Rapid detection method and system for subway contact rail

Abstract

The invention belongs to the technical field of metro contact rail detection, and particularly relates to a metro contact rail rapid detection method and system.The detection method comprises the following steps that 1, a contact rail automatic detection device carrying a 3D camera is used for continuously and dynamically scanning a metro contact rail, and each frame of data of the 3D camera is taken out; combining Z-axis mileage sensor data of the contact rail automatic detection device to form a contact rail three-dimensional point cloud; 2, filtering point clouds of a lower contact type third rail, an upper contact type third rail and a return rail to remove invalid interference data, then carrying out feature multi-data cooperative training, and constructing unified point cloud features of a contact rail insulation support, an expansion joint and an end elbow; 3, when the contact rail automatic detection device moves for measurement, the contact rail surface, the insulation support, the expansion joint and the end elbow are judged in real time according to the point cloud features of the insulation support, the expansion joint and the end elbow of the contact rail, and the conduction height and pull-out measurement values of the components are calculated.

Description

Technical Field

[0001] The invention belongs to the technical field of subway contact rail detection, and in particular relates to a rapid detection method and system for subway contact rails. Background Art

[0002] In recent years, my country's subway rail transit has achieved rapid development. The subway contact rail is an important way to power the subway rail transit. It is very important for the safe running of the train and is an important guarantee for the safe operation of the train. During the long-term operation of the subway line, the subway third rail and return rail are prone to poor contact with the conductive surface, loose screws and falling off. At present, the subway operation and maintenance department uses track rulers and manual inspections to detect the geometric parameters of the contact rails and the defects of the contact rails. The efficiency is very low and the measurement accuracy is unstable. An automated detection technology and system are needed to improve the efficiency of contact rail detection. Summary of the invention

[0003] The purpose of the present invention is to provide a rapid detection method and system for subway contact rails, so as to solve the problems existing in the background technology.

[0004] In order to achieve the above technical objectives, the technical solution adopted by the present invention is as follows:

[0005] A rapid detection method for a subway contact rail, the detection method comprising the following steps:

[0006] Step 1: Use the contact rail automatic detection device equipped with a 3D camera to continuously and dynamically scan the subway contact rail, extract each frame of data from the 3D camera, and combine it with the Z-axis mileage sensor data of the contact rail automatic detection device to form a three-dimensional point cloud of the contact rail;

[0007] Step 2: Filter the point clouds of the lower contact third rail, upper contact third rail, and return rail to remove invalid interference data, and then perform feature multi-data collaborative training to construct unified point cloud features of the contact rail insulation bracket, expansion joint, and end elbow;

[0008] Step 3: When the contact rail automatic detection device is moving for measurement, the point cloud features of the contact rail's insulating bracket, expansion joint and end elbow are used to judge the contact rail surface, insulating bracket, expansion joint and end elbow in real time, calculate the lead height and pull-out measurement values ​​of these components, and further trigger the high-definition industrial camera to take pictures, merge the pictures and mark the corresponding pole number, kilometer mark position, lead height, and pull-out measurement values.

[0009] Step 1: Use the contact rail automatic detection device equipped with a 3D camera to continuously and dynamically scan the subway contact rail, extract each frame of data from the 3D camera, and combine it with the Z-axis mileage sensor data of the contact rail automatic detection device to form a three-dimensional point cloud of the contact rail:

[0010] S1: The contact rail automatic detection device moves along the subway track. The extended robotic arm is equipped with a contact rail detection unit, in which the 3D camera continuously scans the contact rails including the third rail and the return rail;

[0011] S2: Obtain data of different contact rails scanned by the 3D camera in each frame, and convert the 3D camera coordinates into coordinates relative to the rail center;

[0012] S3: For each frame of contact rail data scanned by the 3D camera, the mileage sensor data obtained in real time during the movement of the contact rail automatic detection device is associated to construct a three-dimensional point cloud of the contact rail relative to the center of the rail.

