Longitudinal displacement detection method and equipment for continuously welded rail tracks

Abstract

The invention relates to a method and equipment for detecting longitudinal displacement of a seamless track steel rail, and belongs to the technical field of steel rail detection. The longitudinal displacement detection device comprises a hand-push type support which is used for being arranged on a steel rail and providing support for other modules; the camera module is used for capturing image information of a displacement observation pile and an observation scale on the steel rail; and the centralized control module is used for collecting, processing and transmitting data. A second industrial personal computer configured by the centralized control module is configured to perform illumination adjustment on image data acquired in real time by the camera module in the process of moving along with the hand-push type bracket by using a deep learning algorithm to ensure that an image is clear; centroid coordinates of the observation pile and the observation scale are calculated through an edge detection and contour extraction algorithm; and comparing the images acquired in real time with the images in the basic database, calculating the longitudinal displacement of the steel rail, and updating the database.

Description

Technical Field

[0001] The present invention relates to the technical field of rail detection, and in particular to a method and device for detecting longitudinal displacement of seamless line rails. Background Art

[0002] The seamless track is usually welded from multiple short rails, which reduces a large number of joints, effectively eliminates the impact and noise caused by the joints of traditional tracks, improves the stability and comfort of train operation, and reduces the maintenance costs of locomotives and tracks, extending their service life. In addition, the seamless track has the characteristics of simple structure, convenient maintenance, and long service life, so it is used more and more widely. The seamless track rails are fastened to the sleepers through intermediate fasteners (including anti-climbing devices), and the rail ends are fully clamped by the joint fasteners, so they cannot be freely extended after being locked. The larger free extension amount becomes a smaller restricted extension amount, and the part of the rail that cannot be freely extended is accumulated in the form of temperature stress inside the rail. The greater the change in rail temperature, the greater the longitudinal displacement and internal stress of the rail. When the fastener pressure is not enough to offset the internal stress, the rail will undergo longitudinal displacement, affecting driving safety. In order to ensure operational safety and grasp the trend of rail stress changes, it is necessary to periodically detect the longitudinal displacement of the seamless track rails.

[0003] Since the longitudinal displacement of rails is a key inspection item for seamless lines, the current detection methods are mostly manual patrol measurements or real-time monitoring by fixed-installed sensors. These methods have disadvantages such as low efficiency, high equipment installation costs, and the fixed equipment is not capable of being used across lines.

[0004] CN207826260U discloses a temperature displacement detection device for seamless rails based on image recognition. It includes a detection box installed on both sides of the track. A power supply, a temperature sensor, a camera, a fill light, an embedded central computer, a communication device and an antenna are arranged in the detection box. This patent relates to a temperature displacement detection device for seamless rails based on image recognition. Through a detection box (with a temperature sensor, a camera, a fill light, an industrial computer, etc.) installed at a specific position of the track line, the temperature and displacement changes of the rails are monitored at a fixed point, and the detection results can be transmitted to the monitoring center through a mobile data wireless network.

[0005] CN114942084A discloses a monitoring method and system for the temperature change of a seamless track, which includes the following steps for the installation of supporting data sampling: step 1, selecting a test section of a rail as the installation surface in the tested section of the rail, installing a strain gauge sensor and a rail temperature sensor; step 2, after the sensor is stably installed, connect a signal acquisition instrument at the construction site to collect strain and rail temperature, and set the sampling interval. This patent installs strain gauge sensors and temperature sensors in a specific section of the rail, sets the sampling interval through an external signal acquisition instrument, and realizes the collection of strain and rail temperature.

[0006] Li Kunpeng proposed a rail longitudinal displacement detection system based on image processing technology in the article "Design and Implementation of Rail Longitudinal Displacement Detection System Based on Image Processing" (Li Kunpeng. Design and Implementation of Rail Longitudinal Displacement Detection System Based on Image Processing [D]. Xi'an University of Technology, 2022.), which consists of a mechanical part and a software part. Before acquisition, it is necessary to deploy image acquisition equipment near a specific measurement point, install a marker on the waist of the rail at the measurement point, and bury a reference pile on one side of the track. The image is captured by the image acquisition device, and then the position information of the marker in the image is found by using computer image processing technology. After comparing it with the calibrated image, the longitudinal displacement of the rail within a certain period of time can be obtained.

[0007] The methods for detecting the longitudinal displacement of seamless rails are mainly divided into contact and non-contact measurement methods. Different methods have the following disadvantages:

[0008] (1) Contact measurement method: Through manual on-site inspection, the relative position changes of the displacement observation pile and the rail bottom scale are compared and measured using a square ruler. In addition to requiring a large amount of manpower, it is difficult to accurately control the selection of the geometric center of the observation pile during manual inspection, which may lead to large measurement errors. In addition, ultrasonic sensors, strain gauge sensors, rail temperature sensors, etc. can be installed at the joint position to achieve real-time monitoring of changes in rail stress and temperature. This method requires the deployment of a large number of sensor equipment along the entire line, which has high investment costs and subsequent maintenance costs, and does not have the ability to be applied across lines.

[0009] (2) The non-contact measurement method also requires the placement of corresponding observation marks on the rails and the roadbed to achieve accurate measurement of rail displacement through machine vision technology. However, the image acquisition equipment used in the existing methods is mostly fixed, which limits its applicability in dynamic measurement.

[0010] Therefore, there is an urgent need in the art for a method and device that is more suitable for detecting the longitudinal displacement of seamless line rails.

