A rail wear measurement method, device, medium and equipment

Abstract

The present invention discloses a method, device, medium and equipment for measuring track wear, and the method comprises: obtaining point cloud data of a track to be measured and point cloud data of a standard track; the point cloud data of the track to be measured comprises first contour point cloud data and second contour point cloud data; inputting the contour point cloud data into a pre-trained PointNet++ network to obtain rail waist point cloud data; selecting a standard point from the point cloud data of the standard track, performing point cloud registration using an ICP algorithm, and obtaining a first corresponding point and a second corresponding point of the standard point; creating a projection plane based on the first corresponding point and the second corresponding point, projecting the contour point cloud data onto the projection plane, and obtaining point cloud data of the projection track; performing ICP algorithm registration on the point cloud data of the standard track and the point cloud data of the projection track, and obtaining the measurement result of the wear of the track to be measured. The present invention can obtain the measurement value of track wear, and improves the stability and accuracy of track measurement.

Description

Technical Field

[0001] The invention relates to a method, device, medium and equipment for measuring track wear, and belongs to the technical field of track wear detection. Background Art

[0002] With the rapid development of my country's economy, the demand for logistics and transportation is increasing, the demand for transport trains is also increasing year-on-year, the wear of the tracks carrying trains is gradually increasing, and the demand for detecting track wear is also increasing. It is very important to measure the track wear quickly and accurately.

[0003] At present, due to the rapid development of deep learning technology, especially in the work of component segmentation and scene segmentation, deep learning technology has been widely used in the rail transit industry due to its good stability, high segmentation efficiency, high segmentation accuracy and stronger adaptability to the environment.

[0004] In recent years, the application of deep learning technology in rail wear measurement has increased year by year. However, the conventional method generates track point cloud data through structured light, uses the radius filter algorithm to denoise, and the pass-through filter algorithm to obtain the rail waist. The iterative closest point algorithm is used to align the rail waist, thereby realizing the measurement of rail head wear.

[0005] Since the track is located in a natural light environment, the point cloud data collected by the structured light camera has more impurities and the background of the point cloud data is complex, which can easily lead to poor data stability; it will also lead to instability in the data segmented at the waist and head of the track, resulting in inaccurate registration of the waist by the Iterative Closet Point algorithm. In addition, due to the interference of spring bars, screws and nuts in the background of the point cloud data, the difficulty of track data segmentation increases, and the stability of the Iterative Closet Point algorithm of the waist is reduced, which will lead to an increase in the wear measurement error of the rail head; similarly, when the acquisition equipment is jittery or the laser is tilted, the generated track point cloud data is deformed, and the waist target area extracted by the pass-through filtering algorithm changes, resulting in the failure of the Iterative Closet Point algorithm of the waist, and ultimately resulting in inaccurate wear measurement results of the rail head, which cannot meet the measurement requirements of track wear detection. Summary of the invention

[0006] The purpose of the present invention is to overcome the deficiencies in the prior art and provide a method, device, medium and equipment for measuring track wear, which solves the problems of poor track segmentation stability and low track wear measurement accuracy in the prior art, and can obtain the measured value of track wear, thereby improving the stability and accuracy of track measurement. To achieve the above purpose, the present invention is implemented by adopting the following technical solutions:

[0007] In a first aspect, the present invention provides a rail wear measurement method, comprising:

[0008] Acquire point cloud data of the track to be measured and point cloud data of the standard track; wherein the point cloud data of the track to be measured includes first contour point cloud data generated by the first laser line and second contour point cloud data generated by the second laser line;

[0009] Input the first contour point cloud data and the second contour point cloud data into the pre-trained PointNet++ network respectively to obtain the first contour rail waist point cloud data and the second contour rail waist point cloud data;

[0010] A standard point is selected from the point cloud data of the standard track, and the standard rail waist point cloud data in the point cloud data of the standard track is aligned with the first contour rail waist point cloud data and the second contour rail waist point cloud data by using the ICP algorithm, and the first corresponding point and the second corresponding point of the standard point are obtained in the first contour rail waist point cloud data and the second contour rail waist point cloud data respectively; a projection plane is created based on the first corresponding point and the second corresponding point of the standard point, and the first contour point cloud data and the second contour point cloud data are projected onto the projection plane to obtain the point cloud data of the projection track;

[0011] The point cloud data of the standard track and the point cloud data of the projection track are aligned using the ICP algorithm to obtain the measurement results of the track wear to be measured.

