Methods, devices, electronic equipment and storage media for rail wear detection with full profile registration

Abstract

This application discloses a method, device, electronic equipment, and storage medium for rail wear detection based on full-section profile registration. The method includes the following steps: S1, performing clustering filtering preprocessing on the collected original track profile point cloud data, and then extracting the rail head and rail web region point clouds based on the longitudinal amplitude characteristics of the point cloud combined with the profile curvature characteristics; S2, segmenting the feature arcs of the inner rail web region point cloud to obtain large and small arcs, calculating the initial double centers of the segmented large and small arcs, and fitting the optimal double centers; S3, calculating the inner attitude transformation parameters of the inner profile based on the optimal double center positioning, and calculating the outer attitude transformation parameters of the outer profile of the rail through the inner attitude transformation parameters and system calibration parameters to achieve full-section profile registration; S4, calculating the rail wear value by combining the feature points of the inner and outer sides of the registered measured profile and the standard profile. This application has high detection accuracy, simple calculation, and good real-time detection performance.

Description

Technical Field

[0001] This application relates to the field of rail inspection technology, and in particular, to a method, apparatus, electronic device and storage medium for rail wear detection with full-section profile registration. Background Technology

[0002] As a fast and convenient mode of transportation, the subway has become an indispensable part of public travel. With the continuous increase in subway operating mileage, the safety and reliability of subway tracks have received increasing attention. The frequent and high-frequency operation of trains exacerbates rail wear, directly impacting operational safety. Therefore, real-time and efficient dynamic rail wear detection is of great significance for ensuring subway operational safety and line maintenance.

[0003] Currently, common methods for detecting rail wear in subway systems include static contact testing and dynamic non-contact testing. Static contact testing involves manually measuring rail wear using calipers. This method is susceptible to subjective human error, and the calipers' long-term friction with the rail can cause damage, affecting the accuracy and reliability of the measurement data. Dynamic non-contact testing typically employs a combination of area array cameras and lasers. The principle involves using an area array camera to capture images of the light rays projected onto the test object by the laser. Image processing and coordinate transformation are then used to calculate the actual point cloud coordinates of the rail profile. After profile registration using an algorithm, rail wear is calculated. The testing equipment uses a vehicle as a carrier, moving with the vehicle to achieve dynamic non-contact testing. Compared to static contact testing, dynamic non-contact testing offers advantages such as no contact wear, speed and efficiency, and high precision.

[0004] Existing rail wear detection technologies mainly focus on the extraction and registration of two-dimensional feature point clouds of rail cross sections. Accurate extraction of feature point clouds of rail head and rail web is an important prerequisite for profile registration, but it is easily affected by external interference. Profile registration is the key to ensuring the accuracy of wear detection. Existing technologies for extracting feature point clouds from rail cross-sections typically employ single-condition discrimination based on profile curvature features and regional continuity of concavity and convexity. However, in real-world subway lines with multiple interference factors, these methods are susceptible to interference, prone to misjudgment, and suffer from poor real-time detection. Existing technologies for rail profile registration typically use ICP (Iterative Closest Point) and its improved versions, which are sensitive to initial values ​​and have high computational complexity. Furthermore, existing technologies for calculating rail wear typically calculate wear values ​​based on standard definitions using wear feature points, failing to consider the offset errors between the measured and actual profiles caused by track interference factors (such as oil coatings, lettering, and foreign objects) during dynamic detection. This directly affects the accuracy of wear detection.

[0005] In summary, this invention proposes a wear detection method based on efficient registration of the full-section rail profile, in order to overcome the defects and shortcomings of existing rail wear detection technologies. Summary of the Invention

[0006] This application provides a rail wear detection method based on full-section profile registration to address the technical problems of low detection accuracy, complex calculations, and poor real-time performance of existing rail wear detection technologies.

[0007] This application is achieved through the following solution:

[0008] A method for detecting rail wear with full-section profile registration, comprising the following steps:

[0009] S1. The distance-based clustering filtering algorithm performs clustering and filtering preprocessing on the collected original track profile point cloud data. Then, based on the longitudinal amplitude characteristics of the point cloud combined with the profile curvature characteristics, the point cloud of the rail head and rail web regions is accurately extracted.

[0010] S2. The point cloud feature arcs of the inner rail web region of the rail profile are segmented to obtain large and small arcs. The initial double center is calculated by combining the perpendicular bisector theorem of the circle with the properties of similar triangles. The optimal double center of the segmented large and small arcs is obtained by fitting the initial double center using a nonlinear least squares algorithm based on radius constraints.

[0011] S3. Calculate the inner attitude transformation parameters of the inner profile of the rail based on the optimal double-center positioning obtained by fitting. Calculate the outer attitude transformation parameters of the outer profile of the rail by combining the inner attitude transformation parameters with the system calibration parameters, thereby achieving full-section profile registration between the measured profile and the standard profile.

[0012] S4. Calculate the rail wear value by combining the characteristic points on the inner and outer sides of the rail profile after profile registration with the measured profile and the standard profile.

[0013] Furthermore, before collecting the original orbital profile point cloud data, the following steps are also included:

[0014] S0. The system parameters of the rail wear detection device, consisting of four line laser sensors mounted on the detection beam, are calibrated. This involves transforming each line laser from its original coordinate system to a unified coordinate system, yielding the system calibration parameters, including the transformation rotation angle θ between the original and unified coordinate systems. The origin of a coordinate system is in the X and Z directions. The relative offset Δ x Δz, where the X direction is the rail. The cross-section is in the horizontal direction, and the Z direction is the vertical direction of the rail cross-section height.

[0015] Further, step S1 specifically includes the following steps:

[0016] S11. Based on the distribution characteristics of the measured original profile point cloud, where normal data is dense and outlier interference data is sparse, a distance-based clustering algorithm is used to segment the point cloud to obtain the segmented set S = {P}. i ={(x j , z j )}} i，j∈N Where i represents the index of the segmented point set, and j represents the index of a point in a single point set. The search starts from the first point and stores it in the point set P. i If the distance between the current point and the next point is less than or equal to a set threshold T1, then the next point is stored in the point set P. i If the value is greater than the set threshold T1, then the next point is stored in a new point set P. i+1 And so on, stopping the search once all points have been traversed;

[0017] S12. Set a threshold T2 and calculate the distance between the tail point of the previous point set and the first point of the next point set in set S. If the distance is less than the threshold T2, merge the two point sets into a new point set to avoid dividing the outline of the same feature region into multiple point clouds, which would affect the subsequent feature point cloud extraction.

