Intelligent inspection system and method for smoothness of track in tunnel based on third track

Abstract

The invention relates to the technical field of railway track detection, and discloses an intelligent inspection system and method for smoothness of a track in a tunnel based on a third rail, and the system comprises the independently arranged third rail, a detection robot and a data processing center. The track direction of the third rail railway is arranged outside a track driving section; an image acquisition device is arranged on the detection robot, and the detection robot runs on the third rail and acquires a rail image in the running process; the data processing center receives the track image and analyzes the smoothness of the track; according to the invention, the extra robot driving track is arranged outside the track driving section, so that the detection robot can work around the clock, and continuous monitoring is realized.

Description

Technical Field

[0001] The present invention relates to the field of track detection technology, and more particularly to an intelligent inspection system and method for track smoothness in a tunnel based on a third rail. Background Art

[0002] At present, traditional track status detection mainly relies on high-speed comprehensive inspection trains, and track inspection is generally carried out every half a month. This low-frequency, periodic inspection makes it difficult to timely capture track status fluctuations in a short period of time. When seismic activity is frequent, the track status may change significantly within the inspection interval, and the maintenance department cannot intervene in time before the problem becomes apparent. For railway tunnel sections located in high-intensity earthquake zones or active fault zones, this low-frequency inspection mode is difficult to meet the needs of high-risk operating environments for real-time perception and continuous monitoring of track smoothness.

[0003] In addition, there are some intelligent solutions for tunnel inspection. Although they can achieve a certain degree of automation and intelligent inspection and basic image acquisition, it is difficult to extract sufficient information from the image to achieve accurate anomaly detection.

[0004] Therefore, how to achieve continuous monitoring and improve the accuracy of anomaly detection is an urgent problem that technicians in this field need to solve. Summary of the invention

[0005] In view of this, the present invention provides an intelligent inspection system and method for track smoothness in a tunnel based on a third rail. An additional robot travel track is set outside the track travel section, which enables the inspection robot to work around the clock and achieve continuous monitoring.

[0006] In order to achieve the above object, the present invention adopts the following technical solution:

[0007] An intelligent inspection system for track smoothness in a tunnel based on a third rail, comprising an independently arranged third rail, an inspection robot and a data processing center;

[0008] The third rail railway track direction is arranged outside the track running section;

[0009] The detection robot is provided with an image acquisition device, and the detection robot travels on the third rail and acquires rail images during the travel;

[0010] The data processing center receives the track image and analyzes the track smoothness.

[0011] Preferably, positioning tags are distributed on the third rail; the detection robot travels on the third rail and performs mileage positioning and distance correction through an on-board tag recognition device.

[0012] Preferably, it also includes pattern position coordinates arranged at multiple specific positions of the track; in the data processing, the world coordinates of the track in the image are located according to the pattern coordinates in the track image; the data processing center is also used to obtain track image time series data at specific track positions based on track images collected at multiple time points associated with mileage points; and is used to analyze track smoothness based on the track push time series data.

[0013] Preferably, there is a base station passing by the third rail, and the base station is used to perform maintenance and / or data uploading for the inspection robot passing by.

[0014] Preferably, a wireless communication module is provided in the base station. When the detection robot enters the area where the base station is located, it is connected to the wireless communication module and sends the collected track image to the data processing center through the wireless communication module.

[0015] Preferably, an automatic cleaning module is provided in the base station; after completing the track detection, the detection robot arrives at the base station located at the end of the third rail, and cleans the lens through the automatic cleaning module, and after cleaning, returns to the third rail and enters the next working cycle.

[0016] Preferably, the base station is further provided with a status checking module, and the status checking module detects the degree of lens dirtiness according to the clarity and contrast changes of the markers in the track image.

[0017] Preferably, a charging module is provided in the base station, and the detection robot is connected to the charging module by a contact connection method for power supply.

[0018] A third rail-based intelligent inspection method for track smoothness in a tunnel comprises the following steps:

[0019] Arranging a third rail outside the track running section, and arranging a detection robot in the third rail;

[0020] Receive the track image data collected by the detection robot.

[0021] Preferably, the steps further include:

[0022] Arrange pattern position coordinates at different positions of the track respectively;

[0023] receiving track image data with pattern position coordinates, and confirming the world position of the track according to the pattern position coordinates;

[0024] Generate orbital state time series data at each position according to the orbital world position;

[0025] The track deformation amount within a specific time period is determined according to the track state time series data, and an alarm is issued when the deformation amount exceeds a preset threshold.

