Platform door train-ground linkage method and system for train type detection

Abstract

The invention provides a platform door train-ground linkage method and system for train type detection, and the method comprises the steps: judging whether a moving target exists in an execution area or not based on an image flow collected by image collection equipment in real time, extracting an image of a corresponding frame if the moving target exists, and judging whether the moving target is a train or not based on the image of the corresponding frame; when it is judged that the moving target is a train, the train speed is monitored through a laser speed measuring radar, and whether the train is a station-crossing train or a stopping train is judged based on the train speed; identifying the serial number of the parked train by adopting an image identification unit, determining the corresponding train type of the train based on the serial number of the train, determining the train door position of the train type, performing image acquisition on the train door position of the train type, identifying the opening and closing of the train door, and controlling the opening and closing of the corresponding platform door; the laser speed measuring radar monitors the speed of the parked train, judges whether the parked train leaves the station or not, and sends out a station leaving signal if the parked train leaves the station.

Description

Technical Field

[0001] The present invention relates to the field of rail transportation technology, and in particular to a platform door-train-ground linkage method and system for train model detection. Background Art

[0002] With the rapid development of modern society, rail transit, as a high-capacity, highly efficient public transportation mode, has become increasingly prominent in people's daily travel. Whether it's the subway network within a city or the main railway lines connecting cities and regions, their operations are continuously growing in scale and complexity, becoming an indispensable component of the modern transportation system. Accurately obtaining train model information is crucial in the actual daily operation of rail transit. Train model information is not only related to the precise execution of scheduling plans and operational organization, but also plays a vital role in platform door matching, equipment maintenance, and emergency response.

[0003] However, traditional methods for detecting train models face numerous challenges. For one thing, traditional methods based on manual observation have significant limitations. These methods rely on the experience of staff members to visually identify train models through external features or markings. Due to the significant influence of human factors, these methods are highly subjective and uncertain. Long hours of work can easily lead to visual fatigue, which can lead to misjudgments. Furthermore, during complex environments such as peak hours with dense train traffic or inclement weather, manual observation becomes even more difficult and error-prone, making it difficult to ensure timely and accurate detection. Furthermore, while existing automated detection systems offer improvements over manual methods, their effectiveness remains constrained by technical limitations. For example, detection systems based on a single sensor often only capture limited information, making them incapable of identifying train models in complex environments and prone to errors. Furthermore, these systems often lack effective integration with ground-based systems, making it impossible to adjust platform door positions, optimize scheduling, or handle emergencies based on model detection results. This isolated system architecture hinders comprehensive improvements in rail transit operational efficiency. Summary of the Invention

[0004] In view of this, an embodiment of the present invention provides a platform door-train-ground linkage method for train model detection to eliminate or improve one or more defects in the prior art.

[0005] One aspect of the present invention provides a platform door vehicle-ground linkage method for train vehicle type detection, the method comprising the following steps:

[0006] Based on the image stream collected in real time by the image acquisition device, it is determined whether there is a moving target in the execution area. If there is a moving target, the image of the corresponding frame is extracted, and based on the image of the corresponding frame, it is determined whether the moving target is a train;

[0007] When the moving target is determined to be a train, the train speed is monitored by a laser speed radar, and the train is determined to be a passing train or a stopped train based on its speed;

[0008] For stopped trains, an image recognition unit is used to identify the train number, determine the corresponding train model based on the train number, and determine the door position of the model. The image of the door position of the model is captured, the door opening and closing is identified, and the corresponding platform door is controlled to open and close.

[0009] The laser speed radar monitors the speed of the stopped train and determines whether the stopped train has left the station. If it has left the station, it will send a departure signal.

[0010] By adopting the above scheme, this scheme can carry out comprehensive coordinated processing of trains. When a train enters the station, it can accurately identify the entry and determine whether the vehicle is a stopped train or a non-stop train based on the speed of the train. If it is a non-stop train, the speed is continuously monitored, and a departure signal is issued when the train leaves the station; if it is a stopped train, the model is identified and the position of the door is determined. When the train stops, the opening and closing of the door is identified, and the corresponding platform door is synchronously controlled to open and close, completing the accurate linkage between the train and the platform, and then issuing a departure signal when the train finally leaves the station to ensure the overall operational efficiency of rail transit.

