A method for qualitative identification and quantitative reconstruction of rail damage detection

Abstract

The present application relates to the technical field of rail damage detection, and particularly relates to a rail damage detection qualitative identification and quantitative reconstruction method. The present application comprises a vehicle body, a handrail is fixed on one side of the vehicle body, two parallel aluminum alloy square tubes are arranged in the vehicle body, and a running mechanism is arranged at both ends of the aluminum alloy square tube. The present application adopts automatic force detachable design, is more labor-saving to operate, and realizes rapid qualitative identification through a plurality of types of rail damage databases. The image gray value and the wear depth value are combined to establish a correlation equation, high-precision quantitative reconstruction is realized, the rail damage wear characteristics are more accurately restored, the rail damage image gray processing, histogram equalization and FPN structure are adopted to enhance the image features, improve the detection accuracy and range, the correlation model of the rail damage image gray and the wear depth is fitted through the neural network, the damage image is mapped to a three-dimensional model, and strong support is provided for subsequent evaluation and polishing guidance.

Description

Technical Field

[0001] This invention relates to the field of rail damage detection technology, and in particular to a method for qualitative identification and quantitative reconstruction of rail damage. Background Technology

[0002] In recent years, my country's rail transit construction has developed rapidly, with its operational mileage ranking first in the world. At the same time, the dynamic loads on rails have increased significantly, leading to aggravated rail damage problems such as fatigue cracks, abrasions, and corrugation, directly affecting rail life and train operation safety. Field investigations have revealed a wide variety and large number of rail damage problems in my country. Current rail inspection methods are not mature, with most still relying on manual inspection and hand-pushed rail inspection vehicles. These methods suffer from low efficiency, inaccuracy, high labor costs, and safety hazards. While large-scale integrated rail inspection vehicles offer high inspection efficiency, their inconvenient scheduling, high cost, and downtime limit the demand for rail inspection.

[0003] Current methods for measuring rail damage and wear characteristics typically employ steel rulers and electronic wear measuring scales. However, these methods are susceptible to errors due to operational variations. The rise of machine vision technology has driven the intelligent upgrading of rail damage detection technology, providing a new direction for rail damage detection. Therefore, designing a device capable of rapidly and qualitatively identifying rail damage and proposing a method for quantitative reconstruction of rail damage are of great significance for assessing the current health status of rails and guiding subsequent rail grinding. To this end, we propose a qualitative identification and quantitative reconstruction method for rail damage detection. Summary of the Invention

[0004] The purpose of this invention is to provide a method for qualitative identification and quantitative reconstruction of rail damage detection, so as to solve the problems mentioned in the background art.

[0005] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0006] An automatic power trolley for detecting rail damage includes a trolley body with a push handle fixed to one side. Two parallel aluminum alloy square tubes are installed inside the trolley body, and a traveling mechanism is provided at both ends of each aluminum alloy square tube. The traveling mechanism is used to assist the trolley body in moving on the rail. The traveling mechanism is equipped with a power wheel, a traveling wheel, and a guide wheel, all of which are in close contact with the rail.

[0007] Preferably, two acquisition mechanisms are provided at both ends of the two aluminum alloy square tubes, and the acquisition mechanisms are used to acquire rail damage data.

[0008] Preferably, the acquisition mechanism includes a surface acquisition component and a profile acquisition component. Two acquisition boxes are fixed at both ends of the two aluminum alloy square tubes, and the surface acquisition component and the profile acquisition component are assembled inside the acquisition boxes.

[0009] Preferably, the surface acquisition component includes an LED tube and an industrial camera 1. The LED tube is fixed to the top of the inside of the acquisition box, and the industrial camera 1 is fixed to the top of the inside of the acquisition box at a position offset from the LED tube.

[0010] Preferably, the profile acquisition component includes a laser emitter, a laser bracket, and an industrial camera 2. The laser bracket and the industrial camera 2 are fixed on one side inside the acquisition box, and the laser emitter is fixed at one end of the laser bracket.

[0011] A qualitative identification and quantitative reconstruction method for rail damage detection, applicable to an automated power trolley for rail damage detection, includes the following steps:

[0012] S1: When conducting inspection work on the proposed railway route, use an automatic power trolley for rail damage detection to collect data, build a multi-type rail damage database, and widely collect and record common rail damage types in the line.