[0013] The point clouds of the lower contact third rail, the upper contact third rail, and the return rail are filtered to remove invalid interference data, and then feature multi-data collaborative training is performed to construct unified point cloud features of the contact rail insulation bracket, expansion joint, and end elbow:

[0014] S1: Remove the obvious invalid point cloud outside the tunnel wall and the cable contact rail area of ​​the tunnel wall through edge filtering, and further remove the noise through Gaussian filtering;

[0015] S2: Label the training data sets of the lower contact third rail, upper contact third rail, and return rail respectively, and distinguish the insulating bracket, expansion joint, end elbow, and contact rail surface;

[0016] S3: Use the annotated insulating brackets, expansion joints, end elbows and contact rail surface point cloud features of the lower contact third rail, upper contact third rail and return rail to cross-validate the unannotated lower contact third rail, upper contact third rail and return rail data sample sets. For example, the lower contact third rail feature value verifies the upper contact third rail sample set, and corrects the point cloud features to ensure that the recognition accuracy meets the requirements.

[0017] S4: Introduce a new larger validation set, and use the corrected and annotated insulating brackets, expansion joints, end elbows and contact rail surface point cloud features of the lower contact third rail, upper contact third rail and return rail to cross-validate the unannotated lower contact third rail, upper contact third rail and return rail data sample set, and replace the cross-validation sample set object, such as the lower contact third rail feature value to validate the return rail sample set;

[0018] S5: Compare the accuracy of the point cloud features of the lower contact third rail, upper contact third rail, insulating bracket, expansion joint, end elbow and contact rail surface of the return rail in S4 in the verification set, and take the one with the highest accuracy and reaching the accuracy threshold as the training result point cloud feature. If the accuracy threshold is not reached, repeat S3 and S4.

[0019] When the contact rail automatic detection device moves and measures, the contact rail surface, insulating bracket, expansion joint and end elbow point cloud features are used to judge the contact rail surface, insulating bracket, expansion joint and end elbow in real time, calculate the guide height and pull-out measurement values ​​of these components, and further trigger the high-definition industrial camera to take pictures, merge the pictures and mark the corresponding pole number, kilometer mark position, guide height and pull-out measurement;

[0020] S1: When the contact rail automatic detection device is moving for measurement, the contact rail surface, insulating bracket, expansion joint and end elbow point cloud features are used to judge the contact rail surface, insulating bracket, expansion joint and end elbow in real time;

[0021] S2: Calculate the conduction height and pull-out value of the corresponding insulating bracket, expansion joint and end elbow out of the contact rail surface;

[0022] S3: Synchronously trigger the high-definition industrial camera to take pictures, merge multiple related pictures into one picture, and mark the corresponding pole number, kilometer mark position, guide height, and pull-out measurement on the picture.

[0023] A rapid detection method for a subway contact rail also includes a detection system, including a contact rail detection unit equipped with a 3D camera and a high-definition industrial camera, as well as a mobile detection platform and a computing unit; the contact rail detection unit integrates a 3D camera to measure the geometric parameters and wear of the contact rail, and a high-definition industrial camera to take high-definition photos of the contact rail surface, bracket, fastening screws, and end elbows; the mobile detection platform and the computing unit are basic measurement units that measure the gauge between two rails, the height difference between the two rails, and the moving distance when moving along the rail, a computing storage unit for contact rail measurement parameters and wear calculation, photo control, and image storage, and a human-computer interaction unit for detection result presentation and system control, and the contact rail detection unit exchanges data and signaling with the mobile detection platform and the computing unit through a network port.

[0024] The contact rail detection unit includes a high-precision 3D camera for scanning and measuring the contact rail. The 3D camera performs continuous cross-sectional scanning on the subway contact rail. The 3D camera scanning data is combined with the mileage sensor data of the mobile detection platform and the computing unit to form a three-dimensional point cloud of the contact network. The coordinate data of the 3D camera is converted into the center of the rail surface as the coordinate origin, the pull-out direction as the X-axis, and the direction perpendicular to the rail surface as the Y-axis.

[0025] The contact rail detection unit includes a high-definition industrial camera and fill light for photographing the contact rail surface, bracket, fastening screws, and end elbow, a circuit board for controlling the camera shooting, a control of the camera shooting frequency and the fill light switch, and a multi-network port switch, which is responsible for transmitting the high-definition photos and 3D scanning data of multiple industrial cameras to the mobile detection platform and computing unit through the corresponding network ports in real time for contact rail detection data analysis and calculation, contact rail status image analysis and storage, and defect status analysis.

