Railway passenger platform electronic landmark display method and device
By combining high-definition cameras and multi-line lidar, the electronic landmark display of railway passenger stations is dynamically adjusted, solving the problems of wrong boarding and disorder caused by train stopping position deviations, and achieving accurate carriage number display and orderly passenger boarding.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HENGYUE KONGSEN (SHANGHAI) INFORMATION TECH CO LTD
- Filing Date
- 2026-04-01
- Publication Date
- 2026-06-26
AI Technical Summary
The existing electronic landmark displays at railway passenger stations rely on static preset information and cannot be adjusted in real time, which leads to deviations in train stopping positions, causing passengers to board the wrong train and causing chaos on the platform.
The system uses high-definition cameras to capture train numbers in real time and combines this with multi-line lidar to scan the train's outline. The control module dynamically calculates and updates the carriage numbers on the electronic landmark display screen, enabling precise adjustments.
It enables precise display of train stopping positions, ensuring orderly passenger boarding, maintaining platform order, and adapting to different train models and formations as well as temporary adjustments, without the need for manual intervention.
Smart Images
Figure CN122275967A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of railway transportation technology, specifically to a method and device for displaying electronic landmarks at railway passenger stations. Background Technology
[0002] In railway passenger stations, electronic landmarks are key facilities that guide passengers to quickly find the corresponding carriages. The accuracy of their display directly affects the efficiency and experience of passengers. In the existing technology, the display of electronic landmarks mainly relies on preset train formation information and fixed stopping position data, which has the following shortcomings: (1) High dependence on static preset: The content displayed by electronic landmarks is usually preset according to the scheduling plan before the train arrives at the station. If the actual stopping position of the train deviates from the plan (such as the common error of ±0.5-2 meters), or the train formation is temporarily adjusted (such as adding / removing carriages), the carriage number displayed on the electronic landmark display screen will not match the actual carriage position, resulting in passengers boarding the wrong train. (2) Insufficient dynamic adjustment capability: During the train's entry into the station, the existing electronic landmarks cannot obtain the actual stopping position and carriage length data of the train in real time, and cannot dynamically correct the content displayed on the electronic landmark display screen, which is especially likely to cause platform disorder during peak hours. Summary of the Invention
[0003] The technical problem to be solved by the present invention is to provide a method and device for displaying electronic landmarks on railway passenger platforms, which can dynamically and accurately adjust the carriage number displayed on the electronic landmark display screen according to the real-time stopping position of the train, so as to facilitate passengers to board the train in an orderly manner and maintain good order on the platform.
[0004] The technical solution of this invention is as follows:
[0005] A method for displaying electronic landmarks on railway passenger platform, specifically including the following steps:
[0006] (1) A high-definition camera installed on the railway passenger station platform collects dynamic images of the guidance screen on the railway passenger station platform in real time and sends them to the control module;
[0007] (2) The control module extracts the train number on the guide screen, and then, based on the identified train number, retrieves the standard carriage length and number of carriages corresponding to the train number stored in the data storage module, calculates the total length of the train, and, in conjunction with the length of a single electronic landmark display screen, pre-displays the corresponding train number, carriage number and carriage direction on each electronic landmark display screen.
[0008] (3) During the train's entry into the station and after it stops, the train's side is continuously scanned by a multi-line laser radar deployed along the edge of the railway passenger station platform and sent to the control module in real time. The control module identifies and divides the physical boundaries of each carriage, and then calculates the train's position offset relative to the set reference stopping point after the train stops, as well as the actual length of each carriage.
[0009] (4) The control module calculates the index range of the electronic landmark display screen corresponding to each carriage, thereby updating the carriage number displayed on each electronic landmark display screen.
[0010] The specific steps for the control module to extract the train number from the guidance screen are as follows:
[0011] S21. After receiving the dynamic images of the guide screen captured in real time by the high-definition camera at a fixed frame rate, the control module performs grayscale conversion, histogram equalization, and median filtering noise reduction preprocessing on each frame of the image in sequence.