[0018] S13. Set a threshold T3. If the number of points in a single point set in set S is less than the threshold T3, then filter out the point set to eliminate outlier interference point clouds in the original profile.

[0019] S14. Based on the actual characteristics of the collected profile point cloud, the point cloud in the railhead region contains points with the maximum longitudinal amplitude. The index h of the potential railhead region point set is calculated from this, and the corresponding point set P is quickly extracted from set S. h The extracted railhead point set P is determined by the positional relationship between the railhead and the rail web. h If the rail web region point set index is smaller than the rail head region point set for the outer rail profile, the opposite is true for the inner rail profile. If the rail head point set is extracted correctly, continue to step S15; otherwise, skip to step S16.

[0020] S15. Due to the installation angle of the linear laser, there is a significant discontinuity between the profile rail head and the rail web, and the point sets of the rail head and rail web are adjacent. Therefore, the index w of the point set in the rail web region is calculated, and the corresponding point set P is quickly extracted from set S. w The effective profile point set P is obtained by merging the point sets of the rail head and rail waist in sequence. hw ;

[0021] S16. Utilizing the characteristic that both the railhead and rail waist are curves, calculate the curvature of each point set in set S and the longitudinal amplitude of the first point of each point set. Select the indices h and w corresponding to the point sets in the railhead and rail waist regions respectively, and quickly extract the corresponding railhead point set P from set S. h And the set of waist points Pw The point sets are then merged sequentially to obtain the effective profile point set P. hw If the point sets for the rail head and rail waist areas are not selected, the profile is deemed invalid and discarded.

[0022] Furthermore, step S2 specifically includes the following steps:

[0023] S21. Divide the two large and small circular arcs with radii R350 and R20 on the inner side of the standard 50kg / m rail profile into point cloud segments to obtain the large circular arc point set P. R350 and the set of small arc points P R20 ;

[0024] S22. Fit the two large and small arcs after the division to obtain the optimal double center. When fitting, first use the perpendicular bisector theorem of the circle combined with the properties of similar triangles to calculate a relatively accurate initial double center, and then use the nonlinear least squares method based on radius constraints to fit the optimal double center.

[0025] Furthermore, step S22 specifically includes the following steps:

[0026] S221, at the set of points P on the great circle arc R350 In the above, two points E and F are randomly selected. The origin of the circular arc point cloud coordinate system is set to O, the midpoint of chord EF to G, the center of the fitted circle to H, and line segment GH to the perpendicular bisector of chord EF. The vector is then calculated. sum vector The model:

[0027]

[0028] in That is, the radius of the arc, and the vector is obtained by applying the properties of similar triangles.

[0029]

[0030] The center H of the circle is then determined based on the given conditions:

[0031]

[0032] Assume the coordinates of the fitted circle center H are (x c ,z c ), calculate the point set P R350 ={(x i ,z i )} i∈N The distance from each point to the center H of the fitted circle is related to the circle radius R. 350 The sum of the differences is used as the fitting center error, and the fitting error for this time is set as e:

[0033]

[0034] Where n represents the number of points in the point set;

[0035] Following the above method, a threshold t = 20 is set, and the calculation is repeated t times. The fitting error is then stored in set S. e ={e t} t∈N Find the set S e The minimum fitting error corresponds to the center H of the fitting circle;

[0036] S222. The Levenberg-Marquardt algorithm, based on radius constraints and least squares nonlinear fitting, is used to quickly iteratively calculate the optimal circle center, and the point set P is... R350 ={(x i ,z i )} i∈N Circular radius R 350 Step S321 Fitting the center H(x) c ,z c The target residual function is defined as the input parameter for the algorithm iteration:

[0037]

[0038] Where M represents the number of fitting points, and the optimal center Q of the great circle arc is obtained when F(x) reaches its minimum value. R350 ;

[0039] S223, in the set of small arc points P R20 In the above, two points E and F are randomly selected. The origin of the circular arc point cloud coordinate system is set to O, the midpoint of chord EF to G, the center of the fitted circle to H, and line segment GH to the perpendicular bisector of chord EF. The vector is then calculated. sum vector The model:

[0040]

[0041] in That is, the radius of the arc, and the vector is obtained by applying the properties of similar triangles.

[0042]

[0043] The center H of the circle is then determined based on the given conditions:

[0044]

[0045] Assume the coordinates of the fitted circle center H are (x c ,z c ), calculate the point set P R20 ={(x i ,z i )} i∈NThe distance from each point to the center H of the fitted circle is related to the circle radius R. 20 The sum of the differences is used as the fitting center error, and the fitting error for this time is set as e:

[0046]

[0047] Where n represents the number of points in the point set;

[0048] Following the above method, a threshold t = 20 is set, and the calculation is repeated t times. The fitting error is then stored in set S. e ={e t} t∈N Find the set S e The minimum fitting error corresponds to the center H of the fitting circle;

[0049] S224. The Levenberg-Marquardt algorithm, based on radius constraints and least squares nonlinear fitting, is used to quickly iteratively calculate the optimal circle center, and the point set P is... R20 ={(x i ,z i )} i∈N Circular radius R 20 Step S323 Fitting the center H(x) c ,z c The target residual function is defined as the input parameter for the algorithm iteration:

[0050]

[0051] Where M represents the number of fitting points, and the optimal center Q of the small arc is obtained when F(x) reaches its minimum value. R20 .

[0052] Furthermore, step S3 specifically includes the following steps:

[0053] S31. Assume the optimal center coordinates for fitting the large and small circular arcs are (Cx) 350 ,Cz 350 ) and (Cx 20 ,Cz 20 The coordinates of the center of the large and small arcs corresponding to the profile of a standard 50kg / m rail are (Cx) B350 ,Cz B350 ) and (Cx B20 ,Cz B20 The rotation angle θ, rotation matrix R, and translation matrix T for registering the original profile with the standard profile are calculated, thus completing the registration of the inner profile.

[0054]

[0055] S32. Assume the inner profile registration rotation and translation matrices are R... IT I The rotation matrix for inner profile calibration is R. Ibd The rotation and translation matrices between the unified coordinate system and the standard profile coordinate system for the inner profile are obtained as R. bz T bz :

[0056] R bz =R I *R Ibd T

[0057] T bz =T I

[0058] Assume the calibration rotation and translation matrices of the outer profile are R and R respectively. Obd T Obd Before and after registration, a certain point is (x) i ,z i ) and (x i ',z i This completes the registration of the outer profile.

