A method and system for detecting defects in the finger plates of a single-rail track

Abstract

This invention discloses a method and system for detecting defects in finger plates on a single-track rail, belonging to the field of finger plate defect detection technology. The method involves acquiring image data of the rail surface, preprocessing it, and then inputting it into a deep learning target detection model to extract the rectangular frame region and finger plate category of the finger plate. The rectangular frame region is then segmented to extract the finger plate area. An opening operation is performed on the finger plate area to extract the number of fingers in the finger region; if the number of fingers does not meet the requirement, a finger detection anomaly is determined. Based on the finger plate finger region, the shortest distance from a single finger to the base of an opposing finger is measured; if the distance exceeds the corresponding finger length, a finger detachment is determined. A closing operation is performed on the finger plate finger region, and a collision is determined by judging whether there is an overlapping area. This invention combines image data with deep learning technology to determine whether adjacent finger plates have detection anomalies, finger detachment, or collisions, achieving intelligent rail inspection and improving inspection efficiency.

Description

Technical Field

[0001] This invention relates to the field of finger plate defect detection technology, and in particular to a method and system for detecting defects in finger plates of a single-track rail. Background Technology

[0002] Straddle-type monorail lines consist of multiple track beams with joints between them. If these joints are not properly treated during vehicle operation, problems such as concrete breakage and detachment at the beam ends and uneven vehicle running can occur. Currently, monorail tracks generally use finger plates installed at the wheel-carrying points on the track beam ends to ensure smooth and safe vehicle operation. However, over long-term operation, the track beams can deform and tilt due to external factors, and the fixing bolts for the finger plates can loosen or fall off, failing to secure the finger plates. This can lead to abnormalities such as finger contact and finger detachment. If not addressed promptly, these defects can cause finger plate deformation and detachment, seriously endangering vehicle safety.

[0003] Currently, the detection methods for finger-collision and finger-detachment defects in monorail track fingers still rely on manual inspection, which is not only costly in terms of manpower but also inefficient. To address these issues, there is an urgent need for a method and system for detecting defects in monorail track fingers. Summary of the Invention

[0004] In view of this, this application provides a method and system for detecting defects in the finger plates of a single-track rail, in order to overcome the shortcomings of the existing technology.

[0005] The first aspect of this application provides a method for detecting defects in the finger plates of a single-rail track, including:

[0006] The image acquisition equipment installed on the track inspection vehicle simultaneously acquires three-dimensional images and two-dimensional texture images of the track surface;

[0007] After preprocessing the two-dimensional texture image of the track surface, it is input into a deep learning object detection model to extract the rectangular region of the finger plate and the finger plate category;

[0008] The three-dimensional image is cropped based on the extracted rectangular area of ​​the finger plate, and the cropped three-dimensional image is segmented and cropped using a preset algorithm to extract the finger plate area.

[0009] A morphological opening operation is performed on the finger plate region to extract the finger regions corresponding to two adjacent beams, and the number of fingers in the corresponding finger regions is obtained; if the number of fingers in the corresponding finger regions does not meet the number of fingers corresponding to the finger plate category, the finger detection of the finger plate is determined to be abnormal.

[0010] Based on the finger regions corresponding to the two adjacent beams, the shortest distance from a single finger to the root of the opposing finger is measured. If the shortest distance exceeds the finger length corresponding to the finger type of the finger plate, it is determined that the finger plate finger has dislodged.

[0011] After performing morphological closing operations on the corresponding finger regions of the two adjacent beams, it is determined whether there is an overlapping area between the two finger regions. If so, the width of each finger is measured in the overlapping area. If the measured finger width exceeds the finger width corresponding to the finger type of the finger plate, it is determined that the finger plate fingers are touching. If not, the preset parameters of each finger are measured in the two finger regions. If the measured preset parameters of the fingers meet the set conditions, it is determined that the finger plate fingers are touching.

[0012] In one possible implementation of the first aspect, performing a morphological opening operation on the finger-shaped plate region includes:

[0013] Perform an opening operation on the finger-shaped plate region to separate and extract the base plate regions of the two finger-shaped plates;

[0014] Subtract the finger plate area from the base plate area of ​​the two finger plates to extract all finger areas;

[0015] Calculate the distance from each finger region to the base plate regions of the two finger-shaped plates, assign each finger region to the base plate region that is closer, and complete the extraction of the corresponding finger regions of two adjacent beams.