[0011] In addition, on the one hand, there are differences in understanding among those skilled in the art; on the other hand, the applicant studied a large number of documents and patents when making the present invention, but due to space limitations, not all details and contents are listed in detail. However, this does not mean that the present invention does not have the characteristics of these prior arts. On the contrary, the present invention already has all the characteristics of the prior art, and the applicant reserves the right to add relevant prior art to the background technology. Summary of the invention

[0012] In view of the deficiencies of the prior art, the present invention provides a method and device for detecting the longitudinal displacement of seamless line rails, and in particular, a method and device for detecting the longitudinal displacement of seamless line rails in real time at a low speed, so as to solve at least some of the above-mentioned technical problems.

[0013] The present invention realizes the real-time acquisition of the observation pile and observation scale of the longitudinal displacement of the seamless line rail under low-speed conditions (no more than 15km / h) through the developed seamless line rail longitudinal displacement detection equipment. Operation under low-speed conditions ensures the stability and continuity of image acquisition, reduces the vibration and jitter during the movement of the equipment, and thus improves the image quality and consistency. The equipment adopts machine vision technology and deep learning algorithms to capture images through high-definition linear array cameras and linear laser fill light modules, and uses deep learning algorithms to automatically learn and identify features in the image, such as the contours of observation piles and observation scales. This not only improves the robustness and accuracy of image processing, but also enables the equipment to work stably in various complex environments. The detection accuracy is better than 1mm, which means that the equipment can measure the longitudinal displacement of the rail very accurately, detect small displacement changes in time, prevent potential safety hazards, and ensure the safe operation of the railway. In addition, the design of the hand-push bracket makes the equipment easy for on-site staff to push and run smoothly on the rails, improves the detection efficiency, and reduces the installation time and cost of the equipment. Compared with traditional fixed detection equipment, the equipment of the present invention can be flexibly used on different lines without fixed installation, which significantly reduces the cost of railway operation and maintenance. In summary, the present invention significantly improves the efficiency and accuracy of longitudinal displacement detection of seamless line rails through real-time acquisition under low-speed conditions, combination of machine vision technology and deep learning algorithm, high-precision detection and flexible and portable design, solves many problems existing in traditional detection methods, and provides strong guarantee for the safe operation of railways.

[0014] Specifically, the present invention discloses a longitudinal displacement detection device for seamless line rails, which includes: a hand-push bracket, which is used to be arranged on the rails and provide support for other modules; a camera module, which is used to capture the image information of the displacement observation pile and the observation scale on the rails; and a centralized control module, which is used to realize data collection, processing and transmission. Preferably, the second industrial computer configured in the centralized control module is configured to: use a deep learning algorithm to adjust the illumination of the image data collected in real time by the camera module in the process of following the movement of the hand-push bracket to ensure the clarity of the image; calculate the centroid coordinates of the observation pile and the observation scale by edge detection and contour extraction algorithms; compare the real-time collected image with the image in the basic database, calculate the longitudinal displacement of the rail, and update the database.

[0015] According to a preferred embodiment, the hand-push bracket includes a bracket body and a plurality of wheels connected to the bracket body through supporting legs, and the camera module and the centralized control module are directly or indirectly arranged on the bracket body through corresponding mounting frames, so that when the wheels are driven to rotate, the hand-push bracket can drive the camera module and the centralized control module to move along the extension direction of the rail, and during the movement, the image data collected by the camera module is sent to the centralized control module for data processing to calculate the longitudinal displacement of the rail.

[0016] According to a preferred embodiment, the hand-push bracket includes a height-adjustable handle, and the wheels are driven to rotate by applying external force to the handle. A speed encoder for recording the travel distance and speed of the device is provided on the rotating shaft of one of the wheels. The speed encoder is connected to the pulse port of the centralized control module through a pulse line to continuously send pulse signals to the centralized control module, so that the centralized control module generates timestamps accordingly and allocates these timestamps to the camera module.

[0017] According to a preferred embodiment, the camera module is installed on the wide side of the bracket body to be suspended above the rail, wherein when the hand-push bracket moves along the extension direction of the rail, the camera module can shoot the displacement observation piles and observation scales on the rail at predetermined time intervals based on real-time adjusted shooting parameters to generate a series of continuous image sequences.

[0018] The present invention also discloses a method for detecting longitudinal displacement of seamless line rails, which comprises the following steps:

[0019] Original image acquisition: real-time acquisition of rail image data in a moving manner along the extension direction of the rail;

[0020] Image illumination adjustment based on deep learning: Use deep learning algorithms to adjust the illumination of collected image data to ensure clear images;

[0021] Observation pile and observation scale contour recognition and centroid calculation: Calculate the centroid coordinates of the observation pile and observation scale through edge detection and contour extraction algorithms;

[0022] Real-time calculation of rail longitudinal displacement: Compare the real-time acquired images with the images in the basic database, calculate the longitudinal displacement of the rail, and update the database.

[0023] According to a preferred embodiment, in the step of "image illumination adjustment based on deep learning", the image data is illuminated by an illumination module comprising a deep neural network, which can learn through training a method of estimating illumination components to remove the influence of illumination and restore the inherent properties of the image.

[0024] According to a preferred embodiment, in the step of "observation pile and observation scale contour identification and centroid calculation", the following sub-steps are included:

[0025] Use two 3×3 convolution kernels to convolve the preprocessed image to detect horizontal and vertical edges;

[0026] Calculate the gradient size and direction at each pixel point, and determine the pixel points with gradient values ​​greater than the threshold as edge points by setting a threshold;

[0027] Use the connected region algorithm to connect adjacent edge points to form a complete contour, and filter out contours that meet the characteristics of observation piles and observation scales through set features;

[0028] Calculates the centroid coordinates of the selected contour.

[0029] According to a preferred embodiment, before performing real-time calculation of the longitudinal displacement of the rails, displacement observation piles and observation scales of the entire line are collected as a basic database. The basic database includes initial position information and image data of each observation point, so that during actual testing, pre-collected images of observation piles and observation scales can be retrieved from the basic database to achieve comparison with the test data, wherein the relative relationship of the comparison is the difference between the horizontal and vertical coordinates.

