Method and device for determining serial number of turnout sleeper

Abstract

The invention discloses a turnout sleeper number determination method and device, and the method comprises the steps: obtaining turnout three-dimensional point cloud data, and carrying out the regional division of the turnout three-dimensional point cloud data based on the prior information of a steel rail structure; inputting the turnout three-dimensional point cloud data of different areas into the established three-dimensional point cloud segmentation model, and outputting point cloud data carrying semantic tags; constructing a three-dimensional point cloud segmentation model by taking the hierarchical feature network structure as a first-layer model and the point feature propagation network structure as a second-layer model; inputting the turnout three-dimensional point cloud data of different areas into the hierarchical feature network structure, and outputting point cloud features of the corresponding areas; inputting the point cloud features into a point feature propagation network structure, and outputting point cloud data carrying semantic tags; and determining the serial number of the sleeper according to the point cloud data carrying the semantic tag. According to the method, the positioning precision of the turnout sleeper can be improved, high-precision associated positioning of defect positions is realized, and the safety of railway operation is ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of railway infrastructure, and in particular to a method and device for determining turnout and sleeper numbers. Background Art

[0002] This section is intended to provide a background or context to the embodiments of the invention that are recited in the claims. No statement herein is admitted to be prior art by virtue of its inclusion in this section.

[0003] With the development of the railway economy, ensuring railway safety is crucial. As a key component of the track structure, comprehensive inspection and monitoring of turnouts can promptly identify and repair potential safety hazards, prevent accidents, and ensure the smooth operation of railway transportation. Accurate identification of turnout sleeper numbers affects turnout inspection and positioning accuracy and operational efficiency. Existing technologies are unable to achieve high-precision turnout sleeper number positioning, making it impossible to promptly address track defects, resulting in uncertainties in railway operation safety. Summary of the Invention

[0004] An embodiment of the present invention provides a method for determining turnout sleeper numbers, which is used to improve the accuracy of turnout sleeper positioning, improve the timeliness of track defect treatment, and ensure the safety of railway operation. The method includes:

[0005] Obtain the 3D point cloud data of the turnout. Based on the prior information of the rail structure, divide the 3D point cloud data of the turnout into regions to obtain the rail head point cloud data, rail waist point cloud data, and rail bottom point cloud data.

[0006] Input the three-dimensional point cloud data of turnouts in different areas into the established three-dimensional point cloud segmentation model, and output point cloud data with semantic labels; semantic labels are used to mark the types of turnout components; in the process of establishing the three-dimensional point cloud segmentation model, a hierarchical feature network structure is used as the first layer model and a point feature propagation network structure is used as the second layer model to construct the three-dimensional point cloud segmentation model; input the three-dimensional point cloud data of turnouts in different areas into the hierarchical feature network structure, and output the point cloud features of the corresponding areas; input the point cloud features into the point feature propagation network structure, and output the point cloud data with semantic labels;

[0007] Determine the type and location of turnout components based on semantically labeled point cloud data; the location information of turnout components includes the center position of the sleeper and the tip of the switch rail.

[0008] The sleeper number is determined based on the sleeper center position information and the point rail tip position information.

[0009] The embodiment of the present invention further provides a device for determining turnout sleeper numbers, which is used to improve the accuracy of turnout sleeper positioning, improve the timeliness of track disease treatment, and ensure the safety of railway operation. The device includes:

[0010] The turnout 3D point cloud data region division module is used to obtain the turnout 3D point cloud data. Based on the prior information of the rail structure, the turnout 3D point cloud data is divided into regions to obtain the rail head point cloud data, rail waist point cloud data and rail bottom point cloud data;

[0011] The point cloud data semantic labeling module is used to input the 3D point cloud data of turnouts in different regions into the established 3D point cloud segmentation model and output point cloud data with semantic labels. The semantic labels are used to mark the types of turnout components. In the process of establishing the 3D point cloud segmentation model, the 3D point cloud segmentation model is constructed using a hierarchical feature network structure as the first layer model and a point feature propagation network structure as the second layer model. The 3D point cloud data of turnouts in different regions are input into the hierarchical feature network structure, and the point cloud features of the corresponding regions are output. The point cloud features are input into the point feature propagation network structure, and the point cloud data with semantic labels are output.