[0059]

[0060] At this point, the registration of the entire cross-sectional profile of the rail has been completed.

[0061] Furthermore, step S4 specifically includes the following steps:

[0062] S41. The coordinates of the feature points on both the inner and outer sides of the rail profile after registration, including the inner vertical wear points of the standard and measured profiles, are V(x) and V(x). bv ,z bv ), V'(x v ,z v The coordinates of the inner and outer jaw points of the standard profile are A and B, respectively. in (x bie ,z bie A) out (x boe ,z boe The measured coordinates of the inner and outer jaw points of the profile are A and B, respectively. i ' n (x ie ,z ie ), A' out (x oe ,z oe The coordinates of the wear points on the inner and outer sides of the standard profile are H, respectively. in (x bih ,z bih H out (xboh ,z boh The measured coordinates of the wear points on the inner and outer sides of the profile are H, respectively. i ' n (x ih ,z ih ), H' out (x oh ,z oh );

[0063] S42. Calculate the vertical wear ω of the rail based on the characteristic points on the inner and outer sides of the rail profile. v Side wear ω h Total wear ω a :

[0064]

[0065] This application also provides a rail wear detection device with full-section profile registration, comprising:

[0066] The point cloud preprocessing and feature extraction module is used to perform clustering and filtering preprocessing on the collected original track profile point cloud data using a distance-based clustering filtering algorithm. Then, based on the method of combining the longitudinal amplitude features of the point cloud with the profile curvature features, the point cloud of the rail head and rail web regions is accurately extracted.

[0067] The optimal double-center fitting module is used to segment the point cloud feature arcs of the inner rail web region of the rail profile to obtain large and small arcs, and to obtain the optimal double center of the segmented large and small arcs by fitting the nonlinear least squares algorithm based on radius constraints.

[0068] The full-section profile registration module is used to calculate the inner attitude transformation parameters of the inner profile of the rail based on the optimal double-center positioning obtained by fitting, and to calculate the outer attitude transformation parameters of the outer profile of the rail by combining the inner attitude transformation parameters with the system calibration parameters, so as to realize the full-section profile registration between the measured profile and the standard profile.

[0069] The rail wear value calculation module is used to calculate the rail wear value by combining the feature points on the inner and outer sides of the rail profile after profile registration with the measured profile and the standard profile.

[0070] This application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the rail wear detection method for full-section profile registration.

[0071] This application also provides a storage medium including a stored program that, when the program is executed, controls the device containing the storage medium to perform the steps of the rail wear detection method with full profile registration.

[0072] This application also provides a computer program product, including a computer program or computer-executable instructions, which, when executed by a processor, implement the steps of the rail wear detection method with full profile registration as described above.

[0073] Compared with the prior art, this application has the following advantages:

[0074] (1) When processing the collected original track profile point cloud data, this application adopts distance-based clustering and filtering algorithms to effectively filter out interfering outlier point clouds, making the extracted rail profile data more accurate.

[0075] (2) This application adopts a feature point cloud segmentation algorithm that combines the longitudinal amplitude feature of point cloud with the profile curvature feature to accurately extract and segment the point cloud of the rail head and rail waist regions, ensuring higher accuracy in profile matching.

[0076] (3) This application uses the perpendicular bisector theorem of a circle combined with the properties of similar triangles to first calculate a relatively accurate initial double center, and then uses the nonlinear least squares method based on radius constraints to fit the optimal double center, thereby reducing the sensitivity of the algorithm to the initial value, improving the algorithm iteration efficiency, and improving the real-time performance of the system detection.

[0077] (4) The inner side of the rail profile in this application is registered by double-center positioning through rail web fitting, and the outer side of the profile is registered by combining the inner side attitude transformation parameters with the system calibration parameters, so as to achieve rapid registration of the full-section profile and better real-time detection of the system.

[0078] (5) This application uses the combination of feature points on the inner and outer sides of the rail to calculate the wear value, which effectively eliminates the offset error caused by interference during dynamic detection and has higher wear detection accuracy under multi-state working conditions.

[0079] In addition to the purposes, features, and advantages described above, this application has other purposes, features, and advantages. A further detailed description of this application will be provided below with reference to the figures. Attached Figure Description

[0080] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments of this application and their descriptions are used to explain this application and do not constitute an undue limitation of this application.

[0081] Figure 1 This is a schematic diagram of the rail wear detection method with full-section profile registration according to a preferred embodiment of this application.

[0082] Figure 2 This is a schematic diagram of the rail wear detection device according to a preferred embodiment of this application.

[0083] Figure 3 This is a schematic diagram of the profile of a standard 50kg / m steel rail.

[0084] Figure 4 This is a rendering before full-section profile registration.

[0085] Figure 5 This is the effect diagram after full-section profile registration.

[0086] Figure 6 This is a schematic diagram for standard wear calculation.

[0087] Figure 7 This is a schematic diagram of wear calculation when the profile is offset.

[0088] Figure 8 This is a schematic diagram of a rail wear detection device module with full-section profile registration according to a preferred embodiment of this application. picture.

[0089] Figure 9 This is a schematic diagram of a rail wear detection device module with full-section profile registration according to a preferred embodiment of this application.

[0090] Figure 10 This is a schematic block diagram of an electronic device according to a preferred embodiment of this application.

[0091] Figure 11 This is an internal structural diagram of a computer device according to a preferred embodiment of this application.

[0092] In the diagram: 1. Detection beam; 2. Line laser. Detailed Implementation

[0093] The embodiments of this application are described in detail below with reference to the accompanying drawings; however, this application may be implemented in a variety of different ways as defined and covered below.

[0094] like Figure 1 As shown, a preferred embodiment of this application provides a rail wear detection method based on full-section profile registration, comprising the following steps:

[0095] S1. The distance-based clustering filtering algorithm performs clustering and filtering preprocessing on the collected original track profile point cloud data. Then, based on the longitudinal amplitude characteristics of the point cloud combined with the profile curvature characteristics, the point cloud of the rail head and rail web regions is accurately extracted.

[0096] S2. The point cloud feature arcs of the inner rail web region of the rail profile are segmented to obtain large and small arcs. The initial double center is calculated by combining the perpendicular bisector theorem of the circle with the properties of similar triangles. The optimal double center of the segmented large and small arcs is obtained by fitting the initial double center using a nonlinear least squares algorithm based on radius constraints.