[0030] According to a preferred embodiment, the longitudinal displacement detection method also includes the following steps: detection equipment installation and wiring: placing a hand-push bracket equipped with a camera module and a centralized control module on the rail, and completing the wiring and initial settings of each component to use the installed longitudinal displacement detection equipment for detection.

[0031] According to a preferred embodiment, the longitudinal displacement detection method further includes the following steps: disease classification alarm: according to a preset threshold value, the longitudinal displacement of the rail is classified into health levels, and a real-time alarm is issued for abnormal conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 It is a schematic structural diagram of the longitudinal displacement detection device provided by the present invention at a viewing angle;

[0033] Figure 2 is a schematic structural diagram of the longitudinal displacement detection device provided by the present invention from another viewing angle;

[0034] Figure 3 It is a hardware connection diagram of the longitudinal displacement detection device provided by the present invention;

[0035] Figure 4 is a panel schematic diagram of the centralized control module provided by the present invention;

[0036] Figure 5 It is the main flow chart of the longitudinal displacement detection method provided by the present invention.

[0037] Reference numerals list

[0038] 100: hand-push bracket; 110: bracket body; 111: wide side; 112: long side; 120: supporting leg; 121: rod; 130: wheel; 131: speed encoder; 140: handle; 141: locking knob; 150: first mounting bracket; 160: second mounting bracket; 170: third mounting bracket; 200: camera module; 300: centralized control module; 310: second industrial computer; 400: rail. DETAILED DESCRIPTION

[0039] The following is a detailed description with reference to the accompanying drawings.

[0040] Example 1

[0041] The present invention discloses a device for detecting the longitudinal displacement of a seamless line rail 400, and in particular, a device for detecting the longitudinal displacement of a seamless line rail 400 at a low speed and in real time, wherein the "low speed" referred to in the present invention is a speed not exceeding 15 km / h.

[0042] Preferably, if Figure 1 to Figure 3 As shown, the longitudinal displacement detection device of the present invention may include: a hand-push bracket 100, which is used to be arranged on the rail 400 and provide support for other modules; a camera module 200, which is used to capture the image information of the displacement observation pile and the observation scale on the rail 400; and a centralized control module 300, which is used to realize data collection, processing and transmission. Further preferably, the longitudinal displacement detection device may be configured with a camera module 200 on both sides thereof, respectively, to obtain the image information of the two rails 400.

[0043] Preferably, the hand-push bracket 100 is the basic supporting component of the entire device, and its design takes lightness and stability into consideration so that it can be easily pushed by on-site workers and run smoothly on the rails.

[0044] Preferably, the hand-push bracket 100 may include a bracket body 110, whose overall structure is roughly a rectangular frame, ensuring the lightness and sturdiness of the equipment. Preferably, the bracket body 110 may be made of high-strength aluminum alloy material, and the surface is treated with anti-corrosion to ensure durability for long-term use. Furthermore, the length of the bracket body 110 is slightly smaller than the standard track gauge, so that assembly structures for connecting with other components can be provided on both sides of the length direction of the bracket body 110, wherein the length direction refers to a direction perpendicular to the extension direction of the two paired rails 400, the width direction refers to a direction parallel to the extension direction of the two paired rails 400, and the height direction refers to a direction orthogonal to both the length direction and the width direction.

[0045] Preferably, the two sides in the length direction of the bracket body 110 can be referred to as wide sides 111, and the assembly structure provided at the four corners of the wide sides 111 can be used to achieve a detachable connection with the support leg 120 of the hand-push bracket 100, wherein the support leg 120 can be provided with a rod 121 extending in the height direction, and the rod 121 is provided with a plurality of threaded holes spaced apart in the height direction. Further, the assembly structure can include a first through hole capable of accommodating the rod 121 of the support leg 120, so that the rod 121 of the support leg 120 can pass through the first through hole in the height direction and stay at an appropriate position, wherein the appropriate position can be determined by the user according to the height of the bracket body 110 required by the user, and one of the threaded holes of the rod 121 staying at the appropriate position can be roughly aligned with the second through hole of the assembly structure, so that the bolt of the assembly structure can pass through the second through hole and achieve a detachable connection with the corresponding threaded hole.

[0046] Preferably, the support legs 120 installed at the four corners of the support body 110 can be configured with wheels 130, wherein the wheels 130 are installed in the shaft holes of the support legs 120 through bearings. Preferably, the wheels 130 can be slightly wider than the surface width of the rails 400, so that the wheels 130 can be completely placed on the surface of the rails 400, so that the support body 110 is suspended in the air area between the two pairs of rails 400 through the support legs 120. Preferably, the wheels 130 can rotate along the extension direction of the rails 400, so that the hand-push type support 100 can move along the width direction under the drive of external force. Preferably, a speed encoder 131 can be provided on the rotating shaft of any wheel 130 to collect the rotation information of the wheel 130, so as to obtain the speed data of the hand-push type support 100 when it moves along the rails 400.

[0047] Preferably, the hand-push bracket 100 may include a height-adjustable handle 140, which is made of a stainless steel tube and has an anti-slip surface treatment to ensure that the operator feels comfortable and does not slip when pushing the device. The height adjustment range of the handle 140 is 0.8 meters to 1.2 meters, and the required height is fixed by rotating the locking knob 141 at the bottom of the handle 140 to accommodate operators of different heights. Preferably, the bottom of the handle 140 can be connected to one side of the bracket body 110 in the width direction, that is, it can be connected to the long side 112 of the bracket body 110, wherein the handle 140 can be especially connected to the middle of the long side 112 of the bracket body 110 to ensure uniform force during pushing.