[0011] In some embodiments of the present invention, in the step of determining whether there is a moving target in the execution area based on the image stream captured in real time by the image acquisition device, a filtering algorithm is first used to preprocess the frame image of the image stream to eliminate environmental interference; and the previously set background image of the track area is used as a comparison benchmark, and the current frame image is compared with the background image, and a differential algorithm is used to detect whether there is a moving target in the track area.

[0012] In some embodiments of the present invention, in the step of determining whether a moving target is a train based on the image of the corresponding frame, edge detection and template matching techniques are used to confirm whether the target has a typical geometric outline of a train. When the characteristics of the target meet the characteristics of a train and the differential result exceeds a preset threshold, the train is confirmed.

[0013] In some embodiments of the present invention, in the step of monitoring the train speed by means of a laser speed measuring radar and determining whether the train is a passing train or a stopped train based on the train speed, the speed change of the train is counted in real time. If the speed change of the train is gradually slowing down and the speed drops to zero at the predetermined stop sign, the train is determined to be a stopped train; if the speed change of the train is not gradually slowing down, the train is determined to be a passing train.

[0014] In some embodiments of the present invention, a train number is sprayed on the train, and the step of using an image recognition unit to identify the train number of a stopped train includes:

[0015] Image preprocessing: using image enhancement technology to improve image contrast and clarity, and using deflection correction algorithms to adjust the tilt and angle errors in the image;

[0016] Feature extraction: A convolutional neural network is used to extract the feature map of the train number area. The target detection algorithm is used to generate a candidate frame and lock the area containing the number characters in the feature map.

[0017] Character determination, a classification regression model is used to further analyze the image content within each candidate box and output the corresponding character.

[0018] In some embodiments of the present invention, in the steps of determining the corresponding model of a train based on the train number and determining the door position of the model, a match is performed in the database based on the identified train number to determine the corresponding model of the train and call the door position in the model data.

[0019] In some embodiments of the present invention, the step of determining the corresponding model of the train based on the train number and determining the door position of the model further includes:

[0020] Use laser ranging radar to measure the length of the train and determine the number of vehicles in the train;

[0021] The distance to each formation is determined by laser ranging radar, and the data of the corresponding model of the train determined based on the train number is compared with the length of the train measured by laser ranging radar, the number of formations and the distance to each formation determined by laser ranging radar to determine whether they match.

[0022] In some embodiments of the present invention, the image recognition unit adopts a combination of Faster R-CNN and LSTM models. In the steps of capturing images of the door positions of the vehicle model, identifying the opening and closing of the doors, and controlling the opening and closing of the corresponding platform doors:

[0023] The image recognition unit is trained using pre-set training data. The training data includes previously collected videos of train door opening and closing movements, including images in different lighting, occlusion, and vibration scenarios. The labeling tool LabelImg is used to annotate each frame with the door's bounding box and state, including open, closed, and half-open. A time label is added to each frame in the time series to indicate the time progression. Frame extraction, normalization, and time series segmentation are then performed for LSTM model training. Faster R-CNN is then used to annotate the door position and motion trajectory. A trajectory sequence is constructed from the bounding box coordinates of each frame, and the door's speed and direction are calculated through inter-frame differencing. The time series frame feature vectors extracted by Faster R-CNN are then combined in chronological order and input into the LSTM model for training.

[0024] The image recognition unit identifies the video of the corresponding train door, determines whether the train door is open, closed or half-open, and controls the opening and closing of the corresponding platform door.

[0025] In some embodiments of the present invention, in the step of monitoring the speed of a stopped train by a laser speed measuring radar, determining whether the stopped train has left the station, and issuing a departure signal if the train has left the station:

[0026] The laser speed radar is used to measure the train speed, which gradually increases. After a delay period, when the laser ranging radar detection distances are all greater than the threshold distance, a train departure signal is issued.

[0027] The second aspect of the present invention also provides a platform door-train-ground linkage system for train model detection, the system including a computer device, the computer device including a processor and a memory, the memory storing computer instructions, the processor being used to execute the computer instructions stored in the memory, and when the computer instructions are executed by the processor, the system implements the steps implemented by the method described above.

[0028] The third aspect of the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the platform door-train-ground linkage method for the aforementioned train model detection.

[0029] Additional advantages, objects, and features of the present invention will be described in part in the following description and will become apparent to those skilled in the art after studying the following or may be learned by practice of the present invention. The objects and other advantages of the present invention may be particularly pointed out and attained in the description and drawings.