[0026] The mobile detection platform and computing unit include a portable mobile measurement base, a detachable mobile power supply, a linear displacement sensor for measuring track gauge, a multi-axis tilt sensor for measuring the superelevation of two tracks and monitoring the attitude of the detection equipment, a mileage encoder for measuring mileage and position positioning, and a collection circuit board for collecting and processing sensor data. The detachable mechanical arm equipped with the contact rail detection unit is quickly docked with the mobile measurement base through a quick rotation button.

[0027] The mobile detection platform and computing unit include a high-performance processor, which runs a rapid detection method for subway contact rails; and controls the camera of the contact rail status camera shooting unit, transmits and stores pictures, processes defect states, and manages one pole and one file, and exchanges internal data between the contact rail status picture analysis management system and the contact rail geometric parameter calculation system.

[0028] The mobile detection platform and computing unit include a human-computer interaction part, namely, the viewing and exporting of detection data. The human-computer interaction unit is an output window for the geometric parameter measurement values ​​and status pictures of the subway contact rail. The measurement data is displayed in a waveform diagram and defects are warned in real time.

[0029] The invention discloses a rapid detection method and system for subway contact rails, which are a portable automated measurement method and system for contact rails. The method and system adopt a non-contact continuous measurement method, and can solve the problems of difficult and low efficiency in traditional inspection of contact rail geometric parameter measurement, as well as contact rail surface damage, burns, screw shedding and loose defects. The method and system have the characteristics of easy portability, full system functions and high measurement accuracy.

[0030] The present invention constructs a rapid detection system for subway contact rails, which is suitable for rapid detection of different contact rails such as lower contact third rails, upper contact third rails and return rails. While rapidly measuring the geometric parameter values ​​of the contact rails, the state of the current receiving surface of the contact rails, the state of the insulating bracket and the protective cover can also be photographed and analyzed, greatly improving the measurement and inspection efficiency of the subway contact rails.

[0031] Compared with the prior art, the present invention has the following technical advantages:

[0032] 1. High efficiency: Through automation and continuous dynamic scanning, the detection efficiency of the contact rail can be significantly improved, manual intervention can be reduced, and the detection speed can be accelerated.

[0033] 2. Accuracy: Combining 3D camera and mileage sensor data, the 3D point cloud of the contact rail can be accurately constructed to ensure the accuracy of the detection results. At the same time, through feature multi-data collaborative training and cross-validation, the recognition accuracy can be further improved to ensure accurate judgment of key components such as insulating brackets, expansion joints, and end elbows.

[0034] 3. Comprehensiveness: This method not only detects the geometric parameters and wear of the contact rail, but also uses a high-definition industrial camera to take high-definition photos of the contact rail surface, bracket, fastening screws, end elbows, etc., which can fully reflect the status of the contact rail and help to discover potential problems in a timely manner.

[0035] 4. Real-time: During the detection process, the key components of the contact rail can be judged in real time, and parameters such as the height of the conductor and the pull-out measurement value can be calculated. At the same time, the camera can be triggered to realize the real-time acquisition and processing of the detection data.

[0036] 5. Ease of use: The detection system includes a portable mobile measurement base and a detachable mechanical arm, which are easy to carry and install and can adapt to different detection environments and needs. At the same time, the human-computer interaction unit provides intuitive detection data viewing and data export functions, which is convenient for users to conduct subsequent analysis and processing.

[0037] 6. Safety: Through automated detection, the safety risks brought by manual operation can be reduced. At the same time, high-definition photography and real-time alarm functions can also help to promptly discover and deal with potential safety hazards.

[0038] 7. Data traceability: Data generated during the detection process, including pictures, marking information, measurement values, etc., can be stored and managed to provide data support for subsequent maintenance and decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] The present invention is further illustrated by means of the non-limiting examples given in the accompanying drawings.