[0048] Preferably, a mounting interface for externally connecting other modules (such as a camera module 200 and a centralized control module 300) may also be provided on the wide side 111 of the bracket body 110, wherein these mounting interfaces adopt a standard quick-release design and are fixed by screws for convenient and quick installation and removal. A rubber gasket is provided inside the mounting interface to prevent the camera module 200 and the centralized control module 300 from loosening or shaking during movement. The size and position of each mounting interface are carefully designed to ensure that the camera module 200 and the centralized control module 300 can remain stable after installation and will not interfere with each other. Preferably, the bracket body 110 can be connected to other modules directly or indirectly through a mounting frame, wherein the mounting frame can be connected to the mounting interface of the bracket body 110 so that other modules can be installed on the mounting frame.

[0049] Preferably, the first mounting frame 150 for mounting the camera module 200 may be disposed at the outer region of the wide side 111 of the bracket body 110, and the first mounting frame 150 may be configured as a box-like structure to be engaged in the U-shaped groove of the bracket body 110, wherein a first mounting interface may be disposed in the U-shaped groove of the bracket body 110. Preferably, the second mounting frame 160 for mounting the centralized control module 300 may cover the bracket body 110 and be connected to the inner region of the wide side 111 of the bracket body 110, wherein a second mounting interface is disposed on the inner region, wherein the second mounting frame 160 may include a mounting plane for connecting with the second mounting interface and an operating plane for mounting the centralized control module 300, wherein the operating plane is slightly higher than the mounting plane so that the operating plane is elevated relative to the bracket body 110.

[0050] The hand-push bracket 100 of the present invention is designed with full consideration of various practical needs. It adopts high-strength aluminum alloy materials to achieve a perfect combination of lightweight and durability. The operator can easily push the device to move on the track, reducing labor intensity and improving work efficiency. The height-adjustable design of the handle 140 adapts to operators of different heights and ensures the comfort of the operator when pushing the device. The speed encoder 131 can be directly connected to the centralized control module 300 via a data cable to record the travel distance and speed of the device without additional operation, simplifying the use process. The installation interface adopts a quick-release design and is fixed by screws, which facilitates the rapid installation and removal of the camera module 200 and the centralized control module 300, improving the flexibility and maintenance convenience of the equipment. The second mounting frame 160 in the middle can not only be used by the operator to temporarily place the operating computer, but also other tools or equipment, further enhancing the versatility and convenience of the equipment. These design advantages jointly ensure the efficiency and reliability of the low-speed real-time seamless line rail 400 longitudinal displacement detection equipment in practical applications.

[0051] Preferably, the camera module 200 may include a plurality of high-performance high-definition linear array cameras, which may adopt industrial-grade CMOS sensors with a resolution of 2K pixels and a maximum line frequency of up to 60KHz. The camera housing is made of high-strength aluminum alloy material, which has good heat dissipation and vibration resistance. The front end of the camera is equipped with a high-quality optical lens with a focal length of 50mm and an aperture range of f / 2.8 to f / 16, which can adapt to different lighting conditions. The camera is connected to the power supply port of the centralized control module 300 via a four-core power supply cable to obtain a stable power supply. Data transmission is connected to the network port of the centralized control module 300 via a network cable to ensure high-speed transmission of image data.

[0052] Preferably, in order to ensure that clear images can be obtained under different lighting conditions, the camera module 200 is equipped with a line laser fill light module. The line laser fill light module consists of a group of line laser emitters and a controller, which is installed in front of the camera and parallel to the camera lens. The laser beam emitted by the line laser emitter forms a bright line on the rail 400, which significantly improves the contrast and clarity of the image. The controller is connected to the pulse port of the centralized control module 300 through a pulse line, receives the trigger signal sent by the synchronization control board, and ensures that the laser fill light is synchronized with the camera shooting. This design not only improves the image quality, but also enables accurate displacement detection in low light environments.

[0053] Preferably, the camera module 200 has a built-in high-performance first industrial computer responsible for image processing and preliminary data analysis. The first industrial computer adopts an embedded design and is installed at the rear of the camera module 200. It exchanges heat with the external air through a heat sink to ensure stability and reliability for long-term operation. The first industrial computer is connected to the centralized control module 300 via a network cable to achieve data transmission and command reception. The first industrial computer integrates a deep learning algorithm, which can process the image data collected by the camera in real time, extract the contour information of the observation pile and the observation scale, and calculate the displacement data.

[0054] Preferably, the heat sink installed on the back of the first industrial computer is in close contact with the CPU and GPU of the first industrial computer through thermal grease. The heat sink is made of aluminum alloy material, has good thermal conductivity, and is provided with multiple heat sink fins on the surface to increase the heat dissipation area. A fan is provided at the front end of the heat sink to further improve the heat dissipation effect through air cooling. Both sides of the heat sink are fixed to the housing of the camera module 200 by screws to ensure the stability and reliability of the heat dissipation system. This design effectively solves the heat dissipation problem of the first industrial computer when it works for a long time and ensures the stable operation of the system.

[0055] Preferably, the camera module 200 can be mounted on the first mounting frame 150, wherein the first mounting frame 150 can be made of high-strength aluminum alloy material and fixed to the first mounting interface of the hand-push bracket 100 by four screws. The design of the first mounting frame 150 allows the camera module 200 to be fine-tuned in the vertical direction to ensure that the relative position of the camera lens and the rail 400 is always consistent. A shock-absorbing pad is provided at the front end of the first mounting frame 150 to absorb the vibration generated by the device during movement and protect the camera module 200 from damage.

[0056] Preferably, the power supply and data interface of the camera module 200 are located at the rear of the module and connected to the centralized control module 300 through a four-core power supply cable and a network cable. The four-core power supply cable provides a stable DC power supply, and the network cable is used to transmit image data and control instructions. The interface part adopts a waterproof design, and a sealing ring is provided inside to effectively prevent water vapor and dust from entering, thereby improving the durability and reliability of the device.

