A Rail Transit Fastener Model Identification System and Method

Abstract

This invention discloses a system and method for identifying fastener types in rail transit, relating to the field of fastener maintenance technology. The system comprises: an image acquisition module to acquire initial images of the track, and an image preprocessing module to obtain a reference image; an image analysis module to input the reference image into a preset model for processing, segmenting the fastener components and rail components to obtain segmented images; a positioning module to filter images of some fastener components and rail bases, restoring them to the reference image; a feature acquisition module to obtain reference adjustment lines based on the segmented rail base image, and to obtain feature parameters of some components based on the segmented images of some components, correcting them using the reference adjustment lines to obtain reference feature parameters; and a type identification module to match the reference feature parameters of the selected components with data in a constructed database, outputting the corresponding fastener type. This invention not only achieves efficient identification of fastener types but also ensures high accuracy.

Description

Technical Field

[0001] This invention relates to the field of fastener maintenance technology, and in particular to a fastener model identification system and method for rail transit. Background Technology

[0002] In rail transit systems, fasteners are the core components connecting rails and sleepers. Their main functions are to fix the position of rails, maintain track gauge stability, transfer wheel-rail loads to the track bed, and buffer the vibration and impact of train operation. Fastener inspection is the periodic monitoring of the condition of fasteners through technical means. Its significance and role are directly related to track structure safety, train operation safety, and operational efficiency.

[0003] Traditionally, the dimensions of each track fastener are measured manually. Inspectors need extensive prior knowledge to distinguish different fastener models based on the measured dimensions and determine the safety status of each fastener. However, this method is labor-intensive, inefficient, and lacks precision. With the development of 3D point cloud technology, fastener inspection is now being conducted using this technology. Specifically, the fastener location is located by comparing the detected point cloud image with a priori point cloud image. Then, the 3D point cloud data is analyzed to obtain the fastener's 3D geometric information, thus identifying the fastener model. However, this method has the following drawbacks: processing massive amounts of 3D point cloud data is highly complex, making real-time detection difficult to guarantee; 3D point cloud data is highly susceptible to environmental and equipment limitations, resulting in unstable data quality; separating individual fasteners from massive point clouds is constrained by complex scenarios, leading to low segmentation efficiency; and extracting 3D features from the 3D point cloud data is affected by the fastener's condition, causing instability in the 3D features and insufficient robustness.

[0004] In summary, there is an urgent need for a rail transit fastener type identification system and method to solve the above problems. Summary of the Invention

[0005] In view of this, this application provides a rail transit fastener model identification system and method to address the shortcomings of the existing technology.

[0006] The first aspect of this application provides a rail transit fastener model identification system, including:

[0007] The image acquisition module, through multiple image acquisition devices installed on the track inspection vehicle, acquires initial images including fastener components and rail components;

[0008] The image preprocessing module preprocesses the initial image to obtain a reference image;

[0009] The image analysis module inputs the reference image into a preset model for processing, obtains segmented images of each component in the fastener assembly and stores them in a first preset set, and segments images of each component in the rail assembly and stores them in a second preset set;

[0010] The positioning module selects segmented images of multiple components from the first preset set and segmented images of the rail base from the second preset set, and positions and restores them to the reference image.

[0011] The feature acquisition module obtains a baseline adjustment line based on the segmented image of the rail base in the baseline image; it obtains the feature parameters of the corresponding components based on the segmented images of all the screening components in the baseline image, and corrects the feature parameters of the screening components based on the baseline adjustment line to obtain the baseline feature parameters of the screening components.

[0012] The model identification module constructs a database to store the corresponding fastener size parameters and fastener models. It matches the baseline feature parameters of all filtered components with the size parameters of all fasteners in the database, outputs the corresponding fastener model, and completes the identification of the fastener model.

[0013] In one possible implementation of the first aspect, preprocessing the initial image to obtain a reference image includes:

[0014] The initial image is subjected to denoising and geometric correction to obtain the reference image.

[0015] In one possible implementation of the first aspect, the preset model is the SINet model.