[0012] S22. The perspective transformation method is used to correct the guide screen area to a standard rectangle, and then the YOLOv5 object detection network is used to extract the region of interest of the guide screen layout, thereby selecting the text line containing the train number.
[0013] S23. Input the image from which the text lines have been extracted into the CRNN model to extract and recognize the train number string;
[0014] S24. Multiple frames of images are continuously captured and independently OCR recognition is performed. A majority voting mechanism is adopted, that is, only when the number of frames that output completely identical strings exceeds the set threshold are these completely identical strings used as candidate results.
[0015] S25. Perform a regular expression comparison between the string in the candidate results and the train number-carriage parameter table stored in the data storage module. If the string is matched in the train number-carriage parameter table, the verification is correct. If it is not matched, proceed to steps S22-S24 for re-identification and matching.
[0016] S26. Parse the verified string into a train number in structured data format.
[0017] The CRNN model first uses a CNN layer to extract spatial features of the input image, then uses an RNN layer to obtain contextual information of the sequence, and finally uses a CTC layer for deduplication and decoding to achieve string recognition and conversion.
[0018] The control module identifies and segments the physical boundaries of each carriage, and then calculates the train's stopping position offset relative to the set reference stopping point after the train stops, as well as the actual length of each individual carriage. The specific steps are as follows:
[0019] S31. The point cloud data in the LiDAR point cloud image is aggregated by voxel grid downsampling, and then statistical filtering is used to remove outlier noise to obtain the noise-reduced train side point cloud image.
[0020] S32. Transform the point cloud data in the train side point cloud image to the global three-dimensional coordinate system using the calibrated extrinsic parameter matrix.
[0021] S33. The Random Sample Consensus (RANSAC) method is used to perform plane fitting on the point cloud data, extract the flat side profile of the train, and remove non-train structure interference.
[0022] S34. Perform depth gradient analysis along the longitudinal coordinate system of the train. By setting a depth threshold and a width window, and combining density clustering method, accurately locate the center coordinates of the carriage connection.
[0023] S35. Kalman filtering is introduced to smoothly track the feature points at the head of the train. When the train is detected to have come to a complete stop, the offset of the train's stopping position relative to the set reference stopping point and the actual length of each carriage are calculated.
[0024] The formula for calculating the index range of the electronic landmark display screen corresponding to each carriage is shown in the following formula (1):
[0025] (1);
[0026] In equation (1), This represents the train's offset from its designated stopping position relative to a set reference stopping point. This represents the actual length of the j-th carriage. Represents the length of a single electronic landmark display screen. This indicates rounding up.
[0027] The high-definition camera and the multi-line LiDAR adopt a timestamp synchronization mechanism. That is, the control module connects to the time synchronization module to obtain the absolute time PPS signal. The high-definition camera and the multi-line LiDAR are triggered by timestamp hardware and receive the PPS signal synchronously using the NTP protocol. At the same time, the control module adds a global timestamp to the point cloud data of each frame of OCR recognition result and LiDAR point cloud image, and performs time window matching. Only data with matching timestamps will be used for subsequent string matching and point cloud data processing.
[0028] During the train's stop, when the train undergoes a secondary position adjustment that causes a change in the carriage position, the multi-line lidar continuously scans and triggers the control module to update the carriage numbers displayed on each electronic landmark screen in real time until the train departs.
[0029] When the radar line of sight of the multi-line lidar is briefly obstructed by an obstacle, the control module uses a Kalman filter prediction module to estimate the boundary position by using the transient train speed and acceleration before the obstruction. After the obstruction is removed, the measured trajectory is pulled back through a data association method to ensure smooth correction.
[0030] When the control module fails to perform OCR recognition of multiple consecutive frames of images, it retrieves the train schedule table stored in the data storage module to obtain the planned train number, and matches the planned train number with the total length of the train measured by the multi-line lidar for basic guidance, and simultaneously sends an alarm to the dispatch room.
[0031] An electronic landmark display device for railway passenger stations includes a control module, a high-definition camera, a multi-line lidar, a data storage module, and an electronic landmark display screen, all connected to the control module.
