Inspection method and inspection system for rail robot

Abstract

The invention provides an inspection method and inspection system for a track robot, and the method comprises the steps: enabling the track robot to receive an inspection task, to run along a track based on a path in the inspection task, and to obtain an original track image when the track robot runs to a preset position; the method comprises the steps of obtaining an original track image, intercepting a target area from the original track image based on a preset interception area, rendering an area outside the target area in the original track image to obtain a first rendered image, carrying out binarization processing on the first rendered image based on a preset pixel threshold to obtain a second rendered image, and outputting the second rendered image. Inputting the second rendered image into a preset first processing model; and if the second rendering image is not qualified, obtaining an image area of the target area from the original track image to obtain an area image, performing step processing on the area image to obtain a step segmentation image, and inputting the step segmentation image into a preset second processing model.

Description

Technical Field

[0001] The present invention relates to the technical field of power station inspection, and in particular to an inspection method and an inspection system for a rail robot. Background Art

[0002] A power plant or power station has a variety of large electrical equipment, such as distribution cabinets and distribution boxes. In order to prevent safety hazards caused by equipment failure, the electrical equipment needs to be inspected regularly.

[0003] The ground conditions of some power plants or power stations are relatively complex, which makes it impossible for inspection robots walking on the ground to complete inspection tasks efficiently. In the prior art, there are already rail robots that can move on preset hanging rails. Rail robots are robots that can move along rails and perform various operations. This type of robot can be equipped with various sensors and equipment to perform autonomous navigation, inspection, detection, maintenance and other tasks. The development of rail robots is closely related to the continuous advancement of intelligent technology. With the continuous development of technologies such as artificial intelligence, the Internet of Things, and big data, the intelligence level of rail robots has also been continuously improved, and they can better adapt to various complex environments and tasks.

[0004] However, existing research on rail robots often focuses on the processing of inspection tasks, and lacks the judgment of the road conditions of the rail robots, which leads to safety hazards when the rail robots perform tasks.

[0005] In view of this, the present invention is proposed. Summary of the invention

[0006] The purpose of the present invention is to provide an inspection method and an inspection system for a rail robot. This scheme determines the road conditions through two stages of judgment. The first stage determines whether there are hidden dangers, and the second stage determines the specific type of hidden dangers.

[0007] The present invention provides an inspection method for a rail robot, the method comprising the following steps:

[0008] The track robot receives the inspection task, and drives along the track based on the path in the inspection task, and obtains the original track image when driving to the predetermined position;

[0009] The original track image includes a track area and a non-track area, and a target area is intercepted from the original track image based on a preset interception area, wherein the target area includes at least a part of the track area;

[0010] Rendering an area outside the target area in the original track image to obtain a first rendered image, binarizing the first rendered image based on a preset pixel threshold to obtain a second rendered image, and inputting the second rendered image into a preset first processing model, wherein the first processing model is used to output a determination result of whether the second rendered image is qualified;

[0011] If the second rendered image is unqualified, the image area of ​​the target area is obtained from the original track image to obtain a regional image, the regional image is step-processed to obtain a step-segmented image, and the step-segmented image is input into a preset second processing model, and the second processing model is used to output the track problem category.

[0012] By adopting the above scheme, when the track robot drives to the predetermined position to obtain the original track image, the camera installed on the top of the track robot takes the image in front, which is the original track image. The scheme first renders the area outside the target area, and binarizes the target area to filter out useful information, and obtains a second rendered image with lower complexity but containing useful information, and can quickly determine whether the second rendered image is qualified. If it is unqualified, the scheme further constructs a step segmentation image to accurately determine the type of track problem.

[0013] In some embodiments of the present invention, the area of ​​the target is an area divided from the original track image by a preset clipping contour, and in the step of rendering the area outside the area of ​​the target in the original track image to obtain a first rendered image, the area outside the area of ​​the target in the original track image is rendered to a preset first grayscale value to obtain a first rendered image.

[0014] The above scheme is adopted. In this scheme, the target area is a preset trapezoidal area, the trapezoidal area is an isosceles trapezoid, the upper base length of the trapezoidal area is less than the lower base length. This scheme first presets the target area as a trapezoidal area, which conforms to the actual shooting angle, can effectively capture the target area, and the area outside the target area is usually an area of ​​useless information. The area outside the target area in the original track image is rendered as a preset first grayscale value, which reduces useless data and increases the subsequent calculation amount, thereby improving the processing speed.