[0026] It can be known from the above technical solutions that, compared with the prior art, the present invention discloses an intelligent inspection system and method for track smoothness in tunnels based on the third rail, in which an additional robot travel track is set outside the track travel section, which enables the detection robot to work around the clock and realize continuous monitoring. Combined with the positioning label on the third rail and the pattern position coordinates arranged beside the track, the world position of the track in the track image can be accurately confirmed, and then time series data can be formed to realize accurate anomaly detection. In addition, the present invention sets a base station based on the third rail, which can realize daily maintenance of the detection robot and centralized transmission of collected data. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0028] Figure 1 A schematic structural diagram of a railway tunnel track intelligent inspection system based on a third rail provided in an embodiment of the present invention.

[0029] Figure 2 A schematic diagram of the internal facility layout structure of a base station provided in an embodiment of the present invention.

[0030] Figure 3 Schematic diagram of the installation position of the third rail in an embodiment of the present invention. DETAILED DESCRIPTION

[0031] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0032] Example 1

[0033] like Figure 1 , Figure 2 and Figure 3 The embodiment of the present invention discloses an intelligent inspection system for track smoothness in a tunnel based on a third rail, including an independently arranged third rail, a detection robot and a data processing center.

[0034] The direction of the third rail railway track is set outside the track running section; an image acquisition device is set on the detection robot, and the detection robot runs on the third rail and collects track images during the driving process; the data processing center receives the track image and analyzes the track smoothness.

[0035] The detection robot can use a wheel set that matches the third rail to achieve stable operation. The image acquisition device can use a high-sensitivity camera to shoot toward the light band area on the top and side surfaces of the track.

[0036] In addition, each inspection robot can also be equipped with a data processing unit to perform preliminary processing on the collected images based on the embedded computing module, such as noise reduction, enhancement and edge detection.

[0037] In this embodiment, the third rail is independently arranged and located outside the track running section, such as the tunnel lining or the ditch position, and can continuously reciprocate for inspection during the normal operation of the train, and use an image acquisition device, such as a camera, to capture the light band on the surface of the rail to obtain image data.

[0038] In order to further implement the above technical solution, positioning tags are distributed on the third rail, and the detection robot has an on-board tag recognition device. When the detection robot is driving, mileage positioning and distance correction can be performed through the on-board tag recognition device.

[0039] Among them, the positioning tag and tag identification device can use radio frequency cards (RFID tags) and RFID readers respectively. Each radio frequency card has a unique identification and a known position. When the robot runs to the corresponding position, it can periodically correct its own positioning based on the known RFID position and travel distance information. In this way, even if the positioning error accumulates after long-distance operation, it can be regularly corrected by RFID to keep the positioning accuracy at a high level.

[0040] For example, the third rail in the tunnel has RFID tags regularly distributed along the length (e.g., one every 100 meters). When the detection robot moves, its built-in RFID reader periodically reads the information of the RFID tags passing by. Each tag corresponds to a fixed mileage coordinate. When the robot detects an RFID tag, it can obtain the current absolute position (mileage point). At the same time, the robot can also use its own odometer (wheel encoder) to measure the travel distance and compare the relative displacement with the known RFID position for error correction.

[0041] Furthermore, it also includes pattern position coordinates arranged at multiple specific positions of the track; in data processing, the world coordinates of the track in the image are located according to the pattern coordinates in the track image; the data processing center is also used to associate the track images collected at multiple time points according to the mileage points to obtain the track image time series data at a specific track position; and is used to analyze the track deformation according to the track push time series data.

[0042] Among them, the pattern position coordinates, i.e., the coding pattern, can be a QR code or a specific geometric mark, which is set on the main line track or sleeper and is a fixed measurement point in the world coordinate system. When the robot recognizes the coding pattern in the image, combined with the aforementioned RFID precise positioning capability, the image coordinates can be matched with the world coordinates to accurately determine the actual position of the photographed rail segment in the tunnel. Since the position of the coding pattern is known and does not change, by comparing the detection results of the coding pattern in the image with the RFID positioning information, the system can accurately map the photographed rail features to the world coordinate system.

[0043] For example, coded patterns (such as specific QR codes or high-contrast markers) are arranged at regular intervals on railway tracks or sleepers. The world coordinates of these markers are accurately determined and stored in the system database. When the camera captures one of the coded patterns, the image processing algorithm identifies the marker and calculates its pixel position in the image coordinates.

