Platform safety door and train door collaborative alignment calibration parking positioning method and system

Abstract

The invention provides a platform safety door and train door collaborative alignment calibration parking positioning method and system, and relates to the technical field of rail safety, and the method comprises the steps: obtaining train position and speed information, and determining the position and deceleration of a safety door; images of a platform safety door and a train door are collected and registered to obtain relative position deviation; generating a position adjusting instruction according to the deviation and the deceleration and sending the instruction to a driving system; calculating a real-time alignment error according to the feedback; and when the error is smaller than a threshold, sending a parking instruction to realize stable parking. According to the invention, the train parking precision is improved, the passenger safety is enhanced, and the station operation efficiency is optimized.

Description

Technical Field

[0001] The present invention relates to the field of track safety technology, and in particular to a method and system for coordinated alignment, calibration, and docking positioning of platform safety doors and train doors. Background Art

[0002] With the rapid development of urban rail transit, platform safety gate systems have become standard equipment on modern subway platforms. Their primary function is to isolate the platform from the track area during train entry and exit, ensuring passenger safety. When a train stops, precise alignment of the platform safety gate with the train door is crucial for ensuring convenient and safe passenger boarding and alighting. Traditionally, train stop positioning relies primarily on ground beacons, onboard speed measuring devices, and driver experience. However, with the advancement of autonomous driving technology, the coordinated alignment of trains and platform safety gates has become a research hotspot.

[0003] At present, the alignment technology for trains and platform safety doors still has some obvious shortcomings. Existing positioning methods mainly rely on ground-fixed beacons or track-side equipment, which have high requirements for external environmental conditions and are easily affected by factors such as weather changes and equipment aging, resulting in unstable positioning accuracy. Traditional alignment technology lacks a real-time dynamic calibration mechanism and cannot be accurately adjusted according to the actual operating status of the train and the actual position relationship of the platform safety door, which may cause the door and the safety door to be misaligned after the train stops. Existing technologies generally lack precise alignment solutions based on visual recognition, making it difficult to achieve high-precision docking at the centimeter level. Especially in scenarios with high-density operations and frequent trains entering and exiting stations during peak hours, alignment errors may accumulate and amplify, affecting the efficiency and safety of passengers getting on and off the train. Summary of the Invention

[0004] The embodiments of the present invention provide a method and system for coordinated alignment, calibration, and docking of platform safety doors and train doors, which can solve the problems in the prior art.

[0005] A first aspect of an embodiment of the present invention provides a method for coordinated alignment, calibration, and docking positioning of a platform safety door and a train door, comprising: Determine the location of the platform safety door and the real-time deceleration of the train based on the acquired train head position information and train running speed information; Collecting first image information on both sides of the platform safety door frame and second image information on both sides of the train door frame, and obtaining a relative position deviation value between the platform safety door and the train door through image registration of the first image information and the second image information; generating a train position adjustment instruction including a position compensation parameter according to the relative position deviation value and the real-time deceleration, and sending the train position adjustment instruction to the train automatic driving system; receiving position adjustment feedback information returned by the automatic train driving system based on a preset adjustment step size, calculating a real-time alignment error between the platform safety door and the train door according to the position adjustment feedback information, and sending a stop command to the automatic train driving system when the real-time alignment error is less than a preset error threshold; In response to the stop instruction, the train is smoothly stopped at a preset stop position through closed-loop feedback control of the train braking system.

[0006] Based on the acquired train head position information and train running speed information, the platform safety door position information and the real-time deceleration of the train are determined, including: Collect the position signal output by the Doppler radar sensor at the front end of the train to obtain the train head position information; Collect the speed signal output by the wheel speed sensor to obtain the train running speed information; Establishing a mapping relationship between a station coordinate system and a train coordinate system based on the train locomotive position information, converting the train locomotive position information into the station coordinate system, and obtaining the absolute position coordinates of the train locomotive in the station coordinate system; Calculate the spatial geometric relationship between the train head position and the platform safety door position based on the pre-calibrated platform safety door fixed coordinate information; Correcting and compensating the train running speed information based on the spatial geometric relationship to obtain corrected position information of the platform safety door relative to the train head, and determining the position information of the platform safety door according to the corrected position information; Based on the train running speed information, a sliding time window method is used to calculate the real-time deceleration of the train.