[0016] In one possible implementation of the first aspect, measuring the shortest distance from a single finger to the base of the opposing finger includes:

[0017] Obtain the edge region at the connection between the base plate region and the finger region of any finger-shaped plate;

[0018] A straight line is fitted to the contour line of the edge region near the finger region to obtain an edge straight line;

[0019] Iterate through all pixels within the area of ​​a single finger and calculate the distance to the corresponding edge line of the opposite finger plate. Obtain multiple distance results and store them in a distance set.

[0020] The minimum value in the set of distances is selected as the shortest distance from a single finger to the base of the opposite finger.

[0021] In one possible implementation of the first aspect, performing a morphological closing operation on the corresponding finger regions of the two adjacent beams includes:

[0022] The finger regions corresponding to the two adjacent beams are expanded using structural elements to obtain expanded finger regions;

[0023] The morphological closing operation of the finger regions corresponding to the two adjacent beams is completed by eroding the expanded finger regions using the structural element.

[0024] In one possible implementation of the first aspect, measuring preset parameters for each finger in the two finger regions includes:

[0025] Measure the width of each finger in the two finger regions and the minimum lateral distance between each finger and all fingers on the opposing finger plate.

[0026] In one possible implementation of the first aspect, if the preset parameters of the measured finger meet the set conditions, determining that the finger on the fingertip is touching the finger includes:

[0027] The specified conditions are:

[0028]

[0029] in, To measure the width of each finger, The standard width corresponds to the aforementioned finger-shaped plate category. The maximum threshold that the finger width is allowed to exceed for the finger plate category; For a single finger area, For a single finger area The set of all finger areas of the opposing finger plate. For calculation and The minimum lateral distance between them The minimum threshold for the lateral distance between a single finger area and all finger areas of the opposing finger plate.

[0030] In one possible implementation of the first aspect, the deep learning object detection model adopts the YOLO model.

[0031] In one possible implementation of the first aspect, the method includes:

[0032] Construct a fingerboard type database, and based on the extracted fingerboard categories, obtain the number of fingers, finger length, and finger width of the corresponding fingerboard from the fingerboard type database.

[0033] In one possible implementation of the first aspect, the preset algorithm employs a region growing algorithm.

[0034] A second aspect of this application provides a defect detection system for single-rail track finger plates, comprising:

[0035] The image acquisition module, consisting of an image acquisition device installed on the track inspection vehicle, simultaneously acquires three-dimensional images and two-dimensional texture images of the track surface;

[0036] The preprocessing module preprocesses the two-dimensional texture image of the track surface and then inputs it into the deep learning target detection model to extract the rectangular frame region of the finger plate and the finger plate category;

[0037] The segmentation and cropping module crops the three-dimensional image based on the extracted rectangular area of ​​the finger plate, and uses a preset algorithm to segment and crop the cropped three-dimensional image to extract the finger plate area;

[0038] The finger anomaly detection module performs morphological opening operations on the finger plate area, extracts the finger areas corresponding to two adjacent beams, and obtains the number of fingers in the corresponding finger areas; if the number of fingers in the corresponding finger areas does not meet the number of fingers corresponding to the finger plate category, then the finger detection of the finger plate is determined to be abnormal.

[0039] The finger detachment determination module measures the shortest distance from a single finger to the root of the opposing finger based on the finger regions corresponding to the two adjacent beams. If the shortest distance exceeds the finger length corresponding to the finger type of the finger plate, it determines that the finger plate finger has detached.

[0040] The finger-touch detection module performs morphological closing operations on the corresponding finger regions of two adjacent beams to determine whether there is an overlapping area between the two finger regions. If so, it measures the width of each finger within the overlapping area. If the measured finger width exceeds the finger width corresponding to the finger type of the finger plate, it determines that the finger plate is touching the finger. If not, it measures the preset parameters of each finger in the two finger regions. If the measured preset parameters of the fingers meet the set conditions, it determines that the finger plate is touching the finger.