[0016] In one possible implementation of the first aspect, inputting the reference image into a preset model for processing includes:

[0017] The reference images are input into the SINet model in batches for processing, and the images processed by the SINet model are standardized to obtain the segmented images of each component in the fastener assembly and the segmented images of each component in the rail assembly.

[0018] In one possible implementation of the first aspect, filtering segmented images of multiple components from the first preset set includes:

[0019] In the first preset set, obtain the segmented images of all components in a single fastener assembly to obtain multiple first segmented images;

[0020] Based on prior knowledge, all components in a single fastener assembly are divided into symmetrical components, denoted as the first component, and asymmetrical components, denoted as the second component.

[0021] Determine the axisymmetry line in all first components, and divide the edge regions on both sides of the axisymmetry line into multiple verification regions of equal size. Calculate whether the first segmented images of the corresponding verification regions on both sides of the axisymmetry line overlap. If so, determine that the verification region group is successfully verified.

[0022] If the ratio between the number of groups of the verified regions that are successfully verified in a single first component and the total number of groups of all verified regions exceeds a set value, the first segmentation image of the corresponding first component will be used as the first filtering result.

[0023] Construct a standard image library for all asymmetric components. Iterate through the first segmented image of the second component and the corresponding image in the standard image library. If the first segmented image of the second component overlaps with the image in the standard image library, use the first segmented image of the corresponding second component as the second filtering result.

[0024] Based on the first and second filtering results, segmented images of multiple components are filtered from the first preset set.

[0025] In one possible implementation of the first aspect, obtaining the reference adjustment line based on a segmented image of the rail base in the reference image includes:

[0026] Obtain a segmented image of the rail base in the reference image, and denote it as the second segmented image;

[0027] The edge region image of the second segmented image near the fastener assembly is obtained, and it is determined whether there is an occlusion area in the edge region image. If not, the reference adjustment line is obtained by fitting the contour line of the second segmented image near the fastener assembly. If so, the second segmented image is processed to obtain the reference adjustment line.

[0028] Processing the second segmented image includes:

[0029] Using the top vertices on both sides of the second segmented image as a reference, multiple feature points are determined on both sides of the second segmented image at a preset interval. Feature points on both sides of the second segmented image with equal distances from the top vertices are grouped into a set of feature points, and the perpendicular line of each set of feature points is obtained.

[0030] Determine whether the perpendicular lines of all feature points coincide. If they do, use them as the reference adjustment line. If not, select the perpendicular line with the highest degree of coincidence as the reference adjustment line.

[0031] In one possible implementation of the first aspect, obtaining the feature parameters of the corresponding components based on the segmented images of all selected components in the reference image includes:

[0032] Any filter component is obtained and denoted as the first filter component. At the same time, the contour lines on both sides of the image segmented by the first filter component are obtained and straight line fitting is performed to obtain a set of reference straight lines about the first filter component.

[0033] The distance between the set of reference lines is calculated as a feature parameter of the first screening component.

[0034] In one possible implementation of the first aspect, correcting the feature parameters of the screening component based on the baseline adjustment line includes:

[0035] Obtain a set of baseline lines from the first filtering component, and the corresponding feature parameters are denoted as the first feature parameters;

[0036] Based on the principle that the distance between the reference adjustment line and any one of the reference lines in the set of reference lines remains equal, the corresponding two reference lines are adjusted.

[0037] The distance between a set of baseline lines after adjustment is calculated and used as the baseline feature parameter of the first screening component.

[0038] In one possible implementation of the first aspect, the baseline feature parameters of all selected components are matched with the size parameters of all fasteners in the database, and the corresponding fastener model is output, including:

[0039] Match the baseline feature parameters of all selected components with the size parameters of any fastener corresponding to any component in the database until a match is found.

[0040] The fastener model that is successfully matched will be output as the corresponding fastener model for all filter components.

[0041] The second aspect of this application provides a method for identifying the type of rail transit fasteners, including:

[0042] Initial images of fastener assemblies and rail assemblies are captured using multiple image acquisition devices installed on the rail inspection vehicle.