[0015] In some embodiments of the present invention, the step of performing binarization processing on the first rendered image based on a preset pixel threshold to obtain a second rendered image includes:

[0016] Calculate each pixel point of the target area in the first rendered image one by one, calculate the average pixel value of the pixel points around each pixel point, use the average value as a comparison threshold to compare with the pixel value of the pixel point, and perform binarization processing;

[0017] The calculation of all pixel points in the target area in the first rendered image is completed to obtain a second rendered image.

[0018] In some embodiments of the present invention, in the step of comparing the average value with the pixel value of the pixel point as a comparison threshold and performing binarization processing, if the absolute value of the difference between the pixel value of the pixel point and the comparison threshold is less than a preset difference threshold, the pixel point is rendered as a first grayscale value; if the absolute value of the difference between the pixel value of the pixel point and the comparison threshold is not less than the preset difference threshold, the pixel point is rendered as a second grayscale value.

[0019] By adopting the above scheme, useful information in the first rendered image is further extracted through binarization processing, and useless information is rendered. Under the premise of reducing the amount of calculation, the processing accuracy is improved. The second rendered image of this scheme is only used for preliminary qualified judgment, which requires a higher processing speed. If it is qualified, the inspection task can be continued quickly; if it is unqualified, high-precision judgment can continue.

[0020] In some embodiments of the present invention, the region image is an image of the region of the target in the original track image, and the step of performing step processing on the region image to obtain a step segmentation image includes:

[0021] Dividing the regional image into a plurality of segments in the length direction of the regional image to obtain a plurality of segmented images, each of the segmented images is an isosceles trapezoid, and each of the segmented images has a different length in the length direction of the regional image;

[0022] Extending the lower base width of each of the segmented images to the lower base length of the region image, and extending the upper base of the segmented images in equal proportion;

[0023] The height of the isosceles trapezoid of each segmented image is extended to the height of the segmented image where the lower base of the regional image is located, so as to obtain a step sub-image corresponding to each segmented image.

[0024] In some embodiments of the present invention, the regional image is an image of the region of the target in the original track image, and the step of performing step processing on the regional image to obtain a step segmentation image also includes combining a plurality of the step sub-images based on the positions of the segmented images corresponding to the step sub-images in the regional image to obtain the step segmentation image.

[0025] In some embodiments of the present invention, in the step of combining a plurality of the step sub-images to obtain the step segmentation image, the step sub-images corresponding to a preset number of adjacent segmented images in the regional image are combined to obtain the step segmentation image.

[0026] With the above scheme, due to the viewing angle of the camera, the track is usually photographed as an area that is wide at the bottom and narrow at the top, and the image taken of the far end area is often smaller. This scheme first divides the regional image into multiple segments in the length direction of the regional image to obtain multiple segmented images, and uses the segmented image where the lower base of the regional image is located as the standard to expand the other segmented images to the same size, thereby improving the image processing fineness and ensuring the equal processing accuracy of each segmented image.

[0027] In some embodiments of the present invention, in the step of dividing the regional image into multiple segments in the length direction of the regional image to obtain multiple segmented images, the regional image is divided into multiple segments of different lengths in the length direction of the regional image, and the division length of the segmented image gradually decreases in the direction from the lower base to the upper base of the regional image.

[0028] By adopting the above scheme, the part with higher fineness in the regional image is divided into segmented images of larger areas, and the part with lower fineness in the regional image is divided into segmented images of smaller areas, thereby improving the fineness of the division and improving the fineness of the image through subsequent enlargement.

[0029] In some embodiments of the present invention, both the first processing model and the second processing model are pre-trained graph neural network models.

[0030] Another aspect of the present invention also relates to a patrol robot inspection system, which includes a computer device, wherein the computer device includes a processor and a memory, wherein the memory stores computer instructions, and the processor is used to execute the computer instructions stored in the memory. When the computer instructions are executed by the processor, the system implements the steps implemented by the method.