[0012] In combination with the first aspect, optionally, the pre-trained PointNet++ network is trained by the following steps:

[0013] Obtain point cloud data of the on-site track;

[0014] The generated point cloud data of the track are classified and labeled using semantic segmentation labels. The label value of the semantic segmentation label of the rail waist point cloud data is "1", and the label value of the semantic segmentation label of the rest is "0", thus obtaining the point cloud dataset used to train the PointNet++ network.

[0015] Divide the point cloud dataset into training set, validation set and test set;

[0016] Inputting the training set into a pre-established PointNet++ network, training the pre-established PointNet++ network to obtain a pre-trained PointNet++ network, wherein the pre-trained PointNet++ network learns the distribution characteristics of the rail waist, sets the parameters of the loss function, and outputs a weight file;

[0017] The validation set and the test set are respectively input into the pre-trained PointNet++ network, and the weight file output by the pre-trained PointNet++ network is optimized and evaluated to obtain the weight file of the track based on the PointNet++ network. The weight file of the track based on the PointNet++ network is used to output the track waist point cloud data based on the weight of each point in the weight file of the track based on the PointNet++ network when the point cloud data of the track to be tested is input into the pre-trained PointNet++ network.

[0018] In combination with the first aspect, optionally, obtaining point cloud data of the track includes:

[0019] Using two lasers disposed before and after the track, data of a first laser plane and data of a second laser plane of the track are obtained;

[0020] Using a camera placed above the track, a real-time image of the track is obtained;

[0021] The Zhang Zhengyou calibration method is used to calibrate the internal and external parameters of the camera, and the structured light calibration is performed to obtain the structured light plane equation in the camera coordinate system;

[0022] The data of the first laser plane and the data of the second laser plane are converted into a camera coordinate system to generate point cloud data of the track, wherein the point cloud data of the track includes first contour point cloud data and second contour point cloud data generated by laser lines of two lasers.

[0023] In combination with the first aspect, optionally, the pre-established PointNet++ network includes the following modules:

[0024] Sampling operation module: used to downsample the original point cloud data set to a smaller subset by random sampling or uniform sampling;

[0025] Feature extraction module: It consists of a point cloud feature network and a local feature network, and is used to extract local features and global features from point cloud datasets;

[0026] Hierarchical feature encoding module: used to encode and integrate local features and global features using a recursive hierarchical structure;

[0027] Aggregation module: used to aggregate the outputs of multiple hierarchical feature encoding modules to obtain the global feature expression of the entire point cloud;

[0028] Segmentation module: used to segment the rail waist point cloud data from the original data.

[0029] In combination with the first aspect, optionally, the loss function of the pre-established PointNet++ network is expressed by the following formula:

[0030] L_seg=-∑(y_seg*log(y_pred_seg)) / N_seg (1)

[0031] In formula (1), L_seg is the loss function of the PointNet++ network, y_seg is the label value of the true semantic segmentation label, y_pred_seg is the label value of the predicted semantic segmentation label, and N_seg is the number of points in the point cloud.

[0032] In combination with the first aspect, optionally, performing ICP algorithm registration on the point cloud data of the standard track and the point cloud data of the projection track to obtain the measurement result of the track wear includes:

[0033] The point cloud data of the standard track and the point cloud data of the projection track are aligned by iterative ICP algorithm to obtain a point cloud pair;

[0034] Measure the distance between the center point clouds of the point cloud above the rail head to obtain the measurement result of the wear above the rail head; measure the distance between the center point clouds of the point cloud inside the rail head to obtain the measurement result of the wear inside the rail head;

[0035] The measurement result of rail wear is obtained based on the measurement result of the wear above the rail head and the measurement result of the wear inside the rail head.

[0036] In combination with the first aspect, optionally, the distance between the point clouds in the point cloud pair is calculated by the following formula:

[0037]

[0038] In formula (2), d is the distance between the point clouds in the point cloud pair, P1(x1, y1, z1) is the coordinate of the point cloud in the standard track of the point cloud pair, and P2(x2, y2, z2) is the coordinate of the point cloud in the projection track of the point cloud pair.