[0044] Using the precise robot position obtained through RFID calibration and the camera external parameters (the camera's installation position and posture relative to the robot), the image coordinates can be mapped to the world coordinates to obtain the spatial position of the coded pattern. Since the coded pattern position is fixed and known in the world coordinates, the comparison calculation results can be used as a basis for re-checking and optimizing the positioning accuracy.

[0045] Step 1: At the beginning of system deployment, calibrate the camera to obtain its internal and external parameters: The internal parameters include focal length , principal point coordinates , distortion coefficient The external parameters include the rotation matrix (R) and the translation vector (T), which define the transformation from the robot coordinate system to the world coordinate system.

[0046] Step 2: The robot’s built-in RFID reader reads the tag ID and its preset world coordinate position when it approaches the RFID tag. It uses the read RFID tag position and the known robot’s internal odometer data to calculate and correct the robot’s current position in the world coordinate system, eliminating the positioning deviation caused by the accumulated odometer error.

[0047] Step 3: Use the camera external parameters and the corrected robot position to map the coded pattern position in the image to the world coordinate system. Convert to a point in the camera coordinate system , assuming that the object is in the camera coordinate system (Normalized coordinates). Use camera intrinsic parameters for inverse distortion and normalization:

[0048]

[0049] Among them, K is the camera intrinsic parameter matrix:

[0050]

[0051] Using the camera external parameters (rotation matrix R and translation vector T), the point in the camera coordinate system is converted to the point in the world coordinate system:

[0052]

[0053] is a point in the world coordinate system, is a point in the camera coordinate system.

[0054] Step 4: Identify and position the coding pattern. In the captured image, locate the coding pattern through image processing algorithms such as grayscale, binarization, and noise filtering to determine its precise pixel position in the image coordinate system. Then firstly, the position of the coding pattern in the image coordinate system is Convert to a point in the camera coordinate system , and then use the camera external parameters to convert the point in the camera coordinate system to the point in the world coordinate system Since the coding pattern has a known height in the real world, we can assume that Zw is a zero reference to further simplify the calculation.

[0055]

[0056] Among them, R is the rotation matrix, T is the translation vector, is a point in the world coordinate system, is the normalized point in the camera coordinate system. Combining the known position of the coding pattern in the world coordinate system and the pixel position identified in the image, the actual position of the coding pattern in the world coordinate system is calculated by the above coordinate conversion formula.

[0057] The same conversion method is used to align the center coordinates of the rail light band in the photograph with the world coordinates. In this way, after the center of the rail light band is extracted and curve-fitted, these edge coordinates are no longer just image coordinates, but correspond to real physical mileage coordinates and position coordinates, achieving high-precision conversion from image measurement to world measurement.

[0058] Furthermore, the track deformation is analyzed using time series data, including:

[0059] The first data benchmark is established. During the system startup phase, the track is photographed and processed under the initial conditions where no obvious track state changes occur. Using the RFID tag and the coded pattern positioning information, the captured image is strictly corrected to a unified world coordinate system. Edge detection is performed through image enhancement, Canny or Sobel operator to extract the edge features of the light band on the top surface of the track rail. The least squares method is then used to fit the edge point set to a straight line or a low-order polynomial to obtain the initial benchmark curve for track straightness. At the same time, the maximum deviation, mean square error and other parameters are statistically described to describe the initial state of the track, and stored in the database as a reference benchmark.

[0060] Multi-period data comparison and trend analysis. In subsequent inspections, the robot repeats the same process regularly (e.g., every few hours) to obtain images of the track surface. Under the same markers and coordinate system, the light band center data obtained during the current inspection is compared one-to-one with the initial reference curve. By calculating the vertical deviation change (deviation increment) of the light band center at the corresponding mileage position, the degree of change in track smoothness can be quantified.

[0061] For a specified mileage point i (e.g., mileage points i = 1, 2, 3, etc.), the system has stored the initial position value of the light band center corresponding to the mileage point during the first detection, which is recorded as In this inspection, the current light band center position value is obtained for the same mileage point i, which is recorded as The vertical deviation change (deviation increment) is calculated by comparing the current detection value with the initial reference value, which is defined as follows:

[0062]

[0063] in, The positive or negative value and its size reflect the degree of "uplift / sinking" of the top surface or edge of the track relative to the initial moment. When it continues to increase, it means that the track smoothness is deteriorating.

[0064] For each mileage position i, As the detection time goes by, it is recorded as a time series:

[0065]

[0066] Different mileage positions i each have a time series, which can be stored separately in the database to facilitate subsequent trend analysis and visualization.