[0032] The high-definition camera is installed above the railway passenger platform, facing the guidance screen, and is used to capture dynamic images of the guidance screen;
[0033] The multi-line lidar is suspended along the edge of the platform and deployed at intervals to scan the outline of the train.
[0034] The data storage module is used to store the train number-carriage parameter table and the train schedule table;
[0035] The electronic landmark display screen consists of LED display units arranged at intervals along the white lines of the platform, used to display the train number, carriage number, and carriage direction;
[0036] The control module is used to receive dynamic images from the guidance screen, identify the train number, and send the corresponding train number, carriage number, and carriage direction to each electronic landmark display screen for pre-display; it is also used to receive LiDAR point cloud images from multi-line LiDAR scanning, extract and calculate the overall parking position and the actual length of each carriage, calculate the index range of the electronic landmark display screen corresponding to each carriage, and send the carriage number of each carriage to the electronic landmark display screen of the corresponding index range for display.
[0037] Advantages of this invention:
[0038] (1) The present invention uses a high-definition camera to collect the train number of the train to be entering the station, extracts the pre-stored train number, each carriage number and carriage direction and displays them in advance. After the train enters the station and stops stably, the laser radar scan confirms that the train has actually stopped. The carriage numbers displayed on each electronic landmark display screen are then precisely adjusted, which solves the problem of landmark misalignment caused by parking position deviation, facilitates orderly boarding of passengers and maintains good order on the platform.
[0039] (2) This invention covers the full-cycle display needs of trains before entering the station, during entering the station, after stopping and before departure, and is suitable for temporary stop adjustment scenarios.
[0040] (3) The electronic landmark display method of the present invention supports different train types (ordinary trains, EMUs), different formations (8 cars / 16 cars) and temporary adjustment scenarios, and can automatically adapt and adjust the display without manual intervention. Attached Figure Description
[0041] Figure 1 This is a flowchart of the electronic landmark display method of the present invention.
[0042] Figure 2 This is a schematic block diagram of the electronic landmark display device of the present invention.
[0043] Attached reference numerals: 1-Control module, 2-High-definition camera, 3-Multi-line lidar, 4-Data storage module, 5-Electronic landmark display screen. Detailed Implementation
[0044] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0045] See Figure 1 A method for displaying electronic landmarks on railway passenger platform, specifically including the following steps:
[0046] (1) A high-definition camera 2 installed on the railway passenger station platform collects dynamic images of the guidance screen on the railway passenger station platform in real time and sends them to the control module 1;
[0047] (2) The control module 1 extracts the train number on the guide screen, and then, based on the identified train number, retrieves the standard carriage length and number of carriages corresponding to the train number stored in the data storage module 4, calculates the total length of the train, and, in conjunction with the length of a single electronic landmark display screen 5, pre-displays the corresponding train number, carriage number and carriage direction (e.g., “Carriage number: 5; Carriage direction ← 8 carriage 5 carriage 1 carriage →”).
[0048] The specific steps for control module 1 to extract the train number from the guidance screen are as follows:
[0049] S21. After receiving the dynamic images of the guide screen captured in real time by the high-definition camera 2 at a fixed frame rate, the control module 1 performs grayscale conversion, histogram equalization and median filtering noise reduction preprocessing on each frame of the image in sequence.
[0050] S22. The perspective transformation method is used to correct the guide screen area to a standard rectangle, and then the YOLOv5 object detection network is used to extract the region of interest of the guide screen layout, thereby selecting the text line containing the train number.
[0051] S23. The image from which the text lines are extracted is fed into the CRNN model. The CRNN model first uses a CNN layer to extract the spatial features of the input image, then uses an RNN layer to obtain the contextual information of the sequence, and finally uses a CTC layer to perform deduplication and decoding to realize the recognition and conversion of the string, thereby extracting and recognizing the string of the train number.
[0052] S24. Multiple frames of images are continuously captured and independently OCR recognition is performed. A majority voting mechanism is adopted, that is, only when the number of frames that output completely identical strings exceeds the set threshold are these completely identical strings used as candidate results.