[0030] Those skilled in the art will understand that the purposes and advantages that can be achieved by the present invention are not limited to the above specific descriptions, and the above and other purposes that can be achieved by the present invention will be more clearly understood based on the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] The drawings described herein are used to provide a further understanding of the present invention, constitute a part of this application, and do not constitute a limitation of the present invention.

[0032] Figure 1 A schematic diagram of an embodiment of the platform door-train-ground linkage method for train model detection according to the present invention;

[0033] Figure 2 A schematic diagram of the processing architecture of the platform door-train-ground linkage method for train model detection according to the present invention;

[0034] Figure 3 This is a schematic diagram of the train entry and exit speed curve of the present invention;

[0035] Figure 4 Schematic diagram of the processing flow of train number identification of the present invention. DETAILED DESCRIPTION

[0036] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments and the accompanying drawings. Here, the exemplary embodiments of the present invention and their descriptions are used to explain the present invention, but are not intended to limit the present invention.

[0037] It should also be noted that, in order to avoid obscuring the present invention due to unnecessary details, the accompanying drawings only show structures and / or processing steps closely related to the solutions according to the present invention, while other details that are not closely related to the present invention are omitted.

[0038] Given these challenges, a new approach to train model detection, integrating platform doors with train-to-ground systems, is needed. This approach leverages real-time data and complex models to improve prediction accuracy and practicality. Such a new approach should better meet the needs of modern traffic management and provide more reliable decision support for traffic managers.

[0039] like Figure 1 and 2 As shown, the present invention proposes a platform door vehicle-ground linkage method for train model detection, the method comprising the following steps:

[0040] Step S100, determining whether there is a moving target in the execution area based on the image stream captured in real time by the image acquisition device, and if there is a moving target, extracting an image of the corresponding frame, and determining whether the moving target is a train based on the image of the corresponding frame;

[0041] During specific implementation, the image acquisition device may be a camera.

[0042] Step S200: When the moving target is determined to be a train, the train speed is monitored by a laser speed measuring radar, and the train is determined to be a passing train or a stopped train based on the train speed;

[0043] During implementation, an image recognition unit captures images of the track area near the platform's finished surface and uses motion detection technology to determine if a train is approaching the station. A laser speed radar monitors the slowdown phase to accurately identify a train's arrival. When the train's speed drops to zero, data from the laser ranging radar is simultaneously collected, and a final signal confirming the train's arrival is issued upon verification that the train has stopped.

[0044] Step S300: For a stopped train, an image recognition unit is used to identify the train number, determine the corresponding train model based on the train number, determine the door positions of the model, capture images of the door positions of the model, identify the opening and closing of the doors, and control the opening and closing of the corresponding platform doors;

[0045] Specifically, the train doors open, passengers get on and off, and then the train doors close. During this process, the train door opening and closing actions are identified and the corresponding platform doors are controlled to open and close synchronously with the train doors.

[0046] In step S400, the laser speed measuring radar monitors the speed of the stopped train and determines whether the stopped train has left the station. If so, a departure signal is issued.

[0047] Specifically, when the platform and train doors are closed and the train is ready to depart, the laser speed radar detects the increase in speed and determines when the train has begun to depart the station. The laser ranging radar collects data simultaneously and issues a departure signal after verification.

[0048] By adopting the above scheme, this scheme can carry out comprehensive coordinated processing of trains. When a train enters the station, it can accurately identify the entry and determine whether the vehicle is a stopped train or a non-stop train based on the speed of the train. If it is a non-stop train, the speed is continuously monitored, and a departure signal is issued when the train leaves the station; if it is a stopped train, the model is identified and the position of the door is determined. When the train stops, the opening and closing of the door is identified, and the corresponding platform door is synchronously controlled to open and close, completing the accurate linkage between the train and the platform, and then issuing a departure signal when the train finally leaves the station to ensure the overall operational efficiency of rail transit.

[0049] Specifically, the vehicle-ground linkage system for train model detection mainly consists of an image recognition unit, a lidar detection unit, a main control system, an auxiliary management system, a door control unit, and a three-dimensional protection unit;

[0050] The image recognition unit consists of an image acquisition device and a front-end intelligent recognition terminal, which together complete the image acquisition and the recognition of train entry and exit, train model, marshalling information and train door opening and closing actions.