[0057] The camera module 200 of the present invention takes full account of efficiency, stability and reliability. The high resolution and high frame rate of the high-definition linear array camera ensure the clarity and real-time performance of the image, and can accurately capture the details of the displacement observation pile and observation scale of the rail 400 under low-speed conditions. The introduction of the linear laser fill light module significantly improves the quality and contrast of the image under different lighting conditions, especially in low-light environments, the laser fill light can provide sufficient brightness to ensure the clarity of the image. The high performance and embedded design of the first industrial computer not only realize the real-time processing of the image, but also improve the accuracy of displacement detection through the integrated deep learning algorithm. The combined design of the heat sink and the fan effectively solves the heat dissipation problem of the first industrial computer when it works for a long time, ensuring the stable operation of the system. The fine-tuning function and shock absorption design of the first mounting frame 150 keep the camera module 200 stable during movement, avoiding image blurring caused by vibration. The waterproof design of the power supply and data interface improves the durability and reliability of the equipment, and is suitable for various harsh environments. These design advantages jointly ensure the efficiency and reliability of the camera module 200 in the low-speed real-time seamless line rail 400 longitudinal displacement detection.

[0058] Preferably, the shell of the centralized control module 300 can be made of high-strength aluminum alloy material, which has good heat dissipation performance and vibration resistance. The overall shell is roughly in the shape of a rectangular parallelepiped, and the surface is anodized, which is corrosion-resistant and beautiful. The four corners of the shell are fixed to the second mounting frame 160 of the hand-push bracket 100 by bolts to ensure that the module remains stable during movement. A plurality of heat dissipation holes are provided at the front and rear ends of the shell, and the internal fan exchanges heat with the external air through these heat dissipation holes to effectively reduce the internal temperature. Preferably, a third mounting frame 170 can also be provided on the second mounting frame 160 to semi-enclose the centralized control module 300, thereby protecting the centralized control module 300, and at the same time, a placement plane can be formed on the centralized control module 300, wherein the user can place the operating computer on the placement plane so that the operating computer can not directly contact the centralized control module 300, thereby ensuring sufficient heat dissipation. Preferably, the third mounting frame 170 can be detachably provided on the operating plane of the second mounting frame 160 by bolts.

[0059] Preferably, the synchronous control board arranged in the shell of the centralized control module 300 adopts a multi-layer PCB design and integrates multiple high-performance processors and communication interfaces. The main function of the synchronous control board is to coordinate the work of each submodule to ensure the synchronization of data acquisition and processing. It is connected to the speed encoder 131 and the camera module 200 through the pulse port, receives the pulse signal sent by the speed encoder 131, and converts it into a synchronous control signal, which is transmitted to the camera module 200 to ensure that the camera performs image acquisition at a specific time point. The synchronous control board is also connected to the camera module 200 and the operating computer through the network port to realize data transmission and command reception. Preferably, the synchronous control board, which is one of the core components of the centralized control module 300, can be installed in a central position inside the shell.

[0060] Preferably, the power supply module of the centralized control module 300 is installed on one side inside the shell, and is responsible for providing a stable power supply for the entire centralized control module 300. The power supply module adopts a dual power input design, and can be connected to an external power adapter through the charging port, or connected to a vehicle battery through the vehicle power interface. A voltage stabilizing circuit and an overcurrent protection circuit are provided inside the power supply module to ensure a stable output voltage and prevent overcurrent and short circuit. The power supply module is connected to the camera module 200 through a four-core power supply port to provide the required DC power supply. In addition, the power supply module also displays the current voltage and current status through a small display screen, which is convenient for operators to monitor the power supply status.

[0061] Preferably, the second industrial computer 310 of the centralized control module 300 adopts a high-performance embedded design, equipped with a multi-core processor with a relatively fast computing speed (such as an Intel Core i7 or higher-level processor) and a large-capacity memory, and can process a large amount of image data and complex algorithm calculations. The second industrial computer 310 is connected to the camera module 200 through a network port, receives image data and performs real-time processing. The second industrial computer 310 integrates a deep learning algorithm and image processing software, which can automatically identify the contours of the observation pile and the observation scale, calculate the longitudinal displacement of the rail 400, and store the results in an internal hard disk. The second industrial computer 310 is also connected to the operating computer through a network port, and the processing results are transmitted to the operating computer in real time for display and storage. Preferably, the second industrial computer 310, which is another core component of the centralized control module 300, can be installed at the rear of the inside of the housing.

[0062] Preferably, the heat dissipation module of the centralized control module 300 includes a plurality of heat sinks and a small fan. The heat sink is made of aluminum alloy material, has good thermal conductivity, and is provided with a plurality of fins on the surface to increase the heat dissipation area. The small fan is installed at the front end of the heat sink to further improve the heat dissipation effect through air cooling. This design of the heat dissipation system effectively solves the heat dissipation problem of the second industrial computer 310 and the synchronous control board when working for a long time, and ensures the stable operation of the system.

[0063] Preferably, if Figure 4 As shown, the centralized control module 300 may be provided with a series of interfaces for external connection to other modules or devices, wherein these interfaces include: a charging port: used to connect an external power adapter or a vehicle battery to provide power to the power supply module; a four-core power supply port: used to connect the camera module 200 to provide a DC power supply; a pulse port: used to connect the speed encoder 131 and the camera module 200 to transmit pulse signals and synchronous control signals; a network port: used to connect the camera module 200 and the operating computer to realize data transmission and command reception; a USB port: used for data copying and external storage device connection to facilitate data backup and export.