[0067] Perform low-pass filtering or RANSAC fitting on multi-period data to eliminate accidental error points caused by noise and illumination changes. After accumulating enough time series data, introduce LSTM network deep learning or time series prediction model to predict the trend of track smoothness changes. If the deviation increment is detected in a certain section and shows a continuous increasing trend, it means that the track smoothness of this section is deteriorating. An early warning can be triggered according to a pre-set threshold (such as a high and low deviation of 4mm within a range of 10m), and the early warning information is transmitted back to the control center through the base station so that the operation and maintenance department can make timely decisions.

[0068] In order to further implement the above technical solution, one or more base stations are also provided, and the base stations can be arranged on the third rail, such as at the end of the third rail. When the detection robot passes through the base station, the collected image data is uploaded centrally. In addition, various equipment maintenance facilities are also provided in the base station to realize the daily automatic maintenance of the detection robot; the base station is divided into multiple branches by the third cabinet through the road, which can realize the simultaneous operation of multiple devices and can also be used as a backup machine.

[0069] Specifically, the base station is equipped with a wireless communication module, an automatic cleaning module, a status inspection module and a charging module. When the detection robot enters the area where the base station is located, the base station detects the arrival of the robot through a wireless beacon, connects to the wireless communication module, and sends the collected track image to the data processing center through the wireless communication module.

[0070] A mechanical positioning structure is set in the base station to ensure that the robot stops at the predetermined position in the base station. The mechanical positioning structure is set with a fixed charging port to realize charging after docking. The robot and the base station realize data transmission through Wi-Fi. After the robot stops at the mechanical positioning structure, the detection image data, light band center calculation results, RFID reading records, etc. during this inspection process are transmitted to the base station via WI-FI. The control module of the base station receives this data, temporarily stores it in the local storage, and then sends the data to the control center through the 5G network for back-end analysis and decision-making.

[0071] Subsequently, the robot camera position mechanically positioned by the base station starts the automatic cleaning module to maintain the robot camera lens. The specific steps may include: the liquid spraying device gently sprays the cleaning liquid to dissolve the dust or oil attached to the lens surface, the flexible brush head or cleaning cloth gently wipes the lens surface to ensure that no scratches are left, and the dry air flow blows the lens to ensure that no water stains remain. After the cleaning is completed, the base station can ask the robot to test the clarity of an image. If there are still stains, repeat the cleaning steps.

[0072] Therefore, after data transmission and cleaning, the detection module in the base station performs basic status checks on the robot (such as wheel wear, RFID reader working condition, sensor operation status). If any abnormality is found, the base station will send maintenance reminder information to the control center.

[0073] During the operation of the base station, the control center may send updated instructions or data to the base station (such as revised inspection plans, parameter adjustments, or new weight files for deep learning models). The base station transmits these instructions and data to the robot through the communication interface, so that the robot can work according to the latest strategy in the next round of inspections.

[0074] When data transmission, lens cleaning, battery replenishment and status detection are completed, and the robot receives the necessary command updates, the base station sends a signal to notify the robot to leave. The robot slowly drives out of the base station along the third track and re-enters the inspection track section inside the tunnel.

[0075] The operation process of this embodiment is as follows:

[0076] The first step is to install the system. After the system is installed, the camera is calibrated. Using the calibration plate and known coordinate points, images are taken from multiple angles, and the camera's internal parameters (focal length, principal point, distortion coefficient) and the camera's external parameters relative to the third rail are calculated using the calibration software. The calibration parameters are stored in the robot data processing unit for subsequent distortion correction and perspective correction of the captured images.

[0077] After the calibration is completed, the initial inspection is carried out. The robot starts from the base station, runs at a speed of 10km / h, travels along the third track, and takes track images 10 times per second with a shutter speed of 1 / 300 second. During the shooting process: use the RFID reader to identify the RFID tag passing through and obtain the precise mileage point. Search for the coded pattern in the image. If the pattern is detected, the image coordinates are corrected according to the known world coordinates of the pattern, and then the overall positioning accuracy is ensured in conjunction with the RFID information. The edge detection of the captured track light band is performed, and the upper and lower edge point sets of the light band are extracted using the Canny operator. The straightness reference value of the track at this mileage point is obtained through least squares linear or quadratic curve fitting. After the initial inspection, the system obtains the initial curve (reference curve) of the track straightness within the entire tunnel, as well as the corresponding deviation parameters (such as maximum deviation, average deviation, standard deviation). After passing through the base station, these data are uploaded to the data processing center for archiving as a benchmark for subsequent comparative analysis.