[0012] A switch component position information determination module is used to determine the type and position information of the switch component based on the point cloud data carrying semantic tags; the position information of the switch component includes the position information of the center of the sleeper and the position information of the tip of the switch rail;

[0013] The sleeper number determination module is used to determine the sleeper number according to the sleeper center position information and the point rail tip position information.

[0014] An embodiment of the present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned method for determining the turnout sleeper number when executing the computer program.

[0015] An embodiment of the present invention further provides a computer-readable storage medium storing a computer program, which implements the above-mentioned method for determining the turnout sleeper number when executed by a processor.

[0016] An embodiment of the present invention further provides a computer program product, which includes a computer program. When the computer program is executed by a processor, it implements the above-mentioned method for determining the turnout sleeper number.

[0017] In an embodiment of the present invention, by acquiring three-dimensional point cloud data of the turnout, the three-dimensional point cloud data of the turnout is divided into regions based on prior information of the rail structure to obtain rail head point cloud data, rail waist point cloud data and rail bottom point cloud data; the three-dimensional point cloud data of the turnout in different regions are input into the established three-dimensional point cloud segmentation model, and point cloud data with semantic labels are output; the semantic labels are used to mark the types of turnout components; in the process of establishing the three-dimensional point cloud segmentation model, a three-dimensional point cloud segmentation model is constructed with a hierarchical feature network structure as the first layer model and a point feature propagation network structure as the second layer model; the three-dimensional point cloud data of the turnout in different regions are input into the hierarchical feature network structure, and the point cloud features of the corresponding regions are output; the point cloud features are input into the point feature propagation network structure, and the point cloud data with semantic labels are output; the type and position information of the turnout components are determined according to the point cloud data with semantic labels; the position information of the turnout components includes the center position information of the sleeper and the tip position information of the point rail; the number of the sleeper is determined according to the center position information of the sleeper and the tip position information of the point rail. In the above process, the embodiment of the present invention divides the three-dimensional point cloud data of the turnout into regions based on the prior information of the rail structure, outputs point cloud data with semantic labels by inputting the point cloud data of different regions into the three-dimensional point cloud segmentation model, and identifies the sleeper number according to the point cloud data with semantic labels, thereby achieving high-precision correlation positioning of the defect position, improving the timeliness of track disease treatment, and ensuring the safety of railway operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work. In the drawings:

[0019] Figure 1 Flowchart of a method for determining turnout sleeper numbers in an embodiment of the present invention;

[0020] Figure 2 This is a flow chart of preprocessing point cloud data in an embodiment of the present invention;

[0021] Figure 3 Flowchart of turnout 3D point cloud data in a hierarchical feature network structure according to an embodiment of the present invention;

[0022] Figure 4 Schematic diagram of a device for determining turnout sleeper numbers in an embodiment of the present invention;

[0023] Figure 5 Schematic diagram of a computer device in an embodiment of the present invention. DETAILED DESCRIPTION

[0024] To make the purpose, technical solutions and advantages of the embodiments of the present invention more clear, the embodiments of the present invention are further described in detail below with reference to the accompanying drawings. Here, the exemplary embodiments of the present invention and their descriptions are used to explain the present invention, but are not intended to limit the present invention.

