A new railway line image detection method and system

Abstract

The present application discloses a novel railway line image detection method and system, the line image detection method comprising: a train image acquisition step: after detecting that a train leaves a station according to a station label, the train speed is converted into a pulse signal, and a train image is acquired according to the pulse signal; a train image processing step: the train image is processed to obtain a processed train image; a fault identification result acquisition step: fault identification is performed on the processed train image through a deep learning algorithm to obtain a fault identification result; a train contour data acquisition step: a distortion value is calculated according to gyroscope data and inclination data, and the distortion value is processed to obtain contour data; a line image detection result acquisition step: the line image detection result is output according to the processed train image, the fault identification result and the contour data.

Description

Technical Field

[0001] The present application relates to the field of track detection technology, and in particular to a novel railway line image detection method and system. Background Art

[0002] In recent years, the number of newly built railways has increased year by year. Due to geographical restrictions, the number of long bridges and extra-long tunnels has increased in railway line construction. In order to expand the transportation capacity, the existing railways are also speeding up in an all-round way and developing towards the goal of "high speed and heavy load". The requirements for line conditions after the railway speed increase are getting higher and higher. Due to the continuous increase in traffic density and the increase in the proportion of long bridges and tunnels in the line, the working environment conditions of engineering inspection, track flaw detection, electrical inspection, track geometric parameter detection, etc. have continued to deteriorate. On the one hand, the personal safety of the staff is not guaranteed, and on the other hand, manual work is far from meeting the requirements of the comprehensive development of the existing railways. At the same time, most of the track detection equipment currently in use is still based on special equipment detection instruments with a low degree of automation. Therefore, in the process of track detection, how to improve the degree of automation of track detection equipment has become an urgent problem to be solved. Summary of the invention

[0003] The embodiments of the present application provide a novel railway line image detection method and system, which at least solve the problems of low automation of detection equipment, large number of personnel for manual detection, low detection accuracy, low image processing efficiency, and long data transmission time during track line detection through the present invention.

[0004] The present invention provides a novel railway line image detection method, comprising:

[0005] Train image acquisition step: after detecting that the train has left the station according to the station tag, converting the train speed into a pulse signal, and acquiring the train image according to the pulse signal;

[0006] Train image processing step: processing the train image to obtain a processed train image;

[0007] Fault identification result acquisition step: performing fault identification on the processed train image through a deep learning algorithm to obtain a fault identification result;

[0008] Train profile data acquisition step: calculating a distortion value according to the gyroscope data and the inclination data, and processing the distortion value to obtain profile data;

[0009] Line image detection result acquisition step: output the line image detection result according to the processed train image, the fault identification result and the contour data.

[0010] In the above-mentioned line image detection method, the train image acquisition step comprises:

[0011] Identify the station tag to obtain an identification result, and determine whether the train has entered the track area to be inspected according to the identification result to obtain a determination result;

[0012] The train speed is acquired according to the judgment result, and a power-on instruction is sent when it is detected that the train speed is greater than 0 and the train leaves the station.

[0013] In the above-mentioned line image detection method, the train image acquisition step further comprises:

[0014] After the train speed is converted into the pulse signal, the train image is collected according to the pulse signal and the power-on instruction.

[0015] In the above-mentioned line image detection method, the train image processing step comprises:

[0016] Perform image fusion on the train image to obtain a fused train image.

[0017] In the above-mentioned line image detection method, the train image processing step further comprises:

[0018] The train image after image fusion is classified into data types to obtain a train image after data type classification.

[0019] In the above-mentioned line image detection method, the step of obtaining the fault identification result includes:

[0020] A train image directory is established, and the fault identification results are saved in the same image directory in the form of XML files with the same naming method as the train images after data type classification, and the train images after corresponding data type classification are saved in the same image directory.

[0021] In the above-mentioned line image detection method, the train profile data acquisition step comprises:

[0022] Acquiring the gyroscope data and the inclination data;

[0023] Calculating and obtaining horizontal angle data between the beam and the ground according to the gyroscope data and the inclination angle data;

[0024] After obtaining the distance between the optical center of the camera and the top surface of the track according to the train image and the horizontal angle data after data type classification, the height data of the left and right tracks are obtained according to the distance between the optical center of the camera and the top surface of the track. The relative horizontal superelevation value is obtained by analysis based on inertia and the height data.

[0025] In the above-mentioned line image detection method, the train profile data acquisition step further comprises:

[0026] According to the relative horizontal superelevation value and the base length, the horizontal difference between the two sections is taken, and the distortion value is calculated based on the horizontal difference.

