Steel rail light band abnormity identification method and device

Abstract

The invention provides a steel rail light band abnormity identification method and device, and the method comprises the steps: obtaining a two-dimensional gray image and a three-dimensional point cloud image of the surface of a pixel-level aligned steel rail through a sensor, rapidly positioning a steel rail head region in the three-dimensional point cloud image through a threshold segmentation method, and then carrying out the recognition of the abnormity of the steel rail light band. And searching a corresponding rail head area in the two-dimensional gray level image of the surface of the steel rail to serve as a reference datum for eccentricity calculation of the steel rail light band, and on the basis, calculating the eccentricity degree of the steel rail light band in the reference area. The method can effectively solve the problem of light band eccentricity detection errors caused by rail head positioning deviation in a traditional method, and guarantees the accuracy of steel rail light band eccentricity calculation under complex working conditions.

Description

Technical Field

[0001] The present invention relates to the field of rail transportation technology, and in particular to a method and device for identifying abnormalities in a rail light band. Background Art

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

[0003] Problems such as too narrow, too wide and eccentric rail light bands are the key rail light band defects in the field of rail transportation technology. Currently, the known technology mainly uses on-board linear array cameras to obtain two-dimensional grayscale images of the rail surface under high-speed motion conditions, and identifies abnormal rail light band defects based on machine vision principles.

[0004] The optical properties of the top surface of the rail are complex. The light band area on the top surface of the rail where the wheel and rail frequently contact each other has a strong mirror reflection ability. In the two-dimensional grayscale image, this area is generally visibly bright. In the area on the top surface of the rail where the wheel and rail rarely contact each other, it is generally severely rusted and mainly diffusely reflected. In the two-dimensional grayscale image, this area is generally visibly dark. Therefore, the imaging dynamic range of the top surface of the rail is very large. However, due to the limited dynamic range of the camera, it is difficult to obtain clear images of the light band area and non-light band area of ​​the rail at the same time. In addition, oil stains, foreign matter, damage, etc. on the top surface of the rail are all shown as black patches on the two-dimensional grayscale image, which are difficult to distinguish, resulting in a very high false alarm rate for abnormal recognition of the top surface of the rail. In addition, factors such as changes in vehicle posture, changes in lighting angle, and ambient light interference can also cause the problem of being unable to effectively distinguish between the rail head and the rail bottom. The existing technology only describes the surface state of the rail through a two-dimensional grayscale image. However, the two-dimensional grayscale image cannot accurately reflect the morphological characteristics of the rail light band area and the non-light band area at the same time, and cannot effectively distinguish between the rail head and the rail bottom area, resulting in the failure of the rail light band abnormality recognition detection. Summary of the invention

[0005] The embodiment of the present invention provides a rail light band anomaly recognition method, which is used to effectively solve the problem of light band recognition anomaly detection error caused by the inability to effectively separate the rail head and the rail bottom in the traditional method. The method includes:

[0006] Obtain pixel-level aligned 2D grayscale image and 3D point cloud image of the rail surface;

[0007] Locate the rail head area in the 3D point cloud image;

[0008] According to the rail head area located in the three-dimensional point cloud image, the corresponding rail head area in the two-dimensional grayscale image is found;

[0009] The rail light band area is determined in the corresponding rail head area in the two-dimensional grayscale image;

[0010] Identify abnormalities in the rail light band area.

[0011] The embodiment of the present invention further provides a rail light band abnormality identification device, which is used to effectively solve the problem of light band abnormality identification detection error caused by the inability to effectively separate the rail head and the rail bottom in the traditional method. The device includes:

[0012] An image acquisition module, used for acquiring a two-dimensional grayscale image and a three-dimensional point cloud image of the rail surface aligned at the pixel level;

[0013] A 3D point cloud image positioning module, used to locate the rail head area in the 3D point cloud image;

[0014] A two-dimensional grayscale image positioning module is used to find the corresponding rail head area in the two-dimensional grayscale image according to the rail head area located in the three-dimensional point cloud image;

[0015] A light band area acquisition module is used to determine the rail light band area corresponding to the rail head area in the two-dimensional grayscale image;

[0016] The light band area anomaly recognition module is used to identify anomalies in the rail light band area.