[0097] S3. Calculate the inner attitude transformation parameters of the inner profile of the rail based on the optimal double-center positioning obtained by fitting. Calculate the outer attitude transformation parameters of the outer profile of the rail by combining the inner attitude transformation parameters with the system calibration parameters, thereby achieving full-section profile registration between the measured profile and the standard profile.

[0098] S4. Calculate the rail wear value by combining the characteristic points on the inner and outer sides of the rail profile after profile registration with the measured profile and the standard profile.

[0099] Compared with the prior art, this embodiment has the following beneficial effects:

[0100] (1) In this embodiment, when processing the collected original track profile point cloud data, a distance-based clustering and filtering algorithm is used to effectively filter out interfering outlier point clouds, making the extracted rail profile data more accurate. Confirmed;

[0101] (2) This embodiment adopts a feature point cloud segmentation algorithm that combines the longitudinal amplitude feature of point cloud with the profile curvature feature to accurately extract and segment the point cloud of the rail head and rail waist regions, ensuring higher profile matching accuracy;

[0102] (3) In this embodiment, the perpendicular bisector theorem of the circle combined with the properties of similar triangles is used to first calculate a relatively accurate initial double center, and then the nonlinear least squares method based on radius constraint is used to fit the optimal double center, which reduces the sensitivity of the algorithm to the initial value, improves the algorithm iteration efficiency, and improves the real-time performance of the system detection.

[0103] (4) The inner side of the rail profile in this application is registered by double-center positioning through rail web fitting, and the outer side of the profile is registered by combining the inner side attitude transformation parameters with the system calibration parameters, so as to achieve rapid registration of the full-section profile and better real-time detection of the system.

[0104] (5) This embodiment uses the combination of feature points on the inner and outer sides of the rail to calculate the wear value, effectively eliminating the offset error caused by interference during dynamic detection, and achieving higher wear detection accuracy under multi-state working conditions.

[0105] Preferably, before acquiring the original orbital profile point cloud data, the following steps are also included:

[0106] S0. Install four line laser sensors on the detection beam 1 to form a rail wear detection device (e.g., Figure 2 As shown), the system parameters are calibrated to realize the transformation from the original coordinate system to the unified coordinate system for each line laser 2, and the system calibration parameters are obtained, including the transformation rotation angle θ between the original coordinate system and the unified coordinate system, and the relative offsets Δx and Δz of the origin of the original coordinate system and the unified coordinate system in the X and Z directions, respectively. The X direction is the horizontal direction of the rail cross section, and the Z direction is the vertical direction of the rail cross section height.

[0107] In this embodiment, the rail wear detection device includes a detection beam and four line lasers 2. The detection beam is rigidly connected to the bottom of the engineering vehicle, and the line lasers 2 are fixedly installed below the detection beam at certain angles. Taking a 50kg / m rail as an example, the side of the rail that rubs against the wheel is defined as the inner side, and the non-friction side is defined as the outer side. The detection device is mounted on the rail inspection engineering vehicle and needs to be calibrated when the engineering vehicle is parked on a straight track and in a stationary state. During calibration, the line laser sensors are turned on, a customized calibration rod is placed on the track surface, and the calibration rod is moved so that the laser line coincides with the positioning mark on the calibration rod for positioning and alignment. Calibration is performed through a software calibration program, which mainly realizes the transformation of the original coordinate system of each line laser to a unified coordinate system. The calculated output calibration parameters include the rotation angle θ of the two coordinate systems and the relative offsets Δx and Δz of the origins of the two coordinate systems in the X direction (horizontal direction of the cross section) and Z direction (vertical direction of height), respectively, thus providing necessary preliminary preparation for subsequent detection.

[0108] Preferably, step S1 specifically includes the following steps:

[0109] S11. Based on the distribution characteristics of the measured original profile point cloud, where normal data is dense and outlier interference data is sparse, a distance-based clustering algorithm is used to segment the point cloud to obtain the segmented set S = {P}. i ={(x j , z j )}} i，j∈N Where i represents the index of the segmented point set, and j represents the index of a point in a single point set. The search starts from the first point and stores it in the point set P. i If the distance between the current point and the next point is less than or equal to a set threshold T1, then the next point is stored in the point set P. i If the value is greater than the set threshold T1, then the next point is stored in a new point set P. i+1 And so on, stopping the search once all points have been traversed;

[0110] S12. Set a threshold T2 and calculate the distance between the tail point of the previous point set and the first point of the next point set in set S. If the distance is less than the threshold T2, merge the two point sets into a new point set to avoid dividing the outline of the same feature region into multiple point clouds, which would affect the subsequent feature point cloud extraction.

[0111] S13. Set a threshold T3. If the number of points in a single point set in set S is less than the threshold T3, then filter out the point set to eliminate outlier interference point clouds in the original profile.

[0112] S14. Based on the actual characteristics of the collected profile point cloud, the point cloud in the railhead region contains points with the maximum longitudinal amplitude. The index h of the potential railhead region point set is calculated from this, and the corresponding point set P is quickly extracted from set S. hThe extracted railhead point set P is determined by the positional relationship between the railhead and the rail web. h If the rail web region point set index is smaller than the rail head region point set for the outer rail profile, the opposite is true for the inner rail profile. If the rail head point set is extracted correctly, continue to step S15; otherwise, skip to step S16.

[0113] S15. Due to the installation angle of the linear laser, there is a significant discontinuity between the profile rail head and the rail web, and the point sets of the rail head and rail web are adjacent. Therefore, the index w of the point set in the rail web region is calculated, and the corresponding point set P is quickly extracted from set S. w The effective profile point set P is obtained by merging the point sets of the rail head and rail waist in sequence. hw ;

[0114] S16. Utilizing the characteristic that both the railhead and rail waist are curves, calculate the curvature of each point set in set S and the longitudinal amplitude of the first point of each point set. Select the indices h and w corresponding to the point sets in the railhead and rail waist regions respectively, and quickly extract the corresponding railhead point set P from set S. h And the set of waist points P w The point sets are then merged sequentially to obtain the effective profile point set P. hw If the point sets for the rail head and rail waist areas are not selected, the profile is deemed invalid and discarded.

[0115] Due to the influence of various interference factors on actual track lines, such as reflection from strong light areas, roadbed, fasteners, and foreign objects beside the track, there are outlier interference points in the original rail profile point cloud acquired by line laser, which will affect the accuracy of profile registration and wear calculation. Therefore, this embodiment preprocesses the original profile point cloud to eliminate outlier interference points, making the extracted rail profile data more accurate. At the same time, it quickly identifies the effective profile and extracts the point cloud of the rail head and rail web area, accurately extracting and segmenting the point cloud of the rail head and rail web area to ensure higher profile matching accuracy.