[0041] Its beneficial effects are as follows: This invention discloses a method and system for detecting defects in single-track finger plates. It acquires three-dimensional and two-dimensional texture images of the track surface, preprocesses the two-dimensional texture images, and inputs them into a deep learning target detection model to extract the rectangular frame region and finger plate category of the finger plate. The rectangular frame region of the finger plate is segmented and cropped to extract the finger plate region. An opening operation is performed on the finger plate region to extract the number of fingers in the finger region. If the number of fingers does not meet the corresponding number for the finger plate category, an abnormal finger detection is determined. Based on the finger plate finger region, the shortest distance from a single finger to the root of the opposing finger is measured. If the distance exceeds the finger length corresponding to the finger plate category, a finger is determined to be detached. A closing operation is performed on the finger plate finger region. If there is an overlapping area in the finger region, the finger width within the overlapping area is measured and exceeds the finger width corresponding to the finger plate category, indicating finger contact. If there is no overlapping area in the finger region, the finger contact is determined by measuring the width of each finger and the minimum lateral distance from the opposing finger, and meeting the set conditions. This invention uses image data combined with deep learning technology to determine whether there are detection anomalies, finger detachment, or finger collisions between adjacent finger plates, thereby achieving intelligent track inspection and improving inspection efficiency. Attached Figure Description

[0042] 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.

[0043] Figure 1 This is a schematic flowchart of a method for detecting defects in the finger plates of a single-track rail, provided in an embodiment of this application.

[0044] Figure 2 This is a schematic diagram of the corresponding architecture of a single-track finger plate defect detection method provided in an embodiment of this application;

[0045] Figure 3 This is a schematic diagram of different areas of the finger plate in a single-track finger plate defect detection method provided in an embodiment of this application;

[0046] Figure 4 This is a schematic diagram of the components of a single-track finger plate defect detection system provided in an embodiment of this application. Detailed Implementation

[0047] 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.

[0048] 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.

[0049] Example 1

[0050] In the current technology, the detection method for finger-collision and finger-detachment defects of finger plates is still based on manual inspection, which is not only costly in terms of manpower, but also inefficient.

[0051] Therefore, this application provides a method for detecting defects in the finger plates of a single-rail track, such as... Figure 1 Shown, including:

[0052] The image acquisition equipment installed on the track inspection vehicle simultaneously acquires three-dimensional images and two-dimensional texture images of the track surface;

[0053] After preprocessing the two-dimensional texture image of the track surface, it is input into a deep learning object detection model to extract the rectangular region of the finger plate and the finger plate category;

[0054] The three-dimensional image is cropped based on the extracted rectangular area of ​​the finger plate, and the cropped three-dimensional image is segmented and cropped using a preset algorithm to extract the finger plate area.

[0055] A morphological opening operation is performed on the finger plate region to extract the finger regions corresponding to two adjacent beams, and the number of fingers in the corresponding finger regions is obtained; if the number of fingers in the corresponding finger regions does not meet the number of fingers corresponding to the finger plate category, the finger detection of the finger plate is determined to be abnormal.

[0056] Based on the finger regions corresponding to the two adjacent beams, the shortest distance from a single finger to the root of the opposing finger is measured. If the shortest distance exceeds the finger length corresponding to the finger type of the finger plate, it is determined that the finger plate finger has dislodged.

[0057] After performing morphological closing operations on the corresponding finger regions of the two adjacent beams, it is determined whether there is an overlapping area between the two finger regions. If so, the width of each finger is measured in the overlapping area. If the measured finger width exceeds the finger width corresponding to the finger type of the finger plate, it is determined that the finger plate fingers are touching. If not, the preset parameters of each finger are measured in the two finger regions. If the measured preset parameters of the fingers meet the set conditions, it is determined that the finger plate fingers are touching.

[0058] The image acquisition equipment, mounted on the track inspection vehicle, simultaneously captures both 3D and 2D texture images of the track surface. This equipment utilizes a line-structured light 3D camera, which can simultaneously acquire pixel-aligned 2D texture and 3D images. The camera can then... Figure 2 The architecture of the vision module fully covers all five surfaces of the track beam. In addition, the image acquisition device is equipped with an industrial control computer, which contains a data acquisition module and a data processing module. Historical data is also stored in the industrial control computer. The signal processing module is an embedded board that receives vehicle speed information in real time and outputs sampling signals at fixed frequencies and equal intervals. The 3D camera receives the trigger signals and acquires three-dimensional images and two-dimensional texture images of the track surface at fixed intervals. The data processing module receives and processes the acquired data, saves the detection results to a database, and displays them for feedback to maintenance personnel for on-site repairs. It should be noted that this embodiment only combines... Figure 2 This embodiment illustrates how image acquisition and processing are achieved using the architecture distribution described above. The same function can also be achieved through other devices, and this embodiment does not impose any specific limitations.