[0039] In a second aspect, the present invention provides a rail wear measuring device, comprising:

[0040] An acquisition module is used to acquire point cloud data of the track to be measured and point cloud data of the standard track; wherein the point cloud data of the track to be measured includes first contour point cloud data generated by the first laser line and second contour point cloud data generated by the second laser line;

[0041] Rail waist segmentation module: used to input the first contour point cloud data and the second contour point cloud data into the pre-trained PointNet++ network respectively to obtain the first contour rail waist point cloud data and the second contour rail waist point cloud data;

[0042] Contour correction module: used to select a standard point from the point cloud data of the standard track, and use the ICP algorithm to perform point cloud registration on the standard rail waist point cloud data in the point cloud data of the standard track with the first contour rail waist point cloud data and the second contour rail waist point cloud data, and obtain the first corresponding point and the second corresponding point of the standard point in the first contour rail waist point cloud data and the second contour rail waist point cloud data respectively; create a projection plane based on the first corresponding point and the second corresponding point of the standard point, and project the first contour point cloud data and the second contour point cloud data onto the projection plane to obtain the point cloud data of the projection track;

[0043] Output module: used to perform ICP algorithm registration on the point cloud data of the standard track and the point cloud data of the projection track to obtain the measurement results of the track wear to be measured.

[0044] In a third aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the track wear measurement method as described in the first aspect.

[0045] In a fourth aspect, the present invention provides a device, comprising:

[0046] A memory for storing instructions;

[0047] The processor is used to execute the instructions so that the device performs operations to implement the track wear measurement method as described in the first aspect.

[0048] Compared with the prior art, the track wear measurement method, device, medium and equipment provided by the embodiments of the present invention have the following beneficial effects:

[0049] The present invention obtains point cloud data of a track to be measured and point cloud data of a standard track; the point cloud data of the track to be measured includes first contour point cloud data generated by a first laser line and second contour point cloud data generated by a second laser line; the first contour point cloud data and the second contour point cloud data are respectively input into a pre-trained PointNet++ network to obtain first contour rail waist point cloud data and second contour rail waist point cloud data; the present invention obtains segmented rail waist point cloud data through a pre-trained PointNet++ network, thereby improving the stability and accuracy of segmenting the track from a complex background, and improving the stability of track segmentation and environmental adaptability; the present invention extracts track point cloud data through a deep learning method, thereby not only improving the segmentation accuracy of the track, but also improving the adaptability of the rail wear measurement algorithm to different working conditions;

[0050] The present invention creates a projection plane through the first contour point cloud data and the second contour point cloud data, and corrects the contour of the deformed point cloud data, thereby solving the problem that the generated track point cloud data is deformed due to the jitter of the acquisition device or the tilt of the laser, and improving the measurement accuracy of track wear;

[0051] The present invention matches the corrected projection track with the template track, and then obtains the measurement value of rail wear, thereby improving the stability and accuracy of track measurement;

[0052] The present invention improves the accuracy of track wear measurement by performing ICP algorithm registration on standard track point cloud data and track point cloud data twice. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 is a flow chart of a rail wear measurement method provided in Embodiment 1 of the present invention;

[0054] Figure 2 It is a schematic diagram of a device for obtaining point cloud data of a track in a track wear measurement method provided in Embodiment 1 of the present invention;

[0055] Figure 3 This is a segmentation result diagram of rail waist point cloud data obtained by segmenting the pre-trained PointNet++ network in a rail wear measurement method provided in Example 1 of the present invention.

[0056] Figure 4 It is a projection plane result diagram of a projected track in a track wear measurement method provided in the first embodiment of the present invention;

[0057] Figure 5 This is a track wear matching result diagram in a track wear measurement method provided in Example 1 of the present invention.

[0058] In the figure: 1. First laser; 2. Second laser; 3. Camera; 4. Track. DETAILED DESCRIPTION

[0059] The present invention will be further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the protection scope of the present invention.

[0060] Embodiment 1:

[0061] like Figure 1 As shown, this embodiment provides a rail wear measurement method, including:

[0062] Acquire point cloud data of the track to be measured and point cloud data of the standard track; wherein the point cloud data of the track to be measured includes first contour point cloud data generated by the first laser line and second contour point cloud data generated by the second laser line;

[0063] Input the first contour point cloud data and the second contour point cloud data into the pre-trained PointNet++ network respectively to obtain the first contour rail waist point cloud data and the second contour rail waist point cloud data;

[0064] A standard point is selected from the point cloud data of the standard track, and the standard rail waist point cloud data in the point cloud data of the standard track is aligned with the first contour rail waist point cloud data and the second contour rail waist point cloud data by using the ICP algorithm, and the first corresponding point and the second corresponding point of the standard point are obtained in the first contour rail waist point cloud data and the second contour rail waist point cloud data respectively; a projection plane is created based on the first corresponding point and the second corresponding point of the standard point, and the first contour point cloud data and the second contour point cloud data are projected onto the projection plane to obtain the point cloud data of the projection track;

[0065] The point cloud data of the standard track and the point cloud data of the projection track are aligned using the ICP algorithm to obtain the measurement results of the track wear to be measured.