[0027] The present invention also provides a novel railway line image detection system, which is applicable to the above-mentioned line image detection method, and the line image detection system comprises:

[0028] Train image acquisition unit: After the axle sensor detects that the train has left the station according to the station tag, it converts the train speed into a pulse signal, and the pulse signal triggers the image information acquisition module to collect train images;

[0029] Train image processing unit: the edge computing module processes the train image, and the edge computing module uploads the processed train image to the server;

[0030] Fault identification result acquisition unit: the edge computing module performs fault identification on the processed train image through a deep learning algorithm to obtain a fault identification result;

[0031] A train profile data acquisition unit: a distortion value is calculated based on the gyroscope data and the inclination data, and the edge computing module processes the distortion value to obtain profile data;

[0032] Line image detection result acquisition unit: The software platform displays the line image detection result on the software platform according to the processed train image, the fault identification result and the contour data.

[0033] In the above-mentioned line image detection system, the line image detection result acquisition unit includes:

[0034] The processed train image, the fault identification result and the profile data are uploaded to the software platform through the JSON upload program.

[0035] Compared with the related art, the present invention proposes a new railway line image detection method and system, which is based on the rail vehicle crossbeam. By installing rail profile detection modules, sleeper and ballast bed fastener detection modules, rail wear detection modules, speed sensors, position sensors and other equipment on the track detection vehicle, rails, sleepers, ballast beds and electrical equipment are collected, so that there is no need to design a separate running mechanism, saving equipment costs; for the image acquisition of key parts of rails, fasteners and ballast beds, 2D or 3D line scan cameras are used to combine images to collect component images and position information, and automatic image recognition and track position automatic positioning technologies are used to perform automatic abnormal analysis and graded warning on images and position information, and line alarm information and large-capacity image data are processed in real time and stored in the on-board data storage unit, and point cloud data can be output to collect depth information, which is easy to Eliminating the interference of road pollution such as mud and weeds, it has a higher recognition accuracy for abnormal conditions such as wear and fastener loss; manually confirming the line abnormality alarm information online or analyzing the collected data offline, realizing the maintenance of the track component status and effective monitoring of the inspection operation quality, and improving the early warning capability and network tracking capability of faults; through the wheel and axle sensors, the real-time collection of the travel speed of the rail vehicle is realized, and the real-time switching of the trigger frequency of the image acquisition module is realized; the track area is divided by electronic tags, and then the travel distance of the rail vehicle is calculated according to the encoder value fed back by the wheel sensor, and the precise coordinates of the faulty track are fed back to the maintenance personnel, so the maintenance personnel do not need to inspect the entire track, which improves work efficiency; through the on-board image processing module, the real-time processing of the collected image data during the operation of the rail vehicle is realized.

[0036] Details of one or more embodiments of the present application are set forth in the following drawings and description to make other features, objects, and advantages of the present application more readily apparent. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0038] Figure 1 is a flow chart of a novel railway line image detection method according to an embodiment of the present application;

[0039] Figure 2 is a flowchart of the steps of detecting a new railway line image according to an embodiment of the present application;

[0040] Figure 3 is a schematic diagram of the laser triangulation measurement principle according to an embodiment of the present application;

[0041] Figure 4It is a structural schematic diagram of the rail vehicle network display screen operation simulation system of the present invention;

[0042] Figure 5 is a composition diagram of an image information acquisition module according to an embodiment of the present application;

[0043] Figure 6 is a system topology diagram according to an embodiment of the present application;

[0044] Figure 7 It is an edge computing module according to an embodiment of the present application.

[0045] Wherein, the accompanying drawings are marked as follows:

[0046] Train image acquisition unit: 51;

[0047] Train image processing unit: 52;

[0048] Fault identification result acquisition unit: 53;

[0049] Train profile data acquisition unit: 54;

[0050] Line image detection result acquisition unit: 55;

[0051] Railcar beam: 1;

[0052] Rail detection module: 2;

[0053] Track rail profile detection module: 3;

[0054] Rails: 4;

[0055] Crossbeam supports:5. DETAILED DESCRIPTION

[0056] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application is described and illustrated below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application. Based on the embodiments provided in the present application, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present application.

[0057] Obviously, the drawings described below are only some examples or embodiments of the present application. For ordinary technicians in this field, the present application can also be applied to other similar scenarios based on these drawings without creative work. In addition, it can also be understood that although the efforts made in this development process may be complicated and lengthy, for ordinary technicians in this field related to the disclosure of the present application, some changes in design, manufacturing or production based on the technical content disclosed in the present application are just conventional technical means, and should not be understood as the disclosure of the present application is insufficient.

[0058] Reference to "embodiments" in this application means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those of ordinary skill in the art that the embodiments described in this application may be combined with other embodiments without conflict.