[0040] Figure 1 The process of the present invention is shown in FIG. Figure 1 ;

[0041] Figure 2 The process of the present invention is shown in FIG. Figure 2 . DETAILED DESCRIPTION

[0042] In order to enable those skilled in the art to better understand the present invention, the technical solution of the present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0043] like Figure 1-2 As shown, a rapid detection method for a subway contact rail of the present invention comprises the following steps:

[0044] Step 1: Use the contact rail automatic detection device equipped with a 3D camera to continuously and dynamically scan the subway contact rail, extract each frame of data from the 3D camera, and combine it with the Z-axis mileage sensor data of the contact rail automatic detection device to form a three-dimensional point cloud of the contact rail;

[0045] The contact rail automatic detection device moves along the subway track. The extended robotic arm is equipped with a contact rail detection unit, in which a 3D camera continuously scans the contact rails including the third rail and the return rail. Data of different contact rails scanned by the 3D camera are obtained in each frame, and converted from 3D camera coordinates to coordinates relative to the rail center. The data of the contact rail scanned by the 3D camera in each frame is associated with the mileage sensor data obtained in real time during the movement of the contact rail automatic detection device to construct a three-dimensional point cloud of the contact rail relative to the rail center.

[0046] Step 2: Filter the point clouds of the lower contact third rail, upper contact third rail, and return rail to remove invalid interference data, and then perform feature multi-data collaborative training to construct unified point cloud features of the contact rail insulation bracket, expansion joint, and end elbow;

[0047] Take the point clouds of the lower contact third rail, upper contact third rail, and return rail containing at least 100 consecutive insulating brackets respectively;

[0048] Through edge filtering, the obvious invalid point clouds outside the tunnel wall and the cable contact rail area of ​​the tunnel wall are removed, and through Gaussian filtering, the noise points are further removed;

[0049] The training data sets of the lower contact third rail, upper contact third rail and return rail containing 25 insulating brackets are annotated respectively, and the insulating brackets, expansion joints, end elbows and contact rail surfaces are annotated respectively, and the annotated data are used as the initial point cloud features;

[0050] The labeled insulating brackets, expansion joints, end elbows and contact rail surface point cloud features of the lower contact third rail, upper contact third rail and return rail are used to cross-validate the unlabeled lower contact third rail, upper contact third rail and return rail data sample set containing 25 insulating brackets. For example, the lower contact third rail feature value verifies the upper contact third rail sample set, and corrects the point cloud features to ensure that the recognition accuracy meets the requirements.

[0051] Use the corrected and annotated insulating brackets, expansion joints, end elbows and contact rail surface point cloud features of the lower contact third rail, upper contact third rail and return rail to cross-validate the remaining unannotated lower contact third rail, upper contact third rail and return rail data sample sets containing 50 insulating brackets, and replace the cross-validation sample set objects, such as the lower contact third rail feature value to validate the return rail sample set;

[0052] Compare the accuracy of the point cloud features of the lower contact third rail, upper contact third rail, insulating bracket of the return rail, expansion joint, end elbow and contact rail surface in the verification set in S4, and take the one with the highest accuracy and reaching the accuracy threshold as the training result point cloud feature. If the accuracy threshold is not reached, repeat S3 and S4.

[0053] Step 3: When the contact rail automatic detection device is moving for measurement, the point cloud features of the contact rail's insulating bracket, expansion joint and end elbow are used to judge the contact rail surface, insulating bracket, expansion joint and end elbow in real time, calculate the lead height and pull-out measurement values ​​of these components, and further trigger the high-definition industrial camera to take pictures, merge the pictures and mark the corresponding pole number, kilometer mark position, lead height, and pull-out measurement values.

[0054] When the contact rail automatic detection device is moving for measurement, the contact rail surface, insulating bracket, expansion joint and end elbow point cloud features are used to judge the contact rail surface, insulating bracket, expansion joint and end elbow in real time;

[0055] Calculate the lead height and pull-out values ​​of the contact rail surface of the corresponding insulating bracket, expansion joint and end elbow; synchronously trigger the high-definition industrial camera to take pictures, merge multiple related pictures into one picture, and mark the corresponding pole number, kilometer mark position, lead height, and pull-out measurement on the picture.

[0056] The point clouds of the lower contact third rail, the upper contact third rail, and the return rail are filtered to remove invalid interference data, and then feature multi-data collaborative training is performed to construct unified point cloud features of the contact rail insulation bracket, expansion joint, and end elbow:

[0057] S1: Remove the obvious invalid point cloud outside the tunnel wall and the cable contact rail area of ​​the tunnel wall through edge filtering, and further remove the noise through Gaussian filtering;

[0058] S2: Label the training data sets of the lower contact third rail, upper contact third rail, and return rail respectively, and distinguish the insulating bracket, expansion joint, end elbow, and contact rail surface;

[0059] S3: Use the annotated insulating brackets, expansion joints, end elbows and contact rail surface point cloud features of the lower contact third rail, upper contact third rail and return rail to cross-validate the unannotated lower contact third rail, upper contact third rail and return rail data sample sets. For example, the lower contact third rail feature value verifies the upper contact third rail sample set, and corrects the point cloud features to ensure that the recognition accuracy meets the requirements.