[0051] The laser radar detection unit consists of multiple laser speed measurement radars and laser ranging radars, which are jointly responsible for determining whether a train enters a station, stops, or starts to depart. In addition, the ranging radar is responsible for determining the length of the train formation.

[0052] The main control system, consisting of two core controllers, input / output interfaces, and a human-machine interface (HMI), is responsible for process control and logic operations throughout the system. The two core controllers are redundant, forming a "two-out-of-two" architecture that reduces the possibility of system malfunctions and enhances system reliability and safety. The HMI, located at the platform local control panel (PSL), is responsible for image display and system configuration, allowing station staff to observe the situation between trains and platform doors and control the doors.

[0053] The auxiliary management system consists of a video surveillance and storage server and a system management server, and is responsible for system configuration management, deep learning training, and the storage and backup of system operation information, fault information, and image information.

[0054] The door control unit receives control instructions from the main control system and is responsible for controlling the opening and closing of the corresponding platform door. At the same time, it collects platform door operation information and uploads it to the main control system.

[0055] The three-dimensional protection unit integrates multiple technical methods such as millimeter-wave radar and video analysis. It is responsible for detecting whether there are obstacles in the three-dimensional space between the train and the platform door, and returning the detection results to the main control system. The results will also be displayed on the human-computer interaction interface.

[0056] In some embodiments of the present invention, in the step of determining whether there is a moving target in the execution area based on the image stream captured in real time by the image acquisition device, a filtering algorithm is first used to preprocess the frame image of the image stream to eliminate environmental interference; and the previously set background image of the track area is used as a comparison benchmark, and the current frame image is compared with the background image, and a differential algorithm is used to detect whether there is a moving target in the track area.

[0057] In some embodiments of the present invention, in the step of determining whether a moving target is a train based on the image of the corresponding frame, edge detection and template matching techniques are used to confirm whether the target has a typical geometric outline of a train. When the characteristics of the target meet the characteristics of a train and the differential result exceeds a preset threshold, the train is confirmed.

[0058] Specifically, this solution installs cameras near the height of the platform's finished surface to ensure coverage of key areas of the track area. The cameras are calibrated to capture clear real-time image streams and transmit them to the front-end intelligent recognition terminal in real time. A filtering algorithm is first used to pre-process the image to eliminate interference caused by changes in ambient light, dust, and weather conditions. Using the previously set background image of the track area as a comparison benchmark, the current frame image is compared with the background image, and a differential algorithm is used to detect whether there are moving targets in the track area. The system extracts possible train images by calculating the target's contour area, movement speed, and other features. After detecting a dynamic target, the system will further use edge detection and template matching technology to confirm whether the target has a typical train geometric outline. When the target's dynamic changes, shape characteristics and other indicators meet the train characteristics and the differential result exceeds the threshold set by the system, the system confirms that the train is entering the station.

[0059] In some embodiments of the present invention, in the step of monitoring the train speed by means of a laser speed measuring radar and determining whether the train is a passing train or a stopped train based on the train speed, the speed change of the train is counted in real time. If the speed change of the train is gradually slowing down and the speed drops to zero at the predetermined stop sign, the train is determined to be a stopped train; if the speed change of the train is not gradually slowing down, the train is determined to be a passing train.

[0060] Specifically, this solution uses a laser speed radar to detect speed when confirming that the train is entering the station, and a laser ranging radar to detect distance. The laser speed radar is located on the crossbeam of the platform door, and its purpose is to continuously collect speed data of passing trains in real time facing the track. Passing trains and stopped trains are distinguished based on the speed readings, and the train's entry, stop, and start-up departure behaviors are identified through analysis of speed changes. Compared with passing vehicles, the train shows a characteristic of gradually slowing down in the entry stage until the speed drops to zero at the predetermined stop sign. After completing the process of passengers getting on and off, the train starts and gradually increases its speed. Its speed change curve is as follows Figure 3 shown.

[0061] By monitoring the deceleration phase, a train's arrival can be accurately identified. When the train's speed drops to zero, after a short delay period t, the train can be confirmed to have stabilized at a station. To ensure accurate results, simultaneous LiDAR data acquisition is required. The measured linear distance from the partially acquired LiDAR to the train is less than a threshold value, S1. (If there is no train on the platform, the total output value of the LiDAR will significantly exceed S1.) A final signal indicating a stable train arrival is issued upon verification of the train's arrival.

