Railway section fault detection method and system of low-altitude unmanned aerial vehicle

Abstract

The invention discloses a fault detection method and system for a railway section by a low-altitude unmanned aerial vehicle, and relates to the technical field of fault detection methods, and the method comprises the steps: determining an abnormal section according to the positions of a plurality of abnormal parameters and the environment type of the detected section; the abnormal grades of the abnormal road sections are determined based on the distribution diagrams of the abnormal road sections and the railway road sections, the fault detection path is determined according to the abnormal grades of the abnormal road sections, the driving state diagrams of the railway road sections and the position of the low-altitude unmanned aerial vehicle, and the detection accuracy of the fault detection path is improved. Therefore, a surface anomaly image and an installation anomaly image are determined according to anomaly detection of an abnormal road section; determining the fault content corresponding to the abnormal road section based on the surface abnormal image, the installation abnormal image and the previous abnormal event of the abnormal road section, and determining the real-time maintenance plan of each abnormal road section according to the operation state of each abnormal road section and the fault content of the plurality of abnormal road sections. And the accuracy of the real-time maintenance plan of each abnormal road section is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of fault detection methods, and in particular to a low-altitude unmanned aerial vehicle fault detection method and system for railway sections. BACKGROUND

[0002] With the development of science and technology, railway sections are part of a railway trunk line and are arranged along the laying direction of the railway trunk line. Trains travel along the extension direction of the railway sections and travel different positions of the railway sections at different time periods. The railway sections are generally detected in a corresponding manner within a preset time period, such as every other month or within half a year. In the prior art, the detection of the railway sections is not real-time detection, and online detection of the railway sections cannot be achieved, thereby affecting the accuracy of abnormal sections and making it impossible to reasonably form real-time maintenance plans for the abnormal sections. SUMMARY

[0003] The present application aims to overcome the shortcomings of the prior art, and provides a low-altitude unmanned aerial vehicle fault detection method and system for railway sections.

[0004] The present application provides a low-altitude unmanned aerial vehicle fault detection method for railway sections, which comprises the following steps: determining a plurality of sub-detection sections according to a distribution map of the railway sections, and determining a section parameter combination of each detection section according to online detection of the plurality of sub-detection sections; in each section parameter combination, determining a plurality of abnormal parameters based on traversal of the section parameter combination, and determining an abnormal section according to the positions of the plurality of abnormal parameters and the environmental categories of the detection sections; determining an abnormal level of the plurality of abnormal sections based on the plurality of abnormal sections and the distribution map of the railway sections, and determining a fault detection path according to the abnormal levels of the plurality of abnormal sections, a travel state map of the railway sections, and the position of the low-altitude unmanned aerial vehicle; the low-altitude unmanned aerial vehicle flies along the fault detection path, performs abnormal detection on each abnormal section, and determines a surface abnormal image and an installation abnormal image according to the abnormal detection of the abnormal section; determines the fault content corresponding to the abnormal section based on the surface abnormal image, the installation abnormal image, and the past abnormal events of the abnormal section, and determines a real-time maintenance plan for each abnormal section according to the operating state of each abnormal section and the fault content of the plurality of abnormal sections.

[0005] The present application provides a low-altitude unmanned aerial vehicle fault detection system for railway sections, which is applied to the low-altitude unmanned aerial vehicle fault detection method described above. The low-altitude unmanned aerial vehicle fault detection system for railway sections comprises:

[0006] A section parameter combination module is configured to determine a plurality of sub-detection sections according to a distribution map of the railway sections, and determine a section parameter combination of each detection section according to online detection of the plurality of sub-detection sections.

[0007] an abnormal section module configured to determine a plurality of abnormal parameters based on traversal of each section parameter combination, determine an abnormal section according to positions of the plurality of abnormal parameters and a kind of environment of the detected section;

[0008] a fault detection path module configured to determine an abnormal level of the plurality of abnormal sections based on a distribution map of the plurality of abnormal sections and the railway section, determine a fault detection path according to the abnormal level of the plurality of abnormal sections, the driving state map of the railway section and the position of the low-altitude unmanned aerial vehicle;

[0009] an abnormal image module configured to fly the low-altitude unmanned aerial vehicle along the fault detection path, perform abnormal detection on each abnormal section, and determine a surface abnormal image and an installation abnormal image according to the abnormal detection of the abnormal section;

[0010] a real-time maintenance plan module configured to determine a fault content corresponding to the abnormal section based on the surface abnormal image, the installation abnormal image and a previous abnormal event of the abnormal section, and determine a real-time maintenance plan of each abnormal section according to the running state of each abnormal section and the fault content of the plurality of abnormal sections.

[0011] Compared with the prior art, the present application has the following advantages:

[0012] In the embodiment of the present application, the abnormal level of the plurality of abnormal sections is determined based on the distribution map of the plurality of abnormal sections and the railway section, the fault detection path is determined according to the abnormal level of the plurality of abnormal sections, the driving state map of the railway section and the position of the low-altitude unmanned aerial vehicle, a plurality of abnormal parameters are introduced, and the abnormal section is controlled, the overall consideration of the abnormal level of the plurality of abnormal sections, the driving state map of the railway section and the position of the low-altitude unmanned aerial vehicle is compatible, and the detection accuracy of the fault detection path is improved.

[0013] Therefore, the low-altitude unmanned aerial vehicle flies along the fault detection path, performs abnormal detection on each abnormal section, and determines a surface abnormal image and an installation abnormal image according to the abnormal detection of the abnormal section; the fault content corresponding to the abnormal section is determined based on the surface abnormal image, the installation abnormal image and a previous abnormal event of the abnormal section, and the real-time maintenance plan of each abnormal section is determined according to the running state of each abnormal section and the fault content of the plurality of abnormal sections, the surface abnormal image and the installation abnormal image are introduced, the multi-dimensional abnormal detection of the abnormal section by the unmanned aerial vehicle is ensured, the overall consideration of the running state of each abnormal section and the fault content of the plurality of abnormal sections is realized, and the accuracy of the real-time maintenance plan of each abnormal section is improved. BRIEF DESCRIPTION OF DRAWINGS

[0014] Figure 1is a flowchart of a fault detection method of a low-altitude unmanned aerial vehicle on a railway section in the embodiment of the present application;

[0015] Figure 2 is a structural composition diagram of a fault detection system of a low-altitude unmanned aerial vehicle on a railway section in the embodiment of the present application. DETAILED DESCRIPTION

[0016] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application.

[0017] Please refer to Figure 1 and Figure 2 A fault detection method of a low-altitude unmanned aerial vehicle on a railway section, applied to a fault detection scene of a low-altitude unmanned aerial vehicle on a railway section, comprises the following steps.

[0018] Step S11: determining a plurality of sub-detection sections according to a distribution map of the railway section, and determining a section parameter combination of each detection section according to online detection of the plurality of sub-detection sections;

[0019] Step S12: determining a plurality of abnormal parameters based on traversal of the section parameter combination in each section parameter combination, and determining an abnormal section according to positions of the plurality of abnormal parameters and an environment type of the detection section;

[0020] Step S13: determining an abnormal level of the plurality of abnormal sections based on the plurality of abnormal sections and the distribution map of the railway section, and determining a fault detection path according to the abnormal level of the plurality of abnormal sections, a driving state map of the railway section, and a position of the low-altitude unmanned aerial vehicle;

[0021] Step S14: the low-altitude unmanned aerial vehicle flies along the fault detection path, performs abnormal detection on each abnormal section, and determines a surface abnormal image and an installation abnormal image according to the abnormal detection of the abnormal section;

[0022] Step S15: determining a fault content corresponding to the abnormal section based on the surface abnormal image, the installation abnormal image, and a previous abnormal event of the abnormal section, and determining a real-time maintenance plan of each abnormal section according to an operating state of the abnormal section and the fault content of the plurality of abnormal sections;

[0023] In step S11, the specific steps are as follows:

[0024] S111: collecting a name of the railway section, determining the distribution map of the railway section according to the name of the railway section and a railway database, triggering initial detection of the low-altitude unmanned aerial vehicle according to the distribution map of the railway section, and collecting a plurality of section demarcation points of the railway section based on the initial detection of the unmanned aerial vehicle;

[0025] S112: Determine a plurality of sub-detection sections based on the plurality of section demarcation points and the distribution map of the railway section, and perform online detection on each sub-detection section;

[0026] S113: Collect a plurality of section parameters based on the online detection of each sub-detection section; mark the plurality of section parameters on the corresponding section detection position, and determine a section parameter combination of each detection section according to the plurality of section parameters and the corresponding section detection position.

[0027] In the embodiments of the present application, the name of the collected railway section is used to initiate a query request to a railway database, and the database returns all key geographic and structural information related to the name. The extracted geographic coordinates, line structure, and other information are visualized in geographic information system (GIS) software or unmanned aerial vehicle task planning software to generate a "distribution map" containing the line center line, key structure position, curve radius, slope change, etc. This distribution map is digitized and contains accurate spatial coordinates. The output is a digital geographic spatial data set, usually in the form of vector data (such as line features, point features) or raster data containing coordinate information.

[0028] The task planning software plans the initial detection flight path of the unmanned aerial vehicle; the initial detection flight path covers the entire target railway section and takes into account the flight performance of the unmanned aerial vehicle (such as speed, endurance, turning radius), sensor field of view angle, overlap rate requirement (for subsequent image stitching or three-dimensional modeling), and safety obstacle avoidance (avoiding known tall obstacles). The planned flight path, flight height, flight speed, sensor working mode (such as camera shooting, laser radar scanning), and other parameters are converted into flight task instructions that the unmanned aerial vehicle can understand.

[0029] Through the unmanned aerial vehicle ground control station (GCS) software, the task instructions are uploaded to the unmanned aerial vehicle and the task is started; or through a preset scheduling system, the takeoff and task execution of the unmanned aerial vehicle are automatically triggered at a specified time. The main purpose of this "initial detection" is to perform a rapid and covering data collection on the entire target railway section, focusing on identifying and accurately positioning the "section demarcation points" on the line; the "section demarcation points" usually refer to positions with clear physical or logical boundaries.