[0116] Preferably, step S2 specifically includes the following steps:

[0117] S21. Divide the two large and small circular arcs with radii R350 and R20 on the inner side of the standard 50kg / m rail profile into point cloud segments to obtain the large circular arc point set P. R350 and the set of small arc points P R20 ;

[0118] S22. Fit the two large and small arcs after the division to obtain the optimal double center. When fitting, first use the perpendicular bisector theorem of the circle combined with the properties of similar triangles to calculate a relatively accurate initial double center, and then use the nonlinear least squares method based on radius constraints to fit the optimal double center.

[0119] Figure 3To obtain a schematic diagram of a standard 50kg / m rail profile, it is necessary to segment the point cloud of two large and small arcs with radii R350 and R20 on the inner side of the rail web region and extract the arcs as registration feature point clouds. Point A is the rail jaw point, the large arc R350 is part BC in the diagram, and the small arc R20 is part CD in the diagram. The two arcs are tangent at point C.

[0120] Based on the standard 50kg / m rail dimensions, the distances between points A and C, and between points D and A, can be accurately calculated. From this, the effective profile point set P can be derived. hw Accurate segmentation and extraction of two arc point sets, P and P respectively. R350 and P R20 .

[0121] The optimal double center is obtained by fitting the divided large and small arcs. First, the initial double center is calculated by combining the perpendicular bisector theorem of the circle with the properties of similar triangles. Then, the optimal double center is obtained by fitting the nonlinear least squares method based on radius constraints.

[0122] This embodiment uses the perpendicular bisector theorem of a circle combined with the properties of similar triangles to first calculate a relatively accurate initial double center of the inner rail web region of the standard 50kg / m rail profile. Then, it uses the nonlinear least squares method based on radius constraints to fit the optimal double center of the two large and small arcs, reducing the algorithm's sensitivity to initial values, improving the algorithm's iteration efficiency, and resulting in better real-time detection of the system.

[0123] Specifically, step S22 includes the following steps:

[0124] S221, at the set of points P on the great circle arc R350 In the above, two points E and F are randomly selected. The origin of the circular arc point cloud coordinate system is set to O, the midpoint of chord EF to G, the center of the fitted circle to H, and line segment GH to the perpendicular bisector of chord EF. The vector is then calculated. sum vector The model:

[0125]

[0126] in That is, the radius of the arc, and the vector is obtained by applying the properties of similar triangles.

[0127]

[0128] The center H of the circle is then determined based on the given conditions:

[0129]

[0130] Assume the coordinates of the fitted circle center H are (x c ,z c ), calculate the point set PR350 ={(x i ,z i )} i∈N The distance from each point to the center H of the fitted circle is related to the circle radius R. 350 The sum of the differences is used as the fitting center error, and the fitting error for this time is set as e:

[0131]

[0132] Where n represents the number of points in the point set;

[0133] Following the above method, a threshold t = 20 is set, and the calculation is repeated t times. The fitting error is then stored in set S. e ={e t} t∈N Find the set S e The minimum fitting error corresponds to the center H of the fitting circle;

[0134] S222. The Levenberg-Marquardt algorithm, based on radius constraints and least squares nonlinear fitting, is used to quickly iteratively calculate the optimal circle center, and the point set P is... R350 ={(x i ,z i )} i∈N Circular radius R 350 Step S321 Fitting the center H(x) c ,z c The target residual function is defined as the input parameter for the algorithm iteration:

[0135]

[0136] Where M represents the number of fitting points, and the optimal center Q of the great circle arc is obtained when F(x) reaches its minimum value. R350 ;

[0137] S223, in the set of small arc points P R20 In the above, two points E and F are randomly selected. The origin of the circular arc point cloud coordinate system is set to O, the midpoint of chord EF to G, the center of the fitted circle to H, and line segment GH to the perpendicular bisector of chord EF. The vector is then calculated. sum vector The model:

[0138]

[0139] in That is, the radius of the arc, and the vector is obtained by applying the properties of similar triangles.

[0140]

[0141] The center H of the circle is then determined based on the given conditions:

[0142]

[0143] Assume the coordinates of the fitted circle center H are (x c ,z c ), calculate the point set P R20 ={(x i ,z i )} i∈N The distance from each point to the center H of the fitted circle is related to the circle radius R. 20 The sum of the differences is used as the fitting center error, and the fitting error for this time is set as e:

[0144]

[0145] Where n represents the number of points in the point set;

[0146] Following the above method, a threshold t = 20 is set, and the calculation is repeated t times. The fitting error is then stored in set S. e ={e t} t∈N Find the set S e The minimum fitting error corresponds to the center H of the fitting circle;

[0147] S224. The Levenberg-Marquardt algorithm, based on radius constraints and least squares nonlinear fitting, is used to quickly iteratively calculate the optimal circle center, and the point set P is... R20 ={(x i ,z i )} i∈N Circular radius R 20 Step S323 Fitting the center H(x) c ,z c The target residual function is defined as the input parameter for the algorithm iteration:

[0148]

[0149] Where M represents the number of fitting points, and the optimal center Q of the small arc is obtained when F(x) reaches its minimum value. R20 .

[0150] This embodiment presents a method that uses the perpendicular bisector theorem of a circle combined with the properties of similar triangles to first calculate a relatively accurate initial double center of the inner rail web region of a standard 50kg / m rail profile. Then, it uses a nonlinear least squares method based on radius constraints to fit the optimal double center of the two large and small arcs, reducing the algorithm's sensitivity to initial values, improving the algorithm's iteration efficiency, and enhancing the system's real-time detection performance.

[0151] Preferably, step S3 specifically includes the following steps:

[0152] S31. Assume the optimal coordinates of the two centers of the fitted large and small circular arcs are (Cx) 350 ,Cz 350 ) and (Cx 20 ,Cz 20 The coordinates of the center of the large and small arcs corresponding to the profile of a standard 50kg / m rail are (Cx) B350 ,Cz B350 ) and (Cx B20 ,Cz B20 The rotation angle θ, rotation matrix R, and translation matrix T for registering the original profile with the standard profile are calculated, thus completing the registration of the inner profile.