[0031] In summary, the present invention has the following beneficial effects:

[0032] 1. When the track robot drives to the predetermined position and obtains the original track image, the camera installed on the top of the track robot takes the image in front, which is the original track image. The scheme first renders the area outside the target area, and performs binarization processing on the target area to filter out useful information, and obtains a second rendering image with low complexity but containing useful information, and can quickly determine whether the second rendering image is qualified. If it is unqualified, the scheme further constructs a step segmentation image to accurately determine the type of track problem;

[0033] 2. The target area in this solution is a preset trapezoidal area, which is an isosceles trapezoid, and the upper base length of the trapezoidal area is less than the lower base length. This solution first presets the target area as a trapezoidal area, which conforms to the actual shooting angle, can effectively capture the target area, and the area outside the target area is usually an area of ​​useless information. The area outside the target area in the original track image is rendered as a preset first grayscale value, which reduces useless data and increases the subsequent calculation amount, thereby improving the processing speed;

[0034] 3. This solution further extracts useful information from the first rendered image through binarization processing, and renders useless information, thereby improving processing accuracy while reducing the amount of calculation. The second rendered image of this solution is only used for preliminary qualified judgment, which requires a higher processing speed. If it is qualified, the inspection task can be continued quickly; if it is unqualified, high-precision judgment can be continued;

[0035] 4. Due to the viewing angle of the camera, the track is usually photographed as an area that is wide at the bottom and narrow at the top, and the image taken of the far end area is often smaller. This solution first divides the regional image into multiple segments in the length direction of the regional image to obtain multiple segmented images, and uses the segmented image where the lower base of the regional image is located as the standard to expand the other segmented images to the same size, thereby improving the image processing precision and ensuring the equal processing accuracy of each segmented image. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. 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 paying creative work.

[0037] Figure 1 A schematic diagram of an embodiment of the inspection method for a rail robot according to the present invention;

[0038] Figure 2 A schematic diagram of an implementation method of constructing a step segmentation image for a rail robot inspection method of the present invention;

[0039] Figure 3 A schematic diagram of another implementation of constructing a step segmentation image for the inspection method for a rail robot of the present invention. DETAILED DESCRIPTION

[0040] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Instead, they are merely examples of devices and methods consistent with some aspects of the present invention as detailed in the appended claims.

[0041] The terms used in the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The singular forms "a", "the" and "the" used in the present invention and the appended claims are also intended to include plural forms unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0042] like Figure 1 As shown, the present invention provides an inspection method for a rail robot, the method comprising the steps of:

[0043] Step S100, the rail robot receives an inspection task, and drives along the track based on the path in the inspection task, and acquires an original track image when driving to a predetermined position;

[0044] In a specific implementation process, the rail robot is hoisted on the rail and cooperates with the rail through pulleys or tracks, so that the rail robot moves along the rail.

[0045] In a specific implementation process, the predetermined position may be a track intersection or a predetermined stop point.

[0046] In a specific implementation process, the original track image can be captured by a camera installed on the top of the track robot, so as to capture the track on which the track robot is about to travel.

[0047] Step S200, the original track image includes a track area and a non-track area, and a target area is intercepted from the original track image based on a preset interception area, wherein the target area includes at least a part of the track area;

[0048] In a specific implementation process, the track area is the area where the track is located in the original track image, and the non-track area is the area outside the track area in the original track image.

[0049] Step S300, rendering an area outside the target area in the original track image to obtain a first rendered image, binarizing the first rendered image based on a preset pixel threshold to obtain a second rendered image, inputting the second rendered image into a preset first processing model, and the first processing model is used to output a determination result of whether the second rendered image is qualified;

[0050] In the specific implementation process, this solution renders the useless area as the first gray value, which can effectively remove the influence of the useless area on the determination result.

[0051] Step S400: If the second rendered image is unqualified, an image area of ​​the target area is obtained from the original track image to obtain a regional image, and the regional image is step-processed to obtain a step-segmented image. The step-segmented image is input into a preset second processing model, and the second processing model is used to output a track problem category.

[0052] In a specific implementation process, if the second rendered image is qualified, the track robot continues to travel along the track to complete the inspection task.

[0053] In a specific implementation process, the track problem categories include attachment problems, crack problems and other problems, and the classification layer of the second processing model outputs specific track problem categories.

[0054] During the specific implementation process, when it is determined to be an attachment problem or other problems, the track robot will suspend operation and report the problem to the staff, and the track robot will wait for the start instruction before continuing to run; when it is determined to be a crack problem, the track robot will reduce the speed to the second speed and report the problem to the staff.