[0060] S4: Introduce a new larger validation set, and use the corrected and annotated insulating brackets, expansion joints, end elbows and contact rail surface point cloud features of the lower contact third rail, upper contact third rail and return rail to cross-validate the unannotated lower contact third rail, upper contact third rail and return rail data sample set, and replace the cross-validation sample set object, such as the lower contact third rail feature value to validate the return rail sample set;

[0061] S5: Compare the accuracy of the point cloud features of the lower contact third rail, upper contact third rail, insulating bracket, expansion joint, end elbow and contact rail surface of the return rail in S4 in the verification set, and take the one with the highest accuracy and reaching the accuracy threshold as the training result point cloud feature. If the accuracy threshold is not reached, repeat S3 and S4.

[0062] When the contact rail automatic detection device moves and measures, the contact rail surface, insulating bracket, expansion joint and end elbow point cloud features are used to judge the contact rail surface, insulating bracket, expansion joint and end elbow in real time, calculate the guide height and pull-out measurement values ​​of these components, and further trigger the high-definition industrial camera to take pictures, merge the pictures and mark the corresponding pole number, kilometer mark position, guide height and pull-out measurement;

[0063] S1: When the contact rail automatic detection device is moving for measurement, the contact rail surface, insulating bracket, expansion joint and end elbow point cloud features are used to judge the contact rail surface, insulating bracket, expansion joint and end elbow in real time;

[0064] S2: Calculate the conduction height and pull-out value of the corresponding insulating bracket, expansion joint and end elbow out of the contact rail surface;

[0065] S3: Synchronously trigger the high-definition industrial camera to take pictures, merge multiple related pictures into one picture, and mark the corresponding pole number, kilometer mark position, guide height, and pull-out measurement on the picture.

[0066] A rapid detection method for a subway contact rail also includes a detection system, including a contact rail detection unit equipped with a 3D camera and a high-definition industrial camera, as well as a mobile detection platform and a computing unit; the contact rail detection unit integrates a 3D camera to measure the geometric parameters and wear of the contact rail, and a high-definition industrial camera to take high-definition photos of the contact rail surface, bracket, fastening screws, and end elbows; the mobile detection platform and the computing unit are basic measurement units that measure the gauge between two rails, the height difference between the two rails, and the moving distance when moving along the rail, a computing storage unit for contact rail measurement parameters and wear calculation, photo control, and image storage, and a human-computer interaction unit for detection result presentation and system control, and the contact rail detection unit exchanges data and signaling with the mobile detection platform and the computing unit through a network port.

[0067] The contact rail detection unit includes a high-precision 3D camera for scanning and measuring the contact rail. The 3D camera performs continuous cross-sectional scanning on the subway contact rail. The 3D camera scanning data is combined with the mileage sensor data of the mobile detection platform and the computing unit to form a three-dimensional point cloud of the contact network. The coordinate data of the 3D camera is converted into the center of the rail surface as the coordinate origin, the pull-out direction as the X-axis, and the direction perpendicular to the rail surface as the Y-axis.

[0068] The 3D camera continuously scans the contact rail. The measured values ​​of the contact rail's guide height and pull-out geometric parameters are calculated relative to the rail center. The data returned by the 3D camera is first converted into data relative to the rail center. The converted contact rail scanning data and the encoder value of the mobile detection platform constitute the point cloud to be analyzed and calculated.

[0069] The contact rail detection unit includes a high-definition industrial camera and fill light for photographing the contact rail surface, bracket, fastening screws, and end elbow, a circuit board for controlling the camera shooting, a control of the camera shooting frequency and the fill light switch, and a multi-network port switch, which is responsible for transmitting the high-definition photos and 3D scanning data of multiple industrial cameras to the mobile detection platform and computing unit through the corresponding network ports in real time for contact rail detection data analysis and calculation, contact rail status image analysis and storage, and defect status analysis.