[0030] The UAV performs the initial detection according to the initial detection flight path; the sensors carried (mainly cameras and laser radars) continuously collect data along the railway line to generate a series of images, point clouds or other sensor data. At the same time, computer vision technology is used to analyze the collected images and automatically identify features such as mile markers, boundary markers, curve start and end markers, bridge / tunnel entrances / exits, etc., and record their coordinates. The precise geographic coordinates (latitude, longitude, and elevation) of each identified road segment boundary point and its corresponding boundary type (such as "mile K1050 marker", "straight curve transition point", "bridge start point") are recorded to form a boundary point list. The output is a dataset containing all identified road segment boundary point coordinates and types, usually in the form of GIS point features or coordinate lists. The system realizes the complete process from specifying the target railway line to obtaining its precise geographic information, planning and performing the initial UAV detection, and finally accurately collecting and recording the key physical or logical boundary points (road segment boundary points) along the line.

[0031] Further, multiple road segment boundary points and a distribution map of railway segments are introduced. The road segment boundary points define the key positions on the railway line, and the distribution map provides the overall geographic context and other reference information (such as bridge, tunnel, station location, etc.). The system divides the railway line based on these boundary points. The most direct method is to define the railway segment between two adjacent road segment boundary points as a "sub-detection segment", which is like dividing a line into several segments with marker points.

[0032] First, ensure that these boundary points are sorted in the actual order along the railway line (e.g., from the start point to the end point, in ascending or descending order of mileage), and then the system pairs the sorted adjacent boundary points. For example, if the boundary point list is [P1, P2, P3, P4], the sub-detection segments formed are P1-P2, P2-P3, and P3-P4.

[0033] For each newly generated sub-detection segment, the system assigns attributes, which usually include: a unique identifier, such as "sub-segment_001", "sub-segment_002", etc.; geographic range, recording the precise coordinates of the start and end points of the sub-segment; length, calculating the length of the sub-segment (e.g., the curve distance along the railway line); associated boundary point information, recording which boundary point the sub-segment starts from and which boundary point it ends at (e.g., sub-segment_001 starts from "mile K1050 marker" and ends at "bridge entrance (XX bridge)"); environmental information, if the distribution map contains relevant information, preliminarily labeling the environmental type contained in the sub-segment, such as "bridge segment", "tunnel segment", "straight line segment", "curve segment", etc. The system outputs a "sub-detection segment" list, each sub-segment containing its definition, range, and attribute information.

[0034] A list of multiple sub-detection sections is introduced; real-time, data-driven detection is performed on each sub-detection section to collect the status information of the section. The "online detection" here emphasizes real-time data collection and analysis during the UAV flight. At this time, for each sub-detection section, the UAV plans a detailed flight path to ensure that it can cover all key areas of the section, which includes flying along the track centerline, hovering or flying around key equipment such as signals and catenary pillars, taking specific angle photos of bridges or tunnels, etc. The path planning takes into account the sensor field of view angle of the UAV, the overlap rate requirement, etc. The UAV flies according to the planned path and activates its onboard sensors in real time to collect data.

[0035] The sensors include: high-definition visible light camera: used to take images of the track, ballast, catenary, signal equipment, etc., for visual inspection and surface defect identification; infrared thermal imager: used to detect abnormal heating of track, catenary, insulator, etc. components; LiDAR (Light Detection and Ranging): used to obtain high-precision three-dimensional point cloud data, used to measure track geometry state (track gauge, high-low, direction, level), detect ballast settlement, measure catenary geometry parameters, etc.; multispectral / hyperspectral camera (optional): used to detect material aging, corrosion, etc.

[0036] All collected data (images, point clouds, thermal images, etc.) are accurately labeled with their corresponding geographical location (usually through the GPS / INS system on the UAV) and timestamp, and are associated with the "sub-detection section" currently being detected, which ensures that during subsequent analysis, it can be accurately known which section and location each data point belongs to. For each sub-detection section, output all raw data (images, point clouds, thermal images, etc.) collected during this online detection process and related metadata (such as GPS coordinates, timestamps, sensor parameters, etc.).

[0037] Therefore, each sub-detection section is introduced, and the raw data collected by online detection in each sub-detection section (e.g. sub-sections 001, 002, 003…) include: high-resolution visible light images, infrared thermal imaging images, LiDAR point cloud data, laser ranging data, millimeter wave radar data, IMU (Inertial Measurement Unit) data, GPS / RTK positioning data, etc.

[0038] The "track section parameters" are not the raw data itself, but the characteristic values or indicators extracted from the raw data, which can reflect the current state of the sub-track section. These parameters should be representative and can quantify the health status or potential risks of the track section. Optionally, the clarity of the track, the presence of foreign objects (such as gravel, debris), the condition of the sleepers (whether damaged or dirty), the accumulation of ballast, etc. are extracted from the visible light image; the temperature distribution of the track and sleepers is extracted from the infrared image, indicating current anomalies or component overheating; the geometric parameters of the track, such as track gauge, track height, direction (horizontal displacement), track gauge change rate, etc. are extracted; the position deviation of the sleepers and the geometric shape and accumulation height of the ballast are detected; if there is speed sensor data, the vibration frequency and amplitude are analyzed; combined with GPS / RTK data, the distance of the detection point from the start of the sub-track section is accurately calculated.

[0039] "Multiple track section parameters" are introduced, and accurate position information recorded by the low-altitude unmanned aerial vehicle during detection is collected; each track section parameter is collected at one or more specific positions during detection, and in order to accurately locate the problem later, the parameter must be associated with its collection position; for example, the track gauge deviation at a specific point is +2mm, not the average deviation of the entire sub-track section.

[0040] In the data record or database, a position marker field is added for each track section parameter, which is: accurate GPS coordinates (longitude, latitude, elevation); distance from the start of the sub-track section (e.g. 75 meters from point P1); combined with the mileage marker information; for continuously changing parameters (such as track geometry), a series of sampling points and their corresponding positions are recorded.

[0041] Multiple track section parameters and corresponding track detection positions are introduced, "track section parameter combination" refers to the collection of all parameters (and their position information) belonging to the same sub-detection track section, forming a comprehensive description of the state of the sub-track section. This combination is considered as a "digital portrait" or "health record" of the sub-track section; the system organizes all parameters and their position information belonging to the same sub-detection track section (e.g. sub-track section 001), which is usually achieved through database queries or data structures (such as dictionaries, lists); ultimately, each sub-detection track section corresponds to a "parameter combination" record containing all its parameters and position information; the massive raw data obtained from the unmanned aerial vehicle online detection is converted into structured "track section parameters" with accurate position information, and these parameters are organized by sub-detection track section to form "track section parameter combinations". This combination not only contains the state information of the sub-track section, but also retains the spatial positioning capability.

[0042] In step S12, the specific steps are:

[0043] S121: Real-time monitoring of each road section parameter combination, traversing each road section parameter combination, determining a plurality of to-be-detected road section parameters according to the traversal of each road section parameter combination, and determining a plurality of associated parameters according to the plurality of to-be-detected road section parameters and the associated parameter mapping relationship;

[0044] S122: In the plurality of to-be-detected road section parameters, if a to-be-detected road section parameter exceeds a preset safety parameter threshold, a proportional operation of the to-be-detected road section parameter and the plurality of associated parameters is triggered to determine a corresponding matching coefficient, and if the matching coefficient exceeds a preset matching coefficient threshold, the to-be-detected road section parameter is taken as an abnormal parameter to collect a plurality of abnormal parameters.

[0045] S123: Marking the position corresponding to the abnormal parameter, determining the corresponding detection road section according to the matching of the position corresponding to the abnormal parameter and the distribution map of the railway road section, determining the environment type of the detection road section based on the environment detection of the detection road section, and determining the abnormal road section based on the values, positions and environment types of the plurality of abnormal parameters.

[0046] In the embodiments of the present application, data is transmitted back to the ground station or cloud platform through wired networks, wireless networks (such as 4G / 5G) or satellite communication; the monitoring is continuous and also polling at a set time interval (for example, every 5 minutes, every 10 minutes); not only the arrival of the parameter combination itself in time is monitored, but also the integrity, data format and whether there is a significant abnormality in the key parameters (for example, whether the value is within the physical range, such as whether the temperature suddenly changes to -100℃ or 5000℃); real-time monitoring is the premise of subsequent traversal and analysis; once the data arrives and is confirmed to be valid, the traversal process will be triggered.

[0047] For each "road section parameter combination" that has just been real-time monitored and confirmed to be valid, all "road section parameters" contained in the combination are checked one by one, including quantitative data (such as numerical value, percentage) and qualitative data (such as state description, image feature); not only to view, but also to prepare for the next step of screening "to-be-detected road section parameters"; the traversal process identifies which parameters are the focus of attention and which parameter changes indicate potential problems; usually realized by programming, using a loop structure (such as for loop) to access each record in the parameter list or database.

[0048] Not all the parameters traversed will become "to-be-detected section parameters"; they are screened according to preset rules or strategies; common standards include: importance: some parameters (such as the height of the guide, the pull-out value, and the state of the insulator) are more important than other parameters (such as the brightness of the image background) and are more likely to cause faults, so they are preferentially selected as to-be-detected parameters; variability: parameters with large value changes or in a dynamic change process (such as the instantaneous change of parameters caused by train passing) are paid attention to; historical data comparison: parameters with large deviations compared with historical normal values or recent average values of the section; parameter type: only quantitative parameters are selected, or parameters that are directly used for subsequent proportional operation are selected; from a section parameter combination, a part of the "to-be-detected section parameters" that are considered to need further key analysis and comparison are screened out, and their specific values and position information are retained.

[0049] Correlation parameter mapping relationship This is a pre-established knowledge base or rule base that defines which "to-be-detected section parameters" are associated with or influence which other "section parameters" (i.e., "correlation parameters"); this association is physical (such as the change in the height of the guide affecting the pull-out value) and logical (such as the difference in the state of the insulator affecting the clarity of the region in the image).