[0153]

[0154] S32. Assume the inner profile registration rotation and translation matrices are R... I T I The rotation matrix for inner profile calibration is R. Ibd The rotation and translation matrices between the unified coordinate system and the standard profile coordinate system for the inner profile are obtained as R. bz T bz :

[0155] R bz =R I *R Ibd T

[0156] T bz =T I

[0157] Assume the calibration rotation and translation matrices of the outer profile are R and R respectively. Obd T Obd Before and after registration, a certain point is (x) i ,z i ) and (x i ',z i This completes the registration of the outer profile.

[0158]

[0159] At this point, the full-section profile registration of the rail's inner and outer sides is complete. The effects before and after the full-section profile registration are as follows: Figure 4 and Figure 5 As shown.

[0160] This embodiment of rail full-section profile registration includes inner (wearing side) and outer (non-wearing side) registration. Inner registration is performed using a double-center positioning method on the rail web, while outer registration is performed by combining inner attitude transformation parameters with system calibration parameters, achieving rapid full-section profile registration. The outer profile registration involves three coordinate systems: the original profile coordinate system, the unified coordinate system, and the standard profile coordinate system. Outer profile registration transforms the outer profile point cloud from the original profile coordinate system to the standard profile coordinate system.

[0161] First, the transformation matrix between the original inner and outer profile coordinate systems and the unified coordinate system can be obtained through system calibration, i.e., the calibration matrix;

[0162] Then, based on the obtained inner profile registration matrix and the known inner profile calibration matrix, the transformation matrix between the unified coordinate system and the standard profile coordinate system can be obtained.

[0163] Finally, the original profile of the outer side of the rail is registered using the above transformation relationships.

[0164] In addition, for elevated bridge track lines, guard rails are laid on the inner side of the track to prevent train derailment. Because the guard rails are close to the track rails, the inner profile feature point cloud of the rails is obscured and missing. At this time, the system can automatically switch to prioritize the use of the outer profile point cloud to perform inner profile registration according to step S31. On this basis, the inner profile point cloud is then used to perform outer profile registration according to step S32, thereby completing the full-section profile registration of the rails.

[0165] Preferably, step S4 specifically includes the following steps:

[0166] S41. The coordinates of the feature points on both the inner and outer sides of the rail profile after registration, including the inner vertical wear points of the standard and measured profiles, are V(x) and V(x). bv ,z bv ), V'(x v ,z v The coordinates of the inner and outer jaw points of the standard profile are A and B, respectively. in (x bie ,z bie A) out (x boe ,z boe The measured coordinates of the inner and outer jaw points of the profile are A and B, respectively. i ' n (x ie ,z ie ), A' out (x oe ,z oe The coordinates of the wear points on the inner and outer sides of the standard profile are H, respectively. in (xbih ,z bih H out (x boh ,z boh The measured coordinates of the wear points on the inner and outer sides of the profile are H, respectively. i ' n (x ih ,z ih ), H' out (x oh ,z oh );

[0167] S42. Calculate the vertical wear ω of the rail based on the characteristic points on the inner and outer sides of the rail profile. v Side wear ω h Total wear ω a :

[0168]

[0169] Figure 6 This is a standard definition of wear calculation diagram. Vertical wear is defined as the thickness of the rail tread worn vertically at 1 / 3 of the rail head width from the working edge; lateral wear is defined as the width of the working edge worn horizontally at 16mm below the top of the rail; total wear is defined as the sum of vertical wear and 1 / 2 lateral wear.

[0170] This embodiment takes into account the deviation error between the measured profile and the actual profile caused by track interference during dynamic detection, which directly affects the accuracy of wear detection. It utilizes multiple feature points on both the inner and outer sides of the rail (see...) Figure 7 The wear value is calculated using a combination of methods to ensure the accuracy of wear detection. Specifically:

[0171] The difference between the distance between the lateral wear points on the inner and outer sides of the tested profile and the distance between the lateral wear points on both sides of the standard profile is the lateral wear value. The lateral wear value is calculated by calculating the jaw points on the inner and outer sides of the tested profile and the inner vertical wear point. Since the jaw points on both sides will not be worn, the height difference between the inner vertical wear point and the inner jaw point of the tested profile and the standard profile is used as the vertical wear value.

[0172] Because this embodiment uses multiple feature points on both the inner and outer sides of the rail to calculate vertical wear, lateral wear, and total wear, it effectively eliminates offset errors caused by interference during dynamic detection, resulting in higher wear detection accuracy under multi-condition working conditions.

[0173] like Figure 8 As shown, another preferred embodiment of this application also provides a rail wear detection device with full-section profile registration, comprising:

[0174] The point cloud preprocessing and feature extraction module is used to perform clustering and filtering preprocessing on the collected original track profile point cloud data using a distance-based clustering filtering algorithm. Then, based on the method of combining the longitudinal amplitude features of the point cloud with the profile curvature features, the point cloud of the rail head and rail web regions is accurately extracted.

[0175] The optimal double-center fitting module is used to segment the point cloud feature arcs of the inner rail web region of the rail profile to obtain large and small arcs, and to obtain the optimal double center of the segmented large and small arcs by fitting the nonlinear least squares algorithm based on radius constraints.

[0176] The full-section profile registration module is used to calculate the inner attitude transformation parameters of the inner profile of the rail based on the optimal double-center positioning obtained by fitting, and to calculate the outer attitude transformation parameters of the outer profile of the rail by combining the inner attitude transformation parameters with the system calibration parameters, so as to realize the full-section profile registration between the measured profile and the standard profile.

[0177] The rail wear value calculation module is used to calculate the rail wear value by combining the feature points on the inner and outer sides of the rail profile after profile registration with the measured profile and the standard profile.

[0178] Preferably, such as Figure 9 As shown, the rail wear detection device with full-section profile registration also includes:

[0179] The system parameter calibration module is used to calibrate the system parameters of the rail wear detection device composed of four line laser sensors installed on the detection beam. It realizes the transformation from the original coordinate system to the unified coordinate system of each line laser and obtains the system calibration parameters, including the transformation rotation angle θ between the original coordinate system and the unified coordinate system, and the relative offsets Δx and Δz of the origin of the original coordinate system and the unified coordinate system in the X and Z directions, respectively. The X direction is the horizontal direction of the rail cross section, and the Z direction is the vertical direction of the rail cross section height.

[0180] like Figure 10 As shown, a preferred embodiment of this application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes... The computer program is real-time The rail wear detection method based on full-section profile registration in the above embodiments is now described. step.