[0013] S2: Combining deep learning technology, the rail damage detection algorithm model is deeply trained using a multi-type rail damage database to build an intelligent rail damage detection system for rapid detection, thereby achieving qualitative identification of rail damage.

[0014] S3: By combining the grayscale value of the rail damage image with the rail damage wear depth value, the correlation between image grayscale and depth is explored.

[0015] S4: Establish the rail damage depth-grayscale correlation equation to map the rail damage image into a three-dimensional rail damage model and restore the rail damage with high accuracy.

[0016] Preferably, the construction of the multi-type rail damage database in S1 includes the following steps:

[0017] S11: Compare the effects of images acquired under different lighting conditions, and select the most suitable light source parameters and configuration scheme for the on-site lighting environment;

[0018] S12: Set the light source parameters and configuration scheme into the data acquisition box of the automatic force trolley for rail damage detection;

[0019] S13: Using an automated rail damage detection trolley, a large amount of on-site data was collected, and a database of multiple types of rail damage was built.

[0020] Preferably, the qualitative identification of rail damage in S2 includes the following steps:

[0021] S21: First, combine the rail damage database built in S1, and then compare the rail damage detection methods based on SSD algorithm, Faster R-CNN algorithm and YOLO algorithm.

[0022] S22: Optimize the process of rail damage detection to address issues such as unclear image information and missed detection of small-scale damage.

[0023] S23: Finally, construct a machine vision-based qualitative identification and detection system for rail damage to achieve qualitative identification of rail damage.

[0024] Preferably, step S3 includes a quantitative reconstruction of rail damage, comprising the following steps:

[0025] S31: When quantitatively reconstructing rail damage, effective grayscale information of rail damage images is first obtained based on visual representation.

[0026] S32: The wear measurement method based on depth sensing obtains the actual wear value of rail damage, and then combines neural network fitting to establish a correlation model between image gray level and wear depth;

[0027] S33: Finally, the three-dimensional reconstruction of rail damage is completed based on the correlation model between image grayscale and wear depth.

[0028] It is clear without a doubt that the technical solution described above in this application can solve the technical problem that this application aims to address.

[0029] Meanwhile, through the above technical solutions, the present invention has at least the following beneficial effects:

[0030] 1. This invention adopts an automatic force-disassembly design. The trolley frame is made of aluminum alloy square tube and is connected by bolts, which facilitates disassembly and transportation. The power wheels are powered by hub motors, and the running wheels are omnidirectional wheels to ensure steering in curved sections. The guide wheels are distributed on the inside and outside of the rails to prevent derailment and also meet the travel requirements in variable gauge sections, making operation more labor-saving.

[0031] 2. This invention achieves rapid qualitative identification by combining a multi-type rail damage database with a deep learning-trained detection algorithm model; and establishes a correlation equation by combining image grayscale values ​​with wear depth values ​​to achieve high-precision quantitative reconstruction, thereby more accurately restoring the rail damage and wear characteristics.

[0032] 3. This invention employs techniques such as grayscale processing of rail damage images, histogram equalization, and FPN structure to enhance image features and improve detection accuracy and range. It also utilizes PyQt5 to build a detection system, achieving precise and efficient detection. By fitting a neural network model to correlate the grayscale of rail damage images with wear depth, the damage images are mapped into a three-dimensional model, intuitively displaying damage characteristics and providing strong support for subsequent evaluation and polishing guidance. Attached Figure Description

[0033] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0034] Figure 1 This is a schematic diagram of the overall assembly of the rail damage detection trolley of the present invention;

[0035] Figure 2 This is a schematic diagram of the traveling mechanism structure of the present invention;

[0036] Figure 3 This is a schematic diagram of the data acquisition mechanism of the present invention;

[0037] Figure 4 This is a flowchart illustrating a specific implementation method of the present invention;

[0038] Figure 5 Flowchart for building a multi-type rail damage database for this invention;

[0039] Figure 6 This is a flowchart of the qualitative identification process for rail damage according to the present invention;

[0040] Figure 7 The diagrams are of the SSD algorithm, Faster R-CNN algorithm, and YOLO algorithm of this invention.

[0041] Figure 8 This is a schematic diagram of the rail damage recognition image information enhancement technology route of the present invention;

[0042] Figure 9 The rail damage recognition image recognition range enhancement technology of the present invention is shown in the following diagrams: (a) Schematic diagram of rail damage algorithm optimization model, (b) Schematic diagram of FPN structure.