[0025] Figure 1 Flowchart of a method for determining turnout sleeper numbers in an embodiment of the present invention, the method comprising:

[0026] Step 101: Acquire three-dimensional point cloud data of a turnout, and divide the three-dimensional point cloud data of the turnout into regions based on prior information of the rail structure to obtain rail head point cloud data, rail waist point cloud data, and rail bottom point cloud data;

[0027] Step 102: Input the three-dimensional point cloud data of the turnouts in different regions into the established three-dimensional point cloud segmentation model, and output point cloud data with semantic labels; the semantic labels are used to mark the types of turnout components; in the process of establishing the three-dimensional point cloud segmentation model, a hierarchical feature network structure is used as the first layer model and a point feature propagation network structure is used as the second layer model to construct the three-dimensional point cloud segmentation model; the three-dimensional point cloud data of the turnouts in different regions are input into the hierarchical feature network structure, and the point cloud features of the corresponding regions are output; the point cloud features are input into the point feature propagation network structure, and the point cloud data with semantic labels are output;

[0028] Step 103: determining the type and position information of the switch component based on the point cloud data carrying the semantic label; the position information of the switch component includes the position information of the center of the sleeper and the position information of the tip of the switch rail;

[0029] Step 104: Determine the sleeper number based on the sleeper center position information and the point rail tip position information.

[0030] Each step is described in detail below.

[0031] In step 101, three-dimensional point cloud data of the turnout is obtained, and based on prior information of the rail structure, the three-dimensional point cloud data of the turnout is divided into regions to obtain rail head point cloud data, rail waist point cloud data and rail bottom point cloud data.

[0032] Figure 2 This is a flow chart of preprocessing point cloud data in an embodiment of the present invention. In one embodiment, before inputting the three-dimensional point cloud data of turnouts in different regions into the established three-dimensional point cloud segmentation model, the process includes:

[0033] Step 201, downsampling the three-dimensional point cloud data of the turnout;

[0034] Input the 3D point cloud data of turnouts in different areas into the established 3D point cloud segmentation model, including:

[0035] Step 202: Input the downsampled three-dimensional point cloud data of the turnout into the established three-dimensional point cloud segmentation model.

[0036] In a specific embodiment, the 3D point cloud data of the turnout is pre-processed and downsampled to reduce the data volume and improve the processing efficiency. The 3D point cloud data of the turnout is divided into different areas, and each area is processed with a different downsampling rate. First, the 3D point cloud data set of the turnout is P. According to the prior information of the rail structure, the 3D point cloud data of the turnout is divided into three different areas: the rail head area P head The amount of data is |P head |, rail waist area P waist The amount of data is |P waist |, rail bottom area P base The amount of data is |P base |, then, the downsampling rate is determined based on the ratio of the data volume of each region to the total data volume, for example, the downsampling rate of the track head region is r head Expressed as:

[0037]

[0038] Similarly, the downsampling rate r of the rail waist area and rail bottom area can be obtained waist and r base .

[0039] In step 102, the three-dimensional point cloud data of the turnouts in different areas are input into the established three-dimensional point cloud segmentation model, and the point cloud data with semantic labels are output; the semantic labels are used to mark the types of turnout components; in the process of establishing the three-dimensional point cloud segmentation model, the three-dimensional point cloud segmentation model is constructed with the hierarchical feature network structure as the first layer model and the point feature propagation network structure as the second layer model; the three-dimensional point cloud data of the turnouts in different areas are input into the hierarchical feature network structure, and the point cloud features of the corresponding areas are output; the point cloud features are input into the point feature propagation network structure, and the point cloud data with semantic labels are output.

[0040] In a specific embodiment, 80 sets of collected three-dimensional point cloud data of turnouts were selected as training and test sets. Each set of data contained 5,000 points and was labeled with 10 categories, including basic rails, separated point rails, close-fitting point rails, guard rails, heart rails, wing rails, ballast, sleepers, fastener bolts, and other categories. In the semantic segmentation results, there are certain misjudgments of the structures between ballast, fasteners, and sleepers. A small number of the sides of the wider part of the heart rail will be judged as wing rails. However, based on the semantically labeled point cloud data output by the three-dimensional point cloud segmentation model, the type and location information of the turnout components can be accurately and efficiently identified, providing data support for determining the turnout sleeper numbers.

