Automatic track state inspection system and working method

Abstract

The invention discloses an automatic track state inspection system and a working method. In the system, a tag reading unit reads radio frequency electronic tags on urban tracks through which a train passes in real time; when the inspection control unit analyzes and determines that the radio frequency electronic tag read in real time is an urban rail starting position tag, an inspection instruction for starting inspection is sent to the image acquisition control unit; when receiving the instruction, the image acquisition unit acquires image information of the urban rail through which the train passes in real time; the image acquisition control unit combines the image signal with the pulse signal received from the coding unit according to a timestamp to obtain real-time inspection information of the urban rail; the artificial intelligence processing unit performs data analysis to determine whether the real-time inspection information represents that the urban rail has an inspection problem; and if yes, sending alarm information carrying the inspection problem. The accuracy and efficiency of track state inspection are improved, and the labor cost is reduced.

Description

Technical Field

[0001] The present invention relates to the field of track detection technology, and in particular to an automated track status inspection system and a working method. Background Art

[0003] With the rapid development of urban rail transit, the importance of urban rail line infrastructure inspection has become increasingly prominent. At present, the appearance inspection of the track and its connecting parts mainly adopts two methods: manual inspection and vehicle inspection. Among them, vehicle inspection mostly uses special vehicles equipped with inspection equipment, such as special inspection vehicles, comprehensive inspection trains and other infrastructure inspection equipment to carry out line inspection. Its system structure is mainly based on "optical system-image acquisition industrial computer-processing system". The line image is collected through industrial cameras and auxiliary lighting sources, and the data is encapsulated and stored using industrial computers and image acquisition cards. The recognition of abnormal image status is completed with the help of the intelligent recognition algorithm in the processing system.

[0004] However, in the line inspection of urban rail transit, although vehicle inspection can improve the inspection efficiency of urban rail infrastructure, there are many problems. On the one hand, the spatial size of the detection equipment, the power consumption of the equipment, and the operation mode of the detection system are highly dependent on manual operation, which leads to the need to use special vehicles to carry detection equipment and arrange special personnel to follow the vehicle for inspection. This requires the urban rail operating unit to arrange for the inspection vehicle to go online separately, which greatly increases the operating cost. On the other hand, the use of independent industrial computers for data collection and storage and intelligent identification is complicated to operate, and the efficiency is difficult to meet the high-speed inspection requirements of urban rail lines. Therefore, there is an urgent need for a new track status inspection method to solve these deficiencies in the existing technology.

[0005] The existing inspection method relies on special inspection vehicles and dedicated personnel to follow the vehicle for inspection. The urban rail operating unit needs to arrange for inspection vehicles to go online separately, which increases the operating costs such as vehicle purchase, maintenance and personnel costs. The data collection, storage and intelligent recognition operations based on independent industrial computers are cumbersome, which is difficult to meet the efficiency requirements of high-speed inspection of urban rail lines and cannot efficiently complete the track status inspection task. Summary of the invention

[0006] The embodiment of the present invention provides an automated track state inspection system for automatically inspecting the track state, simplifying the inspection operation process, reducing labor costs, and improving the accuracy and efficiency of the track state inspection. The system includes: a label reading unit, an inspection control unit, an image acquisition control unit, an image acquisition unit, an encoding unit, and an artificial intelligence processing unit;

[0007] The tag reading unit is used to read the radio frequency electronic tags on the urban track that the train passes through in real time after the train carried by the automated track status inspection system runs;

[0008] The inspection control unit is used to analyze the radio frequency electronic tag obtained from the tag reading unit in real time; when the analysis determines that the radio frequency electronic tag read in real time is the starting position tag of the urban track, it sends an inspection instruction to start the inspection to the image acquisition control unit;

[0009] An image acquisition control unit, configured to issue an image acquisition instruction to the image acquisition unit after receiving the inspection instruction issued by the inspection control unit;

[0010] An image acquisition unit, configured to acquire image information of the urban track through which the train passes in real time upon receiving an image acquisition instruction;

[0011] The encoding unit is used to generate a pulse signal along with the train operation when the image acquisition unit is collecting data; the image acquisition control unit is also used to combine the image information received from the image acquisition unit and the pulse signal received from the encoding unit according to the timestamp to obtain the real-time inspection information of the urban rail;

[0012] The artificial intelligence processing unit is used to perform data analysis on the real-time inspection information of the urban rail to determine whether the real-time inspection information indicates that there is an inspection problem in the urban rail; if so, an alarm message carrying the inspection problem is issued.

[0013] The embodiment of the present invention further provides a working method of an automated track state inspection system, which is used to automatically inspect the track state, simplify the inspection operation process, reduce labor costs, and improve the accuracy and efficiency of the track state inspection. The method includes:

[0014] After the train carried by the automatic track status inspection system is running, the tag reading unit reads the radio frequency electronic tags on the urban track through which the train passes in real time;

[0015] The inspection control unit analyzes the radio frequency electronic tag obtained from the tag reading unit in real time; when the analysis determines that the radio frequency electronic tag read in real time is the starting position tag of the urban rail, it sends an inspection instruction to start the inspection to the image acquisition control unit;

[0016] After receiving the inspection instruction issued by the inspection control unit, the image acquisition control unit issues an image acquisition instruction to the image acquisition unit;

[0017] When receiving the image acquisition instruction, the image acquisition unit acquires image information of the urban track through which the train passes in real time;

[0018] When the image acquisition unit is collecting data, the encoding unit generates a pulse signal as the train runs; the image acquisition control unit combines the image information received from the image acquisition unit and the pulse signal received from the encoding unit according to the timestamp to obtain real-time inspection information of the urban rail;

[0019] The artificial intelligence processing unit performs data analysis on the real-time inspection information of the urban rail to determine whether the real-time inspection information indicates that there is an inspection problem in the urban rail; if so, an alarm message carrying the inspection problem is issued.

[0020] The embodiment of the present invention provides an automated track status inspection system, wherein a tag reading unit is used to read the radio frequency electronic tags on the urban track through which the train passes in real time after the train carried by the automated track status inspection system is running; an inspection control unit is used to analyze the radio frequency electronic tags obtained from the tag reading unit in real time; when the analysis determines that the radio frequency electronic tags read in real time are the starting position tags of the urban track, an inspection instruction for starting the inspection is sent to the image acquisition control unit; the image acquisition control unit is used to send an image acquisition instruction to the image acquisition unit after receiving the inspection instruction sent by the inspection control unit; FIG. The image acquisition unit is used to acquire image information of the urban track through which the train passes in real time upon receiving an image acquisition instruction; the encoding unit is used to generate a pulse signal as the train runs when the image acquisition unit is acquiring; the image acquisition control unit is also used to: combine the image information received from the image acquisition unit and the pulse signal received from the encoding unit according to the timestamp to obtain real-time inspection information of the urban track; the artificial intelligence processing unit is used to perform data analysis on the real-time inspection information of the urban track to determine whether the real-time inspection information indicates that there is an inspection problem in the urban track; if so, an alarm message carrying the inspection problem is issued. The inspection system in the embodiment of the present invention is mounted on the urban rail transit operating vehicle, and there is no need to arrange a dedicated inspection vehicle online separately; through the collaborative work of components such as the label reading unit and the inspection control unit, the track status inspection is carried out while the train is running daily, which reduces the vehicle purchase, maintenance and personnel following costs of the operating unit; the inspection system realizes automated operation through the close cooperation of various units, and the process does not require excessive human intervention, reduces cumbersome manual operation steps, and greatly simplifies the inspection operation process; the image acquisition unit can collect high-quality track image information in real time after receiving the instruction, and the pulse signal generated by the encoding unit is accurately synchronized with the image after processing, ensuring the accuracy and real-time nature of data acquisition, avoiding repeated detection or misjudgment caused by inaccurate data acquisition, and improving data acquisition efficiency; the artificial intelligence processing unit can combine the machine learning model to quickly analyze the real-time inspection information, and can efficiently determine whether there is a problem with the track. Compared with the traditional intelligent recognition algorithm, its processing speed is faster and the accuracy is higher. It can complete data analysis in real time during the high-speed operation of the train, meet the high-speed detection requirements of the urban rail urban line, and effectively improve the inspection efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

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

[0022] Figure 1 This is a schematic diagram of the structure of an automated track status inspection system in an embodiment of the present invention;

[0023] Figure 2 A specific example diagram of the equipment structure of an automated track status inspection system in an embodiment of the present invention;

[0024] Figure 3 This is a specific example diagram of the inspection principle of an automated track status inspection system in an embodiment of the present invention;

[0025] Figure 4 This is an example diagram of the workflow of an image acquisition control unit in an automated track status inspection system in an embodiment of the present invention;

[0026] Figure 5 This is an example diagram of the workflow of an artificial intelligence processing unit in an automated track status inspection system in an embodiment of the present invention;

[0027] Figure 6 A schematic diagram of a process for constructing an urban rail inspection training sample set in an embodiment of the present invention;

[0028] Figure 7 The figure is a flowchart of an automatic track status inspection system working method in an embodiment of the present invention. DETAILED DESCRIPTION

[0029] To make the purpose, technical solution and advantages of the embodiments of the present invention more clear, the embodiments of the present invention are further described in detail below in conjunction with the accompanying drawings. Here, the exemplary embodiments of the present invention and their descriptions are used to explain the present invention, but are not intended to limit the present invention.

[0030] The term "and / or" herein only describes an association relationship, indicating that three relationships may exist. For example, A and / or B may represent the following three situations: A exists alone, A and B exist at the same time, and B exists alone. In addition, the term "at least one" herein represents any combination of at least two of any one or more of a plurality of. For example, including at least one of A, B, and C may represent including any one or more elements selected from the set consisting of A, B, and C.

[0031] In the description of this specification, the terms "include", "including", "have", "contain", etc. are all open terms, which mean including but not limited to. The descriptions with reference to the terms "one embodiment", "a specific embodiment", "some embodiments", "for example", etc. mean that the specific features, structures or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures or characteristics described may be combined in a suitable manner in any one or more embodiments or examples. The order of steps involved in each embodiment is used to schematically illustrate the implementation of the present application, and the order of steps is not limited and can be appropriately adjusted as needed.

[0032] The acquisition, storage, use, and processing of data in the technical solution of this application are in compliance with the relevant provisions of national laws and regulations. The information collected in this application is information and data authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure, and application of relevant data are in compliance with the relevant laws, regulations, and standards of relevant countries and regions, and necessary confidentiality measures are taken, which do not violate public order and good customs, and provide corresponding operation entrances for users to choose to authorize or refuse. In addition, this application provides users with corresponding operation entrances for users to choose to agree or refuse the results of automated decision-making. If the user chooses to refuse, they can enter the expert decision-making process.