[0066] The pre-trained PointNet++ network is trained through the following steps:

[0067] Step 1: Obtain point cloud data of the on-site track.

[0068] Use Figure 2 The device for obtaining point cloud data of a track shown in the figure has two lasers arranged in front and behind the track, and the two lasers respectively emit a first laser line and a second laser line. A camera is arranged above the track.

[0069] Step 1.1: Use two lasers disposed before and after the track to obtain data of a first laser plane and a second laser plane of the track.

[0070] Step 1.2: Use a camera placed above the track to obtain a real-time image of the track.

[0071] In this embodiment, the frame rate of the camera is 60fps.

[0072] Step 1.3: Use Zhang Zhengyou calibration method to calibrate the internal and external parameters of the camera, and perform structured light calibration to obtain the structured light plane equation in the camera coordinate system.

[0073] Step 1.4: Convert the data of the first laser plane and the second laser plane into the camera coordinate system to generate point cloud data of the track, wherein the point cloud data of the track includes first contour point cloud data generated by the first laser line and second contour point cloud data generated by the second laser line.

[0074] Preferably, pre-processing operations such as screening, segmentation, and labeling are performed on the collected point cloud data to form a point cloud data set for track wear measurement.

[0075] Step 2: Use semantic segmentation labels to classify and annotate the generated track point cloud data. The label value of the semantic segmentation label of the rail waist point cloud data is "1", and the label value of the semantic segmentation label of the rest is "0", and the point cloud dataset used to train the PointNet++ network is obtained.

[0076] In this embodiment, the labelcloud tool is used to label the point cloud data of the track to generate a label file and original point cloud data.

[0077] Step 3: Divide the point cloud dataset into training set, validation set and test set.

[0078] In this embodiment, the point cloud data set contains 3,000 point cloud data in total, which are divided into a training set, a validation set, and a test set in a ratio of 7:2:1.

[0079] Step 4: Input the training set into the pre-established PointNet++ network, train the pre-established PointNet++ network, and obtain a pre-trained PointNet++ network. The pre-trained PointNet++ network learns the distribution characteristics of the rail waist, sets the parameters of the loss function, and outputs a weight file.

[0080] Specifically, the pre-built PointNet++ network includes the following modules:

[0081] Sampling operation module: used to downsample the original point cloud data set to a smaller subset by random sampling or uniform sampling;

[0082] Feature extraction module: It consists of a point cloud feature network and a local feature network, and is used to extract local features and global features from point cloud datasets;

[0083] Hierarchical feature encoding module: used to encode and integrate local features and global features using a recursive hierarchical structure;

[0084] Aggregation module: used to aggregate the outputs of multiple hierarchical feature encoding modules to obtain the global feature expression of the entire point cloud;

[0085] Segmentation module: used to segment the rail waist point cloud data from the original data.

[0086] Specifically, the loss function of the pre-established PointNet++ network is expressed as follows:

[0087] L_seg=-∑(y_seg*log(y_pred_seg)) / N_seg (1)

[0088] In formula (1), L_seg is the loss function of the PointNet++ network, y_seg is the label value of the true semantic segmentation label, y_pred_seg is the label value of the predicted semantic segmentation label, and N_seg is the number of points in the point cloud.

[0089] Step 5: Input the validation set and the test set into the pre-trained PointNet++ network respectively, optimize and evaluate the weight file output by the pre-trained PointNet++ network, and obtain the weight file of the track based on the PointNet++ network. The weight file of the track based on the PointNet++ network is used to output the track waist point cloud data based on the weight of each point in the weight file of the track based on the PointNet++ network when the point cloud data of the track to be tested is input into the pre-trained PointNet++ network.