[0055] By adopting the above scheme, when the track robot drives to the predetermined position to obtain the original track image, the camera installed on the top of the track robot takes the image in front, which is the original track image. The scheme first renders the area outside the target area, and binarizes the target area to filter out useful information, and obtains a second rendered image with lower complexity but containing useful information, and can quickly determine whether the second rendered image is qualified. If it is unqualified, the scheme further constructs a step segmentation image to accurately determine the type of track problem.

[0056] In some embodiments of the present invention, the area of ​​the target is an area divided from the original track image by a preset clipping contour, and in the step of rendering the area outside the area of ​​the target in the original track image to obtain a first rendered image, the area outside the area of ​​the target in the original track image is rendered to a preset first grayscale value to obtain a first rendered image.

[0057] In a specific implementation process, the intercepted outline is a pre-defined outline.

[0058] In a specific implementation process, the first grayscale value may be 255, which is white; and the second grayscale value may be 0, which is black.

[0059] The above scheme is adopted. In this scheme, the target area is a preset trapezoidal area, the trapezoidal area is an isosceles trapezoid, the upper base length of the trapezoidal area is less than the lower base length. This scheme first presets the target area as a trapezoidal area, which conforms to the actual shooting angle, can effectively capture the target area, and the area outside the target area is usually an area of ​​useless information. The area outside the target area in the original track image is rendered as a preset first grayscale value, which reduces useless data and increases the subsequent calculation amount, thereby improving the processing speed.

[0060] In some embodiments of the present invention, the step of performing binarization processing on the first rendered image based on a preset pixel threshold to obtain a second rendered image includes:

[0061] Calculate each pixel point of the target area in the first rendered image one by one, calculate the average pixel value of the pixel points around each pixel point, use the average value as a comparison threshold to compare with the pixel value of the pixel point, and perform binarization processing;

[0062] Specifically, the surrounding pixel points are pixel points adjacent to the pixel point.

[0063] The calculation of all pixel points in the target area in the first rendered image is completed to obtain a second rendered image.

[0064] In some embodiments of the present invention, in the step of comparing the average value with the pixel value of the pixel point as a comparison threshold and performing binarization processing, if the absolute value of the difference between the pixel value of the pixel point and the comparison threshold is less than a preset difference threshold, the pixel point is rendered as a first grayscale value; if the absolute value of the difference between the pixel value of the pixel point and the comparison threshold is not less than the preset difference threshold, the pixel point is rendered as a second grayscale value.

[0065] By adopting the above scheme, useful information in the first rendered image is further extracted through binarization processing, and useless information is rendered. Under the premise of reducing the amount of calculation, the processing accuracy is improved. The second rendered image of this scheme is only used for preliminary qualified judgment, which requires a higher processing speed. If it is qualified, the inspection task can be continued quickly; if it is unqualified, high-precision judgment can continue.

[0066] like Figure 2As shown, in some embodiments of the present invention, the regional image is an image of the region of the target in the original track image, and the step of performing step processing on the regional image to obtain a step segmentation image includes:

[0067] Step S411, dividing the region image into a plurality of segments in the length direction of the region image to obtain a plurality of segmented images, each of the segmented images is an isosceles trapezoid, and each of the segmented images has a different length in the length direction of the region image;

[0068] In some embodiments of the present invention, the length direction is the direction from the lower base to the upper base of the trapezoid of the regional image.

[0069] Step S412, extending the lower base width of each of the segmented images to the lower base length of the region image, and extending the upper base of the segmented images in equal proportion;

[0070] Step S413: Extend the height of the isosceles trapezoid of each segmented image to the height of the segmented image where the lower base of the regional image is located, so as to obtain a ladder sub-image corresponding to each segmented image.

[0071] By adopting the above scheme, each step sub-image of the scheme is an image of the same size, so that in the final determination process, the calculation model can calculate each image equivalently.

[0072] like Figure 3 As shown, in some embodiments of the present invention, the regional image is an image of the region of the target in the original track image, and the step of performing step processing on the regional image to obtain a step segmentation image also includes step S414, combining multiple step sub-images based on the positions of segmented images corresponding to the step sub-images in the regional image to obtain the step segmentation image.