[0059] The process involves preprocessing the 2D texture data, including denoising filtering and geometric correction, before inputting it into a deep learning object detection model to detect the installation position of the finger plates at the end of the track beam, thereby extracting the rectangular bounding box region and the finger plate category. The deep learning object detection model uses the YOLO model, specifically its one-stage detector mode, which offers faster inference speeds compared to the two-stage detector mode, meeting real-time requirements. The YOLO deep learning model detection process consists of two steps: data acquisition, preprocessing, and annotation; and training the YOLO deep learning model, selecting the model that performs best on the test set for deployment.

[0060] By constructing a fingerboard type database, a mapping relationship is established between different fingerboard types and the corresponding number of fingers, finger length, and finger width. After extracting the fingerboard category, the corresponding number of fingers, finger length, and finger width can be obtained by matching it with the fingerboard type database.

[0061] The process involves cropping the 3D image based on the extracted finger-shaped bounding box region. A region generation algorithm is then used to perform secondary segmentation and cropping on the cropped 3D image, extracting the finger-shaped bounding box region from the segmented region at the center of the bounding box. The region generation algorithm consists of four steps: 1) initializing seed points; 2) adding adjacent pixels from the initialized seed points to a region based on their height difference being less than a certain threshold; 3) for each newly added pixel, checking whether its neighboring pixels meet the region growth criteria; if so, adding it to the current region; 4) stopping the segmentation when no more pixels meet the growth criteria, or when all pixels have been assigned to a certain region.

[0062] The process involves performing a morphological opening operation on the finger-shaped plate region to extract the corresponding finger regions of two adjacent beams and obtain the number of fingers in each region. Specifically, this includes: performing an opening operation on the finger-shaped plate region to separate and extract the base plate regions of the two finger-shaped plates; subtracting the finger-shaped plate region from the base plate regions of the two finger-shaped plates to extract all finger regions; calculating the distance from each finger region to the base plate regions of the two finger-shaped plates and assigning each finger region to the closer base plate region, thus completing the extraction of the corresponding finger regions of two adjacent beams. If the number of fingers in a corresponding finger region does not meet the standard number of fingers for that type of finger-shaped plate, please refer to [the relevant documentation]. Figure 3 If the finger detection is abnormal, the finger detection in the finger plate will be determined, and the abnormal information will be reported to the maintenance personnel for timely handling.

[0063] Specifically, based on the finger regions corresponding to the two adjacent beams, the shortest distance from a single finger to the root of the opposing finger is measured. If the shortest distance exceeds the finger length corresponding to the finger plate category, the finger is determined to be dislodged. Specifically, based on the morphological opening operation on the finger plate region, the edge region connecting the finger plate region and the finger region is extracted. A straight line is fitted to the contour line of the edge region near the finger region to obtain an edge straight line. Each point within the finger region is traversed, and the distance from the point to this edge straight line is calculated. The minimum value is found as the shortest distance from a single finger to the root of the opposing finger. If a coordinate system is constructed, let the equation of the edge straight line be... , , and All are constants. , Determine the direction and slope of the edge line, and , They cannot both be 0. The position of the edge line in the coordinate system is determined; the coordinates of any point P in the finger region in this coordinate system are ( , Then the distance from point P to the edge line L is:

[0064]

[0065] Calculate the distance from any point in the finger area to the edge line using the above formula, and then select the minimum value, which is the distance from the finger area to the edge line.

[0066] The core principle of performing morphological closing operations on the corresponding finger regions of two adjacent beams is to use morphological operations to repair defects in the region, strengthen structural continuity, and provide a more complete and reliable feature basis for subsequent finger region analysis (such as overlap detection and size measurement). The operation logic is to first expand and then erode. That is, the original finger region is expanded by using structural elements (such as small rectangular or circular templates to define the scale of the operation), expanding the boundary of the finger region and filling small gaps, holes or connecting adjacent small fractures inside the region. Then, the expanded region is eroded using the same structural elements to shrink the boundary of the finger region and restore the original size of the finger region, but retains the defect repair area filled in during the expansion stage.