[0033] It should be noted that in the embodiments of the present application, some existing solutions in the industry such as certain software, components, models, etc. may be mentioned. For example, some existing software tools, components, algorithm models or other well-known solutions in other technical fields may be cited. They should be considered as exemplary, and their purpose is only to illustrate the feasibility of the implementation of the technical solution of the present application. These references should be understood as typical examples, and their core purpose is to explain and verify the rationality and feasibility of the implementation of the technical solution proposed in the present application. However, it does not mean that the applicant has or will inevitably use the solution. Such references do not imply that the applicant has actually adopted these existing solutions, or will inevitably adopt these methods in the future during its technical implementation. In other words, these references only serve to illustrate the nature of the description, to help understand the connection and transcendence between the innovation of the present application and the prior art, and do not constitute an endorsement or reliance statement on a specific prior art product.

[0034] With the rapid development of urban rail transit, the importance of urban rail line infrastructure inspection has become increasingly prominent. The appearance inspection of the track and its connecting parts generally adopts two methods: manual inspection and vehicle inspection. The vehicle inspection currently used generally adopts special vehicle-mounted inspection equipment, and conducts line inspection in the form of infrastructure inspection equipment such as special inspection vehicles and comprehensive inspection trains. The current inspection system is mainly based on the system structure of "optical system-image acquisition industrial computer-processing system", which realizes the acquisition of line images through industrial cameras and auxiliary lighting sources, realizes data packaging and storage through industrial computers and image acquisition cards, and completes the recognition of abnormal image conditions through the intelligent recognition algorithm in the processing system.

[0035] For the line inspection of urban rail transit, the vehicle inspection mode can greatly improve the efficiency of urban rail infrastructure inspection. However, due to the space size of the detection equipment, the power consumption of the equipment, and the fact that the operation mode of the detection system still relies heavily on manual operation, it is necessary to use a special vehicle equipped with detection equipment and a dedicated person to follow the vehicle for inspection. This method requires the urban rail operating unit to arrange for the inspection vehicle to go online separately, which has high operating costs. In addition, the use of independent industrial computers for data collection and storage and intelligent identification is complicated, and the efficiency is difficult to meet the high-speed inspection requirements of urban rail lines. Therefore, it is necessary to propose an unmanned track status inspection method that meets the urban rail operating vehicle carrying requirements and higher recognition speed through system reconstruction.

[0036] The purpose of the present invention is to provide a mounted unmanned track status inspection scheme suitable for urban rail transit. In view of the shortcomings of the current track inspection means that require the use of special inspection vehicles and require more manual intervention, a track status inspection method that can be mounted on urban rail transit operating vehicles is provided, which can realize the real-time autonomous collection of track image data according to the vehicle operating line, and complete the data packaging and restart the new inspection task according to the line switching situation. At the same time, the algorithm library built into the AI ​​module can analyze the inspection data in real time and output suspected defects. On the one hand, it can meet the needs of detection during the operation of urban rail transit and reduce the occupancy of skylights. On the other hand, it can realize unmanned detection and reduce the workload of manual labor.

[0037] Specifically, an embodiment of the present invention provides an automated track status inspection system for automatically inspecting track status, simplifying the inspection operation process, reducing labor costs, and improving the accuracy and efficiency of track status inspection. Figure 1 ,The system includes: a label reading unit, an inspection control unit, an image acquisition control unit, an image acquisition unit, an encoding unit and an artificial intelligence processing unit;

[0038] The tag reading unit 101 is used to read the radio frequency electronic tags on the urban track that the train passes through in real time after the train carried by the automated track status inspection system runs;

[0039] The inspection control unit 102 is used to analyze the radio frequency electronic tag obtained from the tag reading unit in real time; when the analysis determines that the radio frequency electronic tag read in real time is the starting position tag of the urban track, it sends an inspection instruction to the image acquisition control unit to start the inspection;

[0040] The image acquisition control unit 103 is used to send an image acquisition instruction to the image acquisition unit after receiving the inspection instruction sent by the inspection control unit;

[0041] An image acquisition unit 104 is used to acquire image information of the urban track through which the train passes in real time upon receiving an image acquisition instruction;

[0042] The encoding unit 105 is used to generate a pulse signal as the train runs when the image acquisition unit is collecting images; the image acquisition control unit is also used to: combine the image information received from the image acquisition unit and the pulse signal received from the encoding unit according to the timestamp to obtain real-time inspection information of the urban rail;

[0043] The artificial intelligence processing unit 106 is used to perform data analysis on the real-time inspection information of the urban rail to determine whether the real-time inspection information indicates that there is an inspection problem in the urban rail; if so, an alarm message carrying the inspection problem is issued.

[0044] The embodiment of the present invention provides an automated track status inspection system, wherein a tag reading unit is used to read the radio frequency electronic tags on the urban track through which the train passes in real time after the train carried by the automated track status inspection system is running; an inspection control unit is used to analyze the radio frequency electronic tags obtained from the tag reading unit in real time; when the analysis determines that the radio frequency electronic tags read in real time are the starting position tags of the urban track, an inspection instruction for starting the inspection is sent to the image acquisition control unit; the image acquisition control unit is used to send an image acquisition instruction to the image acquisition unit after receiving the inspection instruction sent by the inspection control unit; FIG. The image acquisition unit is used to acquire image information of the urban track through which the train passes in real time upon receiving an image acquisition instruction; the encoding unit is used to generate a pulse signal as the train runs when the image acquisition unit is acquiring; the image acquisition control unit is also used to: combine the image information received from the image acquisition unit and the pulse signal received from the encoding unit according to the timestamp to obtain real-time inspection information of the urban track; the artificial intelligence processing unit is used to perform data analysis on the real-time inspection information of the urban track to determine whether the real-time inspection information indicates that there is an inspection problem in the urban track; if so, an alarm message carrying the inspection problem is issued. The inspection system in the embodiment of the present invention is mounted on the urban rail transit operating vehicle, and there is no need to arrange a dedicated inspection vehicle online separately; through the collaborative work of components such as the label reading unit and the inspection control unit, the track status inspection is carried out while the train is running daily, which reduces the vehicle purchase, maintenance and personnel following costs of the operating unit; the inspection system realizes automated operation through the close cooperation of various units, and the process does not require excessive human intervention, reduces cumbersome manual operation steps, and greatly simplifies the inspection operation process; the image acquisition unit can collect high-quality track image information in real time after receiving the instruction, and the pulse signal generated by the encoding unit is accurately synchronized with the image after processing, ensuring the accuracy and real-time nature of data acquisition, avoiding repeated detection or misjudgment caused by inaccurate data acquisition, and improving data acquisition efficiency; the artificial intelligence processing unit can combine the machine learning model to quickly analyze the real-time inspection information, and can efficiently determine whether there is a problem with the track. Compared with the traditional intelligent recognition algorithm, its processing speed is faster and the accuracy is higher. It can complete data analysis in real time during the high-speed operation of the train, meet the high-speed detection requirements of the urban rail urban line, and effectively improve the inspection efficiency.

[0045] In a specific implementation, the tag reading unit is used to read the radio frequency electronic tags on the urban track through which the train passes in real time after the train carried by the automated track status inspection system is running.

[0046] In the embodiment, the tag reading unit is specifically used to: adopt radio frequency identification technology to read the radio frequency electronic tags on the urban track through which the train passes in real time.

[0047] In the above embodiment, the tag reading unit adopts radio frequency identification (RFID) technology, which is a non-contact automatic identification technology based on radio waves. In the urban rail transit inspection scenario, the target object is automatically identified and relevant data is obtained through radio frequency signals. The system pre-sets radio frequency electronic tags along the track, which store data related to the track position, line information, etc.

[0048] When the train is running, the tag reading unit emits a radio frequency signal of a specific frequency, which forms an electromagnetic field in space. When the train passes by the radio frequency electronic tag, the tag is activated by the electromagnetic field and uses its own antenna to reflect the information stored in the chip back to the tag reading unit in the form of a radio frequency signal.

[0049] The working mode of the above tag reading unit lays the foundation for the precise operation of the entire inspection system, ensuring that the system can obtain train location information in a timely manner, and then accurately start and execute corresponding inspection tasks, thereby improving the overall performance and reliability of the inspection system.

[0050] In specific implementation, the inspection control unit is used to analyze the radio frequency electronic tag obtained from the tag reading unit in real time; when the analysis determines that the radio frequency electronic tag read in real time is the starting position tag of the urban rail, an inspection instruction to start the inspection is sent to the image acquisition control unit.

[0051] In the embodiment, the inspection control unit maintains a real-time data communication link with the tag reading unit and receives the radio frequency electronic tag information obtained by the tag reading unit. These radio frequency electronic tags are pre-set along the urban track and contain key identification information related to the track line. After receiving the tag information, the inspection control unit immediately starts the internal data analysis program.

[0052] First, the received RF signal is decoded and converted into a recognizable digital signal format, from which the track position data carried by the tag is extracted. Then, the data is compared with the system's built-in track line database. The database stores detailed layout information of urban tracks, including the starting position tag, characteristic information of different sections, and the direction of the track line. Through precise data comparison operations, the inspection control unit can accurately determine whether the read RF electronic tag is the starting position tag of the urban track.

[0053] When the inspection control unit determines through rigorous data analysis that the real-time read radio frequency electronic tag is the starting position tag of the urban rail, a series of subsequent actions are immediately triggered. It generates an inspection instruction to start the inspection according to the preset system control logic. This inspection instruction contains detailed task start information, such as the initial parameter setting of the inspection task, the start time mark of data collection, and the synchronization signal for the subsequent image acquisition control unit and other related units to work together.

[0054] Subsequently, the inspection control unit accurately sends the inspection instruction to the image acquisition control unit through a high-speed data transmission channel. This transmission process follows a strict data communication protocol to ensure the integrity and accuracy of the instruction and prevent data loss or errors during the transmission process. After receiving the inspection instruction, the image acquisition control unit starts the corresponding image acquisition operation according to the instruction content, thus officially starting the entire track status inspection process.

[0055] Through the accurate analysis of the radio frequency electronic tag information by the inspection control unit and the timely and effective issuance of inspection instructions, precise control of the start of the inspection task is achieved, ensuring that the entire inspection system can accurately carry out its work when the train enters the starting position of the track, providing a reliable starting signal and task scheduling guarantee for subsequent track status detection, and improving the intelligence and work efficiency of the inspection system.