[0062] In some embodiments of the present invention, a train number is sprayed on the train, and the step of using an image recognition unit to identify the train number of a stopped train includes:

[0063] Image preprocessing: using image enhancement technology to improve image contrast and clarity, and using deflection correction algorithms to adjust the tilt and angle errors in the image;

[0064] Feature extraction: A convolutional neural network is used to extract the feature map of the train number area. The target detection algorithm is used to generate a candidate frame and lock the area containing the number characters in the feature map.

[0065] Character determination, a classification regression model is used to further analyze the image content within each candidate box and output the corresponding character.

[0066] In some embodiments of the present invention, in the steps of determining the corresponding model of a train based on the train number and determining the door position of the model, a match is performed in the database based on the identified train number to determine the corresponding model of the train and call the door position in the model data.

[0067] During the specific implementation process, the train number is sprayed on the outside of the train body at the front of the train. After receiving the stable signal of the train entering the station, the image recognition unit begins to collect train image information, obtain the train model and formation information, and transmit it to the main control system. The relevant information is retrieved from the database to know the door position of the current model.

[0068] In some embodiments of the present invention, the step of determining the corresponding model of the train based on the train number and determining the door position of the model further includes:

[0069] Use laser ranging radar to measure the length of the train and determine the number of vehicles in the train;

[0070] The distance to each formation is determined by laser ranging radar, and the data of the corresponding model of the train determined based on the train number is compared with the length of the train measured by laser ranging radar, the number of formations and the distance to each formation determined by laser ranging radar to determine whether they match.

[0071] Specifically, the laser ranging radar measures the train formation based on the straight-line distance from each different installation point to the train, and determines whether it matches the result measured by the image recognition unit.

[0072] Specifically, the train identification of each marshaling is based on the following Table 1, where S and S1 are set thresholds:

[0073] Table 1

[0074] Section 1 and 2 Distance Measurement Sections 7 and 8 Distance Measurement Sections 9 and 10 Ranging Section 16 Ranging Section 17 Ranging Grouping form S~S1 S~S1 >S1 >S1 >S1 First 8 groups >S1 >S1 S~S1 S~S1 >S1 Rear 8 formations S~S1 S~S1 S~S1 S~S1 >S1 16 groups S~S1 S~S1 S~S1 S~S1 S~S1 17 groups >S1 >S1 >S1 >S1 >S1 No train stops

[0075] Specifically, the detection of train formation length mainly relies on the laser ranging radar installed on the platform door crossbeam. The radar collects distance information toward the track in a direction perpendicular to the track. Multiple laser radars are configured on each platform side to monitor the position of each train car respectively. By analyzing the number of blocked laser radars and their distribution, the formation length of the train entering the station can be inferred. When the train stops, the laser radar records the straight-line distance S from the installation location to the train. If there is no train on the platform, the output value of the laser radar will be significantly greater than S. For this reason, a threshold S1 (that is, S plus a certain margin) is set as the judgment standard. By comparing the radar output value with the threshold S1 in real time, the train entry situation can be accurately identified and its length can be measured. In view of the fact that the current train formation is 8-formation, 16-formation and 17-formation, in order to improve the reliability of detection, each detection point is equipped with two sets of laser radars. Based on the effective data of these 8 sets of ranging laser radars, the formation type of the train can be accurately determined, as shown in the following figure. Figure 4 And judge whether it is consistent with the result measured by the image recognition unit when identifying the vehicle model.

[0076] In some embodiments of the present invention, the image recognition unit adopts a combination of Faster R-CNN and LSTM models. In the steps of capturing images of the door positions of the vehicle model, identifying the opening and closing of the doors, and controlling the opening and closing of the corresponding platform doors:

[0077] The image recognition unit is trained using pre-set training data. The training data includes previously collected videos of train door opening and closing movements, including images in different lighting, occlusion, and vibration scenarios. The labeling tool LabelImg is used to annotate each frame with the door's bounding box and state, including open, closed, and half-open. A time label is added to each frame in the time series to indicate the time progression. Frame extraction, normalization, and time series segmentation are then performed for LSTM model training. Faster R-CNN is then used to annotate the door position and motion trajectory. A trajectory sequence is constructed from the bounding box coordinates of each frame, and the door's speed and direction are calculated through inter-frame differencing. The time series frame feature vectors extracted by Faster R-CNN are then combined in chronological order and input into the LSTM model for training.