[0053] S25. Perform a regular expression comparison between the string in the candidate results and the train number-carriage parameter table stored in the data storage module 4. If the string is matched in the train number-carriage parameter table, the verification is correct. If it is not matched, proceed to steps S22-S24 for re-identification and matching.
[0054] S26. Parse the verified string into a train number in structured data format;
[0055] (3) During the train's entry into the station and after it stops, the multi-line lidar 3 deployed along the edge of the railway passenger station platform continuously scans the side of the train and sends the data to the control module 1 in real time. The control module 1 identifies and segments the physical boundaries of each carriage, and then calculates the train's position offset relative to the set reference stopping point after the train stops, as well as the actual length of each carriage. The specific steps are as follows:
[0056] S31. The point cloud data in the LiDAR point cloud image is aggregated by voxel grid downsampling, and then statistical filtering is used to remove outlier noise to obtain the noise-reduced train side point cloud image.
[0057] S32. Transform the point cloud data in the train side point cloud image to the global three-dimensional coordinate system using the calibrated extrinsic parameter matrix.
[0058] S33. The Random Sample Consensus (RANSAC) method is used to perform plane fitting on the point cloud data, extract the flat side profile of the train, and remove non-train structure interference.
[0059] S34. Perform depth gradient analysis along the longitudinal coordinate system of the train. By setting a depth threshold and a width window, and combining density clustering methods (such as the DBSCAN method), accurately locate the center coordinates of the carriage connection (windshield recess).
[0060] S35. Kalman filtering is introduced to smoothly track the feature points at the head of the train. When the train is detected to have come to a complete stop, the offset of the train's stopping position relative to the set reference stopping point and the actual length of each carriage are calculated.
[0061] (4) The control module 1 calculates the index range of the electronic landmark display screen 5 corresponding to each carriage according to the following formula (1), thereby updating the carriage number displayed on each electronic landmark display screen 5;
[0062] (1);
[0063] In equation (1), This represents the train's offset from its designated stopping position relative to a set reference stopping point. This represents the actual length of the j-th carriage. This represents the length of a single electronic landmark display screen 5;
[0064] The platform white lines are 450 meters long, and there are 90 electronic landmark displays (5 in total). Each electronic landmark display (5) is 5 meters long. The distance of the parking position offset is... The actual length of the first carriage Then the index range of the electronic landmark display screen 5 corresponding to the first carriage is: That is, the 2nd to 7th electronic landmark display screens 5 display the carriage number of the first carriage.
[0065] The high-definition camera 2 and the multi-line LiDAR 3 adopt a timestamp synchronization mechanism. That is, the control module 1 connects to the time synchronization module to obtain the absolute time PPS signal, and the high-definition camera 2 and the multi-line LiDAR 3 perform timestamp hardware triggering and synchronously receive the PPS signal using the NTP protocol. At the same time, the control module 1 adds a global timestamp to the OCR recognition result and the point cloud data of the LiDAR point cloud image for each frame, and performs time window matching. Only data with matching timestamps will be used for subsequent string matching and point cloud data processing.
[0066] During train stops, when the train undergoes a secondary position adjustment that causes a change in the position of the carriages, the multi-line lidar 3 continuously scans and triggers the control module 1 to update the carriage numbers displayed on each electronic landmark display screen 5 in real time until the train departs.
[0067] When the radar line of sight of the multi-line lidar 3 is briefly obstructed by an obstacle, the control module 1 uses a Kalman filter prediction module to estimate the boundary position by using the transient train speed and acceleration before the obstruction. After the obstruction is removed, the measured trajectory is pulled back through the data association method to ensure smooth correction. In the overlapping area scanned by multiple multi-line lidars 3, a weighted fusion algorithm is used to dynamically allocate weights according to the signal-to-noise ratio and distance to eliminate blind spots.
[0068] When extreme weather or a malfunction of the guidance screen causes the control module 1 to fail to perform OCR recognition of multiple consecutive frames of images, the train schedule table stored in the data storage module 4 is retrieved to obtain the planned train number. The total length of the train measured by the multi-line lidar 3 is used to match the planned train number for basic guidance, and an alarm is simultaneously sent to the dispatch room.