[0043] Figure 10 This is a schematic diagram of the interface of the rail damage identification system of the present invention;

[0044] Figure 11 This is a flowchart of the quantitative reconstruction process for rail damage according to the present invention;

[0045] Figure 12 This is a detailed schematic diagram of the quantitative reconstruction of rail damage according to the present invention. Detailed Implementation

[0046] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0047] Example 1

[0048] Reference Figure 1-3 An automatic power trolley for detecting rail damage includes a trolley body with a push handle fixed to one side. Two parallel aluminum alloy square tubes are installed inside the trolley body, each with a traveling mechanism at both ends. The traveling mechanism assists the trolley in moving along the rail. The traveling mechanism contains a power wheel, traveling wheels, and guide wheels, all of which are in close contact with the rail. This traveling structure is fundamental to ensuring the trolley's stable movement along the rail, allowing it to travel smoothly along the rail. Specifically, the power wheel, traveling wheels, and guide wheels are in close contact with the rail to ensure the trolley's movement. The power wheel uses a hub motor to provide power output for the trolley's movement on the rail, ensuring it can move at a predetermined speed. The traveling wheels are omnidirectional casters to ensure the trolley can turn on curved sections of the track. The guide wheels are distributed on the inner and outer sides of the rail to prevent the trolley from derailing during operation. Furthermore, a certain gap is maintained between each pair of guide wheels beyond the width of the rail, allowing the trolley to move in sections with varying track gauge. The vehicle body, or trolley frame, is divided into three parts, made of aluminum alloy square tubing, connected by a detachable bolt structure for easy disassembly and transportation. The trolley frame and guide wheel module are fixed to the frame by welding, and the push handle is a foldable design. The power supply provides a stable power supply for the entire trolley's operation; the controller accurately controls the trolley's operating status and can also be adjusted according to different detection targets and rail conditions.

[0049] Two acquisition mechanisms are installed at both ends of two aluminum alloy square tubes to acquire rail damage data. The acquisition mechanisms include a surface acquisition component and a profile acquisition component. Two acquisition boxes are fixed to both ends of the two aluminum alloy square tubes, and the surface acquisition component and profile acquisition component are mounted inside the acquisition boxes. The surface acquisition component includes an LED tube and an industrial camera 1. An LED tube is fixed to the top inside the acquisition box, and the industrial camera 1 is fixed to the top inside the acquisition box, offset from the LED tube. The LED tube provides sufficient illumination for the industrial camera 1 to capture clear images of the rail surface. The industrial camera 1 is fixed directly above the inside of the acquisition box to acquire images of rail surface damage. The profile acquisition component includes a laser emitter, a laser bracket, and an industrial camera 2. A laser bracket and industrial camera 2 are fixed to one side inside the acquisition box. A laser emitter is fixed to one end of the laser bracket. The laser emitter and one end of the laser bracket are placed inside the rail, projecting a laser beam onto the rail surface and the rail web. The industrial camera 2 is positioned close to the inside of the rail via the bracket, responsible for capturing visual image data of the laser-illuminated rail surface.

[0050] Example 2

[0051] like Figure 4 As shown, a qualitative identification and quantitative reconstruction method for rail damage detection, applicable to an automated power trolley for rail damage detection, includes the following steps:

[0052] S1: When conducting inspection work on the proposed railway route, use an automatic power trolley for rail damage detection to collect data, build a multi-type rail damage database, and widely collect and record common rail damage types in the line.

[0053] S2: Combining deep learning technology, the rail damage detection algorithm model is deeply trained using a multi-type rail damage database to build an intelligent rail damage detection system for rapid detection, thereby achieving qualitative identification of rail damage.

[0054] S3: By combining the grayscale value of the rail damage image with the rail damage wear depth value, the correlation between image grayscale and depth is explored.

[0055] S4: Establish a rail damage depth-grayscale correlation equation to map the rail damage image into a three-dimensional rail damage model, accurately reconstruct the rail damage, and thus observe the wear characteristics of rail damage more intuitively and clearly.