[0090] The PointNet++ network pre-trained in this embodiment obtains segmented track waist point cloud data, which improves the stability and accuracy of segmenting the track from the complex background, and improves the stability of track segmentation and the adaptability to the environment.

[0091] This embodiment extracts track point cloud data through a deep learning method, which not only improves the track segmentation accuracy, but also improves the adaptability of the rail wear measurement algorithm to different working conditions.

[0092] like Figure 1 The specific steps of a rail wear measurement method shown include:

[0093] Step 1: Obtain the point cloud data of the track to be tested and the point cloud data of the standard track.

[0094] Use Figure 2 The device for obtaining point cloud data of a track shown in the figure has two lasers arranged in front and behind the track, and the two lasers respectively emit a first laser line and a second laser line. A camera is arranged above the track.

[0095] Step 1.1: Use two lasers disposed before and after the track to obtain data of a first laser plane and a second laser plane of the track.

[0096] Step 1.2: Use a camera placed above the track to obtain a real-time image of the track.

[0097] Step 1.3: Use Zhang Zhengyou calibration method to calibrate the internal and external parameters of the camera, and perform structured light calibration to obtain the structured light plane equation in the camera coordinate system.

[0098] Step 1.4: Convert the data of the first laser plane and the second laser plane into the camera coordinate system to generate point cloud data of the track, wherein the point cloud data of the track includes first contour point cloud data generated by the first laser line and second contour point cloud data generated by the second laser line.

[0099] Step 2: Input the first contour point cloud data and the second contour point cloud data into the pre-trained PointNet++ network respectively to obtain the first contour rail waist point cloud data and the second contour rail waist point cloud data.

[0100] The segmentation module of the pre-trained PointNet++ network is used to segment the rail waist point cloud data from the point cloud data to obtain the following Figure 3 The segmentation result diagram of the rail waist point cloud data shown in the figure respectively includes the first contour rail waist point cloud data and the second contour rail waist point cloud data.

[0101] Step 3.1: Select a standard point from the point cloud data of the standard track.

[0102] In this embodiment, a standard point is selected from the PCL library, and the point where the curvature of the waist of the standard track changes the most is selected as the standard point.

[0103] Step 3.2: Use the ICP algorithm to perform point cloud registration on the standard rail waist point cloud data in the standard track point cloud data with the first contour rail waist point cloud data and the second contour rail waist point cloud data, and obtain the first corresponding point and the second corresponding point of the standard point in the first contour rail waist point cloud data and the second contour rail waist point cloud data respectively.

[0104] Step 3.3: Create a projection plane based on the first corresponding point and the second corresponding point of the standard point, and project the first contour point cloud data and the second contour point cloud data onto the projection plane, such as Figure 4 As shown, the point cloud data of the projection track is obtained.

[0105] In this embodiment, the created projection plane is perpendicular to the line connecting the first corresponding point and the second corresponding point.

[0106] In this embodiment, a projection filter (Project Inliers) algorithm is used in the PCL library to project the first contour point cloud data and the second contour point cloud data onto the projection plane.

[0107] This embodiment creates a projection plane through the first contour point cloud data and the second contour point cloud data, and corrects the contour of the deformed point cloud data, thereby solving the problem of deformation of the generated track point cloud data due to jitter of the acquisition equipment or tilt of the laser, and improving the measurement accuracy of track wear.

[0108] Step 4: Perform ICP algorithm registration on the point cloud data of the standard track and the point cloud data of the projection track to obtain the measurement result of the track wear to be measured.

[0109] Step 4.1: Perform an iterative ICP algorithm to align the point cloud data of the standard track and the point cloud data of the projection track to obtain a point cloud pair.

[0110] In this embodiment, an iterative ICP algorithm is used for registration in the PCL library, and the registration result is as follows: Figure 5 shown.

[0111] This embodiment matches the corrected projection track and the template track, and then obtains the measurement value of rail wear, thereby improving the stability and accuracy of track measurement.

[0112] Step 4.2: Measure the distance between the center point clouds of the point cloud above the rail head to obtain the measurement result of the wear above the rail head; measure the distance between the center point clouds of the point cloud inside the rail head to obtain the measurement result of the wear inside the rail head;

[0113] Specifically, the distance between the point clouds in the point cloud pair is calculated by the following formula:

[0114]

[0115] In formula (2), d is the distance between the point clouds in the point cloud pair, P1(x1, y1, z1) is the coordinate of the point cloud in the standard track of the point cloud pair, and P2(x2, y2, z2) is the coordinate of the point cloud in the projection track of the point cloud pair.