[0041] In one embodiment, the three-dimensional point cloud data of the turnouts in different regions are input into a hierarchical feature network structure, and the point cloud features of the corresponding regions are output, including:

[0042] The three-dimensional point cloud data of the turnout is input into the hierarchical feature network structure for partition sampling and local feature extraction, and the point cloud features of the corresponding area are output.

[0043] In a specific embodiment, the hierarchical feature network structure is composed of multiple abstraction layers (SA), and each abstraction layer (SA) includes a sampling layer (Sampling), a grouping layer (Grouping) and a PointNet layer.

[0044] Figure 3 This is a flow chart of the three-dimensional point cloud data of a turnout in a hierarchical feature network structure according to an embodiment of the present invention. In one embodiment, the three-dimensional point cloud data of the turnout is input into the hierarchical feature network structure for partition sampling and local feature extraction, and the point cloud features of the corresponding area are output, including: the hierarchical feature network structure includes a sampling layer, a grouping layer, and a PointNet layer;

[0045] Step 301: Input the three-dimensional point cloud data of the turnouts in different areas into a sampling layer, and independently sample the three-dimensional point cloud data of the turnouts in different areas through the sampling layer to obtain sampling points in different areas;

[0046] Step 302: input sampling points from different regions into the grouping layer, use the K-nearest neighbor algorithm to search for neighborhood points of the sampling points, and output the generated local point set of sampling points;

[0047] In step 303, the local point set of the sampling points is input into the PointNet layer for local feature extraction to obtain the point cloud features of the corresponding area.

[0048] In one embodiment, the point feature propagation network structure includes a fully connected layer and a ReLU activation function.

[0049] In this specific embodiment, downsampling is performed twice within the hierarchical feature network structure. However, semantic segmentation tasks require features from all sampled points. Therefore, a point feature propagation strategy based on distance interpolation and cross-layer connections is employed. The point feature propagation process: First, a weighted average of the inverses of the k nearest neighbor distances is calculated; then, the interpolated features of the sampled points are concatenated with the skip-connected point features of the set abstraction layer; finally, the concatenated features are passed through the PointNet unit, using fully connected layers and ReLU layers to update the feature vector of each point. This process is repeated until the features are propagated to the original point set.

[0050] In step 103, the type and location of the switch component are determined based on the semantically labeled point cloud data. The switch component location information includes the sleeper center location information and the switch rail tip location information. In step 104, the sleeper number is determined based on the sleeper center location information and the switch rail tip location information.

[0051] In the specific embodiment, when laying and maintaining the turnout, operations are carried out strictly in accordance with the design drawings and technical requirements. The turnout sleeper spacing refers to the distance between the sleepers laid in the turnout. This parameter directly affects the stability, driving safety and comfort of the turnout. Because high-speed trains have higher requirements for line stability, the turnout sleeper spacing is carefully calculated and designed. Taking the SC330 turnout as an example, its sleeper spacing setting is representative. In the SC330 turnout, the spacing between the turnout sleepers at the two traction points is 650 mm, the spacing between the turnout sleepers No. 10 and No. 11 is 607 mm, and the spacing between the remaining turnout sleepers is 600 mm.

[0052] Assume W j，j+1 Indicates the distance between the jth switch tie and the j+1th switch tie, j is the number of the switch tie, and it starts from 1. tip Indicates the position of the tip of the switch rail, T pillow，i Indicates the center position of the i-th switch tie, and the center position of the switch tie is calculated based on each switch tie component. L i N is the distance from the tip of the switch rail to the center of the ith switch tie. pillow，start Indicates the turnout sleeper number corresponding to the tip of the switch rail. This is a known value, usually obtained from the turnout design drawing. N pillow，i Indicates the number of the i-th switch tie.

[0053] The calculation process of the turnout tie number: First, the type of turnout component and the position of the switch rail tip T are determined based on the point cloud data with semantic labels. tip and the center position T of each switch tie pillow，i ; Then, calculate the distance L between the center of each switch tie and the tip of the switch rail i .