[0043] The initial image is preprocessed to obtain a reference image;

[0044] The reference image is input into a preset model for processing to obtain segmented images of each component in the fastener assembly and store them in a first preset set, and segmented images of each component in the rail assembly and store them in a second preset set;

[0045] The segmented images of multiple components selected from the first preset set and the segmented images of the rail base selected from the second preset set are all located and restored to the reference image;

[0046] Based on the segmented image of the rail base in the reference image, a reference adjustment line is obtained; based on the segmented images of all filtering components in the reference image, the feature parameters of the corresponding components are obtained, and based on the reference adjustment line, the feature parameters of the filtering components are corrected to obtain the reference feature parameters of the filtering components.

[0047] A database is constructed to store the corresponding fastener size parameters and fastener models. The baseline feature parameters of all selected components are matched with the size parameters of all fasteners in the database, and the corresponding fastener models are output to complete the identification of fastener models.

[0048] The beneficial effects are as follows: This invention discloses a rail transit fastener model recognition system and method. It acquires initial images of fastener components and rail components through an image acquisition module, and preprocesses them using an image preprocessing module to obtain a reference image. An image analysis module inputs the reference image into a preset model for processing, completing the segmentation of each fastener component and each rail component, and obtaining corresponding segmented images. A positioning module selects images of some components from the segmented images of the fastener components and images of the rails from the segmented images of the rail components, and positions and restores them to the reference image. A feature acquisition module obtains a reference adjustment line based on the segmented image of the rail base in the reference image, and then obtains the feature parameters of the selected components based on the segmented images of the selected components in the reference image. Simultaneously, the feature parameters are corrected using the reference adjustment line to obtain reference feature parameters. A model recognition module matches the reference feature parameters of the selected components with data in a database storing fastener size parameters and corresponding fastener models, outputting the corresponding fastener model, thus completing the fastener model recognition. This invention not only achieves efficient fastener model recognition but also ensures high accuracy. Attached Figure Description

[0049] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0050] Figure 1 This is a schematic diagram of the composition of a rail transit fastener model identification system provided in an embodiment of this application;

[0051] Figure 2 This is an example diagram of a feature acquisition module in a rail transit fastener model identification system provided in an embodiment of this application;

[0052] Figure 3 This is a schematic flowchart of a method for identifying the type of rail transit fasteners provided in an embodiment of this application. Detailed Implementation

[0053] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0054] In this application, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0055] Example 1

[0056] In existing technologies, 3D point cloud technology has been used to detect fasteners. Specifically, the location of the fastener is found by comparing the detected point cloud image with a prior point cloud image. Then, the 3D geometric information of the fastener is obtained by analyzing the 3D point cloud data, thereby completing the identification of the fastener model. However, this method has the following drawbacks: the processing of massive 3D point clouds is highly complex, making it difficult to guarantee the real-time performance of the detection; the 3D point cloud data is highly limited by the environment and equipment, and its quality is unstable; when separating a single fastener from massive point clouds, it is subject to many constraints from complex scenes, resulting in low segmentation efficiency; when extracting the 3D features of the fastener from the 3D point cloud data, the 3D features are unstable due to the condition of the fastener, resulting in insufficient robustness.

[0057] Therefore, this application provides a rail transit fastener model identification system, such as... Figure 1 As shown, it includes:

[0058] The image acquisition module, through multiple image acquisition devices installed on the track inspection vehicle, acquires initial images including fastener components and rail components;

[0059] The image preprocessing module preprocesses the initial image to obtain a reference image;

[0060] The image analysis module inputs the reference image into a preset model for processing, obtains segmented images of each component in the fastener assembly and stores them in a first preset set, and segments images of each component in the rail assembly and stores them in a second preset set;

[0061] The positioning module selects segmented images of multiple components from the first preset set and segmented images of the rail base from the second preset set, and positions and restores them to the reference image.

[0062] The feature acquisition module obtains a baseline adjustment line based on the segmented image of the rail base in the baseline image; it obtains the feature parameters of the corresponding components based on the segmented images of all the screening components in the baseline image, and corrects the feature parameters of the screening components based on the baseline adjustment line to obtain the baseline feature parameters of the screening components.