[0078] During the specific implementation process, videos of the train door opening and closing actions have been collected in the early stage, including complex scenes such as different lighting, occlusion, and vibration. The labeling tool LabelImg is used to label the bounding box and state (open, closed, half-open) of the door for each frame, and a time label is added to each frame in the time series to represent the time process. Frame extraction, normalization, and time series division are then performed for LSTM model training. After that, Faster R-CNN is used to implement the door position and motion trajectory annotation, construct a trajectory sequence from the bounding box coordinates of each frame, and calculate the door speed and direction by inter-frame difference. The time series frame feature vectors extracted by Faster R-CNN are then combined in chronological order, input and trained into the LSTM model. Finally, Faster R-CNN and LSTM model inference is performed. After these preliminary preparations, the image recognition unit LSTM model is used to determine the open or closed state of the door in real time. At the moment the train door opening action is recognized, this signal is transmitted to the main control system;

[0079] The image recognition unit identifies the video of the corresponding train door, determines whether the train door is open, closed or half-open, and controls the opening and closing of the corresponding platform door.

[0080] The image recognition unit identifies train door openings and closings and controls the corresponding platform doors to open and close synchronously with the train. A set of image recognition units was installed at the door locations of carriages 4, 5, 11, and 12, capturing door information from the train body in real time. Video footage of the train door opening and closing motion was previously captured, including complex scenarios such as varying lighting conditions, occlusion, and vibration. The LabelImg annotation tool was used to annotate each frame with the door's bounding box and state (open, closed, or half-open). Time labels were added to each frame in the time series to represent the temporal progression. Frame extraction, normalization, and time series segmentation were then performed for LSTM model training. Faster R-CNN was then used to annotate door positions and motion trajectories. Trajectory sequences were constructed from the bounding box coordinates of each frame, and door speed and direction were calculated through inter-frame differencing. The feature vectors of the time series frames extracted by Faster R-CNN were then combined in chronological order and fed into the LSTM model for training. Finally, the Faster R-CNN and LSTM model inference was performed. After these preliminary preparations, the image recognition unit LSTM model determined the door's open or closed state in real time. The moment a train door opening is detected, this signal is transmitted to the main control system. Since doors on the same side of the train can only open or close simultaneously, all platform doors on the same side corresponding to the train length measured in the previous step will be opened at this time; the same process applies to closing doors.

[0081] In some embodiments of the present invention, image enhancement technology is used to improve the contrast and clarity of the image and enhance the recognition performance under low light or complex background conditions; a deflection correction algorithm is used to adjust the tilt and angle errors in the image so that the numbered area remains horizontally and vertically aligned. After the image preprocessing is completed, the image recognition unit uses a convolutional neural network (CNN) to extract features. CNN extracts the feature map of the train number area through layer-by-layer convolution and pooling operations. The target detection algorithm is used to generate candidate frames and lock the area that may contain numbered characters. These candidate frames are uniformly adjusted to a fixed size through region of interest pooling (ROI), while retaining the key information in the area, providing standardized input for subsequent classification regression. Based on the target detection area, a classification regression model is used to further analyze the image content in each candidate frame. The classification module is responsible for classifying the characters in the image into specific numbers or letters.

[0082] The train's train number is typically painted on the exterior of the train's front end, with the first few characters indicating the train's model. These characters consist of numbers and English letters, with black text on a white background. The image recognition unit captures train images as the train enters the station and preprocesses the raw image data. First, image enhancement techniques are used to improve image contrast and clarity, enhancing recognition performance in low-light or complex background conditions. Second, a deflection correction algorithm is used to adjust for tilt and angular errors in the image, maintaining horizontal and vertical alignment of the train number region. After image preprocessing, the image recognition unit uses a convolutional neural network (CNN) for feature extraction. The CNN extracts feature maps of the train number region through layer-by-layer convolution and pooling operations. An object detection algorithm is then used to generate candidate boxes, identifying areas that may contain the train number characters. These candidate boxes are then resized to a fixed size using region of interest (ROI) pooling, while preserving key information within the region and providing standardized input for subsequent classification and regression. Based on the target detection area, a classification and regression model is used to further analyze the image content within each candidate frame. The classification module is responsible for classifying the characters in the image into specific numbers or letters, while the regression module is used to correct the deviation of the character position and size, thereby improving the recognition accuracy. Finally, the results of classification and regression are converted into numbered text form and transmitted to the main control system. The relevant information is retrieved from the database to know the door position of the current car model. The overall process is as follows Figure 4 shown.