[0059] Unless otherwise defined, the technical terms or scientific terms involved in this application should be understood by people with ordinary skills in the technical field to which this application belongs. The words "one", "a", "a", "the" and the like involved in this application do not indicate a quantitative limitation, and may represent the singular or plural. The terms "include", "comprise", "have" and any of their variations involved in this application are intended to cover non-exclusive inclusions; for example, a process, method, system, product or device that includes a series of steps or modules (units) is not limited to the listed steps or units, but may also include steps or units that are not listed, or may also include other steps or units inherent to these processes, methods, products or devices. The words "connect", "connected", "coupled" and the like involved in this application are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The "multiple" involved in this application refers to two or more. "And / or" describes the association relationship of associated objects, indicating that there may be three relationships, for example, "A and / or B" can represent: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the objects before and after are in an "or" relationship. The terms "first", "second", "third", etc. involved in this application are only used to distinguish similar objects and do not represent a specific ordering of the objects.

[0060] The present invention provides a novel railway line image detection method and system. The present invention realizes the combination of comprehensive detection and key detection in a human-machine combination manner, utilizes the standardized, unified and fixed characteristics of the track group structure, adopts machine vision detection and image recognition and comparison technology, improves the track hidden fault detection capability and the early warning capability of fault generation, enhances the safety prevention level of the railway, and provides an effective means for preventing the track from running with faults.

[0061] The present invention is described below in conjunction with specific embodiments.

[0062] Embodiment 1

[0063] This embodiment also provides a new railway line image detection method. Figures 1 to 3 , Figure 1 is a flow chart of a novel railway line image detection method according to an embodiment of the present application; Figure 2 is a flowchart of the steps of detecting a new railway line image according to an embodiment of the present application;

[0064] Figure 3 Schematic diagram of the laser triangulation measurement principle according to an embodiment of the present application. Figures 1 to 3 As shown, the line image detection method includes:

[0065] Train image acquisition step S1: after detecting that a train has left a station according to a station tag, the train speed is converted into a pulse signal, and a train image is acquired according to the pulse signal;

[0066] Train image processing step S2: processing the train image to obtain a processed train image;

[0067] Fault identification result acquisition step S3: performing fault identification on the processed train image through a deep learning algorithm to obtain a fault identification result;

[0068] Train profile data acquisition step S4: Calculate and obtain distortion value according to gyroscope data and inclination data, process the distortion value, and obtain profile data;

[0069] Line image detection result acquisition step S5: output the line image detection result according to the processed train image, fault identification result and contour data.

[0070] In an embodiment, the train image acquisition step S1 includes:

[0071] Identify the station tag to obtain an identification result, and determine whether the train has entered the track area to be inspected based on the identification result to obtain a determination result;

[0072] The train speed is acquired according to the judgment result, and when it is detected that the train speed is greater than 0 and the train leaves the station, a power-on command is sent;

[0073] After the train speed is converted into a pulse signal, the train image is collected according to the pulse signal and the power-on command.

[0074] In the specific implementation, the station tag is identified to obtain the identification result, and whether the train enters the track area to be inspected is determined according to the identification result to obtain the judgment result; the train speed is acquired according to the judgment result, and when the train speed is detected to be greater than 0 and the train is detected to leave the station, a power-on instruction is sent; in detail, after the train speed is converted into a pulse signal, the train image radio frequency module is collected to obtain the identification result after identifying the station tag according to the pulse signal and the power-on instruction, and the track detection module determines whether the train enters the track area to be inspected according to the identification result, and obtains the judgment result; the axle sensor in the on-board control program collects the train speed according to the judgment result, and when the speed measurement module detects that the train speed is greater than 0 and detects that the train passes the station tag, a power-on instruction is sent to the host computer of the image information acquisition module, and the image information acquisition module starts working; after the axle sensor converts the train speed into a pulse signal, the pulse signal is sent to the FPGA board in the image information acquisition module; after the FPGA board triggers the camera module according to the pulse signal, the camera module collects the train image information and uploads the train image information to the edge computing module;

[0075] The present invention includes a track image information acquisition module and a track status acquisition system, as well as an on-board edge computing module and an on-board information analysis and management module connected via a communication system;

[0076] The image information acquisition module acquires images of the track in real time during the operation of the rail vehicle; the image information acquisition module is connected to the front crossbeam of the rail vehicle, and the image information acquisition module is installed in a position that does not interfere with the obstacle removal device; all camera modules are triggered by an FPGA board, and the mainboard clock realizes synchronous acquisition of all cameras;