[0050] For each "to-be-detected section parameter" screened out, the system queries this mapping relationship to find out other parameters associated with it; these associated parameters are not "to-be-detected" themselves, but they are crucial for evaluating the abnormality degree or authenticity of the to-be-detected parameters, providing operation objects for the proportional operation in the next step S122, so that the abnormality judgment is not only based on the absolute value of a single parameter, but also considers the relative relationship between parameters, improving the accuracy and robustness of the judgment.

[0051] Further, the system compares the current value of each to-be-detected section parameter with its corresponding preset safety parameter threshold value; this threshold value is set according to railway safety specifications, equipment design standards, and historical experience data; threshold setting examples: contact net pull-out value: the standard usually requires within ±150mm; assuming the preset safety threshold is ±200mm (relatively loose, used for preliminary screening); track gauge in track geometry state: the standard usually requires 1435mm±1mm; assuming the preset safety threshold is 1435mm±2mm; insulator surface temperature: if the insulator temperature is significantly higher than the surrounding environment or adjacent insulators, it indicates an abnormality; assuming the preset safety threshold is 15°C higher than the ambient temperature; result judgment: for each to-be-detected parameter, the system will get a Boolean result (yes / no); if the parameter value exceeds the threshold range, it is marked as "yes", indicating that the parameter has a problem and needs further analysis; otherwise, it is marked as "no", indicating that the parameter is within the safe range.

[0052] Only when a certain to-be-detected parameter is judged to be "out of threshold", the proportional operation of the to-be-detected section parameter and multiple associated parameters will be performed, and for that out-of-threshold to-be-detected parameter, the system will take out the multiple associated parameters found for it in step S121. The "proportional operation" needs to be defined specifically, and it is not necessarily a simple division. According to the application scenario, it has the following forms: simple ratio method: to-be-detected parameter value / associated parameter standard value or to-be-detected parameter value / associated parameter current value; for example, when the pull-out value is out of threshold, it is related to the pillar inclination, and the pull-out value out-of-threshold amount / pillar inclination is calculated.

[0053] Through proportional operation, it is attempted to quantify the "fitting degree" or "amplification effect" between the to-be-detected parameter out of threshold and its associated factors; for example, if the pull-out value is out of threshold by a large amount, but at the same time the pillar inclination is also very serious, which means that the problem is more serious or more complex; on the contrary, if the pull-out value is out of threshold, but the pillar is straight, it is only a local small problem.

[0054] The matching coefficient is a comprehensive index calculated based on the proportional operation result, which is used to measure the "matching degree" or "correlation strength" between the to-be-detected parameter out of threshold and its associated parameters. This coefficient aims to judge whether the out-of-threshold phenomenon is "reasonably" caused by the associated factors or is "abnormally" serious; the calculation method of the matching coefficient needs to be defined in advance, which is: simple average or weighted average: average the multiple proportional operation results calculated for a to-be-detected parameter; give different weights to different associated parameters; maximum value method: take the maximum value of all proportional operation results as the matching coefficient; composite function method: use a more complex function (such as exponential, logarithm) to combine the proportional operation results to calculate the matching coefficient; threshold mapping method: according to the different intervals that the proportional operation result falls into, map to a fixed matching coefficient value; the higher the matching coefficient, the stronger the relationship between the to-be-detected parameter out of threshold and its associated factors, or the more "outstanding" the out-of-threshold phenomenon, the more suspicious it is as a real anomaly.

[0055] The preset matching coefficient threshold is an empirical value or a value based on historical data analysis, which represents the "sufficiently suspicious" matching degree threshold that the system considers; below this threshold, even if the parameter is out of threshold, it is considered to be a false alarm or a slight deviation; above this threshold, it is more indicative of a real anomaly.

[0056] The system compares the calculated matching coefficient with a preset matching coefficient threshold; if the matching coefficient > preset matching coefficient threshold: then determine that the to-be-detected parameter is indeed an "abnormal parameter"; the system will record it and trigger the data acquisition or storage mechanism for subsequent analysis and processing; if the matching coefficient ≤ preset matching coefficient threshold: then determine that the to-be-detected parameter is over-standard, but its relationship with the associated parameter is not enough to indicate a serious anomaly, and it is temporarily not marked as an abnormal parameter; at this time, for the overhead line system parameters (such as the pull-out value), a higher threshold is set because a slight over-standard has little impact, but a serious over-standard accompanied by associated problems such as support inclination must be paid attention to; suppose the threshold is 15; for track geometry parameters (such as track gauge), a lower threshold is set because even a small deviation affects the smoothness of driving; suppose the threshold is 3.

[0057] Therefore, the position corresponding to the abnormal parameter is marked, the corresponding detection section is determined according to the matching of the position corresponding to the abnormal parameter and the distribution map of the railway section, the environment type of the detection section is determined based on the environment detection of the detection section, and the abnormal section is determined based on the values, positions and environment types of the detection section of a plurality of abnormal parameters. The overall consideration of the values, positions and environment types of the detection section of a plurality of abnormal parameters is compatible, and the accuracy of the abnormal section is ensured.

[0058] At this time, in step S122, we have identified some abnormal parameters; each abnormal parameter is collected through online detection in a specific sub-detection section, and its original data (such as images taken by a drone, point cloud scanned by a laser) has accurate geographic coordinates or a milepost number relative to the starting point of the railway; these abnormal parameters are accurately associated with their original collection positions and clearly marked.

[0059] The GPS / RTK positioning data of the unmanned aerial vehicle, the inertial navigation system (INS) data, the odometer data, and the registration results with the image / point cloud data; record the accurate latitude and longitude coordinates and / or milepost number (for example, K123+450, indicating 450 meters from the starting point of the railway 123 kilometers) for each abnormal parameter in the database; also directly visualize the mark on the digitized railway distribution map, such as placing a marker point or highlighting the abnormal area.

[0060] Although we have divided the railway into sub-detection sections in S112, the position of an abnormal parameter is located near the boundary of two sub-detection sections, or we need to associate multiple scattered abnormal parameters to a logically more relevant larger area (i.e. detection section) for unified analysis, this step is to use the accurate position of the abnormal parameter and the railway distribution map (containing the boundary information of the sub-detection section) to determine which detection section or sections these abnormal parameters mainly belong to.

[0061] At this time, the location of the abnormal parameter (milepost or coordinate) is compared with the boundary of the sub-detection section; usually, an abnormal parameter is assigned to the sub-detection section where it is located; if multiple abnormal parameters are close to each other, they are grouped into the same detection section for analysis; scattered abnormal points are gathered into logically related section units, facilitating subsequent environmental analysis and comprehensive judgment.

[0062] The environment of the railway section has an important influence on its state and potential faults; for example, the humidity inside a tunnel is higher, causing equipment rusting; vibration on a bridge is more likely to cause loosening; landslides or landslides are more likely to occur on mountain sections; this step uses the environmental data collected previously (in the online detection in S112) or combines geographic information system (GIS) data to determine the type of environment in which the current detection section is located.

[0063] The unmanned aerial vehicle carries environmental sensors (temperature, humidity, wind speed, air pressure, etc.), image analysis (identifying vegetation coverage, soil type, buildings, etc.), laser radar data (identifying terrain, obstacles), and preloaded GIS data (such as tunnels, bridges, rivers, geological zoning, etc. marked on the map); several typical railway environment types are defined in advance, such as: open plain, hilly area, mountainous area, inside tunnel, on bridge, urban area, near river / lake, etc.; comprehensive analysis of environmental detection data and GIS data classifies the detection section into one or several environment types that best fit.

[0064] Optionally, assuming that in the online detection in S112, the unmanned aerial vehicle scans the environment of the section: image analysis shows that most of the section is a bare soil subgrade with a small amount of shrubs on both sides; laser radar data shows that the section has a gentle terrain with no high obstacles; GIS data marks the area as a plain landscape; based on these information, the system determines that the environment type of detection section 003 is: open plain.

[0065] If the detection section is K100+500 to K101+000, and the section contains a long tunnel: image and laser radar data will show the tunnel entrance and internal features; environmental sensors measure higher humidity and lower temperature inside the tunnel; GIS data clearly marks the location of the tunnel; the system will determine the environment type of the detection section as: inside the tunnel (or a composite environment containing the tunnel entrance).

[0066] The system now has comprehensive information about the abnormality: the specific values of the abnormal parameters (severity), their distribution locations on the section, and the environmental background of the section; based on this information, the system needs to determine whether these abnormalities are sufficiently concentrated or severe enough to label the entire detection section (or part of it) as an "abnormal section" that requires special attention.

[0067] The concentration of the anomaly, the severity of the anomaly, and the influence of the environment are introduced. For the concentration of the anomaly, whether a plurality of anomaly parameters are concentrated in a small range of the detection section or are dispersed throughout the entire section. For the severity of the anomaly, whether the numerical value of the anomaly parameter deviates from the safety threshold is slightly exceeded or seriously exceeded. For the influence of the environment, whether the current environment category will exacerbate the anomaly. Some preset rules are needed to assist in judgment. For example, “if more than 3 different types of parameter anomalies are found in the same detection section, or any key parameter (such as track gauge, catenary height) is seriously exceeded, the detection section is marked as an abnormal section” or “if the image features of rockfall risk are found in a mountainous section, even if there is only one, it should be marked as an abnormal section”, one or more “abnormal sections” are determined, and the mileage range, the main anomaly parameter information contained, the environment category, etc. are recorded.

[0068] Optionally, the detection section: the sub-detection section 003 (K100+000 to K100+500); the anomaly parameters: A: the catenary pull-out value 180mm (position K100+320); B: the track gauge 1437mm (position K100+380); the environment category: open plain.

[0069] The system makes a judgment: the number of anomaly parameters: 2; the severity of the anomaly: the pull-out value exceeds more (180mm>150mm standard, also>200mm threshold setting is conservative), the track gauge exceeds less (1437mm>1435mm standard, but does not trigger the anomaly marking of S122 because the matching coefficient is low); the concentration of the anomaly: K100+320 and K100+380, 60 meters apart, not very concentrated in the 500-meter sub-detection section, but also not too dispersed; the influence of the environment: the open plain environment has little influence on such geometric parameter anomalies; the preset rule: the assumption rule is “if 2 or more parameters marked as abnormal by S122 are found in the same sub-detection section, mark the sub-detection section as an abnormal section”.