[0069] See Figure 2 An electronic landmark display device for railway passenger stations includes a control module 1, a high-definition camera 2, a multi-line lidar 3, a data storage module 4, and an electronic landmark display screen 5;
[0070] High-definition camera 2 (high-definition industrial camera, resolution 1920×1080) is installed above the railway passenger platform, facing the guidance screen, to capture dynamic images of the guidance screen, and is connected to control module 1 via Ethernet;
[0071] The multi-line lidar 3 (144-line lidar, ranging range 0.1-200 meters) is suspended along the edge of the platform and deployed at intervals to scan the outline of the train. It is connected to the control module 1 via a CAN bus.
[0072] Data storage module 4 (SSD hard drive, 1TB capacity, built-in MySQL database) is used to store train number-carriage parameter table and train schedule table;
[0073] The electronic landmark display screen 5 is an LED display unit arranged at intervals along the white line of the platform. It is used to display the train number, the number of each carriage and the direction of the carriage. It is connected to the control module 1 via RS485 bus.
[0074] The control module 1 (embedded processor, such as NVIDIA Jetson Xavier) is used to receive dynamic images from the guide screen, identify the train number, and send the corresponding train number, carriage number, and carriage direction to each electronic landmark display screen 5 for pre-display; it is also used to receive LiDAR point cloud images scanned by multi-line LiDAR 3, extract and calculate the overall parking position and the actual length of each carriage, calculate the index range of the electronic landmark display screen 5 corresponding to each carriage, and send the carriage number of each carriage to the electronic landmark display screen 5 of the corresponding index range for display.
[0075] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.
Claims
1. A railway passenger platform electronic sign display method, characterized in that: Specifically, it includes the following steps: (1) A high-definition camera installed on the railway passenger station platform collects dynamic images of the guidance screen on the railway passenger station platform in real time and sends them to the control module; (2) The control module extracts the train number on the guide screen, and then, based on the identified train number, retrieves the standard carriage length and number of carriages corresponding to the train number stored in the data storage module, calculates the total length of the train, and, in conjunction with the length of a single electronic landmark display screen, pre-displays the corresponding train number, carriage number and carriage direction on each electronic landmark display screen. (3) During the train's entry into the station and after it stops, the train's side is continuously scanned by a multi-line laser radar deployed along the edge of the railway passenger station platform and sent to the control module in real time. The control module identifies and divides the physical boundaries of each carriage, and then calculates the train's position offset relative to the set reference stopping point after the train stops, as well as the actual length of each carriage. (4) The control module calculates the index range of the electronic landmark display screen corresponding to each carriage, thereby updating the carriage number displayed on each electronic landmark display screen.
2. The electronic landmark display method for a railway station platform according to claim 1, characterized in that: The specific steps for the control module to extract the train number from the guidance screen are as follows: S21. After receiving the dynamic images of the guide screen captured in real time by the high-definition camera at a fixed frame rate, the control module performs grayscale conversion, histogram equalization, and median filtering noise reduction preprocessing on each frame of the image in sequence. S22. The perspective transformation method is used to correct the guide screen area to a standard rectangle, and then the YOLOv5 object detection network is used to extract the region of interest of the guide screen layout, thereby selecting the text line containing the train number. S23. Input the image from which the text lines have been extracted into the CRNN model to extract and recognize the train number string; S24. Multiple frames of images are continuously captured and independently OCR recognition is performed. A majority voting mechanism is adopted, that is, only when the number of frames that output completely identical strings exceeds the set threshold are these completely identical strings used as candidate results. S25. Perform a regular expression comparison between the string in the candidate results and the train number-carriage parameter table stored in the data storage module. If the string is matched in the train number-carriage parameter table, the verification is correct. If it is not matched, proceed to steps S22-S24 for re-identification and matching. S26. Parse the verified string into a train number in structured data format.
3. The electronic landmark display method for a railway station platform according to claim 2, characterized in that: The CRNN model first uses a CNN layer to extract spatial features of the input image, then uses an RNN layer to obtain contextual information of the sequence, and finally uses a CTC layer for deduplication and decoding to achieve string recognition and conversion.