[0078] During normal operation, trains run continuously. Inspection robots can be used for inspections during the day and at night, and are set to inspect the tunnels once every two hours. After multiple inspections, the data processing center can generate time series data, with clear records of track smoothness changes for each mileage section. Through time series analysis, trend fitting, and LSTM network deep learning prediction models, the track smoothness evolution trend in the future is predicted to help the maintenance team plan maintenance in advance.

[0079] During the inspection, the captured track light band is fitted and the current straightness curve is obtained using the same processing method as the initial benchmark data. This curve is then compared with the initial benchmark curve at the corresponding mileage point. If the deviation increment in a certain section increases significantly (for example, the high and low deviations within a 10m range reach 4mm), an early warning message is sent to the control center after the data is summarized.

[0080] During the inspection process, dust, water vapor, and oil in the tunnel environment are easily attached to the surface of the camera lens, affecting the image quality and recognition accuracy. When the inspection robot enters the base station, it is cleaned. The process is as follows: When the robot approaches the base station, it sends a station signal to the base station through short-range wireless communication. The base station starts the cleaning module, and the robot arm aims the nozzle at the surface of the camera lens and gently sprays the cleaning liquid to dissolve dust and stains.

[0081] The flexible brush attached to the robot arm gently wipes the lens surface, and then blows it dry with dry air to ensure that there are no residual droplets. After cleaning, the detection system evaluates the dirtiness of the lens. If the dirtiness level does not meet the requirements, it will continue to be cleaned. This ensures the stability of image quality during long-term continuous operation. The clearer the image, the more accurate the extraction of the coding pattern and the center of the light band, which is conducive to the continuous maintenance of positioning and measurement accuracy.

[0082] Finally, for the exceptions that may occur during long-term operation:

[0083] If the robot cannot detect an RFID tag, it means there is a problem with the RFID reader or tag. The system will record the anomaly and check the RFID hardware during the next maintenance cycle.

[0084] If the coded pattern cannot be clearly identified for a long time, it may be damaged or covered by dirt, and manual inspection and cleaning or replacement of the pattern sticker is required.

[0085] If the track smoothness deviation increases suddenly and exceeds the warning value, the system will automatically issue an emergency warning message, suggesting that manual inspections and track maintenance be arranged immediately, and if necessary, the train speed should be limited or temporarily stopped to eliminate safety hazards.

[0086] Example 2

[0087] Based on the same inventive concept, an embodiment of the present invention discloses an intelligent inspection method for track smoothness in a tunnel based on a third rail, comprising the following steps:

[0088] A third rail is arranged outside the track running section, and a detection robot is arranged in the third rail; and pattern position coordinates are arranged at different positions of the track.

[0089] Receive the track image data collected by the detection robot and confirm the world position of the track according to the pattern position coordinates in the track image data.

[0090] Generate orbital state time series data at each position based on the orbital world position.

[0091] The track deformation amount within a specific time period is determined according to the track state time series data, and an alarm is issued when the deformation amount exceeds a preset threshold.

[0092] In summary, the present invention has the following advantages:

[0093] (1) High-frequency inspection without interfering with train operation: The present invention deploys an independent third rail on the side wall of the tunnel or in the ditch, allowing the inspection robot to conduct uninterrupted inspections during normal train operation. Compared with the existing technology that relies on nighttime skylights or train downtime for inspection, the present invention does not need to occupy the precious time of the operating line, greatly increases the inspection frequency, and realizes truly continuous and daily track status monitoring.

[0094] (2) Multi-period trend analysis based on initial data: Unlike the traditional method that relies only on single-time detection data to compare with a fixed threshold, the present invention uses the track straightness data obtained from the first detection as the initial benchmark, and the subsequent inspection data is accurately compared with this benchmark. This method can identify subtle changes and long-term trends in track smoothness, providing a basis for early detection of potential problems and hidden dangers. Compared with the static determination method of the prior art, the present invention is more suitable for complex environments such as high-intensity earthquake zones and active fault zones where the track state is prone to fluctuate over time.

[0095] (3) Using RFID and markers to achieve high-precision positioning correction: The present invention distributes RFID tags on the third rail and sets a fixed coding pattern on the rail. Through the dual positioning means of radio frequency cards and image markers, the inspection robot can maintain high-precision position correction during long-distance inspections. Compared with solutions that simply rely on inertial measurement, odometers or single calibration, the present invention can continuously correct positioning errors, ensure that each inspection data has a unified reference coordinate system, and lay a solid foundation for previous data comparisons.