[0063] The model identification module constructs a database to store the corresponding fastener size parameters and fastener models. It matches the baseline feature parameters of all filtered components with the size parameters of all fasteners in the database, outputs the corresponding fastener model, and completes the identification of the fastener model.

[0064] This embodiment provides a rail transit fastener type identification system, specifically including:

[0065] Image Acquisition Module: This module consists of multiple image acquisition devices mounted on the track inspection vehicle. It acquires initial images of the fastener assemblies and rail assemblies. The specific image acquisition devices comprise a light source, an industrial high-speed 3D camera, and a supplementary light source. During image acquisition, the light source and supplementary light source provide illumination. Four high-speed industrial cameras capture initial images of both sides of the two rails (image resolution 3200*1024). It should be noted that fasteners in rail transit are intermediate parts connecting the rails and sleepers, mainly including fastening components, pads, and gauge adjustment components. The specific components vary depending on the type of sleeper. Rail assemblies mainly include the rail body, rail connecting parts, and rail anti-creep components. In this embodiment, the image acquisition devices primarily acquire images of the rail base and some sections of the rail connecting parts.

[0066] The image preprocessing module preprocesses the initial image to obtain a reference image. The preprocessing methods are denoising and geometric correction. Denoising can be done using Gaussian filtering, median filtering, adaptive filtering, etc., which are not specifically limited in this embodiment. Geometric correction is to eliminate image geometric distortion caused by lens distortion or installation deviation and to unify the image coordinate system. It can be done using distortion correction algorithms or perspective transformation algorithms, which are not specifically limited in this embodiment.

[0067] The image analysis module inputs the reference image into the SINet model for processing. SINet is a type of deep learning-based neural network model widely used in image recognition, segmentation, and enhancement. The core idea behind image segmentation is to accurately locate the target region and generate a pixel-level segmentation mask by simulating visual perception mechanisms, combining attention mechanisms and multi-scale feature fusion. After processing by the SINet model, the segmented images of each component in the fastener assembly are stored in a first preset set, and the segmented images of each component in the rail assembly are stored in a second preset set. Furthermore, this embodiment also performs standardization processing on all segmented images. The core objective is to eliminate irrelevant interference factors in the images, improve data quality and consistency, and lay the foundation for subsequent analysis, modeling, or applications.

[0068] The positioning module filters segmented images of multiple components from a first preset set and segmented images of the rail base from a preset set, all of which are then located and restored to a reference image. Existing technologies process massive 3D point clouds of fastener components to obtain corresponding dimensional parameters for fastener model identification. However, this method is limited by the acquisition environment or equipment, and lacks an adjustment reference. Extracting 3D features of the fasteners from the 3D point cloud data is unstable, potentially leading to distortion of subsequent fastener dimensional parameters and affecting fastener model identification. Therefore, this embodiment sets up a positioning module that, when acquiring the feature parameters of the fastener components, first filters the segmented images of the components to ensure the quality and integrity of subsequent image processing. The segmented image of the rail base is then used for subsequent adjustment. The rail itself is used as the adjustment reference because, in rail transit systems, the rail, as the core load-bearing component of the track structure and the direct guiding structure for trains, possesses high stability and accuracy in its geometry and position. Furthermore, the filtering logic for the segmented images of the components is as follows:

[0069] In rail transit, most of the components of the fastener are arranged symmetrically, but there are also a small number of asymmetrical components. Therefore, this embodiment first distinguishes all components based on prior knowledge, and symmetric components are referred to as the first component, while asymmetrical components (some special fastening parts, such as irregularly shaped elastic strips, eccentric fasteners, etc.) are referred to as the second component.

[0070] The axis of symmetry in the first component is determined, and the edge areas on both sides of the axis of symmetry are divided into multiple verification areas of equal size. For example, the edge areas on both sides of the axis of symmetry of the gauge baffle are divided into 10 m x m rectangular areas. If the corresponding verification areas on both sides overlap, the verification area is considered to be successfully verified. If there are 9 groups of verified areas in a single first component, the ratio between the number of groups of verified areas and the number of groups of all verification areas is 0.9, while the set value is 0.8, the first segmented image of the first component is used as the first screening result. Since the fastener model is identified by using the fastener component size parameters, the corresponding width or length parameters of the component are calculated. Therefore, the above screening can ensure that there will be no distortion or large error when obtaining the corresponding component feature parameters in the future.