[0077] The image information acquisition module includes contour acquisition modules divided into 4 groups; the contour acquisition modules are installed on both sides of the module bracket respectively. A single contour detection module adopts a line-surface combination method. A single contour detection module includes a high-speed array camera and an ultra-fine line light source. The laser triangulation method is also used to obtain the size information within the shooting range in real time; the module uses a 25mm lens, the distance between the camera and the laser is 128mm, the angle is 68.5°, and the measurement range is 240mm to 480mm;

[0078] The image information acquisition module includes two sets of track fastener detection modules, which are installed on both sides of the module bracket. A single sleeper fastener detection module mainly includes a high-speed area array camera, a line array camera and two different band line laser sources. The two-dimensional image of a single module adopts a 2K line array camera with a 16mm lens, and the two-dimensional image resolution reaches 0.5mm / pi x; the laser triangulation method is used to obtain the size information of the shooting range in real time;

[0079] The image information acquisition module includes two sets of rail wear detection modules, which are installed on both sides of the module bracket. A single rail wear detection module adopts a line-surface combination method, including a high-speed array camera and a line light source. It also uses laser triangulation to obtain the size information within the shooting range in real time. A single module uses a 16mm lens, the distance between the camera and the laser is 168mm, the angle is 70°, the measurement range is 400-540mm, and the plane image accuracy is 0.5mm. The rail wear detection module and the track bed fastener detection module are installed in the rail detection module.

[0080] Among them, the measurement principle of laser triangulation method is that the focal length of CCD lens is f=EG, the distance between CCD and light source is l=DE, the depth shooting range is d=AC, and the inclination angle between camera and light source is α=∠BED. In the process of designing the range of vehicle for real on-site shooting, when the vehicle parts move within the AC shooting range, according to the ratio relationship of similar triangles, the imaging position of the object on CCD will correspond one by one.

[0081] The axle of the rail vehicle is equipped with an axle sensor. The pulse signal emitted by the axle sensor is processed by the on-board edge computing module and sent to the FPGA board in the image information acquisition module, and a differential pulse signal is sent to trigger the camera and light source; the axle sensor is also used to collect the actual speed of the wheels on the bogie installed on the equipment. At the same time, the on-board edge computing module obtains the speed of all wheels through the on-board MVB signal, and uses the maximum speed method to obtain the reference speed of the current rail vehicle; the control system obtains the distance compensation image information acquisition module of the vehicle idling wheel slip based on the speed curve and the comparison of the speed curve and the acceleration sensor The frequency and track positioning of the group; the axle sensor can obtain the specific position of the current rail vehicle in the section according to the number of pulses counted by the axle sensor, so as to locate the identified fault position. At the same time, the on-board control system can calculate the speed curve of the vehicle operation in combination with the vehicle speed data in the vehicle MVB data, identify the idling wheel slip state of the vehicle, and compensate for image acquisition and fault positioning; an electronic tag is attached to each kilometer of the track that needs to be detected, and a radio frequency module is installed on the rail vehicle. When the rail vehicle passes the electronic tag, the track detection device will identify the current section of the rail vehicle;

[0082] All turnouts are marked with electronic tags. During the system deployment phase, when a train passes through a turnout area, the system will automatically create a template 3D point cloud data file for each turnout. During the system operation phase, each time a train passes through a turnout area marked with an electronic tag, the system uses pattern recognition to compare the current point cloud data with the template data. If there is no abnormality, the system will save the data of this train passing as a new template.

[0083] Since the on-board edge computing module is used to collect and identify image data, the control and power supply of the image information acquisition module are fully integrated, so the system can detect the track status in real time during the operation of the rail vehicle, without the need to upload the data for processing and identification; the trigger frequency of the image information acquisition module changes in real time according to the vehicle speed collected by the axle sensor, and when the speed is equal to 0, it is considered to be stopped.

[0084] In an embodiment, the train image processing step 52 includes:

[0085] Performing image fusion on the train image to obtain a fused train image;

[0086] The data type of the train image after the image fusion is classified to obtain the train image after the data type classification.

[0087] In the specific implementation, the train image is fused to obtain the fused train image; the data type of the fused train image is classified to obtain the data type classified train image; in detail, the edge computing module filters and normalizes the train image information to obtain the processed train image information, and then performs image fusion on the processed train image information, i.e., 2D or 3D data, to obtain the fused train image information; the data type of the fused train image is classified to obtain the image after data processing, and the edge computing module uploads the image to the server.

[0088] In an embodiment, the fault identification result acquisition step 54 includes:

[0089] Performing fault identification on the processed train image through a deep learning algorithm to obtain a fault identification result;

[0090] A train image directory is established, and the fault identification results are saved in the same image directory in the form of XML files with the same naming method as the train images after data type classification, and the train images after corresponding data type classification are saved in the same image directory.