[0017] An embodiment of the present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned method for identifying abnormalities in rail light bands when executing the computer program.

[0018] An embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the above-mentioned method for identifying anomalies in a rail light band is implemented.

[0019] An embodiment of the present invention further provides a computer program product, which includes a computer program. When the computer program is executed by a processor, the above-mentioned method for identifying anomalies in a rail light band is implemented.

[0020] The rail light band anomaly recognition method and device of the embodiment of the present invention first obtains a two-dimensional grayscale image and a three-dimensional point cloud image of the rail surface aligned at the pixel level; then locates the rail head area in the three-dimensional point cloud image; according to the rail head area located in the three-dimensional point cloud image, finds the corresponding rail head area in the two-dimensional grayscale image; determines the rail light band area based on the corresponding rail head area in the two-dimensional grayscale image; finally, performs anomaly recognition on the rail light band area. This method can effectively solve the problem of light band anomaly detection error caused by rail head positioning deviation in traditional methods, and ensure the accuracy of rail light band anomaly detection under complex working conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings:

[0022] Figure 1 Schematic diagram of a rail light band abnormality recognition method in an embodiment of the present invention;

[0023] Figure 2 A schematic diagram of a process for obtaining a two-dimensional grayscale image and a three-dimensional point cloud image of a rail surface aligned at the pixel level in an embodiment of the present invention;

[0024] Figure 3 This is a schematic diagram of a specific example of an integrated 2D / 3D sensor in an embodiment of the present invention;

[0025] Figure 4 It is a schematic diagram of a specific example of a split 2D / 3D sensor in an embodiment of the present invention;

[0026] Figure 5 A schematic diagram of a process for preprocessing a two-dimensional grayscale image and a three-dimensional point cloud image in an embodiment of the present invention;

[0027] Figure 6 It is a schematic diagram of a process of locating a rail head area in a three-dimensional point cloud image in an embodiment of the present invention;

[0028] Figure 7 It is a schematic diagram of a process of searching a rail head region in a two-dimensional grayscale image based on a rail head region in a three-dimensional point cloud image in an embodiment of the present invention;

[0029] Figure 8 A schematic diagram of a process for determining a rail light band area from a rail head area corresponding to a rail in a two-dimensional grayscale image in an embodiment of the present invention;

[0030] Fig. 9 It is a schematic diagram of a process of identifying abnormalities in a rail light band area in an embodiment of the present invention;

[0031] Fig.10 Schematic diagram of the structure of the rail light band abnormality identification device in an embodiment of the present invention. DETAILED DESCRIPTION

[0032] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0033] Those skilled in the art will appreciate that the embodiments of the present invention may be implemented as a system, device, method or computer program product. Therefore, the present invention may be implemented in the following forms, namely: complete hardware, complete software (including firmware, resident software, microcode, etc.), or a combination of hardware and software.

[0034] As mentioned above, due to the limited dynamic range of the camera, it is difficult to obtain clear images of the rail light band area and the non-light band area at the same time; not only that, oil stains, foreign matter, damage, etc. on the top surface of the rail are all shown as black patches on the two-dimensional grayscale image, which are difficult to distinguish, resulting in a very high false alarm rate for abnormal identification of the top surface of the rail. In order to solve this problem, an embodiment of the present invention provides a method for identifying abnormalities in rail light bands. The method first obtains a two-dimensional grayscale image and a three-dimensional point cloud image of the rail surface aligned at the pixel level through a sensor, and quickly locates the rail head area in the three-dimensional point cloud image through methods such as threshold segmentation, and then searches for the corresponding rail head area in the two-dimensional grayscale image of the rail surface as a reference for detecting abnormalities in the rail light band. On this basis, the rail light band area is identified as abnormal.

[0035] Figure 1 Schematic diagram of a method for identifying abnormal rail light bands in an embodiment of the present invention. Figure 1 As shown, the method includes:

[0036] Step 101, obtaining a two-dimensional grayscale image and a three-dimensional point cloud image of the rail surface aligned at the pixel level;

[0037] Step 102: locating the rail head area in the three-dimensional point cloud image;

[0038] Step 103: searching for a corresponding rail head region in the two-dimensional grayscale image according to the rail head region located in the three-dimensional point cloud image;

[0039] Step 104, determining the rail light band area in the corresponding rail head area in the two-dimensional grayscale image;

[0040] Step 105: identify abnormalities in the rail light band area.