[0096] (4) Camera cleaning and maintenance ensure long-term data quality stability: The present invention sets up base stations at the entrance and exit of the tunnel. The base stations are equipped with automatic camera cleaning devices to regularly remove dust, oil stains and water stains attached to the lens. Compared with the existing technology where there is no special cleaning mechanism, resulting in image quality degradation over time, the present invention can ensure image clarity for a long time, thereby improving recognition accuracy, data reliability and the accuracy of trend analysis.

[0097] (5) Adapting to the complex environment of high-intensity earthquake zones and long tunnels: Compared with the existing technologies with relatively single applicable scenarios, the present invention focuses on extremely challenging scenarios such as high-intensity earthquake zones, active fault zones and long tunnels. Through uninterrupted high-frequency detection, multi-period data trend analysis, and the comprehensive application of positioning and cleaning strategies, the present invention can respond more sensitively and proactively to rapid changes in track conditions and potential hidden dangers.

[0098] In summary, compared with the existing technology, the present invention realizes the upgrade from single static detection to uninterrupted dynamic monitoring, the expansion from single threshold judgment to multi-period trend analysis, and the improvement from simple positioning to multi-source precise calibration. It also significantly improves the accuracy, flexibility and foresight of track smoothness detection and operation and maintenance decision-making through camera cleaning and high-quality image assurance.

[0099] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0100] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An intelligent inspection system for track smoothness in tunnels based on the third rail, characterized in that: It includes an independently-installed third rail, detection robots and data processing center; The third rail railway track direction is arranged outside the track running section; The detection robot is provided with an image acquisition device, and the detection robot travels on the third rail and acquires rail images during the travel; The data processing center receives the track image and analyzes the track smoothness.

2. According to claim 1, the intelligent inspection system for track smoothness in tunnels based on the third rail is characterized in that: Positioning tags are distributed on the third rail; the detection robot travels on the third rail and performs mileage positioning and distance correction through the vehicle-mounted tag recognition device.

3. The intelligent inspection system for track smoothness in a tunnel based on the third rail according to claim 1 or 2, characterized in that: It also includes pattern position coordinates arranged at multiple specific positions of the track; in the data processing, the world coordinates of the track in the image are located according to the pattern coordinates in the track image; the data processing center is also used to obtain track image time series data at specific track positions based on the track images collected at multiple time points according to the mileage points; and is used to analyze the track smoothness based on the track push time series data.

4. The intelligent inspection system for track smoothness in tunnels based on the third rail according to claim 1 is characterized in that: There is a base station passing by the third rail, and the base station is used to maintain and / or upload data of the passing detection robot.

5. The intelligent inspection system for track smoothness in tunnels based on the third rail according to claim 4 is characterized in that: A wireless communication module is provided in the base station. When the detection robot enters the area where the base station is located, it is connected to the wireless communication module and sends the collected track image to the data processing center through the wireless communication module.

6. The intelligent inspection system for track smoothness in tunnels based on the third rail according to claim 5, characterized in that: An automatic cleaning module is provided in the base station; after completing the track detection, the detection robot arrives at the base station located at the end of the third rail, and cleans the lens through the automatic cleaning module, and returns to the third rail after cleaning and enters the next working cycle.

7. The intelligent inspection system for track smoothness in tunnels based on the third rail according to claim 5, characterized in that: The base station is also provided with a status inspection module, which detects the degree of lens dirtiness according to the clarity and contrast changes of the markers in the track image.

8. The intelligent inspection system for track smoothness in tunnels based on the third rail according to claim 4 is characterized in that: The base station is provided with a charging module, and the detection robot is connected to the charging module by a contact connection method for power supply.

9. An intelligent inspection method for track smoothness in a tunnel based on a third rail, characterized in that: The following steps are involved: Arranging a third rail outside the track running section, and arranging a detection robot in the third rail; Receive the track image data collected by the detection robot.

10. The intelligent inspection method for track smoothness in a tunnel based on the third rail according to claim 9, characterized in that: The steps also include: Arrange pattern position coordinates at different positions of the track respectively; receiving track image data with pattern position coordinates, and confirming the world position of the track according to the pattern position coordinates; Generate orbital state time series data at each position according to the orbital world position; The track deformation amount within a specific time period is determined according to the track state time series data, and an alarm is issued when the deformation amount exceeds a preset threshold.

Images