[0091] In a specific implementation, a deep learning algorithm is used to perform fault identification on the processed train image to obtain a fault identification result; a train image directory is established, and the fault identification result is in the form of an XML file with the same naming method as the train image after data type classification, and the train image after data type classification corresponding to it is saved in the same image directory; in detail, the edge computing module uses a deep learning algorithm to perform fault identification on the image to obtain a fault identification result, that is, the deep learning algorithm is used to perform intelligent analysis on the precise positioning and abnormal status of the track, fasteners, etc.; after the data identifies the fault, the fault data is uploaded to the ground server in units of kilometers; after the edge computing module completes the processing of the collected image, the directory adopts the naming rule of date / serial number / section number / camera number / image, the server establishes an image directory, the edge computing module uploads the fault identification result to the server, the fault identification result is in the form of an XML file with the same naming method as the picture, and the image corresponding to the fault identification result is saved in the same image directory, and the image is stored in the server by the data acquisition computer; after the on-board control system of the track detection device records the fault content and fault location, a fault record will be generated on the software platform, and the operator can select the fault content to generate a work order to re-inspect the track.

[0092] In an embodiment, the train profile data obtaining step 54 includes:

[0093] Collect and obtain gyroscope data and inclination data;

[0094] According to the gyroscope data and the inclination data, the horizontal angle data between the beam and the ground is calculated;

[0095] According to the train images and horizontal angle data classified by data types, the distance between the camera optical center and the top surface of the track is obtained, and then the height data of the left and right tracks are obtained according to the distance between the camera optical center and the top surface of the track. According to the inertia and height data, the relative horizontal superelevation value is obtained through analysis;

[0096] According to the relative horizontal superelevation value and base length, the horizontal difference between the two sections is taken, and the distortion value is obtained based on the horizontal difference.

[0097] In the specific implementation, after the gyroscope data is acquired through the gyroscope sensor and the inclination data is acquired through the inclination sensor, the edge computing module calculates the horizontal angle data between the beam and the ground according to the gyroscope data and the inclination data; according to the computer code, the png image is first converted into an xyz image according to the camera internal parameters, and then the z-axis coordinates in the tif image obtained from the new xyz image are used to calculate the spatial distance from the point to the surface in real time, that is, according to the horizontal angle data between the image and the beam and the ground, the distance between the camera optical center and the top surface of the track is obtained, and then the height data of the left and right tracks are obtained according to the distance between the camera optical center and the top surface of the track, and the relative horizontal superelevation value is obtained by analysis according to inertia and the height data of the left and right tracks; according to the relative horizontal superelevation value and the base length, the horizontal difference between the two sections is taken by the gyroscope sensor and the inclination sensor, and the distortion value is calculated according to the horizontal difference; the edge computing module processes the distortion value, obtains the contour data in the form of DATA, and then uploads the contour data to the server; the base length is a fixed length of 3m;

[0098] Among them, for the profile data, the ground server can receive track data and draw the track model on the system software interface. The amount of track data for a single frame is 2000 points, and the amount of data for a single kilometer is 2000*1000 points. In order to make the front-end display of the system software platform clear, the back-end will extract the data and divide the data within 1 kilometer into a data packet of 1 meter. If a frame of fault data appears in the packet, the frame of fault data will be uploaded. If multiple frames of fault data appear in the packet, the maximum wear data will be uploaded. If there is no fault data in the packet, only one frame of normal data will be uploaded.

[0099] For the two-dimensional image data of fasteners and sleepers, the control system software will automatically generate an XML file containing the coordinates and fault level of the fault pixel data in the two-dimensional image. Based on the coordinate data, a fault frame can be generated on the front-end interface, making it easy for operators to quickly locate the fault point.

[0100] Among them, track data is divided into real-time data and historical data;

[0101] Real-time data refers to the geometric parameter fault data, rail profile fault data, rail top fault data, and sleeper fastener fault data that appear in the system during real-time detection. This data is uploaded to the server in real time via the wireless network; the loading priority of real-time data on the software platform is higher than that of historical data.

[0102] The historical data refers to all the track data detected by the system. The data volume is large and will be uploaded to the server after the system recognizes that the rail vehicle has returned to the depot.

[0103] In an embodiment, the line image detection result acquisition step S5 includes:

[0104] According to the processed train image, fault identification result and contour data, the line image detection result is output.

[0105] In the specific implementation, the line image detection results are output according to the processed train image, fault identification results and contour data; in detail, the server uploads the processed train image, fault identification results and contour data to the software platform through the JSON upload program, and then displays the image, fault identification results and contour data on the software platform display screen. After the staff clicks the fault confirmation button on the touch screen, a work order is generated; when the on-board control program detects the return to the depot ground label, all the data of the current detection will be uploaded to the ground server in units of kilometers.