[0064] Preferably, if Figure 4 As shown, the centralized control module 300 may be provided with physical buttons for power on / off and emergency stop. Preferably, the centralized control module 300 may be provided with an operation panel for displaying system status, setting parameters and / or viewing real-time data, wherein the operation panel may be an operation computer.

[0065] The centralized control module 300 of the present invention takes full account of efficiency, stability and reliability. The shell is made of high-strength aluminum alloy material, which is not only light but also durable and can effectively protect the internal components from the influence of the external environment. The multi-layer PCB design and high-performance processor of the synchronous control board ensure the synchronization and accuracy of data acquisition and processing, and realize efficient collaborative work through the connection of the pulse port and the network port with the speed encoder 131 and the camera module 200. The dual power input design and voltage stabilizing circuit of the power supply module ensure the stability and safety of the power supply, and the small display screen facilitates the operator to monitor the power supply status. The high-performance embedded design and internally integrated deep learning algorithm of the second industrial computer 310 can process a large amount of image data in real time, improving the accuracy and speed of displacement detection. The multiple heat dissipation design of the heat dissipation system effectively solves the heat dissipation problem during long-term work and ensures the stable operation of the system. The diversified design of the interface module enables the device to flexibly connect various external devices, improving the compatibility and scalability of the system. These design advantages jointly ensure the efficiency and reliability of the centralized control module 300 in the longitudinal displacement detection of the low-speed real-time seamless line rail 400.

[0066] In actual operation, the operator first places the device at the starting point of the road section to be tested and starts the device. As the push-type bracket 100 moves forward, the speed encoder 131 continuously sends pulse signals to the centralized control module 300, and the synchronous control board generates timestamps based on this and assigns these timestamps to the camera module 200. The camera module 200 shoots the displacement observation piles and observation scales on the rail 400 at predetermined time intervals to generate a series of continuous image sequences. These images are transmitted to the second industrial computer 310, and after being processed by the deep learning algorithm, key features such as the position change of the observation piles are extracted from them. Finally, the second industrial computer 310 calculates the longitudinal displacement of the rail 400 based on these features, and displays the results on the screen of the operating panel (such as an operating computer) for the operator to view and record.

[0067] Preferably, when performing the longitudinal displacement detection, the longitudinal displacement detection device may be installed and wired first, wherein the installation and wiring of the longitudinal displacement detection device may include the following steps:

[0068] Place the hand-push bracket 100 on the steel rail 400, and ensure that the wheels 130 at the bottom of the hand-push bracket 100 fit tightly with the steel rail 400, so that the hand-push bracket 100 remains level;

[0069] The camera module 200 and the centralized control module 300 are installed at corresponding positions of the hand-push bracket 100. The camera module 200 is installed on the hand-push bracket 100 through the first mounting bracket 150, and the centralized control module 300 is installed on the hand-push bracket 100 through the second mounting bracket 160;

[0070] A speed encoder 131 is mounted on the rotating shaft of one of the wheels 130 of the hand-push bracket 100 and connected to the pulse port of the centralized control module 300 through a pulse line to record the travel distance and speed of the device;

[0071] Place the operating computer on the placement plane above the centralized control module 300, and ensure that the computer screen faces the operator for easy viewing and operation;

[0072] Connect the three wires (power supply wire, network wire, pulse wire) of the two camera modules 200 to the corresponding interfaces of the centralized control module 300, the power supply wire is connected to the four-core power supply port, the network wire is connected to the network port, and the pulse wire is connected to the pulse port;

[0073] Start the dedicated software on the operating computer and perform the initial system settings, where the settings may include parameters such as the fill light intensity, acquisition frequency, and image resolution of the camera module 200 to ensure that the equipment is in the best working condition before formal testing.

[0074] Preferably, after the longitudinal displacement detection device is installed and wired, the longitudinal displacement detection device can be driven to move along the extension direction (i.e., width direction) of the rail 400 while the camera module 200 is used to collect images, wherein the original image collection may include the following steps:

[0075] Adjust the fill light intensity, acquisition frequency and other parameters of the camera module 200 on the operating computer to ensure that clear images can be obtained under different lighting conditions;

[0076] The hand-push bracket 100 is pushed to move (evenly and slowly) on the rail 400, and the two camera modules 200 collect images of the outer sides of the left and right rails 400 in real time during the movement, wherein the moving speed of the equipment can be limited to no more than 15 km / h;

[0077] As the device moves, the image data collected by the camera module 200 can be transmitted to the centralized control module 300 via the network cable in real time and stored in the internal hard disk of the second industrial computer 310. At the same time, the collected image will be displayed in real time on the operating computer, and the operator can view and adjust the parameters at any time.

[0078] Since the longitudinal displacement observation piles and observation scales of the seamless line rail 400 are more exposed than track structures such as fasteners, rails 400 and roadbeds under the same light intensity, they are more likely to be unclear in photographing the scales and observation piles due to overexposure. Therefore, the second industrial computer 310 in the centralized control module 300 of the present invention can complete image lighting adjustment based on deep learning.

[0079] According to the Retinex theory, there is a relationship between the source image y under different illumination and the intrinsic property z of the object itself that is expected to be obtained: y=z×x, where x represents the illumination component. Furthermore, the intrinsic property z of the object refers to the object's ability to reflect light.

[0080] Preferably, the second industrial computer 310 in the centralized control module 300 of the present invention can learn illumination by introducing a deep neural network, thereby removing the influence of illumination and restoring the inherent properties of the image, wherein its basic unit can be formalized as the following formula:

[0081]

[0082] Among them, x t is the illumination component of the tth iteration, u t is the illumination residual of the tth iteration, H θ (·) is a deep neural network used to estimate the illumination residual u t , θ is the parameter of the deep neural network, including weights and biases, etc., x 0=y is the initial illumination component equal to the source image y, and F(·) represents the illumination module, which mainly includes a deep neural network.