[0056] During specific implementation, the image acquisition control unit is used to send an image acquisition instruction to the image acquisition unit after receiving the inspection instruction sent by the inspection control unit.

[0057] In the embodiment, Figure 4 FIG. 1 is a flowchart of an image acquisition control unit in an automated track status inspection system according to an embodiment of the present invention. The image acquisition control unit, such as Figure 4 As shown, it is specifically used for:

[0058] Step 401: determining target image acquisition parameters corresponding to the type of urban track that the train is traveling through according to the association relationship between the preset urban track type and the image acquisition parameters; the image acquisition parameters include exposure time, gain level, and acquisition frame rate;

[0059] Step 402: issuing an image acquisition instruction carrying the target image acquisition parameters.

[0060] In the above embodiment, the system can pre-establish a comprehensive and detailed database of the association between preset urban track types and image acquisition parameters. The database covers various track types commonly used in urban rail transit, such as straight tracks, curved tracks, turnout area tracks, and tracks with different slopes. For each track type, the best image acquisition parameter combination that matches it is determined through a large number of experiments and actual data acquisition analysis.

[0061] When the train is running, the image acquisition control unit first obtains the type information of the track the train is currently on. This information can be obtained in a variety of ways, such as data interaction with the inspection control unit to obtain the track position and type information determined by the inspection control unit based on the tag reading unit and other positioning systems; or through its own connection with the train control system to obtain the train route and track type data. Once the type of urban track that the train is passing through is determined, the image acquisition control unit searches for the corresponding image acquisition parameters in the association database.

[0062] The determination of exposure time is also closely related to the track type. In straight track areas with relatively stable light, the exposure time can be set to the standard value T1us to obtain clear and accurate images. However, in curved tracks or switch areas, due to changes in the direction of train travel and the complexity of the track structure, light reflection and obstruction are more complicated, which may cause local light to be too strong or too dark. At this time, the image acquisition control unit will shorten the exposure time to T2us based on the real-time monitored light changes to reduce image blur or overexposure caused by light problems, ensuring that the track details in the image are clearly discernible.

[0063] In terms of acquisition frame rate, on a normally operating straight track, the track state changes are relatively stable, and the acquisition frame rate changes in real time according to the train speed, which can not only meet the basic needs of track state monitoring, but also effectively control the amount of data. However, when the train passes through the switch area or special section, due to the rapid changes in the track structure and the increase in potential risks, a higher frequency of image acquisition is required to capture the instantly changing track state. Therefore, the image acquisition control unit will increase the acquisition frame rate to R1 times the original frame rate to ensure that abnormal conditions during the switch conversion process, small displacements at the track connection, and other problems can be discovered in a timely manner.

[0064] In terms of gain level, by changing the gain level, the camera imaging effect can be adjusted appropriately. The greater the gain, the brighter the image; the smaller the gain, the darker the image. When the exposure time is short for high-speed acquisition, the gain level can be appropriately increased to ensure image brightness; when the vehicle speed is low and the exposure time can be set longer, the gain level can be appropriately reduced to reduce noise in the image.

[0065] After determining the target image acquisition parameters corresponding to the type of track the train is currently traveling on, the image acquisition control unit integrates and packages these parameters to generate an image acquisition instruction carrying the target image acquisition parameters. The instruction uses a specific data format to ensure the accuracy and completeness of the instruction content so that the image acquisition unit can accurately parse and perform the corresponding operation.

[0066] The image acquisition control unit accurately determines the image acquisition parameters according to the track type and issues corresponding instructions through its working mechanism, which ensures that the image acquisition unit can obtain high-quality track image information that meets the analysis requirements under different track conditions, providing a solid data foundation for the subsequent artificial intelligence processing unit to perform accurate data analysis and track status assessment, greatly improving the detection accuracy and reliability of the entire inspection system.

[0067] In a specific implementation, the image acquisition unit is used to acquire image information of the urban track through which the train passes in real time upon receiving an image acquisition instruction.

[0068] In the embodiment, the image acquisition unit is specifically used for:

[0069] When receiving the image acquisition instruction, the image information of the urban track through which the train passes is collected in real time through the integrated multiple sets of high-resolution linear array cameras and laser light sources.

[0070] In the above embodiment, the image acquisition unit integrates multiple sets of high-resolution linear array cameras and laser light sources. These components are optimized in structural design to ensure their stability and reliability in the train operation environment. The high-resolution linear array camera has excellent imaging capabilities and can clearly capture the subtle features of the track surface, such as scratches and wear marks on the rail surface, the integrity of the fasteners, and the status of the track bed. This ensures the integrity of the track image acquisition, avoids visual blind spots, and ensures that any possible track defects can be effectively recorded.

[0071] When the image acquisition unit receives the image acquisition instruction, it immediately starts the acquisition program. Multiple groups of high-resolution linear array cameras start working synchronously, and perform continuous scanning image acquisition of the urban track where the train passes according to the image acquisition parameters contained in the instruction, such as exposure time, gain level, acquisition frame rate, etc. During the acquisition process, the imaging component of the camera focuses light on the image sensor, converts the optical signal into an electrical signal, and then converts the electrical signal into a digital signal through the built-in analog-to-digital converter to form the original image data. A laser light source is used to fill the camera light, and the laser strobe is controlled according to the acquisition pulse to ensure the consistency of the effect. The laser power range is 0 to 30W.

[0072] In order to ensure the real-time performance of image acquisition, the image acquisition unit uses a high-speed data transmission interface and an efficient data processing algorithm. The acquired image data is transmitted to the image acquisition control unit in real time through a high-speed data transmission line. During the transmission process, the data undergoes preliminary compression and verification processing to reduce data transmission delays and ensure data integrity. At the same time, the control module inside the image acquisition unit monitors the working status of the camera in real time, including temperature, frame rate, exposure parameters, etc., to ensure that the camera is always in the best working state, and can continuously and stably acquire high-quality track image information. Even when the train is running at high speed, it can accurately capture the instantaneous state changes of the track, providing real-time and reliable image data support for the entire inspection system.

[0073] The image acquisition unit, through its integrated advanced components and efficient workflow, can accurately collect image information of urban tracks where trains pass after receiving instructions, providing a solid visual data foundation for subsequent track condition assessment and problem diagnosis, and effectively ensuring the effectiveness and accuracy of the inspection system's track condition monitoring.

[0074] In a specific implementation, the encoding unit is used to generate a pulse signal as the train runs when the image acquisition unit is collecting data; the image acquisition control unit is also used to combine the image information received from the image acquisition unit and the pulse signal received from the encoding unit according to the timestamp to obtain real-time inspection information of the urban rail.

[0075] In the embodiment, the encoding unit is specifically used for:

[0076] The pulse signal generated by the train running is filtered, shaped and synchronously processed by the signal processing algorithm to obtain the target sampling pulse of the image acquisition unit.

[0077] In the above embodiment, the encoding unit is installed at the axle end of the train, and its core function is to generate a pulse signal as the train runs while the image acquisition unit performs acquisition operations. The generation of the pulse signal is based on the rotation of the train axle, and the rotational motion of the axle is accurately measured by a high-precision sensor. Every time the train axle rotates a certain angle, the sensor generates an electrical pulse signal, which reflects the changes in the position and motion state of the train on the track. For example, a photoelectric encoder is used as a sensor, and the grating disk inside it rotates with the axle, and the light is received by the photoelectric detector through the gaps on the grating disk, thereby generating a periodic pulse signal.

[0078] In order to ensure the accuracy and stability of the pulse signal, the encoding unit is equipped with an advanced signal processing circuit. This circuit filters, shapes and synchronizes the original pulse signal generated by the sensor. In the filtering process, multi-level filtering technology is adopted, such as low-pass filtering to remove high-frequency noise interference, and high-pass filtering to eliminate the influence of low-frequency fluctuations to ensure the purity of the pulse signal. The shaping circuit converts the irregular pulse waveform into a standard square wave signal, so that it has a clear rising edge and falling edge, which is convenient for subsequent signal processing and analysis. The synchronization processing circuit uses a high-precision clock signal as a reference to synchronize the pulse signal in time, ensuring that the pulse signal accurately matches the actual time of the train operation, laying the foundation for the subsequent accurate fusion with image information.

[0079] On the one hand, the image acquisition control unit receives image information from the image acquisition unit. This image information contains visual data of the urban track where the train passes, reflecting the status of various parts such as the track surface, fasteners, and roadbed. On the other hand, it receives the processed pulse signal from the encoding unit. While receiving the data, the image acquisition control unit assigns an accurate timestamp to each set of data. The timestamp is generated based on the high-precision clock inside the system, with an accuracy of microseconds or even higher, ensuring the accurate correspondence between the image information and the pulse signal in the time dimension.

[0080] The image acquisition control unit combines the image information and the pulse signal according to the timestamp. Specifically, the pulse signal is used as the time reference to associate the image information at the same time or in a very short time interval with the pulse signal. For example, when a pulse signal is generated, the image frame collected by the corresponding image acquisition unit at this time is recorded, and the two are combined into a set of real-time inspection data. In this way, the image information reflecting the visual state of the track is organically combined with the pulse signal reflecting the position and running status of the train to form real-time inspection information of urban rail containing rich information. These real-time inspection information can not only intuitively display the appearance of the track, but also accurately determine the location of the track problem and the running status of the train when the problem occurs through the position and speed information carried by the pulse signal, providing a comprehensive and accurate data basis for the subsequent artificial intelligence processing unit to conduct in-depth data analysis and track status evaluation.

[0081] Through the coordinated work of the encoding unit and the image acquisition control unit, the precise fusion of image information and pulse signals is achieved, generating high-value real-time urban rail inspection information, effectively improving the accuracy and reliability of the entire inspection system for track status monitoring, and providing strong guarantees for the safe operation of urban rail transit.

[0082] During specific implementation, the artificial intelligence processing unit is used to perform data analysis on the real-time inspection information of the urban rail to determine whether the real-time inspection information indicates that there is an inspection problem in the urban rail; if so, an alarm message carrying the inspection problem is issued.