[0181] A preferred embodiment of this application also provides a computer device, which may be a terminal or a liveness detection server, and its internal structure diagram may be as follows. Figure 11As shown. The computer device includes a processor, memory, and a network interface connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The network interface is used to communicate with other external computer devices via a network connection. When the computer program is executed by the processor, it implements the steps of the above-described rail wear detection method based on full-section profile registration.

[0182] Those skilled in the art will understand that Figure 11 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0183] A preferred embodiment of this application also provides a storage medium, the storage medium including a stored program, which, when the program is executed, controls the device where the storage medium is located to perform the steps of the rail wear detection method with full profile registration in the above embodiments.

[0184] A preferred embodiment of this application provides a computer program product or computer program that includes computer instructions stored in a computer-readable storage medium. A processor of an electronic device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the electronic device to perform the rail wear detection method for full-section profile registration described in the embodiments of this application.

[0185] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0186] If the functions described in this embodiment are implemented as software functional units and sold or used as independent products, they can be stored in one or more computing device-readable storage media. Based on this... According to the understanding of this application, The part of the example that contributes to the prior art or the part of the technical solution mayThe software product is manifested in the form of a software product stored in a storage medium and includes several instructions to cause a computing device (which may be a personal computer, server, mobile computing device, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0187] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of this application can be implemented in various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0188] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0189] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0190] These computer program instructions can also be loaded onto a computer or other programmable data processing equipment, causing a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented processing order. Reason, thus in computers or Instructions that execute on other programmable devices provide the means to implement the process. Figure 1 indivual Process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0191] Although the preferred embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present application.

[0192] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for detecting rail wear using full-section profile registration, characterized in that, Including the following steps: S1. The distance-based clustering filtering algorithm performs clustering and filtering preprocessing on the collected original track profile point cloud data. Then, based on the longitudinal amplitude characteristics of the point cloud combined with the profile curvature characteristics, the point cloud of the rail head and rail web regions is accurately extracted. S2. The point cloud feature arcs of the inner rail web region of the rail profile are segmented to obtain large and small arcs. The initial double center is calculated by combining the perpendicular bisector theorem of the circle with the properties of similar triangles. The optimal double center of the segmented large and small arcs is obtained by fitting the initial double center using a nonlinear least squares algorithm based on the radius constraint. S3. Calculate the inner attitude transformation parameters of the inner profile of the rail based on the optimal double-center positioning obtained by fitting. Calculate the outer attitude transformation parameters of the outer profile of the rail by combining the inner attitude transformation parameters with the system calibration parameters, thereby achieving full-section profile registration between the measured profile and the standard profile. S4. Calculate the rail wear value by combining the characteristic points on the inner and outer sides of the rail profile after profile registration with the measured profile and the standard profile.

2. The rail wear detection method based on full-section profile registration according to claim 1, characterized in that, Before collecting the original orbital profile point cloud data, the following steps are also included: S0. The system parameters of the rail wear detection device consisting of four line laser sensors installed on the detection beam are calibrated to realize the transformation from the original coordinate system to the unified coordinate system of each line laser, and obtain the system calibration parameters, including the transformation rotation angle θ between the original coordinate system and the unified coordinate system, and the relative offsets Δx and Δz of the origin of the original coordinate system and the unified coordinate system in the X and Z directions, respectively. The X direction is the horizontal direction of the rail cross section, and the Z direction is the vertical direction of the rail cross section height.

3. The rail wear detection method based on full-section profile registration according to claim 1, characterized in that, Step S1 specifically includes the following steps: S11. Based on the distribution characteristics of the measured original profile point cloud, where normal data is dense and outlier interference data is sparse, a distance-based clustering algorithm is used to segment the point cloud to obtain the segmented set S = {P}. i ={(x j , z j )}} i，j∈N Where i represents the index of the segmented point set, and j represents the index of a point in a single point set. The search starts from the first point and stores it in the point set P. i If the distance between the current point and the next point is less than or equal to a set threshold T1, then the next point is stored in the point set P. i If the value is greater than the set threshold T1, then the next point is stored in a new point set P. i+1 And so on, stopping the search once all points have been traversed; S12. Set a threshold T2 and calculate the distance between the tail point of the previous point set and the first point of the next point set in set S. If the distance is less than the threshold T2, merge the two point sets into a new point set to avoid dividing the outline of the same feature region into multiple point clouds, which would affect the subsequent feature point cloud extraction. S13. Set a threshold T3. If the number of points in a single point set in set S is less than the threshold T3, then filter out the point set to eliminate outlier interference point clouds in the original profile. S14. Based on the actual characteristics of the collected profile point cloud, the point cloud in the railhead region contains points with the maximum longitudinal amplitude. The index h of the potential railhead region point set is calculated from this, and the corresponding point set P is quickly extracted from set S. h The extracted railhead point set P is determined by the positional relationship between the railhead and the rail web. h If the rail web region point set index is smaller than the rail head region point set for the outer rail profile, the opposite is true for the inner rail profile. If the rail head point set is extracted correctly, continue to step S15; otherwise, skip to step S16. S15. Due to the installation angle of the linear laser, there is a significant discontinuity between the profile rail head and the rail web, and the point sets of the rail head and rail web are adjacent. Therefore, the index w of the point set in the rail web region is calculated, and the corresponding point set P is quickly extracted from set S. w The effective profile point set P is obtained by merging the point sets of the rail head and rail waist in sequence. hw ; S16. Utilizing the characteristic that both the railhead and rail waist are curves, calculate the curvature of each point set in set S and the longitudinal amplitude of the first point of each point set. Select the indices h and w corresponding to the point sets in the railhead and rail waist regions respectively, and quickly extract the corresponding railhead point set P from set S. h And the set of waist points P w The point sets are then merged sequentially to obtain the effective profile point set P. hw If the point sets for the rail head and rail waist areas are not selected, the profile is deemed invalid and discarded.

4. The rail wear detection method based on full-section profile registration according to claim 1, characterized in that, Step S2 specifically includes the following steps: S21. Divide the two large and small circular arcs with radii R350 and R20 on the inner side of the standard 50kg / m rail profile into point cloud segments to obtain the large circular arc point set P. R350 and the set of small arc points P R20 ; S22. Fit the two large and small arcs after the division to obtain the optimal double center. When fitting, first use the perpendicular bisector theorem of the circle combined with the properties of similar triangles to calculate the initial double center, and then use the nonlinear least squares method based on radius constraints to fit the optimal double center.