[0116] Step 4.3: According to the measurement results of the wear above the rail head and the measurement results of the wear inside the rail head, the measurement results of the rail wear are obtained.

[0117] The embodiment of the present invention improves the accuracy of track wear measurement by performing ICP algorithm registration on standard track point cloud data and track point cloud data twice.

[0118] Embodiment 2:

[0119] An embodiment of the present invention provides a rail wear measuring device, comprising:

[0120] An acquisition module is used to acquire point cloud data of the track to be measured and point cloud data of the standard track; wherein the point cloud data of the track to be measured includes first contour point cloud data generated by the first laser line and second contour point cloud data generated by the second laser line;

[0121] Rail waist segmentation module: used to input the first contour point cloud data and the second contour point cloud data into the pre-trained PointNet++ network respectively to obtain the first contour rail waist point cloud data and the second contour rail waist point cloud data;

[0122] Contour correction module: used to select a standard point from the point cloud data of the standard track, and use the ICP algorithm to perform point cloud registration on the standard rail waist point cloud data in the point cloud data of the standard track with the first contour rail waist point cloud data and the second contour rail waist point cloud data, and obtain the first corresponding point and the second corresponding point of the standard point in the first contour rail waist point cloud data and the second contour rail waist point cloud data respectively; create a projection plane based on the first corresponding point and the second corresponding point of the standard point, and project the first contour point cloud data and the second contour point cloud data onto the projection plane to obtain the point cloud data of the projection track;

[0123] Output module: used to perform ICP algorithm registration on the point cloud data of the standard track and the point cloud data of the projection track to obtain the measurement results of the track wear to be measured.

[0124] Embodiment three:

[0125] An embodiment of the present invention provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the track wear measurement method as described in the first embodiment is implemented.

[0126] Embodiment 4:

[0127] An embodiment of the present invention further provides a device, including:

[0128] A memory for storing instructions;

[0129] The processor is used to execute the instructions so that the device performs operations to implement the track wear measurement method as described in the first embodiment.

[0130] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0131] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0132] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate 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 A function specified in one or more boxes.

[0133] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0134] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for measuring rail wear, characterized in that: include: Acquire point cloud data of the track to be measured and point cloud data of the standard track; wherein the point cloud data of the track to be measured includes first contour point cloud data generated by the first laser line and second contour point cloud data generated by the second laser line; Input the first contour point cloud data and the second contour point cloud data into the pre-trained PointNet++ network respectively to obtain the first contour rail waist point cloud data and the second contour rail waist point cloud data; A standard point is selected from the point cloud data of the standard track, and the standard rail waist point cloud data in the point cloud data of the standard track is aligned with the first contour rail waist point cloud data and the second contour rail waist point cloud data by using the ICP algorithm, and the first corresponding point and the second corresponding point of the standard point are obtained in the first contour rail waist point cloud data and the second contour rail waist point cloud data respectively; a projection plane is created based on the first corresponding point and the second corresponding point of the standard point, and the first contour point cloud data and the second contour point cloud data are projected onto the projection plane to obtain the point cloud data of the projection track; The point cloud data of the standard track and the point cloud data of the projection track are aligned using the ICP algorithm to obtain the measurement results of the track wear to be measured.

2. The rail wear measurement method according to claim 1, characterized in that: The pre-trained PointNet++ network is trained by the following steps: Obtain point cloud data of the on-site track; The generated point cloud data of the track are classified and labeled using semantic segmentation labels. The label value of the semantic segmentation label of the rail waist point cloud data is "1", and the label value of the semantic segmentation label of the rest is "0", thus obtaining the point cloud dataset used to train the PointNet++ network. Divide the point cloud dataset into training set, validation set and test set; Inputting the training set into a pre-established PointNet++ network, training the pre-established PointNet++ network to obtain a pre-trained PointNet++ network, wherein the pre-trained PointNet++ network learns the distribution characteristics of the rail waist, sets the parameters of the loss function, and outputs a weight file; The validation set and the test set are respectively input into the pre-trained PointNet++ network, and the weight file output by the pre-trained PointNet++ network is optimized and evaluated to obtain the weight file of the track based on the PointNet++ network. The weight file of the track based on the PointNet++ network is used to output the track waist point cloud data based on the weight of each point in the weight file of the track based on the PointNet++ network when the point cloud data of the track to be tested is input into the pre-trained PointNet++ network.