[0083] In the embodiment, if the artificial intelligence processing unit determines that the real-time inspection information indicates that there is an inspection problem in the urban rail, an alarm message is immediately generated. The alarm message contains detailed description information of the inspection problem, such as the type name of the defect (such as broken fasteners, foreign objects in the ballast bed, cracks in the rail, etc.), shape characteristics (such as the length, width, direction of the crack, the size and shape of the foreign objects, etc.) and texture characteristics (such as the surface texture changes at the rail scratches). These description information helps the operation and maintenance personnel to quickly and accurately understand the specific situation of the track defect. At the same time, the alarm information also includes position information, which is accurately determined by the corresponding relationship between the pulse signal provided by the encoding unit and the track position, and can be accurate to the specific mileage position of the track, so that the operation and maintenance personnel can quickly locate the location of the problem. In addition, according to the model's assessment of the severity of the problem, the severity assessment level is assigned to the alarm information, which is divided into three levels: mild, moderate and severe. The assessment basis includes the size of the defect, the development trend and the degree of influence on the stability of the track structure, so that the operation and maintenance personnel can prioritize emergency problems according to the severity.

[0084] The generated alarm information is transmitted to relevant personnel and systems in a timely manner through the communication network within the system. On the one hand, the alarm information is sent to the train driver's driving console to remind the driver of track abnormalities in a visual and sound prompting manner. The driver can take corresponding safety measures in a timely manner according to the alarm information, such as slowing down or stopping at a safe place to wait for further processing. On the other hand, the alarm information is transmitted to the train dispatching center and the operation and maintenance management system at the same time. The dispatchers and operation and maintenance personnel can receive the alarm information in real time and arrange maintenance personnel and equipment to go to the scene for processing according to the severity and location of the problem. In addition, the system also supports sending alarm information to relevant management personnel by SMS, email, etc., to ensure that all relevant personnel can obtain information on track inspection problems in a timely manner, so as to achieve rapid response and effective processing of track problems and ensure the safe operation of urban rail transit.

[0085] Through advanced data processing technology and intelligent model analysis, the artificial intelligence processing unit realizes accurate analysis and problem judgment of real-time inspection information of urban rail, and issues comprehensive and accurate alarm information in a timely manner, effectively improving the intelligence level of the inspection system for track status monitoring and emergency response capabilities.

[0086] In one embodiment, Figure 5 FIG. 1 is a flowchart showing an example of a workflow of an artificial intelligence processing unit in an automated track status inspection system according to an embodiment of the present invention. Figure 5 As shown, the artificial intelligence processing unit is specifically used for:

[0087] Step 501: using a model trained by a deep learning algorithm and a large-scale image database, performing data analysis on the real-time inspection information of the urban rail to determine whether the real-time inspection information indicates that there is an inspection problem in the urban rail; the inspection problem includes one or any combination of problems located on the rail fasteners, foreign objects, ballast bed and / or rail surface;

[0088] Step 502: If so, an alarm message carrying description information, location information, and severity assessment level of the inspection problem is issued.

[0089] In one embodiment, the artificial intelligence processing unit is specifically configured to:

[0090] The real-time inspection information of the urban rail is input into the urban rail inspection machine learning model to obtain the analysis result of the real-time inspection information of the urban rail; the urban rail inspection machine learning model is based on a deep learning algorithm, and is obtained by training the machine learning model with an urban rail inspection training sample set; the urban rail inspection training sample set includes: inspection information of urban rails with different histories, and analysis results of the inspection information corresponding to different histories of urban rails; the analysis results of the inspection information of urban rails with different histories include different types of track defect images and images of normal track states; the machine learning model adopts the architecture of a convolutional neural network; wherein the convolutional neural network model is used to extract spatial features in image information.

[0091] In the above embodiment, the deep learning model adopted by the artificial intelligence processing unit is based on a convolutional neural network (CNN) architecture.

[0092] The network is responsible for extracting the spatial features of image information in real-time inspection information of urban rail. It contains multiple convolutional layers. In the first convolutional layer, a larger convolution kernel (such as 7×7 or 9×9) and an appropriate step size (such as a step size of 4) are used to initially capture the macro features in the image, such as the overall shape of the track and the approximate outline of the roadbed. As the network depth increases, the subsequent convolutional layers use smaller convolution kernels (such as 3×3 or 5×5) and a step size of 1 or 2 to extract more detailed local features, such as the subtle texture of fasteners and minor damage on the surface of the rail. Pooling layers are interspersed between convolutional layers. For example, the maximum pooling operation is used, with a pooling kernel size of 2×2 or 3×3 and a step size of 2, which is used to reduce the resolution of the feature map and reduce the amount of data while retaining key feature information and improving computational efficiency. During the model training phase, a large-scale image database is used, which covers images of various track conditions, including normal track conditions and images of different types of track problems (such as damaged track fasteners, foreign objects in the roadbed, scratches or cracks on the rail surface, etc.). The number of images reaches tens of thousands or even more. At the same time, the corresponding pulse signal data is collected, which accurately records the position and speed change information of the train when running under different track conditions.

[0093] Using these rich training samples, the model is trained through deep learning algorithms. During the training process, a reasonable loss function is set, such as the multi-classification cross entropy loss function, which is used to measure the difference between the model prediction results and the true label (track state category). At the same time, considering the accuracy requirements for the location and severity assessment of the inspection problem, the position error term (the mean square error is used to calculate the deviation between the predicted position and the true position) and the severity error term (based on the numerical representation of the severity level, the mean square error is used to calculate the error between the predicted severity and the true severity), and the final loss function is obtained by weighted summation to guide the model to learn the accurate track state feature pattern. The optimization algorithm selects an algorithm with adaptive learning rate (such as the Adam optimization algorithm), which can dynamically adjust the learning rate of each parameter according to the first-order moment estimation and second-order moment estimation of the gradient during the training process, so that the model can converge quickly in the early stage of training and stably approach the optimal solution in the later stage, effectively improving the training efficiency and stability.

[0094] When the artificial intelligence processing unit receives the real-time inspection information of the urban rail, it first preprocesses the image information and pulse signal information. For the image information, a cropping operation is performed to remove the edge areas in the image that are not related to the track status analysis, so that the image focuses on key parts such as track fasteners, roadbeds, and rails. Then the image size is adjusted to a uniform specification (such as 256×256 pixels or 512×512 pixels) to meet the model input requirements. The image is then normalized and the pixel values ​​are mapped to a specific range (such as [0,1]) to speed up the model calculation speed and improve the generalization ability of the model. For the pulse signal, filtering is performed to remove noise interference in the signal to ensure the accuracy of the signal.

[0095] The model output assigns a probability value to each possible inspection problem category (such as normal rail fasteners, loose fasteners, broken fasteners; normal ballast bed, foreign matter on ballast bed; normal rail surface, rail scratches, etc.), indicating the possibility that the real-time inspection information belongs to this category of problems. A reasonable probability threshold is set (such as 0.5 or 0.6, etc., determined according to actual needs and training results). When the probability value of a problem category exceeds the threshold, it is determined that the real-time inspection information represents the existence of this inspection problem.

[0096] Once it is determined that there is an inspection problem, the artificial intelligence processing unit immediately generates an alarm message. The description information in the alarm message records the specific situation of the inspection problem in detail. For track fastener problems, the description information includes the specific state of the fastener (such as looseness, breakage, etc.), the location of the fastener (such as the specific section of the track, the sleeper number, etc.), and the appearance changes of the fastener. For foreign matter problems on the ballast bed, describe the size of the foreign matter and its location in the ballast bed. For rail surface problems, record the location, length, and width of the scratches or cracks in detail.

[0097] The position information is determined by the precise correspondence between the pulse signal provided by the encoding unit and the track position, which can be accurate to the specific mileage position of the track, and the error is controlled within a very small range (such as no more than 0.5 meters or less, set according to the system accuracy requirements), so that maintenance personnel can quickly and accurately locate the problem. The severity assessment level is divided into three levels: mild, moderate and severe. Mild problems such as slightly loose fasteners have little effect on the stability of the track structure and will not affect the normal operation of the train in the short term; moderate problems such as scratches of a certain length and width on the surface of the rail may have a certain impact on the comfort of the train and the service life of the track, and maintenance needs to be arranged in a relatively short time; severe problems such as broken fasteners and cracks on the rails seriously threaten the safety of train operation, and emergency measures must be taken immediately for repair. The assessment is based on a comprehensive consideration of the size of the defect (such as the degree of loose fasteners, the size of foreign matter, the size of rail damage, etc.), the development trend (such as comparing multiple consecutive inspection data to determine whether the defect has an expansion trend) and the degree of influence on the stability of the track structure, and is calculated and determined through pre-set evaluation rules and algorithms.

[0098] Through the above complex and orderly workflow, the artificial intelligence processing unit realizes in-depth analysis and accurate judgment of real-time inspection information of urban rails, and issues comprehensive and detailed alarm information in a timely manner, providing strong technical support for the safe operation of urban rail transit, and effectively improving the intelligence level of rail inspections and the timeliness of problem handling.

[0099] Figure 6 FIG. 1 is a flow chart of constructing a sample set for urban rail inspection training in an embodiment of the present invention. In the embodiment, as shown in FIG. Figure 6 As shown, it also includes: constructing an urban rail inspection training sample set according to the following steps:

[0100] Step 601: extracting images and pulse signal data under different track states and conditions and corresponding analysis results from inspection information of different historical urban tracks and analysis results corresponding to the inspection information of different historical urban tracks;

[0101] Step 602: marking the location, type, and severity of image defects in the extracted data;

[0102] Step 603: perform image cropping, resizing, normalization and feature vectorization processing on the labeled data to obtain an urban rail inspection training sample set.

[0103] In the above embodiment, the first step in constructing the urban rail inspection training sample set is to widely collect the inspection information of different historical urban rails and the corresponding analysis results. Representative samples are screened out from a large number of actual inspection records to ensure that all possible track states and conditions are covered. These inspection information includes image information and pulse signal data collected in different time periods, different line sections, and different train operation conditions. For example, inspection data in different seasons, different traffic flow periods, and after different track maintenance cycles are collected to enrich the diversity of samples.

[0104] Professionals use special annotation tools to annotate the collected image information in detail. The annotation content includes the location of the track defect, which is determined by selecting a box or marking a point to determine the specific coordinate range of the defect in the image; the type of defect, such as loose fasteners, broken fasteners, foreign matter on the ballast bed, rail scratches, etc.; the severity of the defect, which is divided into different levels (such as mild, moderate, and severe) according to pre-set evaluation criteria. The evaluation basis includes the size of the defect and the degree of impact on the stability of the track structure.