[0071] As for the asymmetric component, i.e. the second component, since it accounts for a small proportion, this embodiment constructs a standard image library for all asymmetric components, and iterates through the segmented image of the second component and the corresponding image in the standard image library. If they overlap, the segmented image of the corresponding second component is used as the second filtering result.

[0072] Finally, combining the first and second screening results, segmented images of multiple components are selected from the first preset set, such as the final segmented images of the gauge baffle and the iron pad.

[0073] For details on the feature acquisition module, please refer to [link / reference]. Figure 2 First, the baseline adjustment line is obtained using the segmented image of the rail base in the baseline image. Then, based on the segmented images of all filtering components in the baseline image, the feature parameters of the corresponding components are obtained. These feature parameters are then corrected using the baseline adjustment line to obtain the baseline feature parameters of the filtering components. The processing logic for obtaining the baseline adjustment line using the segmented image of the rail base is as follows:

[0074] Although the rail, as the core load-bearing component of the track structure and the direct guiding structure for trains, possesses high stability and precision in its geometry and position, its rail base area is easily obscured by stones and other debris during long-term daily operation. Therefore, this embodiment first acquires a segmented image of the rail base, denoted as the second segmented image, and then acquires an edge region image to detect whether there is an obscured area. The detection algorithm can use a target detection model (such as YOLO, Faster R-CNN), and the identification of obscured areas is a relatively mature technology in this field, which will not be elaborated on in this embodiment. If there is no obscured area in the rail base edge region image, the contour line of the second segmented image near the fastener assembly is acquired and fitted to obtain the baseline adjustment line, such as... Figure 2The straight line Lr is the baseline adjustment line. If there is an occlusion area in the image of the bottom edge of the track, fitting the contour line may result in some errors. Therefore, in this embodiment, we first determine the two vertices at the top of the second segmented image, and then determine multiple feature points (e.g., 6, covering the collected bottom area of ​​the track as much as possible) at equal intervals on both sides based on the two vertices. The points with equal intervals are a group of feature points. Then, we obtain the perpendicular line of each group of feature points. Theoretically, this is the midline of the second segmented image (bottom area of ​​the track). If all perpendicular lines coincide, they can be directly used as the baseline adjustment line. If they do not coincide completely, the perpendicular line with the highest degree of coincidence is selected as the baseline adjustment line. If there are 5 groups of perpendicular lines that coincide, this can minimize the influence of the occlusion area on the determination of the baseline adjustment line.

[0075] Based on the segmented images of all selected components, the feature parameter processing logic for obtaining the corresponding components is as follows:

[0076] The contour lines on both sides of the segmented image of any filtering component are obtained and fitted with straight lines to obtain a set of reference straight lines. The distances between these reference straight lines are calculated and processed, i.e., the actual size parameters of the component are obtained by acquiring the image resolution, and used as feature parameters of the filtering component. Figure 2 The characteristic parameters of the center gauge baffle are the fitted straight line L. g With L t The distance b between them, the characteristic parameters of the insulating block are the fitted straight line L j With L jr The distance 'a' between them.

[0077] The logic for correcting the feature parameters of the screening components using the baseline adjustment line is as follows:

[0078] Obtain a set of reference lines for the first screening component, and record the corresponding feature parameters as the first feature parameters; based on the fact that the distance between the reference adjustment line and any one of the reference lines in the set of reference lines remains equal (even if the two reference lines remain parallel to the reference adjustment line), adjust the corresponding two reference lines; calculate the distance between the adjusted set of reference lines as the reference feature parameters of the first screening component.

[0079] The model identification module constructs a database to store the corresponding fastener size parameters and fastener models. It matches the baseline feature parameters of all filtered components with the size parameters of all fasteners in the database, outputs the corresponding fastener model, and completes the identification of the fastener model.