[0054] L i =||T pillow，i -T tip ||

[0055] According to the prior information of turnout type and sleeper spacing, determine the turnout sleeper number N corresponding to the tip of the switch rail. pillow，start Finally, based on the cumulative distance and the turnout tie spacing, start from the tip of the switch rail and follow the known spacing W. j，j+1 Accumulate the distances in sequence until the distance to the center position T of each switch tie is found. pillow，i The corresponding distance L i The interval where the switch is located can directly determine the number N of each switchpillow,i .

[0056] The present invention also provides a device for determining the number of a turnout sleeper, as described in the following embodiments. Since the principle of the device is similar to the method for determining the number of a turnout sleeper, the implementation of the device can refer to the implementation of the method for determining the number of a turnout sleeper, and the repeated parts will not be repeated.

[0057] Figure 4 Schematic diagram of a device for determining turnout sleeper numbers according to an embodiment of the present invention, the device comprising:

[0058] The turnout 3D point cloud data region division module 401 is used to obtain the turnout 3D point cloud data and divide the turnout 3D point cloud data into regions based on the prior information of the rail structure to obtain the rail head point cloud data, rail waist point cloud data and rail bottom point cloud data;

[0059] The point cloud data semantic labeling module 402 is used to input the three-dimensional point cloud data of the turnouts in different regions into the established three-dimensional point cloud segmentation model and output point cloud data with semantic labels; the semantic labels are used to mark the types of turnout components; in the process of establishing the three-dimensional point cloud segmentation model, the three-dimensional point cloud segmentation model is constructed using a hierarchical feature network structure as the first layer model and a point feature propagation network structure as the second layer model; the three-dimensional point cloud data of the turnouts in different regions are input into the hierarchical feature network structure, and the point cloud features of the corresponding regions are output; the point cloud features are input into the point feature propagation network structure, and the point cloud data with semantic labels are output;

[0060] The switch component position information determination module 403 is used to determine the type and position information of the switch component based on the point cloud data carrying semantic tags; the position information of the switch component includes the position information of the center of the sleeper and the position information of the tip of the switch rail;

[0061] The sleeper number determination module 404 is used to determine the sleeper number according to the sleeper center position information and the point rail tip position information.

[0062] In one embodiment, a data preprocessing module is further included, specifically configured to:

[0063] Before inputting the three-dimensional point cloud data of turnouts in different areas into the established three-dimensional point cloud segmentation model, the three-dimensional point cloud data of turnouts are downsampled;

[0064] Input the 3D point cloud data of turnouts in different areas into the established 3D point cloud segmentation model, including:

[0065] The downsampled three-dimensional point cloud data of the turnout is input into the established three-dimensional point cloud segmentation model.

[0066] In one embodiment, the point cloud data semantic labeling module 402 is specifically configured to:

[0067] The three-dimensional point cloud data of the turnout is input into the hierarchical feature network structure for partition sampling and local feature extraction, and the point cloud features of the corresponding area are output.

[0068] In one embodiment, the point cloud data semantic labeling module 402 is specifically configured to: a hierarchical feature network structure including a sampling layer, a grouping layer, and a PointNet layer;

[0069] Input the three-dimensional point cloud data of the turnouts in different areas into the sampling layer, and independently sample the three-dimensional point cloud data of the turnouts in different areas through the sampling layer to obtain sampling points in different areas;

[0070] Input the sampling points of different areas into the grouping layer, use the K nearest neighbor algorithm to search for the neighborhood points of the sampling points, and output the generated local point set of sampling points;

[0071] The local point set of the sampling points is input into the PointNet layer for local feature extraction to obtain the point cloud features of the corresponding area.

[0072] In one embodiment, the point feature propagation network structure includes a fully connected layer and a ReLU activation function.