[0083] In some embodiments of the present invention, in the step of monitoring the speed of a stopped train by a laser speed measuring radar, determining whether the stopped train has left the station, and issuing a departure signal if the train has left the station:

[0084] The laser speed radar is used to measure the train speed, which gradually increases. After a delay period, when the laser ranging radar detection distances are all greater than the threshold distance, a train departure signal is issued.

[0085] The laser speed radar measures that the train speed gradually increases. After a short delay period t, the laser ranging radar detects distances that are all greater than S1, and the system issues a train departure signal.

[0086] With the rapid expansion of the rail transit network, the train formation forms are becoming increasingly diversified under the integration of the four networks. The large-scale use of different models has put forward higher requirements on the accuracy, speed, and real-time and reliability of the detection system's linkage with the ground system. The current technical solutions have been unable to adapt to the increasingly complex operating scenarios, and there is an urgent need for a more advanced, comprehensive and efficient train model detection system. The present invention can collect train model information through an image recognition unit, and at the same time use laser speed measurement and ranging radar to ensure that the train stops steadily and collects the formation length information, complete accurate platform door matching, and improve the efficiency of vehicle-ground linkage. It can not only ensure the safe operation of rail transit, but also significantly optimize operational management, and promote the rapid development of rail transit modernization.

[0087] This solution involves rail transit, including railways, urban rail transit and other types of rail transit. It is particularly suitable for the development of multi-vehicle co-line operation under the background of four-network integration. Through machine vision-based OCR (Optical Character Recognition) technology and multi-lidar recognition technology, the vehicle model information and marshaling information of incoming trains are identified, and based on this information, more efficient and reliable linkage between trains and the ground is achieved.

[0088] An embodiment of the present invention also provides a platform door-train-ground linkage system for train model detection, the system including a computer device, the computer device including a processor and a memory, the memory storing computer instructions, the processor being used to execute the computer instructions stored in the memory, and when the computer instructions are executed by the processor, the system implements the steps implemented by the method described above.

[0089] An embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the computer program implements the steps of the platform door-train-ground linkage method for train model detection. The computer-readable storage medium can be a tangible storage medium, such as a random access memory (RAM), internal memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, register, floppy disk, hard disk, removable storage disk, CD-ROM, or any other form of storage medium known in the art.

[0090] It should be understood by those skilled in the art that the various exemplary components, systems and methods described in conjunction with the embodiments disclosed herein can be implemented in hardware, software or a combination of the two. Whether it is specifically performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention. When implemented in hardware, it can be, for example, an electronic circuit, an application specific integrated circuit (ASIC), appropriate firmware, a plug-in, a function card, etc. When implemented in software, the elements of the present invention are programs or code segments that are used to perform the required tasks. The program or code segment can be stored in a machine-readable medium, or transmitted on a transmission medium or a communication link via a data signal carried in a carrier.

[0091] It should be understood that the present invention is not limited to the specific configurations and processes described above and illustrated in the figures. For the sake of brevity, a detailed description of known methods is omitted. In the above embodiments, several specific steps are described and illustrated as examples. However, the method of the present invention is not limited to the specific steps described and illustrated. Those skilled in the art may make various changes, modifications, and additions, or change the order of the steps after understanding the spirit of the present invention.

[0092] In the present invention, features described and / or illustrated for one embodiment may be used in the same or similar manner in one or more other embodiments, and / or combined with or replace features of other embodiments.

[0093] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations to the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.

Claims

1. A platform door vehicle-ground linkage method for train model detection, characterized in that: The steps of the method include: Based on the image stream collected in real time by the image acquisition device, it is determined whether there is a moving target in the execution area. If there is a moving target, the image of the corresponding frame is extracted, and based on the image of the corresponding frame, it is determined whether the moving target is a train; When the moving target is determined to be a train, the train speed is monitored by a laser speed radar, and the train is determined to be a passing train or a stopped train based on its speed; For stopped trains, an image recognition unit is used to identify the train number, determine the corresponding train model based on the train number, and determine the door position of the model. The image of the door position of the model is captured, the door opening and closing is identified, and the corresponding platform door is controlled to open and close. The laser speed radar monitors the speed of the stopped train and determines whether the stopped train has left the station. If it has left the station, it will send a departure signal.