3. The rail wear measurement method according to claim 1 or 2, characterized in that: Get the point cloud data of the track, including: Using two lasers disposed before and after the track, data of a first laser plane and data of a second laser plane of the track are obtained; Using a camera placed above the track, a real-time image of the track is obtained; The Zhang Zhengyou calibration method is used to calibrate the internal and external parameters of the camera, and the structured light calibration is performed to obtain the structured light plane equation in the camera coordinate system; The data of the first laser plane and the data of the second laser plane are converted into a camera coordinate system to generate point cloud data of the track, wherein the point cloud data of the track includes first contour point cloud data and second contour point cloud data generated by laser lines of two lasers.

4. The rail wear measurement method according to claim 2, characterized in that: The pre-built PointNet++ network includes the following modules: Sampling operation module: used to downsample the original point cloud data set to a smaller subset by random sampling or uniform sampling; Feature extraction module: It consists of a point cloud feature network and a local feature network, and is used to extract local features and global features from point cloud datasets; Hierarchical feature encoding module: used to encode and integrate local features and global features using a recursive hierarchical structure; Aggregation module: used to aggregate the outputs of multiple hierarchical feature encoding modules to obtain the global feature expression of the entire point cloud; Segmentation module: used to segment the rail waist point cloud data from the original data.

5. The rail wear measurement method according to claim 2, characterized in that: The loss function of the pre-established PointNet++ network is expressed as follows: L_seg=-∑(y_seg*log(y_pred_seg)) / N_seg (1) In formula (1), L_seg is the loss function of the PointNet++ network, y_seg is the label value of the true semantic segmentation label, y_pred_seg is the label value of the predicted semantic segmentation label, and N_seg is the number of points in the point cloud.

6. The rail wear measurement method according to claim 1, characterized in that: The point cloud data of the standard track and the point cloud data of the projection track are aligned using an ICP algorithm to obtain the measurement result of the track wear, including: The point cloud data of the standard track and the point cloud data of the projection track are aligned by iterative ICP algorithm to obtain a point cloud pair; Measure the distance between the center point clouds of the point cloud above the rail head to obtain the measurement result of the wear above the rail head; measure the distance between the center point clouds of the point cloud inside the rail head to obtain the measurement result of the wear inside the rail head; The measurement result of rail wear is obtained based on the measurement result of the wear above the rail head and the measurement result of the wear inside the rail head.

7. The rail wear measurement method according to claim 6, characterized in that: The distance between the point clouds in the point cloud pair is calculated by the following formula: In formula (2), d is the distance between the point clouds in the point cloud pair, P1(x1, y1, z1) is the coordinate of the point cloud in the standard track of the point cloud pair, and P2(x2, y2, z2) is the coordinate of the point cloud in the projection track of the point cloud pair.

8. A rail wear measuring device, characterized in that: include: An acquisition module is used to acquire point cloud data of the track to be measured and point cloud data of the standard track; wherein the point cloud data of the track to be measured includes first contour point cloud data generated by the first laser line and second contour point cloud data generated by the second laser line; Rail waist segmentation module: used to input the first contour point cloud data and the second contour point cloud data into the pre-trained PointNet++ network respectively to obtain the first contour rail waist point cloud data and the second contour rail waist point cloud data; Contour correction module: used to select a standard point from the point cloud data of the standard track, and use the ICP algorithm to perform point cloud registration on the standard rail waist point cloud data in the point cloud data of the standard track with the first contour rail waist point cloud data and the second contour rail waist point cloud data, and obtain the first corresponding point and the second corresponding point of the standard point in the first contour rail waist point cloud data and the second contour rail waist point cloud data respectively; create a projection plane based on the first corresponding point and the second corresponding point of the standard point, and project the first contour point cloud data and the second contour point cloud data onto the projection plane to obtain the point cloud data of the projection track; Output module: used to perform ICP algorithm registration on the point cloud data of the standard track and the point cloud data of the projection track to obtain the measurement results of the track wear to be measured.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the rail wear measurement method as described in any one of claims 1 to 7 is implemented.

10. A device, characterized in that: include: A memory for storing instructions; The processor is used to execute the instructions so that the device performs the operations of implementing the track wear measurement method as described in any one of claims 1-7.

Images