[0080] Existing technologies rely on processing massive amounts of 3D point cloud data to obtain fastener size parameters for fastener model identification. However, this method is limited by the acquisition environment or equipment, and lacks adjustment benchmarks. Extracting 3D features from the fasteners from the 3D point cloud data is unstable, potentially leading to distortion of the fastener size parameters and affecting fastener model identification. This embodiment, however, constructs an image acquisition module, an image preprocessing module, an image analysis module, a positioning module, a feature acquisition module, and a model identification module. Employing image vision, it not only filters the fastener images to ensure the accuracy of subsequent fastener feature parameter acquisition but also uses segmented images of the rail base to determine benchmark adjustment lines. After obtaining the corresponding fastener feature parameters from the filtered segmented images, the benchmark adjustment lines are used for correction to obtain benchmark feature parameters. Finally, the benchmark feature parameters (which need to be restored to their actual size after obtaining image resolution) are matched with the fastener parameters stored in the database to output the corresponding fastener model, thus completing the fastener model identification.

[0081] In some embodiments, preprocessing the initial image to obtain a reference image includes:

[0082] The initial image is subjected to denoising and geometric correction to obtain the reference image.

[0083] In some embodiments, the preset model is the SINet model.

[0084] In some embodiments, inputting the reference image into a preset model for processing includes:

[0085] The reference images are input into the SINet model in batches for processing, and the images processed by the SINet model are standardized to obtain the segmented images of each component in the fastener assembly and the segmented images of each component in the rail assembly.

[0086] In some embodiments, filtering segmented images of multiple components from the first preset set includes:

[0087] In the first preset set, obtain the segmented images of all components in a single fastener assembly to obtain multiple first segmented images;

[0088] Based on prior knowledge, all components in a single fastener assembly are divided into symmetrical components, denoted as the first component, and asymmetrical components, denoted as the second component.

[0089] Determine the axisymmetry line in all first components, and divide the edge regions on both sides of the axisymmetry line into multiple verification regions of equal size. Calculate whether the first segmented images of the corresponding verification regions on both sides of the axisymmetry line overlap. If so, determine that the verification region group is successfully verified.

[0090] If the ratio between the number of groups of the verified regions that are successfully verified in a single first component and the total number of groups of all verified regions exceeds a set value, the first segmentation image of the corresponding first component will be used as the first filtering result.

[0091] Construct a standard image library for all asymmetric components. Iterate through the first segmented image of the second component and the corresponding image in the standard image library. If the first segmented image of the second component overlaps with the image in the standard image library, use the first segmented image of the corresponding second component as the second filtering result.

[0092] Based on the first and second filtering results, segmented images of multiple components are filtered from the first preset set.

[0093] In some embodiments, obtaining the reference adjustment line based on a segmented image of the rail base in the reference image includes:

[0094] Obtain a segmented image of the rail base in the reference image, and denote it as the second segmented image;

[0095] The edge region image of the second segmented image near the fastener assembly is obtained, and it is determined whether there is an occlusion area in the edge region image. If not, the reference adjustment line is obtained by fitting the contour line of the second segmented image near the fastener assembly. If so, the second segmented image is processed to obtain the reference adjustment line.

[0096] Processing the second segmented image includes:

[0097] Using the top vertices on both sides of the second segmented image as a reference, multiple feature points are determined on both sides of the second segmented image at a preset interval. Feature points on both sides of the second segmented image with equal distances from the top vertices are grouped into a set of feature points, and the perpendicular line of each set of feature points is obtained.

[0098] Determine whether the perpendicular lines of all feature points coincide. If they do, use them as the reference adjustment line. If not, select the perpendicular line with the highest degree of coincidence as the reference adjustment line.

[0099] In some embodiments, obtaining the feature parameters of the corresponding components based on the segmented images of all selected components in the reference image includes:

[0100] Any filter component is obtained and denoted as the first filter component. At the same time, the contour lines on both sides of the image segmented by the first filter component are obtained and straight line fitting is performed to obtain a set of reference straight lines about the first filter component.

[0101] The distance between the set of reference lines is calculated as a feature parameter of the first screening component.