[0070] The high-definition camera unit of the contact rail detection unit includes a high-definition industrial camera and fill light for photographing the contact rail surface, a high-definition industrial camera and fill light for photographing the insulating bracket of the contact rail, and a high-definition industrial camera and fill light for photographing the inside of the contact rail protective cover. The contact rail calculation and measurement control unit controls the camera's shooting frequency and the fill light switch.

[0071] The mobile detection platform and computing unit include a portable mobile measurement base, a detachable mobile power supply, a linear displacement sensor for measuring track gauge, a multi-axis tilt sensor for measuring the superelevation of two tracks and monitoring the attitude of the detection equipment, a mileage encoder for measuring mileage and position positioning, and a collection circuit board for collecting and processing sensor data. The detachable mechanical arm equipped with the contact rail detection unit is quickly docked with the mobile measurement base through a quick rotation button.

[0072] The mobile detection platform and computing unit include a high-performance processor, which runs a rapid detection method for subway contact rails; and controls the camera of the contact rail status camera shooting unit, transmits and stores pictures, processes defect states, and manages one pole and one file, and exchanges internal data between the contact rail status picture analysis management system and the contact rail geometric parameter calculation system.

[0073] The mobile detection platform and computing unit use a high-performance processor with a CPU main frequency of 2.5GHz, 8 cores and 16 processes or more to run the rapid detection method of the subway contact rail; as well as control the camera of the contact rail status camera shooting unit, transmit and store pictures, and process defect status. All contact rail photos will be marked with the pole number and mileage value, and will be linked in real time with the contact rail geometric parameters and the contact rail measurement values ​​output by the contact rail wear calculation system, establishing a one-pole-one-file image-measurement data management relationship for rapid review of the contact rail geometric parameter status and defect conditions.

[0074] The mobile detection platform and computing unit include a human-computer interaction part, namely, the viewing and exporting of detection data. The human-computer interaction unit is an output window for the geometric parameter measurement values ​​and status pictures of the subway contact rail. The measurement data is displayed in a waveform diagram and defects are warned in real time.

[0075] The above embodiments are merely illustrative of the principles and effects of the present invention, and are not intended to limit the present invention. Anyone familiar with the technology may modify or change the above embodiments without violating the spirit and scope of the present invention. Therefore, all equivalent modifications or changes made by those with ordinary knowledge in the technical field without departing from the spirit and technical ideas disclosed by the present invention should still be covered by the claims of the present invention.

Claims

1. A rapid detection method for subway contact rails, characterized in that: The detection method comprises the following steps: Step 1: Use the contact rail automatic detection device equipped with a 3D camera to continuously and dynamically scan the subway contact rail, extract each frame of data from the 3D camera, and combine it with the Z-axis mileage sensor data of the contact rail automatic detection device to form a three-dimensional point cloud of the contact rail; Step 2: Filter the point clouds of the lower contact third rail, upper contact third rail, and return rail to remove invalid interference data, and then perform feature multi-data collaborative training to construct unified point cloud features of the contact rail insulation bracket, expansion joint, and end elbow; Step 3: When the contact rail automatic detection device is moving for measurement, the point cloud features of the contact rail's insulating bracket, expansion joint and end elbow are used to judge the contact rail surface, insulating bracket, expansion joint and end elbow in real time, calculate the lead height and pull-out measurement values ​​of these components, and further trigger the high-definition industrial camera to take pictures, merge the pictures and mark the corresponding pole number, kilometer mark position, lead height, and pull-out measurement values.