[0106] The system software platform adopts B / S architecture. The system software platform interface has the function of displaying the inspection record list. It can retrieve the corresponding track inspection record according to the inspection date and line number of the rail vehicle, which includes the total number of faults, rail appearance faults, geometric parameter faults, and the number of rail appearance faults, and can jump to the corresponding inspection data or images.

[0107] Rail status check: You can drag the scroll bar to check the 2D images of all rail surfaces and turnout sleepers in the searched area;

[0108] The geometric parameter status displays the detection values ​​of track gauge, triangular pit, vertical acceleration, lateral acceleration, and track height in waveform form. The horizontal coordinate of the waveform is the distance under the current kilometer mark in meters, and the vertical coordinate is the fault level threshold of the parameter. The rail profile status interface displays the profile of the rail section at the current position as well as the vertical wear, side wear, and total wear values ​​in the form of a single-frame image.

[0109] Among them, the vehicle control program and the software platform web server use "detection time ID" as the unique identifier of the detection record;

[0110] The system uses an edge computing module to replace the industrial computer to realize edge-side recognition calculation of images, system control, power supply of image information acquisition modules and sensor data processing.

[0111] Embodiment 2

[0112] This embodiment also provides a novel railway line image detection system. Figure 4 It is a structural schematic diagram of the rail vehicle network display screen operation simulation system of the present invention; Figure 5 is a composition diagram of an image information acquisition module according to an embodiment of the present application; Figure 6 is a system topology diagram according to an embodiment of the present application; Figure 7 is an edge computing module according to an embodiment of the present application. Figures 4 to 7As shown, the line image detection system of the invention is applicable to the above-mentioned line image detection method, and the line image detection system includes:

[0113] Train image acquisition unit: After the axle sensor detects that the train has left the station according to the station tag, it converts the train speed into a pulse signal, which triggers the image information acquisition module to collect train images;

[0114] Train image processing unit: The edge computing module processes the train image and uploads the processed train image to the server;

[0115] Fault identification result acquisition unit: The edge computing module uses a deep learning algorithm to perform fault identification on the processed train image and obtain the fault identification result;

[0116] Train profile data acquisition unit: The distortion value is calculated based on the gyroscope data and the inclination data. The edge computing module processes the distortion value to obtain the profile data.

[0117] Line image detection result acquisition unit: The software platform displays the line image detection results on the software platform based on the processed train images, fault identification results and contour data.

[0118] In the specific implementation, the main control module of the high-voltage end of the edge computer integrates a 2 million gate FPGA real-time controller, a 400MHZ CPU, and a 128MB DRAM. It is also equipped with 2 sets of acceleration acquisition modules, a single-channel 102.4k sampling frequency, an analog input range of ±30V, with I EPE excitation, 60V high-voltage isolation, and synchronization between channels. It can be connected to an acceleration sensor that recognizes the posture of the vehicle body. The low-voltage end of the main control module of the high-voltage end of the edge computer is configured with 8 DI / O channels, which can output and measure the frequency of 0~1MHZ, and realize the synchronous triggering of the camera and light source. At the same time, the main control module has a built-in MXM NVIDIA 16GB independent display to realize image processing on the edge side.

[0119] Track inspection equipment can achieve the following inspection points:

[0120] Sleepers: gussets, spring bars, pads, bolts are loose or lost, spikes are pulled out, sleepers are cracked or fail;

[0121] Track bed: slag pollution, track bed compaction, disease, ballast missing;

[0122] Track circuit: cable damage;

[0123] Rails: track profile exceeding limit, rail surface peeling, track cracks, fishplate loss, fastener loss, bolts loose or missing, and rail seams, joints, and welds broken.

[0124] In an embodiment, the train image acquisition unit 51 includes:

[0125] After the axle sensor detects that the train has left the station based on the station tag, it converts the train speed into a pulse signal, which triggers the image information acquisition module to collect the train image.

[0126] In the specific implementation, the image information acquisition module is composed of 2 sets of sleeper and ballast fastener acquisition modules, 4 sets of track rail profile detection modules, and 2 sets of rail wear detection modules; a single sleeper and ballast fastener detection module mainly includes 1 high-speed area array camera, 2 line light sources, and 1 line array camera, and uses laser triangulation to obtain the size information of the shooting range in real time; a single track rail profile detection module mainly includes 1 high-speed area array camera and 1 ultra-fine line light source, and also uses laser triangulation to obtain the size information within the shooting range in real time; the rail wear detection module mainly includes 1 high-speed area array camera and 1 line light source, and also uses laser triangulation to obtain the size information within the shooting range in real time;

[0127] After the front-end image acquisition device uploads the image data to the on-board edge computing module, the on-board edge computing module processes and identifies the data. The processed and identified images will be uploaded to the ground server, and the fault alarm information will be saved in the form of an XML file with the same naming method as the image in a unified file path with the corresponding image, so that it can be queried and retrieved by the upper platform.