[0102] In some embodiments, adjusting the feature parameters of the screening component based on the baseline adjustment line includes:

[0103] Obtain a set of baseline lines from the first filtering component, and the corresponding feature parameters are denoted as the first feature parameters;

[0104] Based on the principle that the distance between the reference adjustment line and any one of the reference lines in the set of reference lines remains equal, the corresponding two reference lines are adjusted.

[0105] The distance between a set of baseline lines after adjustment is calculated and used as the baseline feature parameter of the first screening component.

[0106] In some embodiments, the baseline feature parameters of all selected components are matched with the size parameters of all fasteners in the database, and the corresponding fastener model is output, including:

[0107] Match the baseline feature parameters of all selected components with the size parameters of any fastener corresponding to any component in the database until a match is found.

[0108] The fastener model that is successfully matched will be output as the corresponding fastener model for all filter components.

[0109] Example 2

[0110] Based on the rail transit fastener model identification system provided in Embodiment 1 of this application, correspondingly, Embodiment 2 of this application also provides a rail transit fastener model identification method, such as... Figure 3 As shown, it includes:

[0111] Initial images of fastener assemblies and rail assemblies are captured using multiple image acquisition devices installed on the rail inspection vehicle.

[0112] The initial image is preprocessed to obtain a reference image;

[0113] The reference image is input into a preset model for processing to obtain segmented images of each component in the fastener assembly and store them in a first preset set, and segmented images of each component in the rail assembly and store them in a second preset set;

[0114] The segmented images of multiple components selected from the first preset set and the segmented images of the rail base selected from the second preset set are all located and restored to the reference image;

[0115] Based on the segmented image of the rail base in the reference image, a reference adjustment line is obtained; based on the segmented images of all filtering components in the reference image, the feature parameters of the corresponding components are obtained, and based on the reference adjustment line, the feature parameters of the filtering components are corrected to obtain the reference feature parameters of the filtering components.

[0116] A database is constructed to store the corresponding fastener size parameters and fastener models. The baseline feature parameters of all selected components are matched with the size parameters of all fasteners in the database, and the corresponding fastener models are output to complete the identification of fastener models.

[0117] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computing software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0118] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0119] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A rail transit fastener model identification system, characterized in that, include: The image acquisition module, through multiple image acquisition devices installed on the track inspection vehicle, acquires initial images including fastener components and rail components; The image preprocessing module preprocesses the initial image to obtain a reference image; The image analysis module inputs the reference image into a preset model for processing, obtains segmented images of each component in the fastener assembly and stores them in a first preset set, and segments images of each component in the rail assembly and stores them in a second preset set; The positioning module selects segmented images of multiple components from the first preset set and segmented images of the rail base from the second preset set, and positions and restores them to the reference image. The feature acquisition module obtains a baseline adjustment line based on the segmented image of the rail base in the baseline image; it obtains the feature parameters of the corresponding components based on the segmented images of all the screening components in the baseline image, and corrects the feature parameters of the screening components based on the baseline adjustment line to obtain the baseline feature parameters of the screening components. The model identification module constructs a database to store the corresponding fastener size parameters and fastener models. It matches the baseline feature parameters of all filtered components with the size parameters of all fasteners in the database, outputs the corresponding fastener model, and completes the identification of the fastener model.

2. The rail transit fastener model identification system according to claim 1, characterized in that, Preprocessing the initial image to obtain a reference image includes: The initial image is subjected to denoising and geometric correction to obtain the reference image.

3. The rail transit fastener model identification system according to claim 1, characterized in that, The preset model is the SINet model.

4. The rail transit fastener model identification system according to claim 3, characterized in that, The process of inputting the reference image into a preset model includes: The reference images are input into the SINet model in batches for processing, and the images processed by the SINet model are standardized to obtain the segmented images of each component in the fastener assembly and the segmented images of each component in the rail assembly.