[0105] The annotated image data undergoes a series of preprocessing operations. First, the image is cropped to remove the areas on the edge of the image that are irrelevant to the track status analysis, so that the image focuses on the key parts of the track (such as fasteners, roadbed, rails, etc.). Then the image size is adjusted to a uniform specification (such as 256×256 pixels or 512×512 pixels) to meet the model input requirements. Next, the image is normalized to map the pixel values ​​to a specific range (such as [0,1]) to improve the model calculation efficiency and generalization ability. For the pulse signal data, filtering is performed to remove noise interference, and then feature vectorization is performed to convert it into a vector form suitable for model input. After these preprocessing steps, the image data and pulse signal data and their corresponding annotation information are combined to form an urban rail inspection training sample set, which will be used to train the urban rail inspection machine learning model so that it can accurately identify and evaluate the status of urban rails and provide a reliable analysis basis for actual inspection work.

[0106] Through the urban rail inspection training sample set constructed above and the machine learning model trained based on the sample set, the artificial intelligence processing unit can analyze the real-time inspection information of urban rail more accurately and efficiently, thereby effectively improving the accuracy and reliability of the entire inspection system for track status monitoring.

[0107] In a specific embodiment, the inspection control unit is connected to the train control system in communication to obtain train operation plan information, and the specific operations are as follows:

[0108] When the train operation plan shows that the train is about to enter a specific section such as a curve, turnout, bridge, tunnel, etc., the inspection control unit adjusts the image acquisition parameters and the analysis threshold of the artificial intelligence processing unit according to the pre-set correspondence between the section type and the inspection strategy. For curved sections, due to the change in the curvature of the track at the curve, the image acquisition control unit shortens the exposure time of the image acquisition unit and increases the image acquisition frame rate to prevent the dynamic blur caused by the train turning from affecting the image quality. In terms of analysis thresholds, the artificial intelligence processing unit reduces the detection thresholds of related defect types and improves the detection sensitivity in response to problems such as rail side wear and uneven fastener force that are prone to occur on the track at curves. For example, the detection threshold of rail side wear is lowered to detect potential risks earlier.

[0109] If the train operation plan indicates that the train is about to perform specific tasks such as high-speed driving tests and heavy-load transportation tasks, the inspection control unit will also make corresponding adjustments based on the task type and the inspection strategy mapping table. Before performing the high-speed driving test task, the image acquisition control unit increases the acquisition frame rate of all cameras to the maximum value and sets the exposure time to the minimum value to ensure that clear and continuous track images can be obtained at high speeds; at the same time, the encoding unit increases the sampling frequency of the pulse signal to more accurately record the train position information. For the artificial intelligence processing unit, the strictness of the overall analysis threshold is increased to avoid false alarms caused by factors such as image jitter and noise caused by high-speed operation, such as increasing the confidence threshold for judging the normal track status.

[0110] The inspection control unit also has the function of real-time monitoring of the actual operating status of the train and the matching of the operating plan. If the train is delayed or the route is temporarily changed, the inspection strategy can be readjusted in time according to the new operating conditions to ensure that the track status inspection can be carried out in the best way under various circumstances, effectively improving the inspection efficiency and accuracy.

[0111] When the inspection strategy is adjusted according to the train operation plan, the data interaction and collaborative workflow between the inspection control unit, the image acquisition control unit, and the artificial intelligence processing unit are as follows:

[0112] After receiving the train operation plan update information, the inspection control unit immediately parses the plan content and determines the specific section the train is about to enter or the specific task type to be performed. Then, based on the built-in inspection strategy optimization rule library, it calculates the image acquisition parameter optimization value and artificial intelligence analysis threshold adjustment value for the current situation. These optimization values ​​and adjustment values ​​are encapsulated in a specific data format to form an inspection strategy adjustment instruction package.

[0113] The inspection control unit sends the inspection strategy adjustment instruction package to the image acquisition control unit and the artificial intelligence processing unit at the same time. After receiving the instruction package, the image acquisition control unit parses the parameter optimization values ​​related to image acquisition, such as exposure time, gain level, acquisition frame rate, etc., and updates these new parameter values ​​to the configuration parameters of the image acquisition unit in real time. During the update process, the image acquisition control unit adjusts the parameters in a gradual transition manner to avoid temporary interference of parameter mutations on image acquisition.

[0114] After receiving the instruction packet, the AI ​​processing unit extracts the analysis threshold adjustment value and applies it to the corresponding defect detection algorithm and classification model. At the same time, according to the new threshold adjustment value, some pre-trained model weights are reloaded to adapt to the new detection requirements. During the model weight update process, the AI ​​processing unit uses dynamic memory management technology to ensure the efficiency and stability of the model update and avoid affecting system operation due to excessive memory usage.

[0115] After completing the inspection strategy adjustment, the image acquisition control unit and the artificial intelligence processing unit respectively feedback the adjustment completion signal to the inspection control unit. After receiving the feedback signal, the inspection control unit verifies the entire adjustment process to ensure that the image acquisition parameters and artificial intelligence analysis thresholds have been accurately updated. If the verification fails, the inspection control unit will resend the adjustment instruction and record the error information for subsequent analysis and troubleshooting. Through this rigorous data interaction and collaborative workflow, the accuracy and effectiveness of the inspection strategy adjustment are guaranteed, and the inspection performance of the system in different operating scenarios is further improved.

[0116] In order to ensure the accuracy and reliability of adjusting the inspection strategy according to the train operation plan, the system has the following safeguards:

[0117] The built-in correspondence between the section type and the inspection strategy, the mapping table between the task type and the inspection strategy, and the inspection strategy optimization rule base can all be regularly updated and optimized through the remote monitoring center. Based on a large amount of actual inspection data, historical track status data, and expert experience, the remote monitoring center continuously adjusts these configuration information to adapt to changes in track lines and train operation characteristics. For example, when a track line is upgraded, resulting in changes in the curvature of the curve or adjustments to the track structure of a specific section, the remote monitoring center promptly updates the image acquisition parameter optimization plan and artificial intelligence analysis threshold setting for the corresponding section to ensure that the inspection system can still work efficiently and accurately under the new track conditions.

[0118] During the operation of the train, the inspection control unit continuously collects and analyzes the working data of the image acquisition unit and the artificial intelligence processing unit, such as image quality evaluation indicators, defect detection accuracy, false alarm rate, etc. If it is found that the actual inspection effect deviates from the expected inspection strategy adjustment target, the inspection control unit will automatically start the adaptive adjustment mechanism. This mechanism corrects the image acquisition parameters and artificial intelligence analysis thresholds in the inspection strategy in real time according to the deviation, combined with the current train operation status and track environment information. For example, if it is found that there are many blurred areas in the image acquired by the image acquisition unit on a certain curved road section, the inspection control unit will further fine-tune the camera's exposure time and acquisition frame rate, and appropriately reduce the confidence threshold of the relevant defect detection in the artificial intelligence processing unit in this section to improve the sensitivity of detection and ensure that no potential defects are missed.

[0119] The system also has a simulation test function. Before adjusting the inspection strategy according to the train operation plan each time, the inspection control unit uses historical track data and simulated train operation information to conduct a virtual test of the inspection strategy to be implemented. During the simulation test, the system simulates the image acquisition process of the image acquisition unit and the data analysis process of the artificial intelligence processing unit, predicts the inspection results under the new strategy, and compares them with the expected goals. If the simulation test results are not ideal, the inspection control unit will optimize the strategy adjustment plan until the simulation test achieves satisfactory results. Through this simulation test and adaptive adjustment mechanism, the system can continuously optimize the inspection strategy adjustment process, improve the adaptability and reliability of the system in a complex and changeable train operation environment, and always maintain a high inspection efficiency and accuracy.

[0120] Two specific embodiments are given below to illustrate the specific application of the system of the present invention.

[0121] First embodiment:

[0122] Assume that the automatic track status inspection system is applied in a subway line in a certain city. The subway line contains multiple track types such as straight track, curved track, and switch area, and the train running speed varies in different sections. The operating environment includes different lighting conditions (such as dark in the tunnel, strong light on the ground section, etc.) and different weather conditions (sunny, rainy, etc.).

[0123] The first embodiment provides an automated track status inspection system, and the workflow of each unit is as follows:

[0124] (I) Tag reading unit

[0125] When the train starts and begins to run, the tag reading unit is continuously in operation.

[0126] (II) Inspection control unit

[0127] The inspection control unit receives the radio frequency electronic tag information obtained by the tag reading unit in real time. When the train runs to a specific location, the tag reading unit reads a tag information, and the inspection control unit immediately analyzes it. First, the received radio frequency signal is decoded to obtain the track location identification data contained therein.

[0128] Then, the data is compared with the track line database built into the system. Assuming that the database corresponding to the current position of the train is marked as a possible starting position area, the inspection control unit further makes an accurate judgment. If it is determined that the tag is the starting position tag of the urban rail, an inspection instruction for starting the inspection is generated, which contains information such as the task start time and initial parameter settings, and is sent to the image acquisition control unit.

[0129] (III) Image acquisition control unit

[0130] After receiving the inspection instruction, the image acquisition control unit first obtains the type information of the track the train is currently on. Assuming that the train is currently traveling on a curved track, the image acquisition control unit searches for the target image acquisition parameters corresponding to the curved track based on the association between the preset urban track type and the image acquisition parameters. The camera exposure time is determined to be 8us (shorter than 12us for straight tracks to avoid overexposure or blurring caused by light changes at bends), and the acquisition frame rate is increased to 1.5 times (higher than the acquisition frame rate for straight tracks to capture rapid changes in track conditions at bends).

[0131] Next, the image acquisition control unit encapsulates these target image acquisition parameters into an image acquisition instruction and sends it to the image acquisition unit. Meanwhile, the image acquisition control unit is ready to receive the image information acquired by the image acquisition unit and the pulse signal generated by the encoding unit.

[0132] (IV) Image acquisition unit

[0133] After receiving the image acquisition instruction, the image acquisition unit uses a laser light source to fill in the camera light, and controls the laser strobe according to the acquisition pulse to ensure the consistency of the effect. The laser power range is 0 to 30W. The camera collects the image information of the train passing through the curved track in real time according to the set parameters, and transmits the image data to the image acquisition control unit.

[0134] (V) Coding Unit

[0135] The encoder unit continuously generates pulse signals as the train axle rotates during the operation of the train. For example, when the train axle rotates a certain angle (such as 1 degree), the sensor inside the encoder unit generates an original pulse signal.

[0136] The original pulse signal is first filtered, low-pass filtered to remove high-frequency noise interference, and then converted into a standard square wave signal through a shaping circuit. Finally, the synchronization processing circuit uses a high-precision clock signal (such as a 1MHz clock) to synchronize the pulse signal, ensuring that the pulse signal accurately matches the actual time of the train operation, and obtains the target sampling pulse of the image acquisition unit, which is sent to the image acquisition control unit.

