extract the next single contact line;

[0075] like Figure 2 This process requires calling S2. Based on the results output by S2, the ROI area is extracted. Based on the PCA coarse classification, the contact line points are clustered. The clustering results are fitted with RANSAC lines. After fitting, multiple contact lines or small line segments on the contact lines are formed. The search direction and starting point are updated and the search is repeated until all contact lines are clustered. The clustered contact lines are merged and the extraction results are output. The specific process is as follows:

[0076] In S3, based on the first obtained single contact line, principal component analysis is performed including:

[0077] S301, segmenting the contact network point cloud data based on the contact lines extracted in S2 and the preset upper and lower buffer zones of the point cloud to obtain a region ROI containing all contact lines;

[0078] S302, based on principal component analysis, the contact lines are roughly classified. , calculate the neighborhood point set The covariance matrix of :

[0079]

[0080] Among them, q represents the neighborhood point, is the neighborhood point mean, ; Calculate the eigenvalues ​​of the covariance matrix , according to the characteristic value , select Line feature metrics for:

[0081]

[0082] Classify the points according to the preset line feature threshold to obtain the rough classification results:

[0083]

[0084] Furthermore, when clustering is performed based on the principal component analysis results,

[0085] Clustering classification results based on Euclidean distance Perform clustering and remove categories with too few points to remove noise.

[0086] When extracting contact lines based on clustering results, perform linear fitting on each cluster based on the RANSAC method, redefine the seed point and search direction, and then extract a single contact line according to the method in step 2.

[0087] S4, based on the number of intersection points between different contact lines, the single contact line extraction results are merged to obtain the final contact line extraction result.

[0088] In S4, based on the number of intersection points between different contact lines, the single contact line extraction results are merged to obtain the final contact line extraction result, including: when the number of intersection points between different contact lines exceeds a preset threshold, the single contact line extraction results are merged to obtain the final contact line extraction result.

[0089] Throughout this specification, reference to terms such as "one embodiment," "some embodiments," "examples," "specific examples," or "some examples" means that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0090] Although the embodiments of the present invention have been shown and described above, it should be understood that the above embodiments are illustrative and are not to be construed as limiting the present invention. Those skilled in the art may make changes, modifications, substitutions, and variations to the above embodiments without departing from the principles and intent of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for extracting contact lines of electrified railways based on geometric features, characterized in that: The following steps are involved: S1, obtain the contact network point cloud data and select two seed points on the target contact line; S2, using the selected seed point, determines the starting point and initial search direction, constructs a cylindrical search area based on the starting point and the initial search direction, extracts the contact line points in the search area, uses the contact line points to perform straight line fitting, updates the starting point and search direction based on the obtained straight line direction and the projection point of the contact line on the straight line, and iterates the search until all points on the current single contact line are obtained, completing the first extraction of the single contact line; including: S211, define parameterized cylindrical model; S212, extracting the point set of all points in the cylindrical search domain from the contact network point cloud data ; S213, point set Perform RANSAC straight line fitting on the points in the equation to get the optimal straight line ; S214, based on the optimal straight line and the initial search direction Update the search direction, calculated as follows: in, is the current search direction, is the next search direction, The best straight line obtained by fitting The corresponding direction vector; S215, for any point , and its projection parameters on the line are: in, The optimal straight line Previous point, determine the starting point of the next stage ; S3, based on the first obtained single contact line, perform principal component analysis and clustering, and perform straight line fitting according to the process in S2 based on the clustering results; update the starting point and search direction in S2 with the fitted straight line direction and the projection point of the current single contact line in the straight line direction; use the steps of S2 to extract the next single contact line; S4, based on the number of intersection points between different contact lines, the single contact line extraction results are merged to obtain the final contact line extraction result.

2. The method for extracting contact lines of electrified railways based on geometric features according to claim 1, characterized in that: In S2, the starting point and initial search direction are determined using the selected seed point, including: S201, select two seeds of the target contact line and ; S202, calculate the initial search direction using the following formula ; S203, select one of the seed points As a starting point.

3. The method for extracting contact lines of electrified railways based on geometric features according to claim 2, characterized in that: In S211, the parametric cylindrical model is represented by the following formula: Among them, R 3 Represents a real number in three-dimensional space; is the center point of the cylinder, is the direction of the cylinder axis, is the radius of the cylinder, is the search step length.

4. The method for extracting contact lines of electrified railways based on geometric features according to claim 3, characterized in that: In S3, based on the first obtained single contact line, principal component analysis is performed including: S301, segmenting the contact network point cloud data based on the contact lines extracted in S2 and the preset upper and lower buffer zones of the point cloud to obtain a region ROI containing all contact lines; S302, based on principal component analysis, the contact lines are roughly classified. , calculate the neighborhood point set The covariance matrix of : Among them, q represents the neighborhood point, is the neighborhood point mean, ; Calculate the eigenvalues ​​of the covariance matrix , according to the characteristic value , select Line feature metrics for: Classify the points according to the preset line feature threshold to obtain the rough classification results: in, Indicates the preset line feature threshold.

5. The method for extracting contact lines of electrified railways based on geometric features according to claim 4, characterized in that: In S3, clustering is also performed based on the results of principal component analysis. Clustering classification results based on Euclidean distance Perform clustering and remove categories with too few points to remove noise.

6. The method for extracting contact lines of electrified railways based on geometric features according to claim 4, characterized in that: In S4, it includes: when the number of intersection points between different contact lines exceeds a preset threshold, merging the single contact line extraction results to obtain the final contact line extraction result.

Images