[0041] Depend on Figure 1As can be seen from the process, the embodiment of the present invention can effectively solve the problem of light band abnormality detection error caused by the inability of traditional methods to effectively separate the rail head and the rail bottom by acquiring two-dimensional grayscale images and three-dimensional point cloud images and combining image positioning, abnormality recognition and other methods.

[0042] In the embodiment, a two-dimensional grayscale image and a three-dimensional point cloud image of the rail surface aligned at the pixel level are first obtained. Figure 2 FIG. 1 is a flow chart of obtaining a two-dimensional grayscale image and a three-dimensional point cloud image of a rail surface aligned at the pixel level according to an embodiment of the present invention. Figure 2 As shown, a two-dimensional grayscale image and a three-dimensional point cloud image of the rail surface are obtained by an integrated 2D and 3D sensor or a split 2D and 3D sensor.

[0043] In the embodiment, Figure 3 FIG. 1 is a schematic diagram of a specific example of an integrated 2D / 3D sensor according to an embodiment of the present invention, including an axonometric view and a front view of an integrated 2D / 3D sensor and a rail. Figure 3 As shown, a two-dimensional grayscale image and a three-dimensional point cloud image of the rail surface aligned at the pixel level are obtained by an integrated 2D / 3D sensor: the integrated 2D / 3D sensor includes a planar array 3D camera and a line laser, wherein the planar array 3D camera can simultaneously obtain a two-dimensional grayscale image and a three-dimensional point cloud image of the rail surface, and the two-dimensional grayscale image and the three-dimensional point cloud image of the rail surface are aligned at the pixel level.

[0044] In the embodiment, Figure 4 Schematic diagram of a specific example of a split 2D / 3D sensor in an embodiment of the present invention, including an axonometric view and a front view of a split 2D / 3D sensor and a rail, as shown in FIG. Figure 4 As shown, a two-dimensional grayscale image and a three-dimensional point cloud image of the rail surface with pixel-level alignment can also be obtained by a split 2D / 3D sensor: the split 2D / 3D sensor includes a set of 2D sensors and a set of 3D sensors, wherein the 2D sensor includes a light source (such as an LED or laser light source) and a linear array camera for obtaining a two-dimensional grayscale image of the rail surface; the 3D sensor includes a surface array camera and a linear laser for obtaining a three-dimensional point cloud image of the rail surface. The coordinate transformation parameters between the two sensors are obtained by calibrating the 2D sensor and the 3D sensor in advance, and the pixel-level alignment of the two-dimensional grayscale image and the three-dimensional point cloud image of the rail surface is achieved by the coordinate transformation parameters.

[0045] In an embodiment, after obtaining the two-dimensional grayscale image and the three-dimensional point cloud image of the rail surface aligned at the pixel level, it can also include: filtering and denoising the two-dimensional grayscale image and the three-dimensional point cloud image of the rail surface; subsequently locating the rail head area in the three-dimensional point cloud image after filtering and denoising, and finding the corresponding rail head area in the two-dimensional grayscale image after filtering and denoising.

[0046] Figure 5 FIG. 1 is a schematic diagram of a preprocessing process for a two-dimensional grayscale image and a three-dimensional point cloud image in an embodiment of the present invention. Figure 5 As shown, after obtaining the two-dimensional grayscale image and three-dimensional point cloud image of the rail surface aligned at the pixel level, the obtained images are preprocessed; the two-dimensional grayscale image and the three-dimensional point cloud image of the rail surface are filtered and denoised, and traditional filtering methods such as mean filtering, median filtering, and Gaussian filtering can be used to reduce the noise interference generated during the image acquisition process.