[0137] (VI) Image acquisition control unit (data combination)

[0138] After receiving the image information from the image acquisition unit and the pulse signal from the encoding unit, the image acquisition control unit assigns an accurate timestamp to each set of data. The timestamp is generated based on a high-precision clock inside the system (such as a microsecond level of accuracy).

[0139] Then, the image information at the same time or in a very short time interval is associated with the pulse signal, using the pulse signal as the time reference. For example, when a pulse signal is generated, the image frames collected by the corresponding image acquisition unit are combined to form real-time urban rail inspection information containing image information and pulse signals, and are ready to be transmitted to the artificial intelligence processing unit.

[0140] (VII) Artificial Intelligence Processing Unit

[0141] After receiving the real-time inspection information of the urban rail, the AI ​​processing unit first preprocesses the image information and pulse signal information. For the image information, a cropping operation is performed to remove the areas on the edge of the image that are irrelevant to the track status analysis, so that the image focuses on key parts such as track fasteners, roadbed, and rails. Then the image size is adjusted to 512×512 pixels, and then the image is normalized to map the pixel values ​​to the [0,1] range. For the pulse signal, filtering is performed to remove noise interference.

[0142] The preprocessed image information is input into the convolutional neural network model for spatial feature extraction. The first convolution layer uses a 7×7 convolution kernel with a step size of 4 to extract the macro features of the image; the subsequent convolution layers use a 3×3 convolution kernel with a step size of 1 to extract more detailed local features, and the pooling layer uses a 2×2 maximum pooling kernel with a step size of 2 to reduce the amount of data and retain key features.

[0143] The extracted image spatial feature vector is input into the fully connected layer for classification and judgment. Assuming that the model output shows that the probability of the track fastener being loose is 0.7 (exceeding the set threshold of 0.6), it is determined that the real-time inspection information representation has an inspection problem.

[0144] The artificial intelligence processing unit immediately generates an alarm message, which describes the specific location of the loose fastener and changes in appearance (such as signs of slight displacement of the fastener). The position information is determined by the correspondence between the pulse signal provided by the encoding unit and the track position, accurate to the specific mileage position of the track, such as 5.5 kilometers from the starting station. The severity assessment level is mild, because although the fastener is loose, it does not temporarily affect the normal operation of the train, but maintenance needs to be arranged in the near future. The alarm information is sent to the train driver's console, train dispatching center and operation and maintenance management system to notify relevant personnel to deal with it in time.

[0145] (VIII) System self-test (patrol control unit)

[0146] After the train starts, the inspection control unit first performs a self-test procedure. Functional tests are performed on the hardware devices of each unit, such as checking whether the radio frequency transmission and receiving modules of the tag reading unit are working properly, whether the camera and light source modules of the image acquisition unit can start and collect images normally, whether the sensor and signal processing circuit of the encoding unit can generate and process pulse signals normally, and whether the computing chip and storage unit of the artificial intelligence processing unit are working normally.

[0147] At the same time, the stability of the communication connection between different units is tested. By sending test data packets, check whether the data transmission between the inspection control unit and the label reading unit, the image acquisition control unit is accurate, whether the data interaction between the image acquisition control unit and the image acquisition unit, the encoding unit is stable, and whether the data reception and processing between the artificial intelligence processing unit and the image acquisition control unit is normal. If any abnormality is detected, the inspection control unit immediately records the error information and attempts to automatically repair it (such as restarting related equipment or modules). If the automatic repair fails, the inspection task is stopped, and a system failure alarm is issued to the train driver and operation and maintenance personnel, waiting for manual intervention.

[0148] (IX) Construction of urban rail inspection training sample set (offline operation)

[0149] From the historical inspection records of urban rail transit, we collect inspection information in different time periods (such as daytime and nighttime), different line sections (including straight lines, curves, switches, etc.), different train operation states (acceleration, deceleration, constant speed) and different weather conditions (sunny, rainy, snowy). For example, we screened out 1,000 groups of representative samples from the inspection data of the past year to ensure that all possible track states and conditions are covered.

[0150] Professionals annotate the collected image information. For track fasteners, mark whether they are loose or broken, the degree of looseness or breakage, and the specific location in the image (selected by coordinate boxes); for the roadbed, mark whether there are foreign objects, the type of foreign objects (such as stones, metal blocks, etc.), size (length and width) and location; for the rail surface, mark whether there are scratches or cracks, and the length, width (which can be estimated through image analysis) and shape characteristics of the scratches or cracks. At the same time, the pulse signal data is accurately associated with the image information, and the timestamp and train location information corresponding to the track defect in the pulse signal are annotated.

[0151] The annotated image data is cropped to remove irrelevant areas at the edge of the image, resized to 512×512 pixels, normalized (pixel values ​​are mapped to [0,1]), and then converted into feature vectors. The pulse signal data is also converted into feature vectors after filtering. Finally, the image feature vector and the pulse signal feature vector and their corresponding annotation information are combined to form a training sample set for urban rail inspection. For example, the sample set contains 800 groups of normal track status samples and 200 groups of different types of defect samples, which are used to train the urban rail inspection machine learning model so that it can accurately identify and evaluate the status of urban rails.

[0152] Through the above specific embodiments, the workflow of the automated track status inspection system in actual urban rail transit is demonstrated in detail. The collaborative work of various links, from label reading, inspection control, image acquisition, encoding, data combination to artificial intelligence processing, system self-inspection and training sample set construction, realizes efficient and accurate inspection of the track status, and effectively ensures the safe operation of urban rail transit.

[0153] Second embodiment:

[0154] Figure 2 FIG. 1 is a specific example diagram of the equipment structure of an automated track status inspection system in an embodiment of the present invention. Figure 2 As shown, the second embodiment proposes an automated track status inspection system suitable for urban rail transit, and its hardware equipment includes: 1. an image acquisition processor (i.e., an image acquisition control unit), 2. an integrated control host (i.e., an inspection control unit), 3. an acquisition main control, 4. an image acquisition module (i.e., an image acquisition unit), 5. a label reader (i.e., a label reading unit), 6. an encoder (i.e., an encoding unit), and 7. an AI module (i.e., an artificial intelligence processing unit).

[0155] Among them, 1 is an acquisition and processing host equipped with a high-performance CPU and a large-computing AI module, which can simultaneously perform synchronous compression processing and distribution processing of 50KHz large data streams on 4 image acquisition modules.

[0156] After the vehicle power supply powers on the detection equipment, the image acquisition host 1 and the integrated control host 2 automatically start, and after entering the system, the detection software runs. The acquisition main control 3 completes the initialization of the image acquisition module 4, and the program enters the waiting command state.

[0157] When the vehicle runs to the starting position of the line to be detected and the tag reader 5 detects the start tag, the integrated control host sends a start detection instruction to the acquisition main control 3.

[0158] The acquisition main control 3 starts the detection task according to the received line information and input information such as the detection direction.

[0159] The original pulse signal generated by the encoder 6 as the vehicle axle rotates is processed by the signal synchronization module and converted into a sampling pulse of the image acquisition module 4 to continuously sample the line.

[0160] While generating the acquisition task, the acquisition main control 3 sends the analysis instruction of the preset intelligent recognition project to the AI ​​module 7. The AI ​​module 7 calls the relevant algorithm module from the algorithm library to perform intelligent recognition on the image data, and encapsulates and stores the suspected defects and their corresponding information.

[0161] Through the above process, the integrated control host issues collection start and stop information by receiving control commands such as line electronic tags and locomotive control information in real time. In the image acquisition and processing host, unmanned image acquisition and compression of the current inspection line, intelligent analysis of image data, and storage of original data and defect data can be automatically realized, thereby realizing unmanned inspection of urban rail transit lines.

[0162] Figure 3 FIG. 1 is a specific example diagram of the inspection principle of an automated track status inspection system according to an embodiment of the present invention. Figure 3 As shown, the automated track status inspection system applicable to urban rail transit proposed in the second embodiment includes an on-board part and an off-board part;

[0163] Among them, the on-board equipment is installed in the equipment cabinet or under the seat, including the image acquisition and processing host and the integrated control host; the off-board equipment mainly includes the image acquisition module (such as an imaging component) directly above and on the side of the center line of the rail, the label reader in the center under the car body, the encoder installed at the axle end and the radio frequency electronic tag arranged on the line.

[0164] Therefore, in view of the current situation that urban rail transit operating units find it difficult to arrange special inspection vehicles for online inspections and provide dedicated personnel to operate inspection equipment, the present invention designs an unmanned track status inspection method that can be carried on urban rail transit operating vehicles. The method can automatically perform image acquisition, storage and data analysis according to the vehicle operation conditions and line switching conditions, which is of great significance for solving objective problems on site.

[0165] In specific applications, the image acquisition processor and the integrated control host are powered by the electric passenger car distribution cabinet to complete the automatic startup of the machine and system software, and at the same time complete the communication between the acquisition host and the integrated host.

[0166] After the acquisition master acquisition program is started, the initialization of each image acquisition module under the vehicle is completed and it is in a state of waiting for acquisition tasks.

[0167] The integrated main control reads the RFID tag and locomotive control information in real time. When it reads the line starting position tag, it sends synchronous data to the inspection main control to complete the mileage correction and issue the collection start command.

[0168] After receiving the control command from the integrated master control, the inspection master control starts data collection, captures the vehicle data in real time through the image acquisition module, and accumulates and stores it;

[0169] When the acquisition task is started, the intelligent recognition algorithm call instruction is sent to the AI ​​module synchronously.

[0170] After receiving the instruction, the AI ​​module will analyze the current collected line data in real time according to the preset infrastructure defect type. When the data of this task is traversed, the identification result data will be independently packaged and stored. When the detection line information changes, the integrated main control will make a judgment based on the corresponding line RFID tag and locomotive control information. If the line needs to be switched to restart the task, it will send a collection stop command to the collection main control, and then restart a new collection task. By obtaining the current line information in real time, the inspection system automatically starts collection and intelligent defect identification, thereby realizing unmanned detection on the urban rail train.

[0171] The image acquisition processor is a device that realizes image acquisition and intelligent analysis of the inspection system. It has functions such as imaging component power supply, pulse signal distribution, network switching, etc. It can simultaneously access multiple imaging components to perform acquisition and storage tasks and realize intelligent recognition of image data.

[0172] The integrated control host can realize functions such as positioning synchronization, automatic control, data transmission and integrated management, and is mainly used to realize the collection and control function of the inspection system through RFID and train control information.