[0070] In step S13, the specific steps are:

[0071] S131: collect a plurality of abnormal sections, and determine the section length of the abnormal section by length detection of each abnormal section, and determine the first anomaly parameter according to the section position of the plurality of abnormal sections and the section length of the abnormal section;

[0072] S132: determine the second anomaly parameter according to the section position of the plurality of abnormal sections and the distribution map of the railway section; determine the abnormal level of the plurality of abnormal sections based on the first anomaly parameter, the second anomaly parameter, and the abnormal level mapping relationship;

[0073] S133: Determine the driving state map of the railway section according to the name of the railway section, the railway database and the current time, and mark the driving state of each abnormal section, while collecting the position of the low-altitude unmanned aerial vehicle, determining the fault detection path based on the abnormal level of the multiple abnormal sections, the driving state and the position of the low-altitude unmanned aerial vehicle, at this time, the fault detection path presents the sub-low-altitude flight path of the multiple abnormal sections, and the flight order of each sub-low-altitude flight path is determined according to the sub-low-altitude flight path of the multiple abnormal sections and the driving state of the multiple abnormal sections.

[0074] In the embodiments of the present application, the information of all abnormal sections identified and marked in step S12 (in particular, S123) is obtained; the system needs to extract the identification information of each abnormal section from the previously stored or generated abnormal list, which usually includes the name of the abnormal section, the unique ID, and the accurate geographic coordinate range (start point and end point coordinates) of the abnormal section on the railway section distribution map, which ensures that we have a clear list of all objects that need to be further analyzed.

[0075] For each collected abnormal section, its physical length needs to be accurately calculated, which is determined in the following way: way one (based on coordinates): using the start point and end point geographic coordinates obtained in step S131, using geographic spatial calculation method (such as Haversine formula or more accurate distance calculation under projection coordinate system) to calculate the straight line distance or curve distance along the railway line (if railway line data is available) between two points; for complex curved sections, it needs to be calculated segmentally and then summed up;

[0076] Combine the position information and length information of the abnormal section to generate one or more "first abnormal parameters" for subsequent analysis, the design purpose of this parameter is to quantify some characteristics of the abnormality, which is one dimension of its impact range, severity, or its position importance in the railway network; how to determine the "first abnormal parameter" depends on the design goal, and the following are some definition ways:

[0077] Definition way one: abnormal impact range index; directly use the length of the section itself as the first abnormal parameter; the longer the length, the greater the potential impact; the first abnormal parameter = abnormal section length;

[0078] Definition way two: position weighted length; if the abnormality of some areas (such as near stations, tunnel entrances, bridge sections) is more critical than other areas, different weight coefficients are given to the lengths of sections in different positions; for example, the length of the section near the station * 1.5, the length of the ordinary section * 1.0; the first abnormal parameter = abnormal section length * position weight coefficient.

[0079] The system explicitly identifies the objects to be processed (abnormal sections); secondly, it accurately measures the physical impact range of each abnormality (calculates the length of the section); finally, it combines the length of the section with other factors such as location to generate the "first abnormal parameter", which provides an important quantitative basis for subsequent evaluation of the severity, priority and planning of detection resources for the abnormality.

[0080] Further, a second abnormal parameter is determined according to the section location of the multiple abnormal sections and the distribution map of the railway section; and an abnormal level of the multiple abnormal sections is determined based on the first abnormal parameter, the second abnormal parameter and an abnormal level mapping relationship, which is compatible with the overall consideration of the first abnormal parameter, the second abnormal parameter and the abnormal level mapping relationship, and ensures the accuracy of the abnormal level of the multiple abnormal sections.

[0081] At this time, the spatial location information of the abnormal section in the railway network is used to evaluate the strategic importance or potential impact range of the abnormal section in combination with the understanding of the entire railway distribution map (which usually contains important geographic information such as line grade, station location, hub, bridge and tunnel); the second abnormal parameter aims to quantify this importance or impact, which no longer focuses on the physical properties of the abnormality itself (such as length, represented by the first abnormal parameter), but on the additional risks or values brought by its location in the entire railway system.

[0082] The system analyzes the geographical location of each abnormal section to see if it is close to or crosses important nodes (such as large stations, marshalling yards, hubs), key engineering structures (such as long bridges, tunnels, high fill embankments), or is located in a busy section; it also considers the line grade (main line, branch line) of the section, whether it is close to residential areas or important facilities, etc.; according to the analysis results, a second abnormal parameter value is defined for each abnormal section; the definition method is as follows:

[0083] Based on the proximity of important facilities: a score is defined according to the distance of the abnormal section from the nearest station, hub, bridge, etc. (the closer the distance, the higher the score); based on line grade and location: a score is defined according to the line grade of the abnormal section (main line scores high) and whether it is located in the middle of the section (affects a lot) or at both ends; based on potential impact range: a score is defined considering how large the impact range is if the section fails (for example, whether it will cause the entire section to be shut down).

[0084] The two quantitative indicators (first abnormal parameter and second abnormal parameter) calculated above are combined, and through a pre-set abnormal level mapping relationship (usually a rule base, scoring model or decision tree), a clear abnormal level (for example: first level, second level, third level, or high, medium, low) is finally determined for each abnormal section, which represents the urgency, severity and priority level of the abnormality.

[0085] The abnormality level mapping relationship is predefined, which specifies which abnormality level should be corresponded under different combinations of the first abnormality parameter and the second abnormality parameter; for example, the rule stipulates: “if the first abnormality parameter > 1000 meters and the second abnormality parameter > 8 minutes, then determine as a first-level abnormality; if the first abnormality parameter > 800 meters and the second abnormality parameter > 5 minutes, then determine as a second-level abnormality; otherwise, as a third-level abnormality”; this mapping relationship is adjusted according to actual needs and safety strategies; by comprehensively evaluating the physical scale and strategic position of the abnormality, the abnormality is classified, which provides a clear basis for subsequent resource allocation (such as dispatching unmanned aerial vehicles, arranging maintenance personnel) and decision-making (such as whether to immediately block the line).

[0086] Therefore, the driving state diagram of the railway section is determined according to the name of the railway section, the railway database and the current time, and the driving state of each abnormal section is marked, at the same time, the position of the low-altitude unmanned aerial vehicle is collected, the fault detection path is determined based on the abnormality level of the multiple abnormal sections, the driving state and the position of the low-altitude unmanned aerial vehicle, at this time, the fault detection path presents the sub-low-altitude flight path of the multiple abnormal sections, and the flight order of each sub-low-altitude flight path is determined according to the sub-low-altitude flight path of the multiple abnormal sections and the driving state of the multiple abnormal sections, which is compatible with the overall consideration of the sub-low-altitude flight path of the multiple abnormal sections and the driving state of the multiple abnormal sections, ensures the accuracy of the flight order of each sub-low-altitude flight path, at the same time, multiple abnormal parameters are introduced, and the abnormal section is controlled, which is compatible with the overall consideration of the abnormality level of the multiple abnormal sections, the driving state diagram of the railway section and the position of the low-altitude unmanned aerial vehicle, and improves the detection accuracy of the fault detection path.

[0087] At this time, the actual operation of the railway is obtained; the system needs to know which sections of the railway line are idle (for detection) and which sections have trains passing or will pass (cannot be detected) at the time when unmanned aerial vehicle detection is planned, which needs to query the railway database containing train timetable, current train position, line occupation situation and other information, and combine the current accurate time to generate a dynamic “driving state diagram”, this diagram is a data structure, which marks the state of each section at the query time; at the same time, the system inputs: railway section name list, current time (accurate to seconds); the system queries: railway operation database (contains train timetable, real-time tracking data, dispatching instructions, etc.); the system outputs: a “driving state diagram”, for example, a dictionary or database table, the format is {section ID: state, …}, where the state is “idle”, “occupied (XX trains, estimated passing time YY:ZZ)”, “under maintenance (not allowed to pass)” and the like.

[0088] The "travel state map" generated in the previous step is matched with the previously identified "abnormal section" list, and each abnormal section is labeled as currently detectable or undetectable, which directly relates to which abnormal sections are accessed within the current time window when planning the flight path subsequently; meanwhile, the system inputs: abnormal section list (including section ID / position information), "travel state map" generated in the previous step; the system processes: traverses each abnormal section, and finds the corresponding state in the "travel state map" according to the section ID / position information; the system outputs: updates the abnormal section information, and adds a "current travel state" field.

[0089] The current position of the UAV is the starting point of planning the flight path; the system needs to know the specific coordinates of the UAV at the beginning of planning the path in real time or quasi-real time, so as to calculate the flight distance and time from the current position to each target abnormal section; the system obtains the real-time latitude and longitude coordinates of the UAV through the GPS or other positioning system integrated on the UAV, and this data is usually transmitted back to the ground control station or cloud platform through a wireless link.

[0090] Based on the abnormal levels, travel states of multiple abnormal sections, and the position of the low-altitude UAV, a fault detection path is determined, at this time, the abnormal sections with high levels are preferentially accessed; only the abnormal sections with the state of "idle" can be accessed (the sections with the states of "occupied" or "under repair" are temporarily excluded from the current planning); the path planning needs to start from the current position of the UAV and consider the feasibility of returning to the base or the next take-off and landing point; the goal of the system is to generate a total "fault detection path" including a series of "sub-low-altitude flight paths"; each sub-path corresponds to an abnormal section to be detected.

[0091] For path planning of the "fault detection path", graph theory (such as Dijkstra method) or heuristic method (such as genetic method, ant colony method) is used to solve; a graph is constructed, the nodes include the current position of the UAV, the start / endpoint of all accessible (state "idle") abnormal sections, the base / take-off and landing point; the edges represent the flyable paths, and the weights are the flight distance, the estimated flight time, or the comprehensive cost considering fuel / electricity; the goal of the method is to find a path that covers all abnormal sections with high priority and accessibility, while satisfying the minimization of total cost (time, distance, etc.) or the maximization of priority.