[0067] In this embodiment, after performing morphological closing operations on the corresponding finger regions of two adjacent beams, it is determined whether there is an overlapping region between the two finger regions. The algorithm for this determination can be direct detection based on region intersection operations, indirect determination based on minimum distance, or collision detection based on contour boundaries; this embodiment does not specify a particular algorithm. The reason why this embodiment first determines whether there is an overlapping region in the finger regions is that when fingers in two finger regions touch, there may be overlapping touches and contact touches. Of course, the detection methods for these two situations are different. The specific detection method is as follows:

[0068] If it is determined that there is an overlapping area between two finger areas, the width of each finger is measured within the overlapping area. If the measured finger width exceeds the width of the finger corresponding to the finger type, it is determined that the finger is touching.

[0069] If it is determined that there is no overlapping area between the two finger regions, the width of each finger in the two finger regions is measured first, and then the minimum lateral distance between any finger and the opposite finger is measured. Then, the following conditions are used for judgment:

[0070]

[0071] To measure the width of each finger, The standard width corresponds to the aforementioned finger-shaped plate category. The maximum threshold that the finger width is allowed to exceed for the finger plate category; For a single finger area, For a single finger area The set of all finger areas of the opposing finger plate. For calculation and The minimum lateral distance between them The minimum threshold for the lateral distance between a single finger area and all finger areas of the opposing finger plate.

[0072] When it is determined that there is no overlapping area between two fingers, the core principle of detecting finger touching is to combine the conditions of excessive finger width and excessively close lateral distance between fingers. The essence of finger touching is structural interference caused by abnormal contact or excessive closeness between fingers. If it has been determined that there is no overlapping area between the fingers, then relying solely on excessive width (such as measurement noise) may lead to errors. In this embodiment, by combining the conditions of excessive finger width and excessively close lateral distance between fingers, when the condition of excessive finger width is met, it indicates that the fingers are suspected of overlapping (indirect evidence). When the condition of excessively close lateral distance between fingers is met, it indicates that the spatial position of the fingers is interfered (direct evidence). Under dual verification, it can be determined that the two fingers have actually made contact or are seriously too close, that is, it is determined to be finger touching.

[0073] Furthermore, in this embodiment, the finger status data of the inspection and detection finger plate is saved to the database. After the inspection is completed, the historical data of the finger plate status is analyzed offline to determine whether there are any potential abnormal finger plates, so as to avoid risks in advance.

[0074] This embodiment provides a method for detecting defects in the finger plates of a single-rail track, by employing...

[0075] In some embodiments, performing a morphological opening operation on the finger-shaped plate region includes:

[0076] Perform an opening operation on the finger-shaped plate region to separate and extract the base plate regions of the two finger-shaped plates;

[0077] Subtract the finger plate area from the base plate area of ​​the two finger plates to extract all finger areas;

[0078] Calculate the distance from each finger region to the base plate regions of the two finger-shaped plates, assign each finger region to the base plate region that is closer, and complete the extraction of the corresponding finger regions of two adjacent beams.

[0079] In some embodiments, measuring the shortest distance from a single finger to the base of the finger includes:

[0080] Obtain the edge region at the connection between the base plate region and the finger region of any finger-shaped plate;

[0081] A straight line is fitted to the contour line of the edge region near the finger region to obtain an edge straight line;

[0082] Iterate through all pixels within the area of ​​a single finger and calculate the distance to the corresponding edge line of the opposite finger plate. Obtain multiple distance results and store them in a distance set.

[0083] The minimum value in the set of distances is selected as the shortest distance from a single finger to the base of the opposite finger.

[0084] In some embodiments, performing a morphological closing operation on the corresponding finger regions of two adjacent beams includes:

[0085] The finger regions corresponding to the two adjacent beams are expanded using structural elements to obtain expanded finger regions;

[0086] The morphological closing operation of the finger regions corresponding to the two adjacent beams is completed by eroding the expanded finger regions using the structural element.

[0087] In some embodiments, measuring preset parameters for each finger in two finger regions includes:

[0088] Measure the width of each finger in the two finger regions and the minimum lateral distance between each finger and all fingers on the opposing finger plate.

[0089] In some embodiments, if the preset parameters of the measured finger meet the set conditions, determining that the finger plate touches the finger includes:

[0090] The specified conditions are:

[0091]

[0092] in, To measure the width of each finger, The standard width corresponds to the aforementioned finger-shaped plate category. The maximum threshold that the finger width is allowed to exceed for the finger plate category; For a single finger area, For a single finger area The set of all finger areas of the opposing finger plate. For calculation and The minimum lateral distance between them The minimum threshold for the lateral distance between a single finger area and all finger areas of the opposing finger plate.