2. The platform door and vehicle-ground linkage method for train model detection according to claim 1 is characterized in that: In the step of determining whether there is a moving target in the execution area based on the image stream collected in real time by the image acquisition device, a filtering algorithm is first used to pre-process the frame image of the image stream to eliminate environmental interference; The previously set background image of the track area is used as a comparison benchmark, and the current frame image is compared with the background image. A differential algorithm is used to detect whether there is a moving target in the track area.

3. The platform door and vehicle-ground linkage method for train model detection according to claim 1 is characterized in that: In the step of determining whether the moving target is a train based on the image of the corresponding frame, edge detection and template matching technology are used to confirm whether the target has a typical geometric outline of a train. When the characteristics of the target meet the characteristics of a train and the differential result exceeds a preset threshold, the train is confirmed.

4. The platform door and vehicle-ground linkage method for train model detection according to claim 1 is characterized in that: In the step of monitoring the train speed by means of a laser speed measuring radar and determining whether the train is a passing train or a stopped train based on the train speed, the speed change of the train is counted in real time. If the speed change of the train is gradually slowing down and the speed drops to zero at the predetermined stop sign, the train is determined to be a stopped train; if the speed change of the train is not gradually slowing down, the train is determined to be a passing train.

5. The platform door and vehicle-ground linkage method for train model detection according to claim 1 is characterized in that: The train number is sprayed on the train, and the step of using the image recognition unit to identify the train number of the stopped train includes: Image preprocessing: using image enhancement technology to improve image contrast and clarity, and using deflection correction algorithms to adjust the tilt and angle errors in the image; Feature extraction: A convolutional neural network is used to extract the feature map of the train number area. The target detection algorithm is used to generate a candidate frame and lock the area containing the number characters in the feature map. Character determination, a classification regression model is used to further analyze the image content within each candidate box and output the corresponding character.

6. The platform door and vehicle-ground linkage method for train model detection according to any one of claims 1 to 5, characterized in that: In the step of determining the corresponding model of the train based on the train number and determining the door position of the model, matching is performed in the database based on the identified train number, the corresponding model of the train is determined, and the door position in the model data is called.

7. The platform door and vehicle-ground linkage method for train model detection according to claim 6 is characterized in that: The steps of determining the corresponding model of the train based on the train number and determining the door position of the model of the train also include: Use laser ranging radar to measure the length of the train and determine the number of vehicles in the train; The distance to each formation is determined by laser ranging radar, and the data of the corresponding model of the train determined based on the train number is compared with the length of the train measured by laser ranging radar, the number of formations and the distance to each formation determined by laser ranging radar to determine whether they match.

8. The platform door and vehicle-ground linkage method for train model detection according to claim 1 is characterized in that: The image recognition unit uses a combination of Faster R-CNN and LSTM models to capture images of the door positions of the vehicle model, identify the opening and closing of the doors, and control the opening and closing of the corresponding platform doors: The image recognition unit is trained using pre-set training data. The training data includes previously collected videos of train door opening and closing movements, including images in different lighting, occlusion, and vibration scenarios. The labeling tool LabelImg is used to annotate each frame with the door's bounding box and state, including open, closed, and half-open. A time label is added to each frame in the time series to indicate the time progression. Frame extraction, normalization, and time series segmentation are then performed for LSTM model training. Faster R-CNN is then used to annotate the door position and motion trajectory. A trajectory sequence is constructed from the bounding box coordinates of each frame, and the door's speed and direction are calculated through inter-frame differencing. The time series frame feature vectors extracted by Faster R-CNN are then combined in chronological order and input into the LSTM model for training. The image recognition unit identifies the video of the corresponding train door, determines whether the train door is open, closed or half-open, and controls the opening and closing of the corresponding platform door.

9. The platform door and vehicle-ground linkage method for train model detection according to claim 1, characterized in that: In the step of monitoring the speed of a stopped train by a laser speed measuring radar and determining whether the stopped train has left the station, and issuing a departure signal if the train has left the station: The laser speed radar is used to measure the train speed, which gradually increases. After a delay period, when the laser ranging radar detection distances are all greater than the threshold distance, a train departure signal is issued.

10. A platform door vehicle-ground linkage system for train model detection, characterized in that: The system includes a computer device, which includes a processor and a memory. The memory stores computer instructions. The processor is used to execute the computer instructions stored in the memory. When the computer instructions are executed by the processor, the system implements the steps implemented by the method according to any one of claims 1 to 9.

Images