[0092] Optionally, assuming the current time is 10:10 AM, we only consider the abnormal sections A and C in the state of "free"; abnormal level: A is level 2, C is level 3; UAV position: 116.3912, 39.9087; path 1: UAV position > abnormal section A > abnormal section C > return to base; path 2: UAV position > abnormal section C > abnormal section A > return to base; since the level of abnormal section A (level 2) is higher than that of abnormal section C (level 3), the algorithm will tend to choose path 1, i.e., first fly to the abnormal section A with a high level; the fault detection path is: UAV position > sub-low-altitude flight path 1 (covering A) > sub-low-altitude flight path 2 (covering C) > return to base; at this time, sub-low-altitude flight paths 1 and 2 have been determined.

[0093] The system rechecks the current latest driving state (there will be a short change), and finally determines the execution order of each sub-low-altitude flight path in the total path according to the abnormal level (high level first), distance / time cost (when the levels are the same, the closer / less time-consuming one is preferred), and driving state (ensure that the target abnormal section is indeed in the "free" state when executing the sub-path).

[0094] Optionally, path 1 is UAV position > A > C > return to base; during the UAV flying to A, the state of A remains "free"; after the UAV completes the detection of A and flies to C, the state of C remains "free" (for example, check whether the G1234 train has passed B); the final flight order is to execute sub-low-altitude flight path 1 (detect A) first, and then execute sub-low-altitude flight path 2 (detect C).

[0095] In step S14, the specific steps are:

[0096] S141: The low-altitude UAV obtains the fault detection path and flies along the fault detection path, and the low-altitude UAV detects each abnormal section in real time during the low-altitude flight, at this time, the low-altitude UAV detects each abnormal section;

[0097] S142: In the abnormal detection of each abnormal section, the first camera of the UAV collects the surface image of the abnormal section, determines the surface defect of the abnormal section according to the detection of the surface image of the abnormal section, and determines the surface abnormal image according to the synthesis of the defect area and the defect position of each surface defect;

[0098] S143: The second camera of the UAV is arranged on one side of the first camera and dynamically captures the installation position of the abnormal section, determines the installation defect of the abnormal section according to the shooting of the installation position of the abnormal section, and determines the installation abnormal image according to the synthesis of the defect area and the defect position of each installation defect.

[0099] In the embodiments of the present application, before or shortly after the UAV takes off, it needs to receive the "fault detection path" planned by step S13 from the ground control station, a preset database, or through wireless communication. This path usually exists in the form of a series of precise geographic coordinate points (waypoints), each containing location (latitude, longitude, altitude), flight height, flight speed, hovering time, etc. The flight control system on the UAV receives and analyzes these path data, stores them in memory as the basis for subsequent autonomous flight. The process of obtaining the path needs to ensure the integrity and accuracy of the data.

[0100] After obtaining the path, the UAV starts its flight control system and begins autonomous flight according to the stored path instructions. The flight control system automatically controls the attitude (pitch, roll, yaw) and power (throttle) of the UAV based on the current GPS position, target waypoint position, set flight height and speed, guiding the UAV to fly precisely from one waypoint to the next. The entire flight process is maintained at a predetermined low altitude (e.g., 5-15 meters from the track surface) to ensure that the camera can obtain clear and detailed ground images, while avoiding obstacles and complying with airspace management regulations. The flight control system continuously monitors the flight state and makes fine adjustments as necessary to respond to environmental disturbances such as wind.

[0101] While performing low-altitude flight, the sensor system on the UAV (mainly the camera, but also including laser radar, infrared sensors, etc., but according to the context, here mainly refers to the camera) is activated to continuously or on-demand detect the railway section below. When the flight path passes through a pre-marked "abnormal section", the system will automatically trigger detailed detection of the current section below according to the instructions in the path planning (such as hovering after reaching a certain waypoint, adjusting the camera angle, etc.) or through real-time image recognition technology. This process involves using the camera to capture high-resolution images or videos, which are stored or analyzed in real-time for subsequent analysis (such as surface and installation defect identification in S142 and S143). The main task of the UAV itself at this step is to perform physical proximity and observation, collecting raw data.

[0102] Further, when the UAV flies over a marked abnormal section (e.g., K101153 section identified by GPS coordinates or milepost number) according to the low-altitude flight path planned in S13, the system automatically triggers the first camera (usually a high-resolution, zoom-capable visible light camera) to start working.

[0103] The UAV needs to maintain a stable, level flight attitude, flying at a preset constant height (e.g., 5-10 meters above the track surface) and speed (e.g., 5-8 kilometers per hour); the first camera will shoot vertically downward (Nadir) or at a slightly tilted angle, ensuring that the image covers the entire abnormal area and a certain range around it; the shooting frequency and resolution will be dynamically adjusted according to the length and importance of the abnormal section, ensuring that no details are missed; for example, for shorter and more serious abnormalities, the flight speed will be reduced and the shooting resolution will be increased.

[0104] An overlapping shooting strategy is adopted to ensure that there is 30%-50% overlap between adjacent images, which is crucial for subsequent image stitching and three-dimensional reconstruction (if needed); at the same time, the exact shooting time, GPS coordinates, camera pose (pitch, yaw, roll angle), and other metadata of each image are recorded to accurately locate the defects; the system automatically detects the ambient lighting conditions; if the lighting is insufficient or there are strong shadows, it will automatically switch to the infrared camera (if equipped) to assist in judgment, or adjust the flight time to avoid unfavorable lighting; at the same time, it will try to avoid low-altitude hovering in strong winds to ensure image clarity.

[0105] The raw images collected will first be pre-processed, including denoising (such as using Gaussian filtering), contrast enhancement (such as histogram equalization), color correction (if needed), etc., to improve the accuracy of subsequent detection; usually relying on advanced computer vision and machine learning algorithms; the system will load a pre-trained deep learning model (such as a CNN-based image segmentation or object detection model) for railway surface defects (such as rail corrugation, peeling, side wear, fish scales, ballast disease, sleeper damage, etc.).

[0106] The pre-processed images are input into the model; the model automatically scans the images and identifies areas that match the pre-set feature patterns; for example, for rail corrugation, the model will look for wavy textures with specific wavelengths and amplitudes; for peeling, it will look for block-shaped areas with clear edges and different colors / textures from the surrounding; each detected area will be classified into a specific defect type (such as "rail corrugation" or "sleeper crack") and marked on the image with different colored or shaped bounding boxes (Bounding Boxes) or segmentation masks (Segmentation Masks), while outputting a confidence score (indicating how confident the model is that this is the defect of that type); for defects with low confidence or complexity, the system will mark them for review and confirmation by human engineers, improving the accuracy of detection.

[0107] Therefore, the second camera of the unmanned aerial vehicle is arranged on one side of the first camera, and dynamically captures the installation position of the abnormal section. The installation defects of the abnormal section are determined according to the shooting of the installation position of the abnormal section. The installation abnormal image is determined according to the synthesis of the defect area and the defect position of each installation defect. The overall consideration of the synthesis of the defect area and the defect position of each installation defect is compatible, and the accuracy of the installation abnormal image is ensured.

[0108] At this time, in order to realize multi-angle and multi-aspect observation, the unmanned aerial vehicle usually carries multiple cameras; the second camera is physically installed beside the first camera, but usually has different viewing angles or focal lengths. This layout design enables the second camera to capture different pictures from the first camera (which is usually mainly aligned with the track surface), and is particularly suitable for observing the support and fixing system of the track, that is, the so-called "installation position". For example, the second camera is arranged to be slightly inclined, or at a fixed angle with the first camera, so as to better observe the connection of the track and the sleeper, the state of the fastener, the side surface of the sleeper, and the contact between the sleeper and the ballast.

[0109] When the unmanned aerial vehicle flies along the preset fault detection path and passes through the abnormal section previously marked by the system, the second camera will automatically start or adjust the angle, and begin to continuously shoot the installation position of this section. "Dynamic capture" here means that the camera will adjust the focal length, angle or shooting frequency as the unmanned aerial vehicle flies, to ensure that the details of the installation structure can be clearly recorded. It focuses on the "skeleton" and "joints" of the track, that is, the parts that support and fix the track.

[0110] The unmanned aerial vehicle control system analyzes the installation position image captured by the second camera, which usually involves image processing and pattern recognition technology; the system will look for predefined defect patterns, such as by analyzing the texture changes, color abnormalities, geometric shape deviations (such as deformation of fastener holes, direction of sleeper cracks) in the image to determine whether there are installation defects; for example, the system will set a rule: if the image brightness or shape of a certain fastener in the image differs from the surrounding normal fasteners by more than a threshold value, or detects linear or network structural changes on the surface of the sleeper, it is determined that there is an installation defect.

[0111] After determining the installation defects, the system needs to intuitively superimpose these defect information onto the original installation position image to generate a new "installation abnormal image", which includes accurately marking the position of each defect on the image (such as using rectangular boxes, circles or arrow marks), and labeling the type of defect with text; at the same time, the system will also record or display the area size affected by the defect (such as the length, width of the crack, or the range of the loose area), which is the process of "synthesizing" the "area" and "position" information of the defect onto the original image.

[0112] Optionally, assuming that the first camera (Camera A) is a wide-angle camera mainly aimed at the track center line downward to shoot the track surface, used to observe the surface defects such as cracks and peeling of the track surface; then, the second camera (Camera B) is installed on the left side of Camera A and slightly tilted downward to the left, or at a small angle (such as 15 degrees) with Camera A, so that when the UAV flies over the track, Camera A mainly shoots the installation position; when the UAV flies to the previously marked abnormal section (for example, K101150 to K101155), the second camera (B) will continuously shoot the fastener system of this section of track, which will capture the details such as whether the tie side, fastener is loose, the base plate is displaced, the rail spike is missing or broken, etc.; the image includes a close-up of the tie and rail joint, as well as the ballast state around the tie.