[0083] In the present invention, the illumination module refers to a component or algorithm used to estimate and adjust image illumination.

[0084] Preferably, during the training phase, the deep neural network H θ (·) Gradually optimize the illumination residual u by learning from a large number of images under different lighting conditions t The estimation accuracy.

[0085] Preferably, after the module is trained, the intrinsic property z of the object to be acquired can be inferred based on the illumination component x estimated by the illumination module F(·) and the source image y under different illuminations. Specifically, the image y can be input first and processed by the illumination module F(·) to obtain the illumination component x; the intrinsic property z of the object itself can be calculated using the formula z=y / x, that is, the image after the influence of illumination is removed, thereby reducing the overexposure of the image caused by exposure problems.

[0086] Preferably, the second industrial computer 310 in the centralized control module 300 of the present invention can perform edge detection on the processed image, wherein the edge detection may include:

[0087] Convolution operation: Use two 3×3 convolution kernels (one for horizontal direction and one for vertical direction) to convolve the image. The horizontal convolution kernel is used to detect horizontal edges, and the vertical convolution kernel is used to detect vertical edges.

[0088] Gradient calculation: Calculate the magnitude and direction of the gradient at each pixel. The magnitude of the gradient indicates the edge strength of the point, and the direction of the gradient indicates the edge direction of the point.

[0089] Threshold processing: Set a threshold and determine the pixels whose gradient values ​​are greater than the threshold as edge points.

[0090] Preferably, the second industrial computer 310 in the centralized control module 300 of the present invention can connect adjacent edge points using a connected region algorithm according to the edge detection result to form a complete contour. Further, the second industrial computer 310 can filter out contours that meet the characteristics of the observation pile and the observation scale by characteristics such as area and perimeter.

[0091] Preferably, the second industrial computer 310 in the centralized control module 300 of the present invention can perform centroid calculation on the screened contours to calculate the centroid of all points on the contours by using the following formula:

[0092]

[0093] Among them, x i ,yi are the horizontal and vertical coordinates of each point on the contour, and N is the total number of points on the contour.

[0094] Preferably, the second industrial computer 310 in the centralized control module 300 of the present invention can realize the real-time calculation of the longitudinal displacement of the rail 400 according to the calculated centroid of all points on the contour. Preferably, before the real-time calculation of displacement is performed, the equipment can be used to collect the displacement observation piles and observation scales of the entire line as a basic database, wherein the basic database includes the initial position information and image data of each observation point. Further, in actual testing, the pictures of the observation piles and observation scales collected in advance can be called from the basic database, the centroid of the observation pile in the collected data and the centroid of the observation pile at the position in the database are overlapped, and the position relationship of the observation scale is compared to achieve comparison with the test data, and the displacement can be calculated. Since the test data is compared with the basic database data, the factors such as the field of view, the acquisition angle, and the acquisition distance have not changed, so the observation pile is used as the reference point, and the displacement is calculated in real time by comparing the relative relationship of the displacement observation scale, and the database is updated to ensure the accuracy and timeliness of the data, wherein the relative relationship is the difference between the horizontal and vertical coordinates. Preferably, the updated data continues to be stored in the internal hard disk of the second industrial computer 310, and can be transmitted to the operating computer through the network port for viewing and further processing.

[0095] Preferably, the second industrial computer 310 in the centralized control module 300 of the present invention can set different thresholds according to the standard value of the longitudinal displacement of the rail 400, and divide the longitudinal displacement of the rail 400 into five health levels: healthy, good, attention, deterioration, and danger. Further, the second industrial computer 310 can judge the health level of the longitudinal displacement of the rail 400 at each observation point according to the longitudinal displacement value of the rail 400 calculated in real time. Preferably, a real-time alarm is given for the longitudinal displacement of the rail 400 at the attention, deterioration, and danger levels, wherein the alarm method includes a sound alarm, flashing lights, and a pop-up window prompt on the operating computer. Preferably, the alarm information can be recorded on the operating computer, and a detailed detection report can be generated for subsequent analysis and processing.

[0096] Example 2

[0097] This embodiment is a further improvement on Embodiment 1, and the repeated contents will not be repeated here.

[0098] The present invention discloses a method for detecting the longitudinal displacement of a seamless line rail 400, and in particular, a method for detecting the longitudinal displacement of a seamless line rail 400 in a low-speed and real-time manner. The method can be implemented using the longitudinal displacement detection device described in Example 1.

[0099] Preferably, if Figure 5As shown, the longitudinal displacement detection method of the present invention may include the following steps:

[0100] Original image acquisition: acquiring image data of the rail 400 in real time in a moving manner along the extension direction of the rail 400;

[0101] Image illumination adjustment based on deep learning: Use deep learning algorithms to adjust the illumination of collected image data to ensure clear images;

[0102] Observation pile and observation scale contour recognition and centroid calculation: Calculate the centroid coordinates of the observation pile and observation scale through edge detection and contour extraction algorithms;

[0103] Real-time calculation of the longitudinal displacement of the rail 400: Compare the real-time acquired images with the images in the basic database, calculate the longitudinal displacement of the rail 400, and update the database.

[0104] Preferably, the longitudinal displacement detection method of the present invention may also include the installation and wiring of the detection equipment, that is, placing the hand-push bracket 100 equipped with the camera module 200 and the centralized control module 300 on the rail 400, and completing the wiring and initial settings of each component, so as to use the installed longitudinal displacement detection equipment for detection.

[0105] Preferably, after the longitudinal displacement of the rail 400 is calculated, a disease classification alarm step may be performed, wherein the step may classify the longitudinal displacement of the rail 400 into health levels according to a preset threshold value and issue a real-time alarm for abnormal conditions.