[0128] The on-board edge module uses an edge computing module with image processing capabilities to process and identify images uploaded by the front-end acquisition device, and uploads the identified data to the data storage server in real time; the image information acquisition module integrates the use of FPGA boards to process communication signals of devices such as speed sensors, and control the triggering of the image acquisition module.

[0129] The electronic tags are installed on the sleepers by adhesive. The interval between the electronic tags is set at 1km, that is, the track to be detected is divided into sections of 1km. When the rail vehicle travels to a certain section, the electronic tag reader will send instructions to the host computer; the precise distance of the rail vehicle's current position is measured by the axle sensor, that is, the photoelectric rotary encoder.

[0130] The wheel axle sensor mainly adopts a photoelectric rotary encoder, which is installed on the axle and rotates synchronously with the axle, outputting 1000 pulses every week. The counter is used to count the output pulses. The counter value can be converted into the vehicle's travel distance according to the coefficient and the wheel diameter. According to the vehicle's travel distance and the electronic tag attached to the track, the exact position of the current vehicle and the fault point can be accurately calculated, which is convenient for manual replacement of accessories and tracks; the wheel axle sensor is also used to trigger the image acquisition module; the wheel axle sensor converts the speed of the train into a differential pulse signal to trigger the camera; the combination of the camera trigger frequency and the vehicle speed can capture continuous images of the track. This method can achieve the acquisition of continuous images without affecting the restart of the rail vehicle after parking.

[0131] In an embodiment, the line image detection result acquisition step 55 includes:

[0132] The processed train images, fault identification results and contour data are uploaded to the software platform through the JSON upload program.

[0133] In the specific implementation, the system is equipped with a UPS uninterruptible power supply device to ensure the reliability of power supply and eliminate the interference with image acquisition when the rail car has abnormal power supply or power outage.

[0134] The rail vehicle is equipped with a wireless Internet of Things module. After generating fault information in the server, it sends a fault information alarm signal to the ground server, and uploads fault pictures and data in JSON and DATA formats to the monitoring center server for display on the system platform.

[0135] The software platform adopts B / S architecture, and the maintenance department can view it through the WEB terminal. The system image browsing terminal can query fault pictures and track information by location; the interface has a fault marking tool to manage fault pictures and accurately locate the fault location, and automatically generate a detection report with fault type, location, abnormal part, etc.

[0136] The data storage server is equipped with a large-capacity hard disk to ensure that image information and data files can be stored for more than three months for retrospective query of fault points.

[0137] In summary, during the detection process, the present invention only needs to be equipped with one system operator to recheck the detection results. After confirming the abnormal status, the track is repaired according to the positioning position. There is no need to inspect the entire track, which improves the detection efficiency and reduces the number of personnel for manual track inspection; it uses the rail car crossbeam for installation, and there is no need to specially design special track inspection equipment for installation, which saves equipment costs; 3D and 2D image technology are used to detect key technical indicators of the track. Compared with the detection scheme based on grayscale contrast and color 2D contrast, the present invention uses deep analysis to improve the detection accuracy and the scalability of the algorithm; compared with the method of uploading data to the monitoring center for image processing, real-time image processing and recognition are performed on the on-board server, which improves the image processing efficiency. The system can greatly save the data transmission time and has good compatibility with the local area network with limited bandwidth. Compared with the special track detection equipment, the present invention can be combined with other working vehicles to simplify the on-track detection operation process of the operating personnel to generate the detection results on the software platform, report them to the quality and track maintenance department, and then perform maintenance operations on the abnormal sections at the fixed point. Compared with the traditional track detector which needs to be manually pushed at a low speed, the travel speed of the equipment of the present invention is not limited. The control board developed based on ARM is small in size, light in weight and highly integrated, and has high compatibility with vehicles of freight railways, especially heavy-duty railways.

[0138] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the protection scope of the attached claims.