[0007] Collecting first image information of both sides of the platform safety door frame and second image information of both sides of the train door frame, and obtaining a relative position deviation value between the platform safety door and the train door through image registration of the first image information and the second image information, including: Collecting first image information on both sides of the platform safety door frame and second image information on both sides of the train door frame; Performing motion compensation on the first image information according to the train running speed, adaptively adjusting the image contrast parameter based on the ambient light intensity, extracting edge contour features of the platform safety door frame, and obtaining a platform safety door feature point set; Converting the second image information into a frequency domain, suppressing environmental interference through adaptive bandpass filtering, extracting edge contour features of the train door frame, and obtaining a train door feature point set; Based on the platform safety door feature point set and the train door feature point set, the feature points are spatially reprojected using a homography matrix transformation to establish a correspondence between the two sets of feature point sets; According to the correspondence between the feature point sets and in combination with the platform safety door reference coordinates, a three-dimensional coordinate transformation equation group is established, and the relative position deviation value between the platform safety door and the train door is obtained by solving the three-dimensional coordinate transformation equation group.

[0008] Based on the platform safety door feature point set and the train door feature point set, the feature points are spatially reprojected using a homography matrix transformation to establish a correspondence between the two sets of feature point sets, including: Feature classification is performed on the platform safety door feature point set and the train door feature point set respectively, and the neighborhood grayscale distribution information of each feature point is encoded into a feature description vector to obtain the platform safety door feature description vector set and the train door feature description vector set; Calculating a Euclidean distance matrix between feature points based on the platform safety door feature description vector set and the train door feature description vector set; Performing initial matching of feature points according to the Euclidean distance matrix, selecting the feature point pair with the smallest distance in the Euclidean distance matrix, and generating an initial feature point correspondence set; Constructing a homography matrix based on the corresponding set of the initial feature points, taking the set of feature description vectors as prior knowledge, iteratively optimizing the transformation parameters of the homography matrix through maximum a posteriori probability estimation, and performing spatial reprojection on the feature points; The reprojection error is calculated based on the coordinates of the feature points after spatial reprojection and the coordinates of the original feature points before spatial reprojection. The feature point pairs with a reprojection error greater than a preset threshold are removed from the initial feature point correspondence set, and the correspondence between the platform safety door feature point set and the train door feature point set is established.

[0009] Generating a train position adjustment instruction including a position compensation parameter according to the relative position deviation value and the real-time deceleration, and sending the train position adjustment instruction to the train automatic driving system, comprising: Decomposing the relative position deviation value into a lateral deviation component and a longitudinal deviation component, and calculating a train position adjustment compensation coefficient based on the lateral deviation component and the longitudinal deviation component; collecting a train wheel-rail adhesion coefficient, determining a train braking parameter according to the real-time deceleration and the train wheel-rail adhesion coefficient, and combining the train braking parameter with the train position adjustment compensation coefficient to generate a position compensation parameter; Building a position adjustment priority queue based on the position compensation parameters, dynamically adjusting the execution order of instructions in the position adjustment priority queue according to the numerical values ​​of the lateral deviation component and the longitudinal deviation component, and generating a train position adjustment instruction containing priority information; The train position adjustment instruction is sent to the train automatic driving system through the on-board communication network.