[0093] In some embodiments, the deep learning object detection model employs the YOLO model.

[0094] In some embodiments, the method includes:

[0095] Construct a fingerboard type database, and based on the extracted fingerboard categories, obtain the number of fingers, finger length, and finger width of the corresponding fingerboard from the fingerboard type database.

[0096] In some embodiments, the preset algorithm employs a region growing algorithm.

[0097] Example 2

[0098] Based on the single-track finger plate defect detection method provided in Embodiment 1 of this application, correspondingly, Embodiment 2 of this application also provides a single-track finger plate defect detection system, such as... Figure 4 Shown, including:

[0099] The image acquisition module, consisting of an image acquisition device installed on the track inspection vehicle, simultaneously acquires three-dimensional images and two-dimensional texture images of the track surface;

[0100] The preprocessing module preprocesses the two-dimensional texture image of the track surface and then inputs it into the deep learning target detection model to extract the rectangular frame region of the finger plate and the finger plate category;

[0101] The segmentation and cropping module crops the three-dimensional image based on the extracted rectangular area of ​​the finger plate, and uses a preset algorithm to segment and crop the cropped three-dimensional image to extract the finger plate area;

[0102] The finger anomaly detection module performs morphological opening operations on the finger plate area, extracts the finger areas corresponding to two adjacent beams, and obtains the number of fingers in the corresponding finger areas; if the number of fingers in the corresponding finger areas does not meet the number of fingers corresponding to the finger plate category, then the finger detection of the finger plate is determined to be abnormal.

[0103] The finger detachment determination module measures the shortest distance from a single finger to the root of the opposing finger based on the finger regions corresponding to the two adjacent beams. If the shortest distance exceeds the finger length corresponding to the finger type of the finger plate, it determines that the finger plate finger has detached.

[0104] The finger-touch detection module performs morphological closing operations on the corresponding finger regions of two adjacent beams to determine whether there is an overlapping area between the two finger regions. If so, it measures the width of each finger within the overlapping area. If the measured finger width exceeds the finger width corresponding to the finger type of the finger plate, it determines that the finger plate is touching the finger. If not, it measures the preset parameters of each finger in the two finger regions. If the measured preset parameters of the fingers meet the set conditions, it determines that the finger plate is touching the finger.

[0105] The specific principles and execution processes of each unit in the single-track finger plate defect detection system disclosed in Embodiment 2 of this application are the same as those in the single-track finger plate defect detection method disclosed in Embodiment 1 of this application. Please refer to the corresponding parts in the single-track finger plate defect detection method disclosed in Embodiment 1 of this application, and they will not be repeated here.

[0106] 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.

[0107] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0108] 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 method for detecting defects in the finger plates of a single-rail track, characterized in that, include: The image acquisition equipment installed on the track inspection vehicle simultaneously acquires three-dimensional images and two-dimensional texture images of the track surface; After preprocessing the two-dimensional texture image of the track surface, it is input into a deep learning object detection model to extract the rectangular region of the finger plate and the finger plate category; The three-dimensional image is cropped based on the extracted rectangular area of ​​the finger plate, and the cropped three-dimensional image is segmented and cropped using a preset algorithm to extract the finger plate area. A morphological opening operation is performed on the finger plate region to extract the finger regions corresponding to two adjacent beams, and the number of fingers in the corresponding finger regions is obtained; if the number of fingers in the corresponding finger regions does not meet the number of fingers corresponding to the finger plate category, the finger detection of the finger plate is determined to be abnormal. Based on the finger regions corresponding to the two adjacent beams, the shortest distance from a single finger to the root of the opposing finger is measured. If the shortest distance exceeds the finger length corresponding to the finger type of the finger plate, it is determined that the finger plate finger has dislodged. After performing morphological closing operations on the corresponding finger regions of the two adjacent beams, it is determined whether there is an overlapping area between the two finger regions. If so, the width of each finger is measured in the overlapping area. If the measured finger width exceeds the finger width corresponding to the finger type of the finger plate, it is determined that the finger plate fingers are touching. If not, the preset parameters of each finger are measured in the two finger regions. If the measured preset parameters of the fingers meet the set conditions, it is determined that the finger plate fingers are touching.