[0113] For the images of the K101150 to K101155 section shot by Camera B, the system analyzes; at K101152, the image analysis algorithm detects that the image of the screw cap of a rail spike is darker than other screw caps, or the edge is blurred, combined with the clarity information of the image, it is judged that the rail spike is loose; at K101153, the algorithm detects that there is a long and slightly dark linear structure in the image of the tie side, which is different from the normal texture, and it is judged that it is a tie crack, so the system determines the specific installation defects: the rail spike loosening at K101152 and the tie crack at K101153; for the two installation defects identified before (the rail spike loosening at K101152 and the tie crack at K101153), the system will generate an installation anomaly image on the original image shot by Camera B; at K101152, a rectangular frame will be superimposed on the image to enclose the loose screw cap, and a label "Rail Spike Loosening K101152" will be attached; at K101153, a marker (such as an arrow or a curve) will be superimposed on the image to point to the position of the tie crack, and a label "Tie Crack K101153" will be attached; the final generated image with markers is the "installation anomaly image", which not only contains the original appearance of the installation structure, but also clearly points out the specific installation defects and their accurate positions, providing clear visual reference and positioning information for subsequent maintenance personnel.

[0114] In step S15, the specific steps are:

[0115] S151: determining the past abnormal events of the abnormal section based on the abnormal tracing of each abnormal section, and determining the surface abnormal features based on the identification of the surface abnormal image, determining the installation abnormal features based on the identification of the installation abnormal image, and determining the fault content corresponding to the abnormal section according to the matching of the surface abnormal features, the installation abnormal features and the past abnormal events of the abnormal section;

[0116] S152: Collect a running schedule of the railway section, and determine a running state of each abnormal section according to the running schedule of the railway section, a current time and a current position of the railway train;

[0117] S153: Determine a repair sequence and a repair time of each abnormal section based on the running state of each abnormal section and the fault content of the plurality of abnormal sections, and determine a real-time repair plan of each abnormal section according to the repair sequence, the repair time of each abnormal section and the repair measure corresponding to each abnormal section.

[0118] In the embodiment of the present application, the past abnormal events of the abnormal section are determined based on the abnormal tracing of each abnormal section, the surface abnormal features are determined based on the recognition of the surface abnormal image, the installation abnormal features are determined based on the recognition of the installation abnormal image, and the fault content corresponding to the abnormal section is determined according to the matching of the surface abnormal features, the installation abnormal features and the past abnormal events of the abnormal section, which is compatible with the overall consideration of the matching of the surface abnormal features, the installation abnormal features and the past abnormal events of the abnormal section, and ensures the accuracy of the fault content corresponding to the abnormal section.

[0119] At this time, the system will access a historical database or a maintenance record system, which stores all detection reports, maintenance records, accident records and the like of the railway section in the past; for each abnormal section (for example, a section with a track mileage mark of K101153) detected at present, the system will query all events recorded in the past period (for several months, one year or longer, depending on the design of the system) of the section.

[0120] These “past abnormal events” include: defect type: such as rail crack, sleeper damage, ballast settlement, fastener loosening, signal device failure and the like; occurrence time: the date when the event was first recorded or found; treatment measure: the repair or treatment method taken at that time, such as replacing sleepers, grinding rails, filling ballast, tightening bolts and the like; treatment result: whether the repair is effective or not, whether it recurs or not; frequency: the frequency of the same type or similar type of event occurring in the section; understanding the historical “health status” of the section, judging whether the current abnormality is “new” or “recurrence”, and whether there is some “inertia problem” or not.

[0121] The surface anomaly image is introduced, and the system will use image recognition algorithms (which are based on deep learning models) to analyze this image; the algorithm will try to identify specific patterns, shapes, colors, or textures in the image and classify them into known defect types; "surface anomaly features" usually include: defect type: as mentioned earlier, rail corrugation, peeling, scratches, rust, rail head crushing, etc.; defect location: coordinates in the image, mapping to the exact location on the actual track (e.g., how many meters from the starting point, on which side of the rail); defect size: length, width, depth (if it can be estimated through multiple images or specific algorithms); defect severity: based on size, shape, or comparison with standard specifications, a preliminary assessment of its severity (such as slight, moderate, severe).

[0122] The installation anomaly image is introduced, and image recognition or sensor data analysis algorithms are used; for installation anomalies, it is necessary to detect whether the components are missing, loose, misaligned, rusty, damaged, etc.; "installation anomaly features" include: anomaly type: loose or missing fasteners, cracked or displaced sleepers, uneven or lost ballast, expanded or reduced track gauge, damaged insulating joints, etc.; anomaly location: specific location on the track structure; anomaly state: such as "severe looseness", "slight rust", "complete absence", etc.

[0123] Obtain anomaly information related to the installation of track structures and their associated equipment, which is associated with surface defects (for example, loose fasteners cause rail position deviation, which in turn causes uneven wear).

[0124] The system will use a pre-set rule base or machine learning model to "match"; for example, whether the currently detected "rail corrugation" and "loose fasteners" are related to "loose fasteners (February 2025)" in the historical record, and whether loose fasteners are the cause of the aggravation of this corrugation; whether similar problems (such as corrugation and fastener problems) have repeatedly occurred on this section of track in the past, which is related to the design, materials, train load, or environmental factors of this section of track; the system infers the causal chain based on the railway engineering knowledge base; for example, "loose fasteners > unstable rail position > uneven stress when train passes > accelerate rail corrugation"; "fault content" is a more in-depth and specific diagnostic result that describes the root cause or main feature combination of the anomaly, which is a specific fault name or a descriptive phrase, and it is a valuable fault diagnosis conclusion that is extracted from multiple dimensions of information for maintenance decision-making.

[0125] Optionally, assuming the UAV detects an anomaly on track K101153 section; the system traces back and finds: November 2024: records of "mild ballast settlement", which has been "partially filled"; February 2025: records of "loose fasteners (left rail)", which has been "tightened"; April 2025: records of "rail surface scratches", which has been "polished and repaired"; these historical records show that this section has experienced a variety of minor problems recently, but all have been addressed.

[0126] For the K101153 section, the system analyzes the surface anomaly image generated by S142 and identifies: defect type: rail corrugation (a periodic rail surface irregularity); defect location: 4.5 meters from the start of the section, on the descending line (assuming); defect size: wavelength about 30 cm, wave depth about 0.3 mm; defect severity: moderate (judged according to preset severity threshold, e.g. wave depth over 0.2 mm is moderate).

[0127] The UAV also takes an installation state image at the K101153 section; after system analysis, it identifies: anomaly type: loose fasteners; anomaly location: also at 4.5 meters from the start, corresponding to the location of the rail corrugation above, the fasteners (e.g. a certain type of spring bar) below are loose; anomaly state: severe loosening (spring bar has lost pre-tightening force, clearly loose).

[0128] The system matches and analyzes the information of the K101153 section: surface anomaly: rail corrugation (moderate); installation anomaly: severe loosening of fasteners; historical events: recent records of loose fasteners and rail scratches, and ballast settlement.

[0129] The current fastener loosening is directly related to the historical record of fastener problems; the loosening of fasteners leads to unstable rail position, which in turn causes or exacerbates corrugation; although there has been ballast settlement in history, this detection does not show obvious settlement, so the main reason for the corrugation is more likely to be installation problems; the determined fault content is: "uneven rail wear (corrugation) caused by loose fasteners", this conclusion is more instructive than simply saying "there is corrugation" or "there is a fastener problem", it indicates the priority direction of maintenance - first the fastener loosening problem needs to be addressed, then the rail needs to be polished.

[0130] Further, the operation plan table of the railway section is collected, and the running state of each abnormal section is determined according to the operation plan table of the railway section, the current time and the current position of the railway train, which is compatible with the overall consideration of the operation plan table of the railway section, the current time and the current position of the railway train, ensuring the accuracy of the running state of each abnormal section.

[0131] At this point, the system needs to access the railway dispatch center or operation management system to obtain the latest train operation diagram (timetable), which specifies in detail which trains will run on which lines at what times during a specific time period; the operation schedule usually includes: train number: such as G1234 (high-speed train), K567 (ordinary passenger train), and cargo 123 (freight train); running line: the name or number of the line the train will pass through; passing stations: stations where the train is scheduled to stop; arrival and departure times: planned times at each station or specific location; speed limit: speed regulations that exist in certain sections; the operation schedule needs to be real-time or near real-time, as train times can change due to weather, scheduling adjustments, etc.

[0132] The system needs to obtain the current accurate time, which is usually from a Network Time Protocol (NTP) server or a system's built-in real-time clock, and synchronize it to ensure accuracy; the current time is the basis for determining the status of the train plan; by comparing it with the time points in the operation schedule, it can determine whether the planned events (such as train arrival, passing) are about to occur, are currently in progress, or have already passed.

[0133] The system needs to access the train tracking system (such as GPS-based, beacon-based, or communication-based positioning system), which can provide real-time updated train location information; for each train in operation, the system receives its current precise geographic location (e.g., kilometer marker from the starting point, such as K101200) and running status (e.g., in motion, in stop); the current location information of the train is crucial for determining the real-time occupancy of the line, as it reveals whether the train is currently located on an abnormal section or is about to enter that section.

[0134] The system now has three pieces of information: the operation schedule, the current time (14:25), and the current location of each train; for each abnormal section (e.g., K101153), the system makes the following judgments: check if the current location of any train is located on the K101153 section; if so, the current status of the section is "occupied" or "train passing"; check the operation schedule to see which trains are scheduled to pass through K101153 shortly after the current time (e.g., within the next 5-10 minutes); if the schedule shows that a train is about to enter, the status of the section is "about to be occupied" or "approaching passing".

[0135] If there is no train currently occupying the section, no train is going to enter, and the plan does not show any train passing through the section in a short time, the section status is "free" or "detectable / repairable"; if the abnormal section is located in a station, it also needs to consider whether there is a train stopping at the station; for example, K567 is stopping at K101000 station, if K101153 contains part of the station, the status is "occupied".

[0136] By integrating the static operation plan and dynamic train position information, a real-time operation status label (such as "free", "occupied", "soon to be occupied") is provided for each abnormal section, which is crucial and directly determines whether further detection (such as detailed image acquisition in S14) or repair work on the section is carried out at the current time; for example, if the system judges that K101153 is currently "free", then the UAV or repair team can safely enter the area to work; if it is judged as "occupied" or "soon to be occupied", it needs to wait until the train passes before operating, or adjust the work plan, which ensures that detection and repair activities do not interfere with normal railway operation, and also ensures the safety of the operating personnel.