[0173] The acquisition master control is the core functional module of the image acquisition processor, which adopts a high-efficiency and low-power CPU and can provide a high-performance multi-channel camera acquisition control computing platform for the inspection system.

[0174] The AI ​​module is based on an embedded system, supports an optimized deep learning framework and image processing library, and is mainly used to realize image recognition functions with high-performance AI reasoning capabilities.

[0175] The image acquisition module integrates a laser light source and a high-definition linear array scanning camera module, which can continuously acquire images of the target object and achieve basically consistent imaging quality around the clock.

[0176] The existing track status inspection method usually adopts the installation method of carrying a special inspection vehicle. Data collection and processing analysis require inspection personnel to operate a special industrial computer to complete, which cannot meet the objective situation of short window time of urban rail transit and difficult locomotive operation. The working method of the automated track status inspection system proposed in the present invention has the conditions for installation on urban rail transit operating vehicles, and the equipment can independently complete the inspection work such as line collection and line change, data intelligent analysis, etc., which greatly improves the efficiency of urban rail line inspection.

[0177] The automated track status inspection system and working method of the present invention mainly have the following three beneficial effects:

[0178] 1. Mounted unmanned collection

[0179] 1. Advanced onboard hardware design

[0180] The compact chassis design based on CPCIE high-speed bus structure is adopted. This design greatly improves the integration of the system, makes the layout of each component more compact and reasonable, and reduces the physical space occupied by the system. At the same time, it significantly enhances the reliability of the system and reduces the risk of system abnormalities caused by loose hardware connections or failures. The integrated image acquisition module is a major feature of this hardware, in which the high-resolution line array camera and laser light source adopt synchronous trigger control technology. The high-resolution line array camera can capture the details of the track surface and provide rich image data for subsequent accurate analysis. The laser light source provides the camera with stable and sufficient lighting, ensuring that the camera can obtain clear and high-quality images under various lighting conditions, especially in low-light environments. The synchronous trigger control technology of the two ensures the accuracy and consistency of image acquisition, so that the acquired images can accurately reflect the real-time status of the track.

[0181] 2. Efficient unmanned collection mechanism

[0182] An all-weather unmanned data collection system has been built, which has a powerful adaptive imaging parameter adjustment capability. Whether it is during a sunny day, a dim night, or in severe weather conditions, the system can automatically adjust the camera's imaging parameters, such as exposure time, gain, etc., according to the real-time changes in ambient light, thereby ensuring that the imaging quality remains consistent. In addition, the system realizes automatic synchronous control of multiple cameras and multiple light sources. Multiple cameras work together to shoot the track from different angles, and multiple light sources provide lighting synchronously to ensure that all parts of the track can be fully illuminated. In this way, high-speed synchronous acquisition of multi-channel data is achieved, which greatly improves the efficiency and integrity of data acquisition and provides a comprehensive and accurate data basis for subsequent track status analysis.

[0183] 2. Autonomous inspection control

[0184] 1. Perfect control protocol and function realization

[0185] A set of proprietary control protocols has been built, which provides a solid rule basis for the automatic operation of the entire inspection system. Based on this protocol, the system can realize the automatic start and stop function, and can start or stop the inspection task according to the running status of the train and the track line conditions without human intervention. When the track line changes, such as the train switches from one line to another, the system can automatically identify and perform the line change operation to ensure the continuity and accuracy of the inspection work. At the same time, the system can automatically encapsulate and store the detection data, and organize and store the collected image data, pulse signal data, etc. in a specific format to facilitate subsequent data management and analysis. The intelligent recognition algorithm can run automatically with the support of this protocol. Once the inspection task is started, the algorithm begins to analyze the collected data in real time and quickly identify possible problems with the track. In addition, the system also has the function of defect identification results and report export, which can present the defect information obtained from the analysis in the form of intuitive reports, and provide detailed track status reports for operation and maintenance personnel to facilitate their accurate decision-making.

[0186] 2. Accurate multi-source mileage calibration

[0187] In order to ensure the accuracy of inspection data and location information, the system uses multi-source mileage calibration technology. Among them, the high-precision photoelectric encoder can accurately measure the rotation angle and speed of the train axle, and calculate the train's mileage and position changes by analyzing the axle rotation data. RFID technology obtains specific location information of the track by reading the radio frequency electronic tags pre-set on the track, providing a reference point for mileage calibration. In addition, the real-time mileage information in the locomotive control signal is also included in the calibration system. Combining the mileage information from these three sources, the system can perform accurate calibration calculations, effectively eliminating the errors that may be caused by a single mileage measurement method, thereby improving the accuracy of the entire inspection system in judging the track position and status.

[0188] 3. Real-time intelligent analysis

[0189] 1. Optimized hardware and algorithm architecture

[0190] In terms of hardware, an embedded artificial intelligence module is used, which has obvious advantages over traditional GPUs. Its low power consumption can effectively reduce the overall energy consumption of the system and meet the needs of the inspection system to run for a long time on the train. At the same time, it has high computing performance and can quickly process large amounts of image data and complex algorithm operations, ensuring that the system can still complete data analysis tasks in real time during the high-speed operation of the train, meeting the strict requirements of automated inspection for hardware efficiency. At the algorithm level, the intelligent recognition algorithm covers the recognition functions of a variety of track defects, including the detection of common problems such as loose fasteners, broken rails, rail scratches, cracks, and foreign objects on the roadbed. By learning and analyzing a large number of different types of defect images, the algorithm can accurately identify the characteristic patterns of these defects and achieve accurate judgment of the track status.

[0191] 2. Efficient multi-operator fusion and high-speed analysis capabilities

[0192] The multi-operator fusion technology is used to organically combine a variety of different algorithm operators. On the basis of ensuring the recognition accuracy, multiple items are analyzed in real time at the same time. For example, when analyzing a track image, the system can simultaneously detect multiple key items such as the fastener status, rail surface quality, and roadbed conditions, greatly improving the analysis efficiency. This multi-operator fusion technology enables the system to have a maximum recognition speed of 160km / h. Even when the train is running at high speed, it can promptly detect possible problems on the track, providing strong technical support for ensuring the safe operation of urban rail transit.

[0193] Of course, it is understandable that the above detailed process may have other variations, and the relevant variations should all fall within the protection scope of the present invention.

[0194] The embodiment of the present invention provides an automated track status inspection system, wherein a tag reading unit is used to read the radio frequency electronic tags on the urban track through which the train passes in real time after the train carried by the automated track status inspection system is running; an inspection control unit is used to analyze the radio frequency electronic tags obtained from the tag reading unit in real time; when the analysis determines that the radio frequency electronic tags read in real time are the starting position tags of the urban track, an inspection instruction for starting the inspection is sent to the image acquisition control unit; the image acquisition control unit is used to send an image acquisition instruction to the image acquisition unit after receiving the inspection instruction sent by the inspection control unit; FIG. The image acquisition unit is used to acquire image information of the urban track through which the train passes in real time upon receiving an image acquisition instruction; the encoding unit is used to generate a pulse signal as the train runs when the image acquisition unit is acquiring; the image acquisition control unit is also used to: combine the image information received from the image acquisition unit and the pulse signal received from the encoding unit according to the timestamp to obtain real-time inspection information of the urban track; the artificial intelligence processing unit is used to perform data analysis on the real-time inspection information of the urban track to determine whether the real-time inspection information indicates that there is an inspection problem in the urban track; if so, an alarm message carrying the inspection problem is issued. The inspection system in the embodiment of the present invention is mounted on the urban rail transit operating vehicle, and there is no need to arrange a dedicated inspection vehicle to go online separately; through the collaborative work of components such as the label reading unit and the inspection control unit, the track status inspection is carried out while the train is running daily, which reduces the vehicle purchase, maintenance and personnel following costs of the operating unit; the inspection system realizes automated operation through the close cooperation of various units, and the process does not require excessive human intervention, reduces cumbersome manual operation steps, and greatly simplifies the inspection operation process; the image acquisition unit can collect high-quality track image information in real time after receiving the instruction, and the pulse signal generated by the encoding unit is accurately synchronized with the image after processing, ensuring the accuracy and real-time nature of data acquisition, avoiding repeated detection or misjudgment caused by inaccurate data acquisition, and improving data acquisition efficiency; the artificial intelligence processing unit can combine the machine learning model to quickly analyze the real-time inspection information, and can efficiently determine whether there is a problem with the track. Compared with the traditional intelligent recognition algorithm, its processing speed is faster and the accuracy is higher. It can complete data analysis in real time during the high-speed operation of the train, meet the high-speed detection requirements of the urban rail urban line, and effectively improve the inspection efficiency.

[0195] As described above, the embodiment of the present invention designs an automated track status inspection system, including multiple collaborative components such as a tag reading unit, an inspection control unit, an image acquisition control unit, an image acquisition unit, an encoding unit, and an artificial intelligence processing unit. Each component is powered by the electric passenger car distribution cabinet, realizing the automatic startup and collaborative communication of the system. For example, the image acquisition processor and the integrated control host automatically start and complete related communications after power is turned on, laying the foundation for subsequent work.

[0196] The image acquisition processor integrates functions such as imaging component power supply, pulse signal distribution, and network switching. It can simultaneously connect to multiple imaging components, realizing the integration of image acquisition and intelligent analysis functions, improving the system's integration and reliability, and reducing the space occupied by the equipment.

[0197] The tag reading unit in the embodiment of the present invention adopts radio frequency identification technology to read the radio frequency electronic tag information on the track in real time. The inspection control unit analyzes and determines the train position based on the tag information. When the line starting position tag is read, the inspection command is accurately sent to the image acquisition control unit, realizing the automatic start of the inspection task.

[0198] The image acquisition control unit determines and sends instructions containing parameters such as appropriate exposure time, gain level, and acquisition frame rate to the image acquisition unit based on the relationship between the preset urban rail type and the image acquisition parameters. The image acquisition unit collects high-quality rail image information according to the instructions through an integrated high-resolution linear array camera and laser light source. The encoding unit generates a pulse signal when the train is running, and obtains an accurate sampling pulse after processing to ensure that the image acquisition is synchronized with the train position. The image acquisition control unit combines the image information and pulse signal into real-time inspection information according to the timestamp, realizing the effective integration and accurate acquisition of multi-source data.