Claims

1. A novel railway line image detection method, characterized in that: The line image detection method comprises: Train image acquisition steps: the image information acquisition module is connected to the front crossbeam of the rail vehicle, and after the train is detected to leave the station according to the station label, the image information acquisition module is used to collect the image of the track in real time during the operation of the rail vehicle; the image information acquisition module includes contour acquisition modules divided into 4 groups; the image information acquisition module includes two sets of track bed fastener detection modules, and the laser triangulation method is used to obtain the size information of the shooting range in real time; the image information acquisition module includes two sets of rail wear detection modules; the train speed is converted into a pulse signal, and the image information acquisition module is triggered according to the pulse signal to collect the train image; Train image processing step: processing the train image to obtain a processed train image; Fault identification result acquisition step: performing fault identification on the processed train image through a deep learning algorithm to obtain a fault identification result; The train profile data acquisition step includes: collecting and acquiring gyroscope data and inclination data; calculating and acquiring horizontal angle data between the beam and the ground according to the gyroscope data and the inclination data; acquiring the distance between the optical center of the camera and the top surface of the track according to the train image and the horizontal angle data after data type classification, acquiring the height data of the left and right tracks according to the distance between the optical center of the camera and the top surface of the track, and acquiring the relative horizontal superelevation value through analysis according to inertia and the height data; acquiring the horizontal difference between the two sections according to the relative horizontal superelevation value and the base length, and acquiring the distortion value according to the horizontal difference, and processing the distortion value to acquire the profile data; Line image detection result acquisition step: output the line image detection result according to the processed train image, the fault identification result and the contour data.

2. The line image detection method according to claim 1, characterized in that: The train image acquisition step comprises: Identify the station tag to obtain an identification result, and determine whether the train has entered the track area to be inspected according to the identification result to obtain a determination result; The train speed is acquired according to the judgment result, and a power-on instruction is sent when it is detected that the train speed is greater than 0 and the train leaves the station.

3. The line image detection method according to claim 2, characterized in that: The train image acquisition step further includes: After the train speed is converted into the pulse signal, the train image is collected according to the pulse signal and the power-on instruction.

4. The line image detection method according to claim 3, characterized in that: The train image processing step comprises: Perform image fusion on the train image to obtain a fused train image.

5. The line image detection method according to claim 4, characterized in that: The train image processing step also includes: The train image after image fusion is classified into data types to obtain a train image after data type classification.

6. The line image detection method according to claim 5, characterized in that: The step of obtaining the fault identification result comprises: A train image directory is established, and the fault identification results are saved in the same image directory in the form of an XML file with the same naming method as the train images after data type classification, and the train images after data type classification corresponding to the fault identification results are saved in the same image directory.

7. The line image detection method according to claim 6, characterized in that: The train profile data acquisition step comprises: Acquiring the gyroscope data and the inclination data; Calculating and obtaining horizontal angle data between the beam and the ground according to the gyroscope data and the inclination angle data; After obtaining the distance between the optical center of the camera and the top surface of the track according to the train image and the horizontal angle data after data type classification, the height data of the left and right tracks are obtained according to the distance between the optical center of the camera and the top surface of the track. The relative horizontal superelevation value is obtained by analysis based on inertia and the height data.

8. The line image detection method according to claim 7, characterized in that: The train profile data acquisition step also includes: According to the relative horizontal superelevation value and the base length, the horizontal difference between the two sections is taken, and the distortion value is calculated based on the horizontal difference.

9. A line image detection system, characterized in that: The line image detection system comprises: Train image acquisition unit: The image information acquisition module is connected to the front crossbeam of the rail vehicle. After the axle sensor detects that the train has left the station according to the station label, the image information acquisition module is used to collect images of the track in real time during the operation of the rail vehicle; the image information acquisition module includes contour acquisition modules divided into 4 groups; the image information acquisition module includes two sets of track bed fastener detection modules, which use laser triangulation to obtain the size information of the shooting range in real time; the image information acquisition module includes two sets of rail wear detection modules; the train speed is converted into a pulse signal, and the pulse signal triggers the image information acquisition module to collect train images; Train image processing unit: the edge computing module processes the train image, and the edge computing module uploads the processed train image to the server; Fault identification result acquisition unit: the edge computing module performs fault identification on the processed train image through a deep learning algorithm to obtain a fault identification result; Train profile data acquisition unit: collects and obtains gyroscope data and inclination data; calculates and obtains horizontal angle data between the beam and the ground according to the gyroscope data and the inclination data; obtains the distance between the camera optical center and the track top surface according to the train image and the horizontal angle data after data type classification, obtains the height data of the left and right tracks according to the distance between the camera optical center and the track top surface, and obtains the relative horizontal superelevation value through analysis according to inertia and the height data; obtains the horizontal difference between the two sections according to the relative horizontal superelevation value and the base length, and obtains the distortion value according to the horizontal difference, and the edge computing module processes the distortion value to obtain profile data; Line image detection result acquisition unit: The software platform displays the line image detection result on the software platform according to the processed train image, the fault identification result and the contour data.

10. The line image detection system according to claim 9, characterized in that: The line image detection result acquisition unit comprises: The processed train image, the fault identification result and the profile data are uploaded to the software platform through the JSON upload program.

Images