2. The rapid detection method for subway contact rail according to claim 1 is characterized in that: The point clouds of the lower contact third rail, the upper contact third rail, and the return rail are filtered to remove invalid interference data, and then feature multi-data collaborative training is performed to construct unified point cloud features of the contact rail insulation bracket, expansion joint, and end elbow: S1: Remove the obvious invalid point cloud outside the tunnel wall and the cable contact rail area of ​​the tunnel wall through edge filtering, and further remove the noise through Gaussian filtering; S2: Label the training data sets of the lower contact third rail, upper contact third rail, and return rail respectively, and distinguish the insulating bracket, expansion joint, end elbow, and contact rail surface; S3: Use the annotated insulating brackets, expansion joints, end elbows and contact rail surface point cloud features of the lower contact third rail, upper contact third rail and return rail to cross-validate the unannotated lower contact third rail, upper contact third rail and return rail data sample sets. For example, the lower contact third rail feature value verifies the upper contact third rail sample set, and corrects the point cloud features to ensure that the recognition accuracy meets the requirements. S4: Introduce a new larger validation set, and use the corrected and annotated insulating brackets, expansion joints, end elbows and contact rail surface point cloud features of the lower contact third rail, upper contact third rail and return rail to cross-validate the unannotated lower contact third rail, upper contact third rail and return rail data sample set, and replace the cross-validation sample set object, such as the lower contact third rail feature value to validate the return rail sample set; S5: Compare the accuracy of the point cloud features of the lower contact third rail, upper contact third rail, insulating bracket, expansion joint, end elbow and contact rail surface of the return rail in S4 in the verification set, and take the one with the highest accuracy and reaching the accuracy threshold as the training result point cloud feature. If the accuracy threshold is not reached, repeat S3 and S4.

3. A rapid detection method for subway contact rail according to claim 2, characterized in that: When the contact rail automatic detection device moves and measures, the contact rail surface, insulating bracket, expansion joint and end elbow point cloud features are used to judge the contact rail surface, insulating bracket, expansion joint and end elbow in real time, calculate the guide height and pull-out measurement values ​​of these components, and further trigger the high-definition industrial camera to take pictures, merge the pictures and mark the corresponding pole number, kilometer mark position, guide height and pull-out measurement; S1: When the contact rail automatic detection device is moving for measurement, the contact rail surface, insulating bracket, expansion joint and end elbow point cloud features are used to judge the contact rail surface, insulating bracket, expansion joint and end elbow in real time; S2: Calculate the conduction height and pull-out value of the corresponding insulating bracket, expansion joint and end elbow out of the contact rail surface; S3: Synchronously trigger the high-definition industrial camera to take pictures, merge multiple related pictures into one picture, and mark the corresponding pole number, kilometer mark position, guide height, and pull-out measurement on the picture.

4. A rapid detection method for subway contact rail according to any one of claims 1 to 3, further comprising a detection system, characterized in that: It includes a contact rail detection unit equipped with a 3D camera and a high-definition industrial camera, as well as a mobile detection platform and a computing unit; the contact rail detection unit integrates a 3D camera to measure the contact rail's geometric parameters and wear, and a high-definition industrial camera to take high-definition photos of the contact rail surface, bracket, fastening screws, and end elbows; the mobile detection platform and the computing unit are basic measurement units for measuring the gauge between two rails, the height difference between the two rails, and the moving distance when moving along the rails, a computing and storage unit for contact rail measurement parameters and wear calculation, photo control, and image storage, as well as a human-computer interaction unit for detection result presentation and system control. The contact rail detection unit exchanges data and signals with the mobile detection platform and the computing unit through a network port.

5. A rapid detection system for subway contact rails according to claim 4, characterized in that: The contact rail detection unit includes a high-precision 3D camera for scanning and measuring the contact rail. The 3D camera performs continuous cross-sectional scanning on the subway contact rail. The 3D camera scanning data is combined with the mileage sensor data of the mobile detection platform and the computing unit to form a three-dimensional point cloud of the contact network. The coordinate data of the 3D camera is converted into the center of the rail surface as the coordinate origin, the pull-out direction as the X-axis, and the direction perpendicular to the rail surface as the Y-axis.

6. The rapid detection system for subway contact rail according to claim 4 is characterized in that: The contact rail detection unit includes a high-definition industrial camera and fill light for photographing the contact rail surface, bracket, fastening screws, and end elbow, a circuit board for controlling the camera shooting, a control of the camera shooting frequency and the fill light switch, and a multi-network port switch, which is responsible for transmitting the high-definition photos and 3D scanning data of multiple industrial cameras to the mobile detection platform and computing unit through the corresponding network ports in real time for contact rail detection data analysis and calculation, contact rail status image analysis and storage, and defect status analysis.

7. A rapid detection system for subway contact rails according to claim 4, characterized in that: The mobile detection platform and computing unit include a high-performance processor, which runs a rapid detection method for subway contact rails; and controls the camera of the contact rail status camera shooting unit, transmits and stores pictures, processes defect states, and manages one pole and one file, and exchanges internal data between the contact rail status picture analysis management system and the contact rail geometric parameter calculation system.

Images