[0056] Example 3

[0057] Further optimizations to Example 1, specifically, such as... Figure 5 As shown, the construction of a multi-type rail damage database in S1 includes the following steps:

[0058] S11: Compare the effects of images acquired under different lighting conditions to select the most suitable light source parameters and configuration scheme for the on-site lighting environment; when detecting rail damage, first establish a database of multiple types of rail damage. Because the rail surface has been in service for a long time, the reflection phenomenon on the working surface is serious, so a lighting experiment is used to determine the image acquisition environment;

[0059] S12: Set the light source parameters and configuration scheme into the data acquisition box of the automatic force trolley for rail damage detection;

[0060] S13: Using an automated rail damage detection trolley, a large amount of on-site data was collected, and a database of multiple types of rail damage was built.

[0061] Example 4

[0062] Example 3 was further optimized, specifically, as follows: Figure 5-10 As shown, the qualitative identification of rail damage in S2 includes the following steps:

[0063] S21: First, combining the rail damage database built in S1, we then compare rail damage detection methods based on the SSD algorithm, Faster R-CNN algorithm, and YOLO algorithm; S22: We optimize the rail damage detection process to address issues such as unclear image information and missed detection of small-area damage; S23: Finally, we construct a machine vision-based rail damage qualitative recognition and detection system to achieve qualitative recognition of rail damage; First, on the data-augmented rail damage database, we compare methods based on the SSD algorithm, Faster R-CNN algorithm, and YOLO algorithm. The average detection accuracy of rail damage detection methods using R-CNN and YOLO algorithms is evaluated. Secondly, grayscale processing of rail damage images makes the texture features of rail damage more prominent. Histogram equalization is proposed to enhance tonal differences in the images, effectively highlighting rail damage features that can be accurately located and detected across different ranges, thus improving visual perception under reflective interference. Then, an FPN structure is added to the rail damage detection model to maximize the acquisition of rail damage feature information during detection. Finally, a machine vision-based qualitative identification and detection system for rail damage is built using PyQt5. A target detection algorithm is selected, and detection indicators such as confidence level, IOU parameters, and delayed activation are set to achieve accurate and efficient detection of rail damage and save the data.

[0064] Example 5

[0065] Example 2 was further optimized, specifically, as follows: Figure 11-12 As shown, S3 includes a quantitative reconstruction of rail damage, comprising the following steps:

[0066] S31: When quantitatively reconstructing rail damage, effective grayscale information of rail damage images is first obtained based on visual representation.

[0067] S32: The wear measurement method based on depth sensing obtains the actual wear value of rail damage, and then combines neural network fitting to establish a correlation model between image gray level and wear depth;

[0068] S33: Finally, based on the correlation model between image grayscale and wear depth, the three-dimensional reconstruction of rail damage is completed. When extracting the grayscale value of rail damage, the image detection and correction are first completed through edge detection, line segment extraction and image correction to ensure that the rail is in the center of the corrected image. Then, the image is located and segmented through threshold segmentation and binarization segmentation to complete the extraction of the rail damage surface. Then, the image is denoised by median filtering to eliminate the inherent interference of the image. Finally, the image is enhanced by multi-scale Retinex method to improve the visual quality of damage and color restoration, thus completing the accurate extraction of the grayscale value of the rail damage image. When extracting the wear depth value of rail damage, the camera and laser plane measurement system are first calibrated to eliminate system errors and restore image coordinates. Laser images of the rail profile are acquired using a data acquisition trolley, and a series of image processing steps are performed, including binarization extraction, perspective transformation, and morphological operations. When extracting the binarized contour of the laser stripes, the effects of skeleton extraction and maximum value detection methods are compared to obtain the actual rail profile. Finally, the Iterative Closest Point (ICP) algorithm and the Normal Distribution Transform (NDT) algorithm are compared to compare the actual rail profile with the standard rail profile, completing the measurement of the wear depth of rail damage. Finally, a neural network is used to fit the grayscale and wear depth of the rail damage image, establishing a grayscale-depth correlation equation for rail damage. This maps the rail damage image to a three-dimensional model of rail damage, allowing for a more intuitive and clear observation of the wear characteristics of the rail damage.