5. A rail transit fastener model identification system according to claim 1, characterized in that, Filtering segmented images of multiple components from the first preset set includes: In the first preset set, obtain the segmented images of all components in a single fastener assembly to obtain multiple first segmented images; Based on prior knowledge, all components in a single fastener assembly are divided into symmetrical components, denoted as the first component, and asymmetrical components, denoted as the second component. Determine the axisymmetry line in all first components, and divide the edge regions on both sides of the axisymmetry line into multiple verification regions of equal size. Calculate whether the first segmented images of the corresponding verification regions on both sides of the axisymmetry line overlap. If so, determine that the verification region group is successfully verified. If the ratio between the number of groups of the verified regions that are successfully verified in a single first component and the total number of groups of all verified regions exceeds a set value, the first segmentation image of the corresponding first component will be used as the first filtering result. Construct a standard image library for all asymmetric components. Iterate through the first segmented image of the second component and the corresponding image in the standard image library. If the first segmented image of the second component overlaps with the image in the standard image library, use the first segmented image of the corresponding second component as the second filtering result. Based on the first and second filtering results, segmented images of multiple components are filtered from the first preset set.

6. The rail transit fastener model identification system according to claim 1, characterized in that, Based on the segmented image of the rail base in the reference image, obtaining the reference adjustment line includes: Obtain a segmented image of the rail base in the reference image, and denote it as the second segmented image; The edge region image of the second segmented image near the fastener assembly is obtained, and it is determined whether there is an occlusion area in the edge region image. If not, the reference adjustment line is obtained by fitting the contour line of the second segmented image near the fastener assembly. If so, the second segmented image is processed to obtain the reference adjustment line. Processing the second segmented image includes: Using the top vertices on both sides of the second segmented image as a reference, multiple feature points are determined on both sides of the second segmented image at a preset interval. Feature points on both sides of the second segmented image with equal distances from the top vertices are grouped into a set of feature points, and the perpendicular line of each set of feature points is obtained. Determine whether the perpendicular lines of all feature points coincide. If they do, use them as the reference adjustment line. If not, select the perpendicular line with the highest degree of coincidence as the reference adjustment line.

7. A rail transit fastener model identification system according to claim 6, characterized in that, Based on the segmented images of all selected components in the benchmark image, the feature parameters of the corresponding components are obtained, including: Any filter component is obtained and denoted as the first filter component. At the same time, the contour lines on both sides of the image segmented by the first filter component are obtained and straight line fitting is performed to obtain a set of reference straight lines about the first filter component. The distance between the set of reference lines is calculated as a feature parameter of the first screening component.

8. A rail transit fastener model identification system according to claim 7, characterized in that, Based on the aforementioned baseline adjustment line, the correction of the feature parameters of the screening component includes: Obtain a set of baseline lines from the first filtering component, and the corresponding feature parameters are denoted as the first feature parameters; Based on the principle that the distance between the reference adjustment line and any one of the reference lines in the set of reference lines remains equal, the corresponding two reference lines are adjusted. The distance between a set of baseline lines after adjustment is calculated and used as the baseline feature parameter of the first screening component.

9. A rail transit fastener model identification system according to claim 1, characterized in that, Match the baseline feature parameters of all selected components with the size parameters of all fasteners in the database, and output the corresponding fastener models, including: Match the baseline feature parameters of all selected components with the size parameters of any fastener corresponding to any component in the database until a match is found. The fastener model that is successfully matched will be output as the corresponding fastener model for all filter components.

10. A method for identifying the model of rail transit fasteners, characterized in that, include: Initial images of fastener assemblies and rail assemblies are captured using multiple image acquisition devices installed on the rail inspection vehicle. The initial image is preprocessed to obtain a reference image; The reference image is input into a preset model for processing to obtain segmented images of each component in the fastener assembly and store them in a first preset set, and segmented images of each component in the rail assembly and store them in a second preset set; The segmented images of multiple components selected from the first preset set and the segmented images of the rail base selected from the second preset set are all located and restored to the reference image; Based on the segmented image of the rail base in the reference image, a reference adjustment line is obtained; Based on the segmented images of all filtering components in the reference image, the feature parameters of the corresponding components are obtained, and the feature parameters of the filtering components are corrected based on the reference adjustment line to obtain the reference feature parameters of the filtering components. A database is constructed to store the corresponding fastener size parameters and fastener models. The baseline feature parameters of all selected components are matched with the size parameters of all fasteners in the database, and the corresponding fastener models are output to complete the identification of fastener models.

Images