[0137] Therefore, based on the operation status of each abnormal section and the fault content of multiple abnormal sections, the repair sequence and repair time of each abnormal section are determined, and according to the repair sequence, repair time and repair measures corresponding to each abnormal section, the real-time repair plan of each abnormal section is determined, which is compatible with the overall consideration of the repair sequence, repair time and repair measures corresponding to each abnormal section, ensuring the accuracy of the real-time repair plan of each abnormal section, and at the same time, the surface anomaly image and installation anomaly image are introduced, ensuring the multi-dimensional anomaly detection of the abnormal section by the UAV, realizing the overall consideration of the operation status of each abnormal section and the fault content of multiple abnormal sections, and improving the accuracy of the real-time repair plan of each abnormal section.

[0138] At this time, the operation status of each abnormal section and the fault content of multiple abnormal sections are introduced, the operation status of each abnormal section: this is the result of S152, which tells us which sections are currently free, which are occupied, and which are soon to be occupied; for example, abnormal section A is currently free, abnormal section B is occupied, and abnormal section C is soon to be occupied; the fault content of multiple abnormal sections: this is the result of S151, which tells us what the problem is for each abnormal section; for example, abnormal section A is rail wear, abnormal section B is switch point wear, and abnormal section C is fastener loosening.

[0139] Determining the repair sequence is usually a prioritization process; the system will decide which to repair first and which to repair later according to the following factors: at this time, fault severity / emergency: which fault, if not handled in time, will lead to more serious consequences (such as the risk of derailment, traffic interruption)? For example, turnout point rail wear (B) is usually more urgent than fastener loosening (C) because it directly affects the safety of train turning; rail corrugation (A) affects comfort and track life, but the urgency is between the two.

[0140] Running status: which section is currently or will be idle, taking advantage of the available "window" time (i.e. the period of line interruption for maintenance)? For example, if A is currently idle, while B and C need to wait, the priority of A will be increased;

[0141] Resource availability: is there a specific maintenance equipment or professional only suitable for a certain type of fault? For example, a grinding car is only suitable for handling rail corrugation (A);

[0142] Influence range: the type and number of trains affected by the fault; for example, the priority of the fault affecting high-speed trains is higher;

[0143] Once the repair sequence is determined, it needs to be combined with the running status to specifically arrange the repair time; for the section with high priority, if it is currently idle, repair immediately; if it is not currently idle, it needs to be predicted according to the train operation plan when the next available "window" time window; for example, if section B is occupied, it needs to be checked when the train passes, and then when the line will be idle again; the repair time also needs to consider the time required for the maintenance work itself; for example, grinding the rail (A) takes 1 hour, and replacing the fastener (C) takes 30 minutes, a preliminary repair sequence (such as A>C>B) and a preliminary repair time arrangement (such as A: start immediately; C: after the train passes at 3 pm; B: the window time from 2 am to 4 am tomorrow) are generated for each abnormal section.

[0144] The repair sequence of each abnormal section, the repair time, and the repair measures corresponding to each abnormal section are introduced, the repair sequence and the repair time are the results of S153, which tells us which to repair first, which to repair later, and approximately when to repair.

[0145] The repair measures corresponding to each abnormal section: according to the fault content determined by S151 and the railway maintenance specifications, it is clear what specific maintenance method needs to be taken for each fault; for example, rail corrugation requires a track grinding car; turnout point rail wear requires replacing the point rail or professional grinding; fastener loosening requires torque wrench tightening or replacing new fasteners.

[0146] Convert the previous decision into a detailed, executable action plan, at this time, further refine the preliminary maintenance time, specify the exact starting and ending time points, taking into account the operation preparation and cleaning time; according to the maintenance measures, specify which personnel, equipment, materials and tools are needed; for example, handling C section needs fastener maintenance team, wrench, spare fasteners; handling A section needs grinding car and its operators; handling B section needs turnout maintenance experts, replacement equipment, etc. Develop specific operation procedures and safety procedures for the maintenance work of each section; for example, fastener maintenance: set up protection > check > tighten / replace > check torque > remove protection; If multiple maintenance tasks need to be performed simultaneously or sequentially, the transfer time of personnel and equipment between different locations needs to be considered; for example, after completing section C, how does the fastener team quickly move to section A; integrate all this information into a clear maintenance schedule and distribute it to the relevant maintenance teams and dispatch center, output a detailed real-time maintenance plan, including each abnormal section: priority, specific maintenance time (start and end), required personnel, required equipment, required materials, specific operation steps, safety precautions, etc.

[0147] By combining the fault diagnosis results, real-time line state and maintenance resource requirements, the priority and schedule of maintenance work are scientifically arranged, and detailed execution plans are developed, which not only ensures that maintenance work can be carried out in a timely and effective manner, maximizes the impact of faults on railway transportation, but also improves the utilization efficiency of maintenance resources and ensures the safety of operation; for example, through this step, the railway department knows exactly "who" is going to "where" at "what time" with what "method" to fix "what problem", so that maintenance work can be carried out in an orderly manner.

[0148] Optionally, abnormal sections: K101153 (A: rail corrugation), K101200 (B: switch rail wear), K101300 (C: fastener loosening); running state (S152 result): A: idle; B: occupied (freight train is passing); C: about to be occupied (K567 train is about to enter); fault content (S151 result): A: rail corrugation, affecting comfort, medium urgency; B: switch rail wear, affecting turning safety, high urgency; C: fastener loosening, causing track deformation, high urgency.

[0149] Assess urgency: B (switch) and C (fastener) are more urgent than A (worn rail); combine operating status: A section is currently free; although the urgency is medium, it should be handled immediately; estimated time for grinding is 1 hour; B section is occupied; very urgent, but must wait for the passing of the freight train; estimated time for the passing of the freight train is about 30 minutes; 30 minutes is not enough to complete the switch repair (which takes several hours), but preliminary inspection or temporary reinforcement can be done; main repair needs to be scheduled at the next long window time; C section is about to be occupied; very urgent, needs to wait for K567 to pass; assume that the line is free for 1 hour after K567 passes; 1 hour is enough to handle the fastener looseness (replace or tighten); repair sequence: C > A > B (handle the immediately handleable urgent issue C first, then the immediately handleable A, and finally the must-wait and complex repair B); repair time: C: start immediately after waiting for K567 to pass, estimated time is 30-60 minutes; A: start immediately after C is handled, use the remaining free time or schedule later, estimated time is 1 hour; B: schedule at the next long window time (e.g., 1:00-5:00 am) during the night or early morning of the same day for thorough repair.

[0150] Repair sequence and time (S153-1 result): C (immediately after K567 passes, about 30-60 minutes), A (immediately after C is completed, about 1 hour), B (1:00-5:00 am next morning); repair measures: C (fastener looseness): tighten or replace the fastener, need wrench, torque detector, spare fasteners; A (steel rail wear): track grinding, need track grinding car and its operation and maintenance personnel; B (switch point rail wear): replace the point rail, need switch repair team, lifting equipment, new point rail, switch adjustment tools.

[0151] Determine real-time repair plan:

[0152] C section (K101300): estimated time for K567 to pass is 14:50; repair time: 14:55-15:25 (assume 30 minutes are needed); required resources: fastener repair team (2 people), wrench, torque detector, spare fasteners; operation steps: 14:50 confirm that the train has passed > 14:55 set up operation protection > check loose fasteners > tighten or replace > check torque > 15:25 remove protection > restore normal line;

[0153] A section (K101153): repair time: 15:25-16:25 (immediately after C section, use the remaining free time); required resources: track grinding car 1, grinding car operation and maintenance personnel; operation steps: 15:20 grinding car arrives and prepares > 15:25 sets up operation protection > grinding car starts operation > 16:15 grinding is completed > 16:25 removes protection > restores normal line;

[0154] B section (K101200): Maintenance time: next day 1:00-5:00 (long window time); Required resources: turnout maintenance team (5 people), lifting equipment, new point rail, turnout adjustment tool, safety protection personnel; Operation steps: personnel and equipment arrive nearby on standby before 23:00 > enter the operation area and prepare at 0:50 > set up operation protection at 1:00 > remove old point rail > install new point rail > adjust turnout geometry > check and confirm multiple times > remove protection at 4:50 > restore normal line at 5:00.

[0155] Generated real-time maintenance plan (abstract):

[0156] Task 1: K101300 fastener maintenance: Time: today 14:55-15:25; Personnel: AA, BB; Equipment: wrench, torque instrument; Materials: 8 sets of M24 high-strength bolts; Notes: must evacuate before train passes, pay attention to adjacent line trains during operation;

[0157] Task 2: K101153 track grinding: Time: today 15:25-16:25; Personnel: grinding car team; Equipment: track grinding car No. 01; Notes: set grinding parameters according to class A grinding, check rail temperature before and after operation;

[0158] Task 3: K101200 point rail replacement: Time: tomorrow 1:00-5:00; Personnel: turnout maintenance team; Equipment: lifting equipment, turnout adjustment tool; Materials: 1 set of P60-1 / 12 movable point frog switch new point rail; Notes: all preparation work must be completed before the window, the replacement process must strictly follow the technical regulations, and the geometry size must be measured multiple times after replacement to ensure compliance.

[0159] Please refer to Figure 2 , Figure 2 is a structure composition diagram of the low-altitude unmanned aerial vehicle fault detection system for the railway section in the embodiment of the application; the low-altitude unmanned aerial vehicle fault detection system for the railway section comprises:

[0160] A section parameter combination module 21 is configured to determine a plurality of sub-detection sections according to a distribution map of the railway section, and determine a section parameter combination of each detection section according to online detection of the plurality of sub-detection sections;

[0161] An abnormal section module 22 is configured to determine a plurality of abnormal parameters based on traversal of the section parameter combination in each section parameter combination, and determine an abnormal section according to positions of the plurality of abnormal parameters and an environment type of the detection section;

[0162] The fault detection path module 23 is configured to determine abnormality levels of the abnormal sections based on the abnormal sections and the distribution map of the railway sections, and determine the fault detection path according to the abnormality levels of the abnormal sections, the driving state map of the railway sections and the position of the low-altitude UAV;

[0163] The abnormal image module 24 is configured to make abnormality detection on each abnormal section when the low-altitude UAV flies along the fault detection path, and determine the surface abnormal image and the installation abnormal image according to the abnormality detection of the abnormal sections.