[0069] In summary:

[0070] This invention addresses the following technical problems: Current rail inspection methods are immature, mostly relying on manual inspection and hand-pushed rail inspection vehicles, which suffer from low efficiency, inaccuracy, high labor costs, and safety hazards. While large-scale integrated rail inspection vehicles offer high inspection efficiency, their scheduling is inconvenient, costly, and subject to downtime, limiting the demand for rail inspection. Current methods for measuring rail damage and wear characteristics typically employ steel rulers and electronic wear measuring scales, whose results are easily affected by the operating method, leading to errors. By adopting the technical solutions of the above embodiments and through the aforementioned settings, this application can certainly solve the above-mentioned technical problems and simultaneously achieve the following technical effects:

[0071] 1. This invention adopts an automatic force-disassembly design. The trolley frame is made of aluminum alloy square tube and is connected by bolts, which facilitates disassembly and transportation. The power wheels are powered by hub motors, and the running wheels are omnidirectional wheels to ensure steering in curved sections. The guide wheels are distributed on the inside and outside of the rails to prevent derailment and also meet the travel requirements in variable gauge sections, making operation more labor-saving.

[0072] 2. This invention achieves rapid qualitative identification by combining a multi-type rail damage database with a deep learning-trained detection algorithm model; and establishes a correlation equation by combining image grayscale values ​​with wear depth values ​​to achieve high-precision quantitative reconstruction, thereby more accurately restoring the rail damage and wear characteristics.

[0073] 3. This invention employs techniques such as grayscale processing of rail damage images, histogram equalization, and FPN structure to enhance image features and improve detection accuracy and range. It also utilizes PyQt5 to build a detection system, achieving precise and efficient detection. By fitting a neural network model to correlate the grayscale of rail damage images with wear depth, the damage images are mapped into a three-dimensional model, intuitively displaying damage characteristics and providing strong support for subsequent evaluation and polishing guidance.

[0074] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection, an electrical connection, or a connection that allows communication between them; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components, unless otherwise explicitly limited. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0075] Obviously, the embodiments described above are merely some embodiments of the present invention, not all embodiments. The accompanying drawings show preferred embodiments of the present invention, but do not limit the patent scope of the present invention. The present invention can be implemented in many different forms; rather, these embodiments are provided to provide a more thorough and complete understanding of the disclosure of the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing specific embodiments, or make equivalent substitutions for some of the technical features. Any equivalent structures made using the content of this specification and drawings, directly or indirectly applied to other related technical fields, are similarly within the patent protection scope of this invention.

Claims

1. A method for qualitative identification and quantitative reconstruction of rail damage detection, characterized in that, Includes the following steps: S1: When conducting inspection work on the proposed railway route, use an automatic power trolley for rail damage detection to collect data, build a multi-type rail damage database, and extensively collect and record common rail damage types in the line. S2: Combining deep learning technology, the rail damage detection algorithm model is deeply trained using a multi-type rail damage database to build an intelligent rail damage detection system for rapid detection, thereby achieving qualitative identification of rail damage. S3: By combining the grayscale value of the rail damage image with the rail damage wear depth value, the correlation between image grayscale and depth is explored. S4: Establish the rail damage depth-grayscale correlation equation to map the rail damage image into a three-dimensional rail damage model and restore the rail damage with high accuracy.

2. The method for qualitative identification and quantitative reconstruction of rail damage detection according to claim 1, characterized in that, The construction of the multi-type rail damage database in S1 includes the following steps: S11: Compare the effects of images acquired under different lighting conditions, and select the most suitable light source parameters and configuration scheme for the on-site lighting environment; S12: Set the light source parameters and configuration scheme into the data acquisition box of the automatic force trolley for rail damage detection; S13: Using an automated rail damage detection trolley, a large amount of on-site data was collected, and a database of multiple types of rail damage was built.

3. The method for qualitative identification and quantitative reconstruction of rail damage detection according to claim 2, characterized in that, The qualitative identification of rail damage in S2 includes the following steps: S21: First, combine the rail damage database built in S1, and then compare the rail damage detection methods based on SSD algorithm, Faster R-CNN algorithm and YOLO algorithm. S22: Optimize the process of rail damage detection to address issues such as unclear image information and missed detection of small-scale damage. S23: Finally, construct a machine vision-based qualitative identification and detection system for rail damage to achieve qualitative identification of rail damage.

4. The method for qualitative identification and quantitative reconstruction of rail damage detection according to claim 1, characterized in that, S3 includes a quantitative reconstruction of rail damage, comprising the following steps: S31: When quantitatively reconstructing rail damage, effective grayscale information of rail damage images is first obtained based on visual representation. S32: The wear measurement method based on depth sensing obtains the actual wear value of rail damage, and then combines neural network fitting to establish a correlation model between image gray level and wear depth; S33: Finally, the three-dimensional reconstruction of rail damage is completed based on the correlation model between image grayscale and wear depth.

Images