[0047] In an embodiment, searching for the corresponding rail head region in the two-dimensional grayscale image after filtering and denoising may include: enhancing the image contrast and highlighting the light band region of the two-dimensional grayscale image after filtering and denoising to obtain the two-dimensional grayscale image after image enhancement, and searching for the corresponding rail head region in the two-dimensional grayscale image after image enhancement. That is, preprocessing the two-dimensional grayscale image and the three-dimensional point cloud image also includes enhancing the two-dimensional grayscale image of the rail surface after filtering and denoising. For example, for the two-dimensional grayscale image of the rail surface after filtering and denoising, the image contrast is enhanced by formula (1), the light band region is highlighted, and the two-dimensional grayscale image of the rail surface after image enhancement is obtained.

[0048] G out =Round((G in -G mean )*α)+G in (1)

[0049] Among them, G in is the pixel grayscale value of the input two-dimensional grayscale image of the rail surface, G mean is the pixel grayscale value of the two-dimensional grayscale image of the rail surface after mean filtering, G out is the pixel grayscale value of the output two-dimensional grayscale image of the rail surface, α is the scale factor, and the value range may be, for example, 0 to 15, and Round() is a rounding operation.

[0050] In the embodiment, the rail head region is located in the three-dimensional point cloud image. Figure 6 FIG. 1 is a schematic diagram of a process for locating a rail head region in a three-dimensional point cloud image according to an embodiment of the present invention. Figure 6 As shown, in the embodiment of the present invention, three methods can be used to locate the rail head area.

[0051] In one example, for the three-dimensional point cloud image of the rail surface after filtering and denoising, the height threshold method is used to locate the rail head area, which includes: in the three-dimensional point cloud image, according to the preset height threshold, the top surface of the rail is segmented from the three-dimensional point cloud image to determine the rail head area.

[0052] Since the top surface of the rail is higher than the rest of the track, in the three-dimensional point cloud image of the rail surface, the top surface of the rail can be segmented from the three-dimensional point cloud image of the rail surface by setting a certain height threshold. The specific method is shown in formula (2):

[0053] R 3d ={(r, c)|H(r, c)≥H t} (2)

[0054] Where (r, c) represents the row and column coordinates of a point in the three-dimensional point cloud image of the rail surface, H(r, c) represents the height value at (r, c), and H t is the height threshold. For example, 50 mm below the top of the rail can be taken as the height threshold. 3d Represents the rail head area in the 3D point cloud image.

[0055] In another example, for the 3D point cloud image of the rail surface after filtering and denoising, the watershed method is used to locate the rail head area, which includes:

[0056] The three-dimensional point cloud image of the rail surface is binarized to obtain a binary image. The background area is determined by morphological dilation operation, and the foreground area is determined by distance transformation. The remaining part is the uncertain area. The determined foreground area is processed by connected component processing to obtain a marked image. The watershed algorithm is applied to the original image according to the marked image, and the marked image is updated to obtain the rail head area R 3d .

[0057] In another example, for the three-dimensional point cloud image of the rail surface after filtering and denoising, the region growing method is used to locate the rail head area, which includes: preliminarily segmenting the three-dimensional point cloud image according to a preset height threshold to obtain candidate areas; performing morphological opening operation on the candidate areas; calculating the rectangularity of the candidate areas after the morphological opening operation, and screening out the rail head area according to the rectangularity calculation result:

[0058] First, the three-dimensional point cloud image of the rail surface is preliminarily segmented. For any point in the three-dimensional point cloud image of the rail surface that satisfies formula (3), it can be determined as the same area, and several candidate areas are obtained.

[0059] R candidate ={(r,c)|abs(H1(r,c)-H2(r,c)≤H t1} (3)

[0060] Among them, R candidate is the candidate region, H1(r, c) and H2(r, c) represent the height values ​​of two adjacent points, abs() represents the absolute value operation, N t1is the height threshold, and the value range may be, for example, 5 to 20.

[0061] Then, a morphological opening operation is performed on the candidate region to further eliminate noise interference;

[0062] Finally, the rectangularity of the candidate area is calculated, and the rail head area R is selected by the rectangularity parameter. 3d .