[0073] An embodiment of the present invention further provides a computer device, Figure 5 This is a schematic diagram of a computer device in an embodiment of the present invention. The computer device 500 includes a memory 510, a processor 520, and a computer program 530 stored in the memory 510 and executable on the processor 520. When the processor 520 executes the computer program 530, the method for determining the turnout sleeper numbering is implemented.

[0074] An embodiment of the present invention further provides a computer-readable storage medium storing a computer program, which implements the above-mentioned method for determining the turnout sleeper number when executed by a processor.

[0075] An embodiment of the present invention further provides a computer program product, which includes a computer program. When the computer program is executed by a processor, it implements the above-mentioned method for determining the turnout sleeper number.

[0076] In an embodiment of the present invention, by acquiring three-dimensional point cloud data of the turnout, the three-dimensional point cloud data of the turnout is divided into regions based on prior information of the rail structure to obtain rail head point cloud data, rail waist point cloud data and rail bottom point cloud data; the three-dimensional point cloud data of the turnout in different regions are input into the established three-dimensional point cloud segmentation model, and point cloud data with semantic labels are output; the semantic labels are used to mark the types of turnout components; in the process of establishing the three-dimensional point cloud segmentation model, a three-dimensional point cloud segmentation model is constructed with a hierarchical feature network structure as the first layer model and a point feature propagation network structure as the second layer model; the three-dimensional point cloud data of the turnout in different regions are input into the hierarchical feature network structure, and the point cloud features of the corresponding regions are output; the point cloud features are input into the point feature propagation network structure, and the point cloud data with semantic labels are output; the type and position information of the turnout components are determined according to the point cloud data with semantic labels; the position information of the turnout components includes the center position information of the sleeper and the tip position information of the point rail; the number of the sleeper is determined according to the center position information of the sleeper and the tip position information of the point rail. In the above process, the embodiment of the present invention divides the three-dimensional point cloud data of the turnout into regions based on the prior information of the rail structure, outputs point cloud data with semantic labels by inputting the point cloud data of different regions into the three-dimensional point cloud segmentation model, and identifies the sleeper number according to the point cloud data with semantic labels, thereby achieving high-precision correlation positioning of the defect position, improving the timeliness of track disease treatment, and ensuring the safety of railway operation.

[0077] It will be understood by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0078] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts 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, a special-purpose computer, an embedded processor, or other programmable data processing device to produce 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 flowcharts and / or block diagrams. 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.

[0079] 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.

[0080] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0081] The specific embodiments described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for determining turnout sleeper numbers, characterized in that: include: Obtain the 3D point cloud data of the turnout. Based on the prior information of the rail structure, divide the 3D point cloud data of the turnout into regions to obtain the rail head point cloud data, rail waist point cloud data, and rail bottom point cloud data. Input the three-dimensional point cloud data of turnouts in different areas into the established three-dimensional point cloud segmentation model, and output point cloud data with semantic labels; semantic labels are used to mark the types of turnout components; in the process of establishing the three-dimensional point cloud segmentation model, a hierarchical feature network structure is used as the first layer model and a point feature propagation network structure is used as the second layer model to construct the three-dimensional point cloud segmentation model; input the three-dimensional point cloud data of turnouts in different areas into the hierarchical feature network structure, and output the point cloud features of the corresponding areas; input the point cloud features into the point feature propagation network structure, and output the point cloud data with semantic labels; Determine the type and location of turnout components based on semantically labeled point cloud data; the location information of turnout components includes the center position of the sleeper and the tip of the switch rail. The sleeper number is determined based on the sleeper center position information and the point rail tip position information.

2. The method according to claim 1, wherein Before inputting the 3D point cloud data of turnouts in different areas into the established 3D point cloud segmentation model, it includes: Downsampling the turnout 3D point cloud data; Input the 3D point cloud data of turnouts in different areas into the established 3D point cloud segmentation model, including: The downsampled three-dimensional point cloud data of the turnout is input into the established three-dimensional point cloud segmentation model.