[0073] In the specific implementation process, the step segmentation image obtained by combining the step sub-images is an image with jagged edges. Since the segmentation is relatively fine, the combined image can be equivalently regarded as a rectangular image, and the judgment stability is guaranteed by the regular image.

[0074] In some embodiments of the present invention, in the step of combining a plurality of the step sub-images to obtain the step segmentation image, the step sub-images corresponding to a preset number of adjacent segmented images in the regional image are combined to obtain the step segmentation image.

[0075] In the specific implementation process, the preset number can be 3, 4 or 5, etc., and the stepped segmentation images obtained by each combination are input into the second processing model. On the one hand, the accuracy of the judgment can be improved. On the other hand, by judging the stepped segmentation image with problems, the corresponding segmented image can be determined, and then the location of the problem can be accurately determined, which is convenient for the staff to locate and can also be used for subsequent reference and maintenance.

[0076] With the above scheme, due to the viewing angle of the camera, the track is usually photographed as an area that is wide at the bottom and narrow at the top, and the image taken of the far end area is often smaller. This scheme first divides the regional image into multiple segments in the length direction of the regional image to obtain multiple segmented images, and uses the segmented image where the lower base of the regional image is located as the standard to expand the other segmented images to the same size, thereby improving the image processing fineness and ensuring the equal processing accuracy of each segmented image.

[0077] In some embodiments of the present invention, in the step of dividing the regional image into multiple segments in the length direction of the regional image to obtain multiple segmented images, the regional image is divided into multiple segments of different lengths in the length direction of the regional image, and the division length of the segmented image gradually decreases in the direction from the lower base to the upper base of the regional image.

[0078] By adopting the above scheme, the part with higher fineness in the regional image is divided into segmented images of larger areas, and the part with lower fineness in the regional image is divided into segmented images of smaller areas, thereby improving the fineness of the division and improving the fineness of the image through subsequent enlargement.

[0079] In some embodiments of the present invention, both the first processing model and the second processing model are pre-trained graph neural network models.

[0080] In the specific implementation process, the graph neural network model is a neural network model applied to the graph structure. It processes and analyzes the graph structure data by processing the nodes in the graph and the relationship between them. In the graph neural network model, the relationship between nodes can be represented by edges, and the edges can carry the characteristic information of the nodes. This network model can simulate the interaction between nodes, so as to conduct a deeper analysis and understanding of the graph structure. Common graph neural network models include adjacency matrix, degree matrix, Laplace matrix, etc. These matrices can represent the relationship between nodes and can be used to calculate indicators such as distance and similarity between nodes. In the graph neural network model, each node can interact with its neighbor nodes and connected edges to update its own state. This interaction process can be achieved through local transfer functions and global transfer functions. The local transfer function mainly processes the relationship between the current node and its neighbor nodes, while the global transfer function processes the structural information of the entire graph. In general, the graph neural network model is an effective tool for graph structure analysis.

[0081] An embodiment of the present invention also provides an inspection robot inspection system, which includes a computer device, wherein the computer device includes a processor and a memory, wherein the memory stores computer instructions, and the processor is used to execute the computer instructions stored in the memory. When the computer instructions are executed by the processor, the system implements the steps implemented by the method.

[0082] The embodiment of the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the aforementioned model building method based on network architecture search. The computer-readable storage medium may be a tangible storage medium, such as a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a floppy disk, a hard disk, a removable storage disk, a CD-ROM, or any other form of storage medium known in the technical field.

[0083] It should be understood by those skilled in the art that the exemplary components, systems and methods described in conjunction with the embodiments disclosed herein can be implemented in hardware, software or a combination of the two. Whether it is performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention. When implemented in hardware, it can be, for example, an electronic circuit, an application specific integrated circuit (ASIC), appropriate firmware, a plug-in, a function card, etc. When implemented in software, the elements of the present invention are programs or code segments used to perform the required tasks. The program or code segment can be stored in a machine-readable medium, or transmitted on a transmission medium or a communication link via a data signal carried in a carrier.

[0084] It should be clear that the present invention is not limited to the specific configuration and processing described above and shown in the figures. For the sake of simplicity, a detailed description of the known method is omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present invention is not limited to the specific steps described and shown, and those skilled in the art can make various changes, modifications and additions, or change the order between the steps after understanding the spirit of the present invention.