[0063] In the embodiment, the corresponding rail head region in the two-dimensional grayscale image is found according to the rail head region located in the three-dimensional point cloud image. Figure 7 FIG. 1 is a flow chart of searching the rail head region in the two-dimensional grayscale image according to the rail head region in the three-dimensional point cloud image in an embodiment of the present invention. Figure 7 As shown in FIG. 1 , since the two-dimensional grayscale image of the rail surface and the three-dimensional point cloud image are aligned at the pixel level, the corresponding rail head area can be located in the two-dimensional grayscale image through area mapping according to the rail head area located in the three-dimensional point cloud image. For example, R 3d The corresponding area quickly locates the rail head area, thereby obtaining the reference benchmark for rail light band anomaly recognition, that is, the rail head area in the two-dimensional grayscale image, denoted as R 2d .

[0064] In the embodiment, the rail light band area is determined in the rail head area corresponding to the two-dimensional grayscale image. Figure 8 FIG. 4 is a flow chart of determining a rail light band area from a rail head area corresponding to a rail in a two-dimensional grayscale image in an embodiment of the present invention. Figure 8 As shown, the rail head area corresponding to the rail in the two-dimensional grayscale image determines the rail light band area, including: dividing the light band candidate area and the non-light band candidate area according to a preset light band division threshold in the rail head area corresponding to the rail in the two-dimensional grayscale image; performing a morphological opening operation on the light band candidate area to obtain the rail light band area.

[0065] For example, in the two-dimensional grayscale image of the rail surface, R 2d The area is divided into the light band candidate area and the non-light band candidate area by formula (4):

[0066]

[0067] Among them, R1 represents the light band selection area, R2 represents the non-light band selection area, and G is R 2d The gray value of any pixel in the area, G t is the light band threshold.

[0068] Due to the interference of factors such as rail surface dirt and falling blocks, the light band selection area may contain some internal holes. If not processed, it will affect the subsequent light band width calculation. For the light band selection area, the internal holes are eliminated through morphological opening operation, and finally the light band area R is obtained. band .

[0069] In the embodiment, abnormality identification is performed on the rail light band area. Fig. 9 FIG. 1 is a schematic diagram of a process for identifying anomalies in a rail light band area according to an embodiment of the present invention. Fig. 9 As shown, light band anomaly recognition includes light band width anomaly recognition and light band eccentricity anomaly recognition; light band width anomaly recognition includes: counting the number of row pixels in the rail light band area to obtain light band width data, performing mean filtering on the light band width data, and performing light band width anomaly recognition on the filtered light band width data through a light band width threshold.

[0070] For example, when identifying light band width anomalies, we first assume that the light band area R band Occupies m rows of the image, for R band The number of pixels in the i-th row in the i , we can get m data representing the light band width. Perform mean filtering on the light band width data to eliminate noise interference. For the filtered light band width data, use formula (5) to determine whether the light band width is abnormal (too narrow or too wide).

[0071]

[0072] Among them, N t1 N is the optical bandwidth over-width management threshold, t2 is the light band too narrow management threshold. When the number of row pixels of the light band is greater than the light band too wide threshold N t1 , it means that the light band is too wide; when the number of row pixels of the light band is less than the light band too narrow threshold N t2 , it means the light band is too narrow.

[0073] The light band eccentricity anomaly recognition includes: counting the starting column index and the ending column index of the rail light band area and the rail head area, calculating the eccentricity value, and using the light band eccentricity management threshold to identify the light band eccentricity anomaly based on the eccentricity value obtained after calculation.

[0074] For example, when identifying light band eccentricity abnormality, firstly, for the light band area R band The starting column index is recorded as S1 and the ending column index is recorded as S2. 2d The starting column index of the row is recorded as S3, and the ending column index is recorded as S4. The eccentricity value O of the row is calculated by formula (6): i .

[0075] Oi =(S1+S2-S3-S4)*β (6)

[0076] Where β is a proportional factor, and its value range is 0 to 10. Formula (7) is used to determine whether the light band is abnormally eccentric.

[0077]

[0078] Among them, O t is the light band eccentricity management threshold, abs() represents taking the absolute value, when the absolute value of the light band eccentricity value is greater than the light band eccentricity management threshold O t When the absolute value of the eccentricity of the light band is less than the light band eccentricity management threshold O t , it means the light band is centered.