[0106] It should be noted that the above-mentioned specific embodiments are exemplary, and those skilled in the art can come up with various solutions inspired by the disclosure of the present invention, and these solutions also belong to the disclosure scope of the present invention and fall within the protection scope of the present invention. Those skilled in the art should understand that the present invention specification and its drawings are illustrative and do not constitute a limitation of the claims. The scope of protection of the present invention is defined by the claims and their equivalents. The present invention specification contains multiple inventive concepts, such as "preferably" or "according to a preferred embodiment", which means that the corresponding paragraph discloses an independent concept, and the applicant reserves the right to file a divisional application based on each inventive concept. Throughout the text, the features guided by "preferably" are only an optional method and should not be understood as a must-have setting. Therefore, the applicant reserves the right to abandon or delete the relevant preferred features at any time.

Claims

1. A seamless railway rail (400) longitudinal displacement detection device, characterized in that: It includes: A hand-push bracket (100) is used to be arranged on the rail (400) and provide support for other modules; A camera module (200) is used to capture image information of a displacement observation pile and an observation scale on a rail (400); The centralized control module (300) is used to collect, process and transmit data. The second industrial computer (310) configured for the centralized control module (300) is configured to: use a deep learning algorithm to adjust the illumination of image data collected in real time by the camera module (200) while following the movement of the hand-push bracket (100) to ensure image clarity; calculate the centroid coordinates of the observation pile and the observation scale by edge detection and contour extraction algorithms; compare the real-time collected image with the image in the basic database, calculate the longitudinal displacement of the rail (400), and update the database.

2. The longitudinal displacement detection device according to claim 1, characterized in that: The hand-push bracket (100) comprises a bracket body (110) and a plurality of wheels (130) connected to the bracket body (110) via supporting legs (120); the camera module (200) and the centralized control module (300) are directly or indirectly arranged on the bracket body (110) via corresponding mounting frames, so that when the wheels (130) are driven to rotate, the hand-push bracket (100) can drive the camera module (200) and the centralized control module (300) to move along the extension direction of the rail (400), and during the movement, the image data collected by the camera module (200) is sent to the centralized control module (300) for data processing, so as to calculate the longitudinal displacement of the rail (400).

3. The longitudinal displacement detection device according to claim 1 or 2, characterized in that: The hand-push bracket (100) comprises a height-adjustable handle (140), and the wheels (130) are driven to rotate by applying an external force to the handle (140). A speed encoder (131) for recording the travel distance and speed of the device is arranged on the rotating shaft of one of the wheels (130). The speed encoder (131) is connected to the pulse port of the centralized control module (300) through a pulse line to continuously send pulse signals to the centralized control module (300), so that the centralized control module (300) generates time stamps accordingly, and distributes these time stamps to the camera module (200).

4. The longitudinal displacement detection device according to any one of claims 1 to 3, characterized in that: The camera module (200) is mounted on the wide side (111) of the support body (110) so as to be suspended above the rail (400), wherein when the hand-push support (100) moves along the extension direction of the rail (400), the camera module (200) can photograph the displacement observation piles and observation scales on the rail (400) at predetermined time intervals based on real-time adjusted photographing parameters to generate a series of continuous image sequences.

5. A method for detecting longitudinal displacement of a seamless railway rail (400), characterized in that: It includes the following steps: Original image acquisition: acquiring image data of the steel rail (400) in real time in a moving manner along the extension direction of the steel rail (400); Image illumination adjustment based on deep learning: Use deep learning algorithms to adjust the illumination of collected image data to ensure clear images; Observation pile and observation scale contour recognition and centroid calculation: Calculate the centroid coordinates of the observation pile and observation scale through edge detection and contour extraction algorithms; Real-time calculation of the longitudinal displacement of the rail (400): Compare the real-time collected image with the image in the basic database, calculate the longitudinal displacement of the rail (400), and update the database.

6. The longitudinal displacement detection method according to claim 5, characterized in that: In the "Deep Learning-Based Image Illumination Adjustment" step, the image data is illuminated by an illumination module that includes a deep neural network. The deep neural network can learn through training to estimate the illumination component to remove the influence of illumination and restore the inherent properties of the image.

7. The longitudinal displacement detection method according to claim 5 or 6, characterized in that: The step of "observation pile and observation scale contour identification and centroid calculation" includes the following sub-steps: Use two 3×3 convolution kernels to convolve the preprocessed image to detect horizontal and vertical edges; Calculate the gradient size and direction at each pixel point, and determine the pixel points with gradient values ​​greater than the threshold as edge points by setting a threshold; Use the connected region algorithm to connect adjacent edge points to form a complete contour, and filter out contours that meet the characteristics of observation piles and observation scales through set features; Calculates the centroid coordinates of the selected contour.

8. The longitudinal displacement detection method according to any one of claims 5 to 7, characterized in that: Before the real-time calculation of the longitudinal displacement of the rail (400) is performed, the displacement observation piles and observation scales of the entire line are collected to serve as a basic database, which includes the initial position information and image data of each observation point, so that during actual testing, the pre-collected images of the observation piles and observation scales can be called out from the basic database to achieve comparison with the test data, wherein the relative relationship of the comparison is the difference between the horizontal and vertical coordinates.

9. The longitudinal displacement detection method according to any one of claims 5 to 8, characterized in that: It also includes the following steps: Installation and wiring of detection equipment: a push-type bracket (100) equipped with a camera module (200) and a centralized control module (300) is placed on a rail (400), and wiring and initial settings of various components are completed, so as to use the installed longitudinal displacement detection equipment for detection.

10. The longitudinal displacement detection method according to any one of claims 5 to 9, characterized in that: It also includes the following steps: Disease classification alarm: The longitudinal displacement of the rail (400) is classified into health levels according to a preset threshold value, and a real-time alarm is issued for abnormal conditions.

Images