4. The electronic landmark display method for a railway station platform according to claim 2, wherein: The control module identifies and segments the physical boundaries of each carriage, and then calculates the train's stopping position offset relative to the set reference stopping point after the train stops, as well as the actual length of each individual carriage. The specific steps are as follows: S31. The point cloud data in the LiDAR point cloud image is aggregated by voxel grid downsampling, and then statistical filtering is used to remove outlier noise to obtain the noise-reduced train side point cloud image. S32. Transform the point cloud data in the train side point cloud image to the global three-dimensional coordinate system using the calibrated extrinsic parameter matrix. S33. The Random Sample Consensus (RANSAC) method is used to perform plane fitting on the point cloud data, extract the flat side profile of the train, and remove non-train structure interference. S34. Perform depth gradient analysis along the longitudinal coordinate system of the train. By setting a depth threshold and a width window, and combining density clustering method, accurately locate the center coordinates of the carriage connection. S35. Kalman filtering is introduced to smoothly track the feature points at the head of the train. When the train is detected to have come to a complete stop, the offset of the train's stopping position relative to the set reference stopping point and the actual length of each carriage are calculated.
5. The electronic landmark display method for a railway station platform according to claim 4, characterized in that: The formula for calculating the index range of the electronic landmark display screen corresponding to each carriage is shown in the following formula (1): （1）； In formula (1), represents the parking position offset amount of the train with respect to the set reference parking point, represents the actual length of the jth car, represents the length of a single electronic landmark display screen, represents the ceiling.
6. The electronic landmark display method for a railway station platform according to claim 4, wherein: The high-definition camera and the multi-line LiDAR adopt a timestamp synchronization mechanism. That is, the control module connects to the time synchronization module to obtain the absolute time PPS signal. The high-definition camera and the multi-line LiDAR are triggered by timestamp hardware and receive the PPS signal synchronously using the NTP protocol. At the same time, the control module adds a global timestamp to the point cloud data of each frame of OCR recognition result and LiDAR point cloud image, and performs time window matching. Only data with matching timestamps will be used for subsequent string matching and point cloud data processing.
7. The method of claim 1, wherein: During the train's stop, when the train undergoes a secondary position adjustment that causes a change in the carriage position, the multi-line lidar continuously scans and triggers the control module to update the carriage numbers displayed on each electronic landmark screen in real time until the train departs.
8. The method of claim 1, wherein: When the radar line of sight of the multi-line lidar is briefly obstructed by an obstacle, the control module uses a Kalman filter prediction module to estimate the boundary position by using the transient train speed and acceleration before the obstruction. After the obstruction is removed, the measured trajectory is pulled back through a data association method to ensure smooth correction.
9. The method of claim 1, wherein: When the control module fails to perform OCR recognition of multiple consecutive frames of images, it retrieves the train schedule table stored in the data storage module to obtain the planned train number, and matches the planned train number with the total length of the train measured by the multi-line lidar for basic guidance, and simultaneously sends an alarm to the dispatch room.
10. An apparatus for implementing the method of claim 1, wherein: It includes a control module, a high-definition camera, a multi-line lidar, a data storage module, and an electronic landmark display screen, all connected to the control module. The high-definition camera is installed above the railway passenger platform, facing the guidance screen, and is used to capture dynamic images of the guidance screen; The multi-line lidar is suspended along the edge of the platform and deployed at intervals to scan the outline of the train. The data storage module is used to store the train number-carriage parameter table and the train schedule table; The electronic landmark display screen consists of LED display units arranged at intervals along the white lines of the platform, used to display the train number, carriage number, and carriage direction; The control module is used to receive dynamic images from the guidance screen, identify the train number, and send the corresponding train number, carriage number, and carriage direction to each electronic landmark display screen for pre-display; it is also used to receive LiDAR point cloud images from multi-line LiDAR scanning, extract and calculate the overall parking position and the actual length of each carriage, calculate the index range of the electronic landmark display screen corresponding to each carriage, and send the carriage number of each carriage to the electronic landmark display screen of the corresponding index range for display.