[0079] The present invention also provides a rail light band abnormality identification device, as described in the following embodiments. Since the principle of the device to solve the problem is similar to that of the rail light band abnormality identification method, the implementation of the device can refer to the implementation of the rail light band abnormality identification method, and the repeated parts will not be repeated.

[0080] Fig.10 FIG. 1 is a schematic diagram of the structure of a device for identifying abnormal rail light bands in an embodiment of the present invention. Fig.10 As shown, the device comprises:

[0081] An image acquisition module 1001 is used to acquire a two-dimensional grayscale image and a three-dimensional point cloud image of the rail surface aligned at the pixel level;

[0082] A three-dimensional point cloud image positioning module 1002 is used to locate the rail head area in the three-dimensional point cloud image;

[0083] A two-dimensional grayscale image positioning module 1003 is used to find the corresponding rail head area in the two-dimensional grayscale image according to the rail head area located in the three-dimensional point cloud image;

[0084] The light band area acquisition module 1004 is used to determine the rail light band area corresponding to the rail head area in the two-dimensional grayscale image;

[0085] The light band area abnormality identification module 1005 is used to identify abnormalities in the rail light band area.

[0086] In one embodiment, the image acquisition module 1001 is specifically used for:

[0087] An integrated 2D and 3D sensor is used to obtain a two-dimensional grayscale image and a three-dimensional point cloud image of the rail surface that are aligned at the pixel level. The integrated 2D and 3D sensor includes a planar array 3D camera and a line laser. The planar array 3D camera simultaneously obtains a two-dimensional grayscale image and a three-dimensional point cloud image of the rail surface, and the two-dimensional grayscale image and the three-dimensional point cloud image are aligned at the pixel level.

[0088] In one embodiment, the image acquisition module 1001 is specifically used for:

[0089] A two-dimensional grayscale image and a three-dimensional point cloud image of a rail surface aligned at the pixel level are obtained by means of split 2D and 3D sensors; wherein the split 2D and 3D sensors include a set of 2D sensors and a set of 3D sensors; the 2D sensor includes a light source and a linear array camera for obtaining a two-dimensional grayscale image of the rail surface; the 3D sensor includes a surface array camera and a linear laser for obtaining a three-dimensional point cloud image of the rail surface; the coordinate transformation parameters between the two sensors are obtained by calibrating the 2D sensor and the 3D sensor in advance, and the pixel level alignment of the two-dimensional grayscale image and the three-dimensional point cloud image of the rail surface is achieved by the coordinate transformation parameters.

[0090] In one embodiment, the image acquisition module 1001 is further used for:

[0091] Filter and denoise the two-dimensional grayscale image and three-dimensional point cloud image of the rail surface;

[0092] The three-dimensional point cloud image positioning module 1002 is specifically used for:

[0093] Locate the rail head area in the filtered and denoised three-dimensional point cloud image;

[0094] The two-dimensional grayscale image positioning module 1003 is specifically used for:

[0095] Find the corresponding rail head area in the filtered and denoised two-dimensional grayscale image.

[0096] In one embodiment, the two-dimensional grayscale image positioning module 1003 is specifically used to:

[0097] For the two-dimensional grayscale image after filtering and denoising, the image contrast is enhanced, the light band area is highlighted, and the two-dimensional grayscale image after image enhancement is obtained. The corresponding rail head area in the two-dimensional grayscale image after image enhancement is found.

[0098] In one embodiment, the 3D point cloud image positioning module 1002 is specifically used to:

[0099] In the three-dimensional point cloud image, the top surface of the rail is segmented from the three-dimensional point cloud image according to a preset height threshold to determine the rail head area.

[0100] In one embodiment, the 3D point cloud image positioning module 1002 is specifically used to:

[0101] The three-dimensional point cloud image is binarized to obtain a binary image. The background area is determined by morphological dilation operation, and the foreground area is determined by distance transformation. The remaining part is the uncertain area. The determined foreground area is processed by connected component processing to obtain a labeled image. The watershed algorithm is applied to the original image according to the labeled image, and the labeled image is updated to obtain the rail head area.