5. The rail wear detection method based on full-section profile registration according to claim 4, characterized in that, Step S22 specifically includes the following steps: S221, at the set of points P on the great circle arc R350 In the above, two points E and F are randomly selected. The origin of the circular arc point cloud coordinate system is set to O, the midpoint of chord EF to G, the center of the fitted circle to H, and line segment GH to the perpendicular bisector of chord EF. The vector is then calculated. sum vector The model: in That is, the radius of the arc, and the vector is obtained by applying the properties of similar triangles. The center H of the circle is then determined based on the given conditions: Assume the coordinates of the fitted circle center H are (x c ,z c ), calculate the point set P R350 ={(x i ,z i )} i∈N The distance from each point to the center H of the fitted circle is related to the circle radius R. 350 The sum of the differences is used as the fitting center error, and the fitting error for this time is set as e: Where n represents the number of points in the point set; Following the above method, a threshold t = 20 is set, and the calculation is repeated t times. The fitting error is then stored in set S. e ={e t } t∈N Find the set S e The minimum fitting error corresponds to the center H of the fitting circle; S222. The Levenberg-Marquardt algorithm, based on radius constraints and least squares nonlinear fitting, is used to quickly iteratively calculate the optimal circle center, and the point set P is... R350 ={(x i ,z i )} i∈N Circular radius R 350 Step S321 Fitting the center H(x) c ,z c The target residual function is defined as the input parameter for the algorithm iteration: Where M represents the number of fitting points, and the optimal center Q of the great circle arc is obtained when F(x) reaches its minimum value. R350 ; S223, in the set of small arc points P R20 In the above, two points E and F are randomly selected. The origin of the circular arc point cloud coordinate system is set to O, the midpoint of chord EF to G, the center of the fitted circle to H, and line segment GH to the perpendicular bisector of chord EF. The vector is then calculated. sum vector The model: in That is, the radius of the arc, and the vector is obtained by applying the properties of similar triangles. The center H of the circle is then determined based on the given conditions: Assume the coordinates of the fitted circle center H are (x c ,z c ), calculate the point set P R20 ={(x i ,z i )} i∈N The distance from each point to the center H of the fitted circle is related to the circle radius R. 20 The sum of the differences is used as the fitting center error, and the fitting error for this time is set as e: Where n represents the number of points in the point set; Following the above method, a threshold t = 20 is set, and the calculation is repeated t times. The fitting error is then stored in set S. e ={e t } t∈N Find the set S e The minimum fitting error corresponds to the center H of the fitting circle; S224. The Levenberg-Marquardt algorithm, based on radius constraints and least squares nonlinear fitting, is used to quickly iteratively calculate the optimal circle center, and the point set P is... R20 ={(x i ,z i )} i∈N Circular radius R 20 Step S323 Fitting the center H(x) c ,z c The target residual function is defined as the input parameter for the algorithm iteration: Where M represents the number of fitting points, and the optimal center Q of the small arc is obtained when F(x) reaches its minimum value. R20 .

6. The rail wear detection method based on full-section profile registration according to claim 1, characterized in that, Step S3 specifically includes the following steps: S31. Assume the optimal center coordinates for fitting the large and small circular arcs are (Cx) 350 ,Cz 350 ) and (Cx 20 ,Cz 20 The coordinates of the center of the large and small arcs corresponding to the profile of a standard 50kg / m rail are (Cx) B350 ,Cz B350 ) and (Cx B20 ,Cz B20 The rotation angle θ, rotation matrix R, and translation matrix T for registering the original profile with the standard profile are calculated, thus completing the registration of the inner profile. S32. Assume the inner profile registration rotation and translation matrices are R... I T I The rotation matrix for inner profile calibration is R. Ibd The rotation and translation matrices between the unified coordinate system and the standard profile coordinate system for the inner profile are obtained as R. bz T bz : R bz =R I *R Ibd T T bz =T I Assume the calibration rotation and translation matrices of the outer profile are R and R respectively. Obd T Obd Before and after registration, a certain point is (x) i ,z i ) and (x i ',z i This completes the registration of the outer profile. At this point, the registration of the entire cross-sectional profile of the rail has been completed.

7. The rail wear detection method based on full-section profile registration according to claim 1, characterized in that, Step S4 specifically includes the following steps: S41. The coordinates of the feature points on both the inner and outer sides of the rail profile after registration, including the inner vertical wear points of the standard and measured profiles, are V(x) and V(x). bv ,z bv ), V'(x v ,z v The coordinates of the inner and outer jaw points of the standard profile are A and B, respectively. in (x bie ,z bie A) out (x boe ,z boe The measured coordinates of the inner and outer jaw points of the profile are A and B, respectively. i ' n (x ie ,z ie ), A' out (x oe ,z oe The coordinates of the wear points on the inner and outer sides of the standard profile are H, respectively. in (x bih ,z bih H out (x boh ,z boh The measured coordinates of the wear points on the inner and outer sides of the profile are H, respectively. i ' n (x ih ,z ih ), H' out (x oh ,z oh ); S42. Calculate the vertical wear ω of the rail based on the characteristic points on the inner and outer sides of the rail profile. v Side wear ω h Total wear ω a :

8. A rail wear detection device with full-section profile registration, characterized in that, include: The point cloud preprocessing and feature extraction module is used to perform clustering and filtering preprocessing on the collected original track profile point cloud data using a distance-based clustering filtering algorithm. Then, based on the method of combining the longitudinal amplitude features of the point cloud with the profile curvature features, the point cloud of the rail head and rail web regions is accurately extracted. The optimal double-center fitting module is used to segment the point cloud feature arcs of the inner rail web region of the rail profile to obtain large and small arcs, and to obtain the optimal double center of the segmented large and small arcs by fitting the nonlinear least squares algorithm based on radius constraints. The full-section profile registration module is used to calculate the inner attitude transformation parameters of the inner profile of the rail based on the optimal double-center positioning obtained by fitting, and to calculate the outer attitude transformation parameters of the outer profile of the rail by combining the inner attitude transformation parameters with the system calibration parameters, so as to realize the full-section profile registration between the measured profile and the standard profile. The rail wear value calculation module is used to calculate the rail wear value by combining the feature points on the inner and outer sides of the rail profile after profile registration with the measured profile and the standard profile.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the rail wear detection method with full profile registration as described in any one of claims 1 to 7.

10. A storage medium comprising a stored program that, when the program is executed, controls a device containing the storage medium to perform the steps of the rail wear detection method for full profile registration as described in any one of claims 1 to 7.

Images