3. The method according to claim 1, wherein Input the 3D point cloud data of turnouts in different areas into the hierarchical feature network structure, and output the point cloud features of the corresponding areas, including: The three-dimensional point cloud data of the turnout is input into the hierarchical feature network structure for partition sampling and local feature extraction, and the point cloud features of the corresponding area are output.

4. The method according to claim 3, wherein The three-dimensional point cloud data of the turnout is input into the hierarchical feature network structure for partition sampling and local feature extraction, and the point cloud features of the corresponding area are output, including: the hierarchical feature network structure includes a sampling layer, a grouping layer and a PointNet layer; Input the three-dimensional point cloud data of the turnouts in different areas into the sampling layer, and independently sample the three-dimensional point cloud data of the turnouts in different areas through the sampling layer to obtain sampling points in different areas; Input the sampling points of different areas into the grouping layer, use the K nearest neighbor algorithm to search for the neighborhood points of the sampling points, and output the generated local point set of sampling points; The local point set of the sampling points is input into the PointNet layer for local feature extraction to obtain the point cloud features of the corresponding area.

5. The method according to claim 1, wherein The point feature propagation network structure includes a fully connected layer and a ReLU activation function.

6. A device for determining turnout sleeper numbers, characterized in that: include: The turnout 3D point cloud data region division module is used to obtain the turnout 3D point cloud data. Based on the prior information of the rail structure, the turnout 3D point cloud data is divided into regions to obtain the rail head point cloud data, rail waist point cloud data and rail bottom point cloud data; The point cloud data semantic labeling module is used to input the 3D point cloud data of turnouts in different regions into the established 3D point cloud segmentation model and output point cloud data with semantic labels. The semantic labels are used to mark the types of turnout components. In the process of establishing the 3D point cloud segmentation model, the 3D point cloud segmentation model is constructed using a hierarchical feature network structure as the first layer model and a point feature propagation network structure as the second layer model. The 3D point cloud data of turnouts in different regions are input into the hierarchical feature network structure, and the point cloud features of the corresponding regions are output. The point cloud features are input into the point feature propagation network structure, and the point cloud data with semantic labels are output. A switch component position information determination module is used to determine the type and position information of the switch component based on the point cloud data carrying semantic tags; the position information of the switch component includes the position information of the center of the sleeper and the position information of the tip of the switch rail; The sleeper number determination module is used to determine the sleeper number according to the sleeper center position information and the point rail tip position information.

7. The device according to claim 6, characterized in that It also includes a data preprocessing module, specifically used for: Before inputting the three-dimensional point cloud data of turnouts in different areas into the established three-dimensional point cloud segmentation model, the three-dimensional point cloud data of turnouts are downsampled; Input the 3D point cloud data of turnouts in different areas into the established 3D point cloud segmentation model, including: The downsampled three-dimensional point cloud data of the turnout is input into the established three-dimensional point cloud segmentation model.

8. The device according to claim 6, wherein The point cloud data semantic labeling module is specifically used for: The three-dimensional point cloud data of the turnout is input into the hierarchical feature network structure for partition sampling and local feature extraction, and the point cloud features of the corresponding area are output.

9. The device according to claim 8, wherein The point cloud data semantic labeling module is specifically used for: the hierarchical feature network structure includes the sampling layer, the grouping layer and the PointNet layer; Input the three-dimensional point cloud data of the turnouts in different areas into the sampling layer, and independently sample the three-dimensional point cloud data of the turnouts in different areas through the sampling layer to obtain sampling points in different areas; Input the sampling points of different areas into the grouping layer, use the K nearest neighbor algorithm to search for the neighborhood points of the sampling points, and output the generated local point set of sampling points; The local point set of the sampling points is input into the PointNet layer for local feature extraction to obtain the point cloud features of the corresponding area.

10. The device according to claim 6, wherein The point feature propagation network structure includes a fully connected layer and a ReLU activation function.

11. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the method according to any one of claims 1 to 5 is implemented.

12. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 5 is implemented.

13. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 5 is implemented.

Images