[0102] In one embodiment, the 3D point cloud image positioning module 1002 is specifically used to:

[0103] The three-dimensional point cloud image is preliminarily segmented according to a preset height threshold to obtain a candidate area;

[0104] Perform morphological opening operation on the candidate region;

[0105] The rectangularity of the candidate area after the morphological opening operation is calculated, and the rail head area is screened out according to the rectangularity calculation result.

[0106] In one embodiment, the two-dimensional grayscale image positioning module 1003 is specifically used to:

[0107] According to the rail head area located in the three-dimensional point cloud image, the corresponding rail head area is located in the two-dimensional grayscale image through area mapping.

[0108] In one embodiment, the light band region acquisition module 1004 is specifically used to:

[0109] In the rail head area corresponding to the two-dimensional grayscale image, the light band candidate area and the non-light band candidate area are divided according to the preset light band division threshold;

[0110] Perform morphological opening operation on the light band candidate area to obtain the rail light band area.

[0111] In one embodiment, the light band area abnormality identification module 1005 is specifically used to:

[0112] The number of row pixels in the rail light band area is counted to obtain the light band width data, and mean filtering is performed on the light band width data. For the filtered light band width data, the light band width is abnormally identified through the light band width threshold.

[0113] In one embodiment, the light band area abnormality identification module 1005 is specifically used to:

[0114] The starting column index and the ending column index of the rail light band area and the rail head area are counted, and the eccentricity value is calculated. The eccentricity of the light band is abnormally identified through the light band eccentricity management threshold value.

[0115] An embodiment of the present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned contact network geometric parameter detection data filtering method when executing the computer program.

[0116] An embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the above-mentioned contact network geometric parameter detection data filtering method is implemented.

[0117] An embodiment of the present invention further provides a computer program product, which includes a computer program. When the computer program is executed by a processor, the above-mentioned contact network geometric parameter detection data filtering method is implemented.

[0118] The rail light band anomaly recognition method and device of the embodiment of the present invention first obtains a two-dimensional grayscale image and a three-dimensional point cloud image of the rail surface aligned at the pixel level; then locates the rail head area in the three-dimensional point cloud image; according to the rail head area located in the three-dimensional point cloud image, finds the corresponding rail head area in the two-dimensional grayscale image; determines the rail light band area based on the corresponding rail head area in the two-dimensional grayscale image; finally, performs anomaly recognition on the rail light band area. This method can effectively solve the problem of light band anomaly detection error caused by rail head positioning deviation in traditional methods, and ensure the accuracy of rail light band anomaly detection under complex working conditions.

[0119] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0120] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0121] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0122] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0123] The specific embodiments described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for identifying rail light band anomalies, characterized in that: include: Obtain pixel-level aligned 2D grayscale image and 3D point cloud image of the rail surface; Locate the rail head area in the 3D point cloud image; According to the rail head area located in the three-dimensional point cloud image, the corresponding rail head area in the two-dimensional grayscale image is found; The rail light band area is determined in the corresponding rail head area in the two-dimensional grayscale image; Identify abnormalities in the rail light band area.

2. The method according to claim 1, characterized in that Obtain pixel-level aligned 2D grayscale images and 3D point cloud images of the rail surface, including: An integrated 2D and 3D sensor is used to obtain a two-dimensional grayscale image and a three-dimensional point cloud image of the rail surface that are aligned at the pixel level. The integrated 2D and 3D sensor includes a planar array 3D camera and a line laser. The planar array 3D camera simultaneously obtains a two-dimensional grayscale image and a three-dimensional point cloud image of the rail surface, and the two-dimensional grayscale image and the three-dimensional point cloud image are aligned at the pixel level.

3. The method according to claim 1, characterized in that Obtain pixel-level aligned 2D grayscale images and 3D point cloud images of the rail surface, including: A two-dimensional grayscale image and a three-dimensional point cloud image of a rail surface aligned at the pixel level are obtained by means of split 2D and 3D sensors; wherein the split 2D and 3D sensors include a set of 2D sensors and a set of 3D sensors; the 2D sensor includes a light source and a linear array camera for obtaining a two-dimensional grayscale image of the rail surface; the 3D sensor includes a surface array camera and a linear laser for obtaining a three-dimensional point cloud image of the rail surface; the coordinate transformation parameters between the two sensors are obtained by calibrating the 2D sensor and the 3D sensor in advance, and the pixel level alignment of the two-dimensional grayscale image and the three-dimensional point cloud image of the rail surface is achieved by the coordinate transformation parameters.