2. The method for detecting defects in the finger plates of a single-rail track according to claim 1, characterized in that, Performing a morphological opening operation on the finger-shaped plate region includes: Perform an opening operation on the finger-shaped plate region to separate and extract the base plate regions of the two finger-shaped plates; Subtract the finger plate area from the base plate area of ​​the two finger plates to extract all finger areas; Calculate the distance from each finger region to the base plate regions of the two finger-shaped plates, assign each finger region to the base plate region that is closer, and complete the extraction of the corresponding finger regions of two adjacent beams.

3. The method for detecting defects in the finger plates of a single-rail track according to claim 2, characterized in that, Measuring the shortest distance from a single finger to the base of the opposite finger includes: Obtain the edge region at the connection between the base plate region and the finger region of any finger-shaped plate; A straight line is fitted to the contour line of the edge region near the finger region to obtain an edge straight line; Iterate through all pixels within the area of ​​a single finger and calculate the distance to the corresponding edge line of the opposite finger plate. Obtain multiple distance results and store them in a distance set. The minimum value in the set of distances is selected as the shortest distance from a single finger to the base of the opposite finger.

4. The method for detecting defects in the finger plates of a single-rail track according to claim 1, characterized in that, Performing a morphological closing operation on the corresponding finger regions of two adjacent beams includes: The finger regions corresponding to the two adjacent beams are expanded using structural elements to obtain expanded finger regions; The morphological closing operation of the finger regions corresponding to the two adjacent beams is completed by eroding the expanded finger regions using the structural element.

5. The method for detecting defects in the finger plates of a single-rail track according to claim 1, characterized in that, The preset parameters for measuring each finger in the two finger areas include: Measure the width of each finger in the two finger regions and the minimum lateral distance between each finger and all fingers on the opposing finger plate.

6. The method for detecting defects in the finger plates of a single-rail track according to claim 5, characterized in that, If the preset parameters for measuring the finger meet the set conditions, then determining that the finger touches the finger on the finger shape plate includes: The specified conditions are: To measure the width of each finger, The standard width corresponds to the aforementioned finger-shaped plate category. The maximum threshold that the finger width is allowed to exceed for the finger plate category; For a single finger area, For a single finger area The set of all finger areas of the opposing finger plate. For calculation and The minimum lateral distance between them The minimum threshold for the lateral distance between a single finger area and all finger areas of the opposing finger plate.

7. The method for detecting defects in the finger plates of a single-rail track according to claim 1, characterized in that, The deep learning object detection model uses the YOLO model.

8. The method for detecting defects in the finger plates of a single-rail track according to claim 1, characterized in that, Also includes: Construct a fingerboard type database, and based on the extracted fingerboard categories, obtain the number of fingers, finger length, and finger width of the corresponding fingerboard from the fingerboard type database.

9. The method for detecting defects in the finger plates of a single-rail track according to claim 1, characterized in that, The preset algorithm uses a region growing algorithm.

10. A defect detection system for finger plates of a single-rail track, characterized in that, include: The image acquisition module, consisting of an image acquisition device installed on the track inspection vehicle, simultaneously acquires three-dimensional images and two-dimensional texture images of the track surface; The preprocessing module preprocesses the two-dimensional texture image of the track surface and then inputs it into the deep learning target detection model to extract the rectangular frame region of the finger plate and the finger plate category; The segmentation and cropping module crops the three-dimensional image based on the extracted rectangular area of ​​the finger plate, and uses a preset algorithm to segment and crop the cropped three-dimensional image to extract the finger plate area; The finger anomaly detection module performs morphological opening operations on the finger plate area, extracts the finger areas corresponding to two adjacent beams, and obtains the number of fingers in the corresponding finger areas; if the number of fingers in the corresponding finger areas does not meet the number of fingers corresponding to the finger plate category, then the finger detection of the finger plate is determined to be abnormal. The finger detachment determination module measures the shortest distance from a single finger to the root of the opposing finger based on the finger regions corresponding to the two adjacent beams. If the shortest distance exceeds the finger length corresponding to the finger type of the finger plate, it determines that the finger plate finger has detached. The finger-touch detection module performs morphological closing operations on the corresponding finger regions of two adjacent beams to determine whether there is an overlapping area between the two finger regions. If so, it measures the width of each finger within the overlapping area. If the measured finger width exceeds the finger width corresponding to the finger type of the finger plate, it determines that the finger plate is touching the finger. If not, it measures the preset parameters of each finger in the two finger regions. If the measured preset parameters of the fingers meet the set conditions, it determines that the finger plate is touching the finger.

Images