[0164] The real-time maintenance plan module 25 is configured to determine the fault content corresponding to each abnormal section based on the surface abnormal image, the installation abnormal image and the past abnormal events of the abnormal sections, and determine the real-time maintenance plan of each abnormal section according to the driving state of each abnormal section and the fault content of the abnormal sections.

[0165] Any combination of the technical features of the above embodiments is possible. In order to make the description simple, all the combinations of the technical features in the above embodiments are not described, however, as long as the combinations of the technical features do not exist, they should be considered as the scope of the present disclosure.

Claims

1. A method for detecting faults in a railway track section by a low-altitude unmanned aerial vehicle, characterized in that, The method comprises the following steps: determining a plurality of sub-detection sections according to a distribution map of the railway section, determining a section parameter combination of each detection section according to online detection of the plurality of sub-detection sections; in each section parameter combination, determining a plurality of abnormal parameters based on traversal of the section parameter combination, determining an abnormal section according to the position of the plurality of abnormal parameters and the environment type of the detection section; determining an abnormal level of the plurality of abnormal sections based on the plurality of abnormal sections and the distribution map of the railway section, determining a fault detection path according to the abnormal level of the plurality of abnormal sections, the driving state map of the railway section and the position of the low-altitude unmanned aerial vehicle; the low-altitude unmanned aerial vehicle flies along the fault detection path, performs abnormal detection on each abnormal section, and determines a surface abnormal image and an installation abnormal image according to the abnormal detection of the abnormal section; determining the fault content corresponding to the abnormal section based on the surface abnormal image, the installation abnormal image and the previous abnormal event of the abnormal section, and determining a real-time maintenance plan of each abnormal section according to the operating state of each abnormal section and the fault content of the plurality of abnormal sections.

2. The low-altitude drone method for detecting faults in a railway section according to claim 1, characterized in that, The method comprises the following steps: collecting the name of the railway section, determining the distribution map of the railway section according to the name of the railway section and the railway database, triggering the initial detection of the low-altitude unmanned aerial vehicle according to the distribution map of the railway section, and collecting a plurality of section boundary points of the railway section based on the initial detection of the unmanned aerial vehicle; determining a plurality of sub-detection sections based on the plurality of section boundary points and the distribution map of the railway section, and performing online detection on each sub-detection section; collecting a plurality of section parameters based on the online detection of each sub-detection section, marking the plurality of section parameters corresponding to the section detection position, and determining the section parameter combination of each detection section according to the plurality of section parameters and the corresponding section detection position.

3. The low altitude unmanned aerial vehicle based method for detection of faults in railway track sections as claimed in claim 1 wherein, The method comprises the following steps: real-time monitoring each section parameter combination, traversing each section parameter combination, determining a plurality of to-be-detected section parameters according to the traversal of each section parameter combination, and determining a plurality of associated parameters according to the plurality of to-be-detected section parameters and the associated parameter mapping relationship; in the plurality of to-be-detected section parameters, if the to-be-detected section parameter exceeds the preset safety parameter threshold, triggering the proportional operation of the to-be-detected section parameter and the plurality of associated parameters to determine the corresponding matching coefficient, if the matching coefficient exceeds the preset matching coefficient threshold, regarding the to-be-detected section parameter as an abnormal parameter, and collecting a plurality of abnormal parameters; marking the position corresponding to the abnormal parameter, determining the corresponding detection section according to the matching of the position corresponding to the abnormal parameter and the distribution map of the railway section, determining the environment type of the detection section based on the environment detection of the detection section, and determining the abnormal section based on the numerical value, position and environment type of the detection section of the plurality of abnormal parameters.

4. The low altitude unmanned aerial vehicle based method for detection of faults in railway track sections as claimed in claim 1 wherein, The abnormality level of the multiple abnormal sections is determined based on the multiple abnormal sections and the distribution map of the railway sections, the fault detection path is determined according to the abnormality level of the multiple abnormal sections, the driving state map of the railway sections and the position of the low-altitude unmanned aerial vehicle, and the fault detection path comprises: The multiple abnormal sections are collected, the length of each abnormal section is detected to determine the section length of the abnormal section, and the first abnormality parameter is determined according to the section position of the multiple abnormal sections and the section length of the abnormal section; The second abnormality parameter is determined according to the section position of the multiple abnormal sections and the distribution map of the railway sections, and the abnormality level of the multiple abnormal sections is determined based on the first abnormality parameter, the second abnormality parameter and the abnormality level mapping relationship.

5. The method for fault detection of railway track sections by low altitude UAV as claimed in claim 4 wherein, The abnormality level of the multiple abnormal sections is determined based on the multiple abnormal sections and the distribution map of the railway sections, the fault detection path is determined according to the abnormality level of the multiple abnormal sections, the driving state map of the railway sections and the position of the low-altitude unmanned aerial vehicle, and the fault detection path comprises: The driving state map of the railway section is determined according to the name of the railway section, the railway database and the current time, and the driving state of each abnormal section is marked, the position of the low-altitude unmanned aerial vehicle is collected, the fault detection path is determined based on the abnormality level of the multiple abnormal sections, the driving state and the position of the low-altitude unmanned aerial vehicle, at this time, the fault detection path presents multiple abnormal section sub-low-altitude flight paths, and the flight order of each sub-low-altitude flight path is determined according to the multiple abnormal section sub-low-altitude flight paths and the driving state of the multiple abnormal sections.

6. The low altitude unmanned aerial vehicle based method for detection of faults in railway track sections as claimed in claim 1 wherein, The low-altitude unmanned aerial vehicle flies along the fault detection path, detects the abnormality of each abnormal section, determines the surface abnormality image and the installation abnormality image according to the abnormality detection of the abnormal section, and comprises: The low-altitude unmanned aerial vehicle obtains the fault detection path and flies along the fault detection path, detects each abnormal section in real time during the low-altitude flight, at this time, the low-altitude unmanned aerial vehicle detects the abnormality of each abnormal section.

7. The low-altitude drone method for detecting faults in a railway section according to claim 6, characterized in that, The low-altitude unmanned aerial vehicle flies along the fault detection path, detects the abnormality of each abnormal section, determines the surface abnormality image and the installation abnormality image according to the abnormality detection of the abnormal section, and comprises: In the abnormality detection of each abnormal section, the first camera of the unmanned aerial vehicle collects the surface image of the abnormal section, determines the surface defect of the abnormal section according to the detection of the surface image of the abnormal section, and determines the surface abnormality image according to the synthesis of the defect area and the defect position of each surface defect; The second camera of the unmanned aerial vehicle is arranged on one side of the first camera and dynamically captures the installation position of the abnormal section, determines the installation defect of the abnormal section according to the shooting of the installation position of the abnormal section, and determines the installation abnormality image according to the synthesis of the defect area and the defect position of each installation defect.

8. The low altitude unmanned aerial vehicle based method for detection of faults in railway track sections as claimed in claim 1 wherein, The fault content corresponding to the abnormal section is determined based on the surface abnormality image, the installation abnormality image and the previous abnormal event of the abnormal section, the real-time maintenance plan of each abnormal section is determined according to the running state of each abnormal section and the fault content of the multiple abnormal sections, and the fault content of the multiple abnormal sections is determined according to the running state of each abnormal section and the fault content of the multiple abnormal sections. The past abnormal events of the abnormal section are determined based on the abnormal tracing of each abnormal section, the surface abnormal features are determined based on the identification of the surface abnormal images, the installation abnormal features are determined based on the identification of the installation abnormal images, and the fault content corresponding to the abnormal section is determined according to the matching of the surface abnormal features, the installation abnormal features and the past abnormal events of the abnormal section.

9. The low-altitude drone method for detecting faults in a railway section according to claim 8, characterized in that, The fault content corresponding to the abnormal section is determined based on the surface abnormal images, the installation abnormal images and the past abnormal events of the abnormal section, the running state of each abnormal section and the fault content of the multiple abnormal sections are determined, and the real-time maintenance plan of each abnormal section is further determined. The running plan table of the railway section is collected, the running state of each abnormal section is determined according to the running plan table of the railway section, the current time and the current position of the railway train, and the maintenance sequence and the maintenance time of each abnormal section are determined based on the running state of each abnormal section and the fault content of the multiple abnormal sections. The low-altitude unmanned aerial vehicle fault detection system for railway sections is applied to the low-altitude unmanned aerial vehicle fault detection method for railway sections as claimed in any one of claims 1-9, and the low-altitude unmanned aerial vehicle fault detection system for railway sections comprises:

10. A low altitude unmanned aerial vehicle based system for detection of faults in a railway track section, characterized in that, A section parameter combination module is configured to determine multiple sub-detection sections according to the distribution map of the railway section, and determine the section parameter combination of each detection section according to the online detection of the multiple sub-detection sections. An abnormal section module is configured to determine multiple abnormal parameters based on the traversal of the section parameter combination in each section parameter combination, and determine the abnormal section according to the position of the multiple abnormal parameters and the environment category of the detection section. A fault detection path module is configured to determine the abnormal level of the multiple abnormal sections based on the multiple abnormal sections and the distribution map of the railway section, and determine the fault detection path according to the abnormal level of the multiple abnormal sections, the driving state map of the railway section and the position of the low-altitude unmanned aerial vehicle. An abnormal image module is configured to fly along the fault detection path by the low-altitude unmanned aerial vehicle, detect the abnormality of each abnormal section, and determine the surface abnormal image and the installation abnormal image according to the abnormal detection of the abnormal section. A real-time maintenance plan module is configured to determine the fault content corresponding to the abnormal section based on the surface abnormal image, the installation abnormal image and the past abnormal events of the abnormal section, and determine the real-time maintenance plan of each abnormal section according to the running state of each abnormal section and the fault content of the multiple abnormal sections. ​

Images