[0085] In the present invention, features described and / or illustrated for one embodiment may be used in the same or similar manner in one or more other embodiments, and / or combined with features of other embodiments or replace features of other embodiments.

[0086] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the embodiments of the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A patrol inspection method for a rail robot, characterized in that: The steps of the method include: The track robot receives the inspection task, and drives along the track based on the path in the inspection task, and obtains the original track image when driving to the predetermined position; The original track image includes a track area and a non-track area, and a target area is intercepted from the original track image based on a preset interception area, wherein the target area includes at least a part of the track area; Rendering an area outside the target area in the original track image to obtain a first rendered image, binarizing the first rendered image based on a preset pixel threshold to obtain a second rendered image, and inputting the second rendered image into a preset first processing model, wherein the first processing model is used to output a determination result of whether the second rendered image is qualified; If the second rendered image is unqualified, the image area of ​​the target area is obtained from the original track image to obtain a regional image, the regional image is step-processed to obtain a step-segmented image, and the step-segmented image is input into a preset second processing model, and the second processing model is used to output the track problem category.

2. The inspection method for a rail robot according to claim 1, characterized in that: The target area is an area divided from the original track image by a preset clipping contour. In the step of rendering the area outside the target area in the original track image to obtain a first rendered image, the area outside the target area in the original track image is rendered to a preset first grayscale value to obtain the first rendered image.

3. The inspection method for a rail robot according to claim 1, characterized in that: The step of performing binarization processing on the first rendered image based on a preset pixel threshold to obtain a second rendered image includes: Calculate each pixel point of the target area in the first rendered image one by one, calculate the average pixel value of the pixel points around each pixel point, use the average value as a comparison threshold to compare with the pixel value of the pixel point, and perform binarization processing; The calculation of all pixel points in the target area in the first rendered image is completed to obtain a second rendered image.

4. The inspection method for a rail robot according to claim 3, characterized in that: In the step of using the average value as a comparison threshold to compare with the pixel value of the pixel point and performing binarization processing, if the absolute value of the difference between the pixel value of the pixel point and the comparison threshold is less than a preset difference threshold, the pixel point is rendered as a first grayscale value; if the absolute value of the difference between the pixel value of the pixel point and the comparison threshold is not less than the preset difference threshold, the pixel point is rendered as a second grayscale value.

5. The inspection method for a rail robot according to any one of claims 1 to 4, characterized in that: The regional image is an image of the region of the target in the original track image, and the step of performing step processing on the regional image to obtain a step segmentation image includes: Dividing the regional image into a plurality of segments in the length direction of the regional image to obtain a plurality of segmented images, each of the segmented images is an isosceles trapezoid, and each of the segmented images has a different length in the length direction of the regional image; Extending the lower base width of each of the segmented images to the lower base length of the region image, and extending the upper base of the segmented images in equal proportion; The height of the isosceles trapezoid of each segmented image is extended to the height of the segmented image where the lower base of the segmented image is located, so as to obtain a step sub-image corresponding to each segmented image.

6. The inspection method for a rail robot according to claim 5, characterized in that: The regional image is an image of the region of the target in the original track image. The step of performing step processing on the regional image to obtain a step segmentation image also includes combining multiple step sub-images based on the positions of segmented images corresponding to the step sub-images in the regional image to obtain the step segmentation image.

7. The inspection method for a rail robot according to claim 6, characterized in that: In the step of combining a plurality of the step sub-images to obtain the step segmentation image, the step sub-images corresponding to a preset number of adjacent segmented images in the region image are combined to obtain the step segmentation image.

8. The inspection method for a rail robot according to claim 7, characterized in that: In the step of dividing the regional image into multiple segments in the length direction of the regional image to obtain multiple segmented images, the regional image is divided into multiple segments with different lengths in the length direction of the regional image, and the division length of the segmented image gradually decreases in the direction from the lower base to the upper base of the regional image.

9. The inspection method for a rail robot according to claim 1, characterized in that: The first processing model and the second processing model are both pre-trained graph neural network models.

10. A patrol robot inspection system, characterized in that: The system includes a computer device, which includes a processor and a memory, wherein the memory stores computer instructions, and the processor is used to execute the computer instructions stored in the memory. When the computer instructions are executed by the processor, the system implements the steps implemented by the method described in any one of claims 1 to 9.

Images