4. The method according to claim 1, characterized in that After obtaining the pixel-level aligned 2D grayscale image and 3D point cloud image of the rail surface, it also includes: Filter and denoise the two-dimensional grayscale image and three-dimensional point cloud image of the rail surface; Locate the rail head area in the 3D point cloud image, including: Locate the rail head area in the filtered and denoised three-dimensional point cloud image; Find the corresponding rail head area in the 2D grayscale image, including: Find the corresponding rail head area in the filtered and denoised two-dimensional grayscale image.

5. The method according to claim 4, characterized in that Find the corresponding rail head area in the filtered and denoised two-dimensional grayscale image, including: For the two-dimensional grayscale image after filtering and denoising, the image contrast is enhanced, the light band area is highlighted, and the two-dimensional grayscale image after image enhancement is obtained. The corresponding rail head area in the two-dimensional grayscale image after image enhancement is found.

6. The method according to claim 1, characterized in that Locate the rail head area in the 3D point cloud image, including: In the three-dimensional point cloud image, the top surface of the rail is segmented from the three-dimensional point cloud image according to a preset height threshold to determine the rail head area.

7. The method according to claim 1, characterized in that Locate the rail head area in the 3D point cloud image, including: The three-dimensional point cloud image is binarized to obtain a binary image. The background area is determined by morphological dilation operation, and the foreground area is determined by distance transformation. The remaining part is the uncertain area. The determined foreground area is processed by connected component processing to obtain a labeled image. The watershed algorithm is applied to the original image according to the labeled image, and the labeled image is updated to obtain the rail head area.

8. The method according to claim 1, characterized in that Locate the rail head area in the 3D point cloud image, including: The three-dimensional point cloud image is preliminarily segmented according to a preset height threshold to obtain a candidate area; Perform morphological opening operation on the candidate region; The rectangularity of the candidate area after the morphological opening operation is calculated, and the rail head area is screened out according to the rectangularity calculation result.

9. The method according to claim 1, characterized in that According to the rail head area located in the 3D point cloud image, the corresponding rail head area in the 2D grayscale image is found, including: According to the rail head area located in the three-dimensional point cloud image, the corresponding rail head area is located in the two-dimensional grayscale image through area mapping.

10. The method according to claim 1, characterized in that The rail head area corresponding to the two-dimensional grayscale image determines the rail light band area, including: In the rail head area corresponding to the two-dimensional grayscale image, the light band candidate area and the non-light band candidate area are divided according to the preset light band division threshold; Perform morphological opening operation on the light band candidate area to obtain the rail light band area.

11. The method according to claim 1, characterized in that Abnormal identification of rail light band area, including: The number of row pixels in the rail light band area is counted to obtain the light band width data, and mean filtering is performed on the light band width data. For the filtered light band width data, the light band width is abnormally identified through the light band width threshold.

12. The method according to claim 1, characterized in that Abnormal identification of rail light band area, including: The starting column index and the ending column index of the rail light band area and the rail head area are counted, and the eccentricity value is calculated. The eccentricity of the light band is abnormally identified through the light band eccentricity management threshold value.

13. A rail light band abnormality recognition device, characterized in that: include: An image acquisition module, used for acquiring a two-dimensional grayscale image and a three-dimensional point cloud image of the rail surface aligned at the pixel level; A 3D point cloud image positioning module, used to locate the rail head area in the 3D point cloud image; A two-dimensional grayscale image positioning module is used to find the corresponding rail head area in the two-dimensional grayscale image according to the rail head area located in the three-dimensional point cloud image; A light band area acquisition module is used to determine the rail light band area corresponding to the rail head area in the two-dimensional grayscale image; The light band area anomaly recognition module is used to identify anomalies in the rail light band area.

14. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 12 is implemented.

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

16. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 12 is implemented.

Images