[0199] The artificial intelligence processing unit uses a model trained with deep learning algorithms and large-scale image databases to perform data analysis on real-time inspection information. The model is based on an architecture that combines convolutional neural networks with recurrent neural networks. It can effectively extract spatial features in image information and timing features in pulse signals, and accurately determine whether there are inspection problems on the track. Once a problem is found, an alarm message containing a detailed description of the inspection problem (such as defect type, shape, and texture features), location information (accurate to specific mileage locations), and severity assessment levels can be issued, achieving intelligent and efficient detection of track status and accurate positioning and assessment of problems, reducing manual intervention and improving inspection efficiency and accuracy. At the same time, the system also has functions such as self-inspection, data storage and transmission, remote communication and upgrading, which further ensure the stability and adaptability of the system and comprehensively solve the problems existing in the existing technology.

[0200] The present invention also provides an automatic track state inspection system working method, as described in the following embodiments. Since the principle of solving the problem by the method is similar to that of the automatic track state inspection system, the implementation of the method can refer to the implementation of the automatic track state inspection system, and the repeated parts will not be repeated.

[0201] Figure 7 The figure is a flow chart of an automated track status inspection system working method in an embodiment of the present invention. The embodiment of the present invention also provides an automated track status inspection system working method, which is used to automatically inspect the track status, simplify the inspection operation process, reduce labor costs, and improve the accuracy and efficiency of the track status inspection. Figure 7 As shown, the method includes:

[0202] Step 701: After the train carried by the automated track status inspection system is running, the tag reading unit reads the radio frequency electronic tags on the urban track that the train passes through in real time;

[0203] Step 702: the inspection control unit analyzes the radio frequency electronic tag obtained from the tag reading unit in real time; when the analysis determines that the radio frequency electronic tag read in real time is the starting position tag of the urban rail, an inspection instruction to start the inspection is sent to the image acquisition control unit;

[0204] Step 703: after receiving the inspection instruction sent by the inspection control unit, the image acquisition control unit sends an image acquisition instruction to the image acquisition unit;

[0205] Step 704: upon receiving the image acquisition instruction, the image acquisition unit acquires image information of the urban track through which the train passes in real time;

[0206] Step 705: When the image acquisition unit is collecting data, the encoding unit generates a pulse signal as the train runs; the image acquisition control unit combines the image information received from the image acquisition unit and the pulse signal received from the encoding unit according to the timestamp to obtain real-time inspection information of the urban rail;

[0207] Step 706: The artificial intelligence processing unit performs data analysis on the real-time inspection information of the urban rail to determine whether the real-time inspection information indicates that there is an inspection problem in the urban rail; if so, an alarm message carrying the inspection problem is issued.

[0208] In one embodiment, the tag reading unit uses radio frequency identification technology to read the radio frequency electronic tags on the urban track through which the train passes in real time.

[0209] In one embodiment, the image acquisition control unit determines the target image acquisition parameters corresponding to the type of urban track that the train passes through according to the association relationship between the preset urban track type and the image acquisition parameters; the image acquisition parameters include exposure time, gain level, and acquisition frame rate;

[0210] An image acquisition instruction carrying the target image acquisition parameters is issued.

[0211] In one embodiment, when receiving the image acquisition instruction, the image acquisition unit collects image information of the urban track through which the train passes in real time through integrated multiple groups of high-resolution linear array cameras and laser light sources.

[0212] In one embodiment, the encoding unit filters, shapes and synchronizes the pulse signal generated by the train operation through a signal processing algorithm to obtain a target sampling pulse of the image acquisition unit.

[0213] In one embodiment, the artificial intelligence processing unit uses a model trained by a deep learning algorithm and a large-scale image database to perform data analysis on the real-time inspection information of the urban rail to determine whether the real-time inspection information indicates that there is an inspection problem in the urban rail; the inspection problem includes one or any combination of problems located on the rail fasteners, foreign objects, the ballast bed and / or the rail surface;

[0214] If so, an alarm message is issued that carries the description information, location information, and severity assessment level of the inspection problem.

[0215] In one embodiment, the artificial intelligence processing unit inputs the real-time inspection information of the urban rail into the urban rail inspection machine learning model to obtain analysis results of the real-time inspection information of the urban rail; the urban rail inspection machine learning model is based on a deep learning algorithm, and is obtained by training the machine learning model with an urban rail inspection training sample set; the urban rail inspection training sample set includes: inspection information of different historical urban rails, and analysis results of the inspection information corresponding to different historical urban rails; the analysis results of the inspection information of different historical urban rails include different types of track defect images and images of normal track conditions; the machine learning model adopts a convolutional neural network architecture; wherein the convolutional neural network model is used to extract spatial features in image information.

[0216] In one embodiment, it further includes:

[0217] The urban rail inspection training sample set is constructed as follows:

[0218] Extracting images and pulse signal data and corresponding analysis results under different track states and conditions from the inspection information of different historical urban tracks and the analysis results of the inspection information of different historical urban tracks;

[0219] Annotate the location, type, and severity of image defects in the extracted data;

[0220] The labeled data is subjected to image cropping, resizing, normalization and feature vectorization to obtain the urban rail inspection training sample set.

[0221] In one embodiment, the inspection control unit executes a self-test program after the train carried by the automated track status inspection system is started; the self-test program is used to perform functional tests on the hardware equipment and software system of each unit, and to perform stability tests on the communication connections between different units.

[0222] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0223] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0224] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0225] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0226] The specific embodiments described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. An automated track status inspection system, characterized in that: include: Tag reading unit, inspection control unit, image acquisition control unit, image acquisition unit, encoding unit and artificial intelligence processing unit; The tag reading unit is used to read the radio frequency electronic tags on the urban track that the train passes through in real time after the train carried by the automated track status inspection system runs; The inspection control unit is used to analyze the radio frequency electronic tag obtained from the tag reading unit in real time; when the analysis determines that the radio frequency electronic tag read in real time is the starting position tag of the urban track, it sends an inspection instruction to start the inspection to the image acquisition control unit; An image acquisition control unit, configured to issue an image acquisition instruction to the image acquisition unit after receiving the inspection instruction issued by the inspection control unit; An image acquisition unit, configured to acquire image information of the urban track through which the train passes in real time upon receiving an image acquisition instruction; The encoding unit is used to generate a pulse signal along with the train operation when the image acquisition unit is collecting data; the image acquisition control unit is also used to combine the image information received from the image acquisition unit and the pulse signal received from the encoding unit according to the timestamp to obtain the real-time inspection information of the urban rail; The artificial intelligence processing unit is used to perform data analysis on the real-time inspection information of the urban rail to determine whether the real-time inspection information indicates that there is an inspection problem in the urban rail; if so, an alarm message carrying the inspection problem is issued.

2. The system according to claim 1, characterized in that The tag reading unit is specifically used to use radio frequency identification technology to read the radio frequency electronic tags on the urban track where the train passes in real time.

3. The system according to claim 1, characterized in that The image acquisition control unit is specifically used for: According to the association between the preset urban track type and the image acquisition parameters, the target image acquisition parameters corresponding to the type of urban track that the train passes through are determined; the image acquisition parameters include exposure time, gain level, and acquisition frame rate; An image acquisition instruction carrying the target image acquisition parameters is issued.

4. The system according to claim 1, characterized in that The image acquisition unit is specifically used for: When receiving the image acquisition instruction, the image information of the urban track through which the train passes is collected in real time through the integrated multiple sets of high-resolution linear array cameras and laser light sources.

5. The system according to claim 1, wherein: Coding unit, specifically used for: The pulse signal generated by the train running is filtered, shaped and synchronously processed by the signal processing algorithm to obtain the target sampling pulse of the image acquisition unit.

6. The system according to claim 1, wherein: Artificial intelligence processing unit, specifically used for: Using a model trained by a deep learning algorithm and a large-scale image database, data analysis is performed on the real-time inspection information of the urban rail to determine whether the real-time inspection information indicates that there are inspection problems in the urban rail; the inspection problems include one or any combination of problems located on the rail fasteners, foreign objects, ballast bed and / or rail surface; If so, an alarm message is issued that carries the description information, location information, and severity assessment level of the inspection problem.

7. The system according to claim 1, characterized in that Artificial intelligence processing unit, specifically used for: The real-time inspection information of the urban rail is input into the urban rail inspection machine learning model to obtain an analysis result of the real-time inspection information of the urban rail; the urban rail inspection machine learning model is based on a deep learning algorithm and is obtained by training the machine learning model with an urban rail inspection training sample set; The urban rail inspection training sample set includes: inspection information of different historical urban rails, and analysis results of the inspection information corresponding to different historical urban rails; the analysis results of the inspection information of different historical urban rails include different types of track defect images and images of normal track conditions; the machine learning model adopts a convolutional neural network architecture; wherein the convolutional neural network model is used to extract spatial features in image information.

8. The system of claim 1, wherein: Also includes: The urban rail inspection training sample set is constructed as follows: Extracting images and pulse signal data and corresponding analysis results under different track states and conditions from the inspection information of different historical urban tracks and the analysis results of the inspection information of different historical urban tracks; Annotate the location, type, and severity of image defects in the extracted data; The labeled data is subjected to image cropping, resizing, normalization and feature vectorization to obtain the urban rail inspection training sample set.

9. The system according to claim 1, characterized in that The inspection control unit is also used for: After the train carried by the automated track status inspection system is started, a self-checking program is executed; the self-checking program is used to perform functional tests on the hardware equipment and software system of each unit, and to perform stability tests on the communication connections between different units.

10. A working method of an automated track status inspection system, characterized in that: Applied to the automated track status inspection system as claimed in any one of claims 1 to 9, the method comprises: After the train carried by the automatic track status inspection system is running, the tag reading unit reads the radio frequency electronic tags on the urban track through which the train passes in real time; The inspection control unit analyzes the radio frequency electronic tag obtained from the tag reading unit in real time; when the analysis determines that the radio frequency electronic tag read in real time is the starting position tag of the urban rail, it sends an inspection instruction to start the inspection to the image acquisition control unit; After receiving the inspection instruction issued by the inspection control unit, the image acquisition control unit issues an image acquisition instruction to the image acquisition unit; When receiving the image acquisition instruction, the image acquisition unit acquires image information of the urban track through which the train passes in real time; When the image acquisition unit is collecting data, the encoding unit generates a pulse signal as the train runs; the image acquisition control unit combines the image information received from the image acquisition unit and the pulse signal received from the encoding unit according to the timestamp to obtain real-time inspection information of the urban rail; The artificial intelligence processing unit performs data analysis on the real-time inspection information of the urban rail to determine whether the real-time inspection information indicates that there is an inspection problem in the urban rail; if so, an alarm message carrying the inspection problem is issued.

Images