[0010] Receiving position adjustment feedback information returned by the automatic train driving system based on a preset adjustment step size, calculating a real-time alignment error between the platform safety door and the train door according to the position adjustment feedback information, and sending a stop instruction to the automatic train driving system when the real-time alignment error is less than a preset error threshold, including: receiving position adjustment feedback information returned by the train automatic driving system based on a preset adjustment step size; Establishing a coordinate mapping relationship between the platform coordinate system and the train coordinate system based on the position adjustment feedback information, mapping the train motion parameters to the platform coordinate system, and obtaining the motion state parameters of the train in the platform coordinate system; constructing a train motion compensation function according to the motion state parameters, calculating a compensation gain coefficient of the train motion compensation function according to track curvature information, performing compensation correction on train position information based on the compensation gain coefficient, and calculating a lateral alignment error and a longitudinal alignment error between a platform safety door and a train door in a platform coordinate system; Constructing a position deviation matrix according to the lateral alignment error and the longitudinal alignment error, and calculating a correction parameter of the preset adjustment step size based on the motion state parameter; The position deviation matrix is ​​decomposed into a transverse projection component and a longitudinal projection component in the orthogonal projection space, and a preset error threshold is adjusted according to the ratio of the projection components. When both the transverse projection component and the longitudinal projection component are less than the adjusted preset error threshold, a stop command is sent to the train automatic driving system.

[0011] Decomposing the position deviation matrix into a transverse projection component and a longitudinal projection component in an orthogonal projection space, adjusting a preset error threshold according to a ratio of the projection components, and sending a stop instruction to the train automatic driving system when both the transverse projection component and the longitudinal projection component are less than the adjusted preset error threshold, including: Extract the eigenvectors from the position deviation matrix and construct a biorthogonal basis matrix; Mapping the biorthogonal basis matrix to a Riemannian manifold space to construct a nonlinear orthogonal projection space; Decomposing the position deviation matrix into a transverse position component and a longitudinal position component in the nonlinear orthogonal projection space, and calculating a projection mapping relationship between the transverse position component and the longitudinal position component; Constructing a Riemann curvature tensor based on the projection mapping relationship, and calculating the principal curvatures of the transverse position component and the longitudinal position component using the Riemann curvature tensor; constructing a Riemannian metric tensor according to the principal curvature, solving a geodesic equation in the Riemannian manifold space based on the Riemannian metric tensor, and using the shortest path length of the geodesic equation as a dynamic adjustment coefficient of a preset error threshold; When the lateral position component is less than the preset error threshold after adjustment by the dynamic adjustment coefficient and the longitudinal position component is less than the preset error threshold after adjustment by the dynamic adjustment coefficient, a stop instruction is sent to the train automatic driving system.

[0012] A second aspect of an embodiment of the present invention provides a system for coordinated alignment, calibration, and docking of platform safety doors and train doors, comprising: The first unit is used to determine the position information of the platform safety door and the real-time deceleration of the train based on the acquired train head position information and train running speed information; The second unit is configured to collect first image information of both sides of the platform safety door frame and second image information of both sides of the train door frame, and obtain a relative position deviation value between the platform safety door and the train door through image registration of the first image information and the second image information; a third unit, configured to generate a train position adjustment instruction including a position compensation parameter according to the relative position deviation value and the real-time deceleration, and send the train position adjustment instruction to a train automatic driving system; a fourth unit, configured to receive position adjustment feedback information returned by the automatic train driving system based on a preset adjustment step length, calculate a real-time alignment error between the platform safety door and the train door according to the position adjustment feedback information, and send a stop command to the automatic train driving system when the real-time alignment error is less than a preset error threshold; The fifth unit is used to respond to the stop instruction and make the train stop smoothly at a preset stop position through closed-loop feedback control of the train braking system.

[0013] According to a third aspect of an embodiment of the present invention, an electronic device is provided, including: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0014] According to a fourth aspect of an embodiment of the present invention, a computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the method described above is implemented.

[0015] The beneficial effects of this application are as follows: The present invention achieves precise alignment of the door and the platform door by performing real-time image acquisition and registration of the positions of the platform safety door and the train door, avoiding the positioning deviation problem caused by factors such as track wear and wheel wear in traditional positioning methods, and improving the parking positioning accuracy.

[0016] The present invention utilizes the train's real-time deceleration information and position deviation value to dynamically generate position compensation parameters, and combines this with a preset adjustment step size for closed-loop feedback control, enabling the train to adaptively adjust its parking position according to actual conditions, thereby realizing intelligent control of automatic calibration of parking positioning and enhancing the system's adaptability and stability. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 The figure is a flow chart of a method for coordinated alignment, calibration, and docking positioning of platform safety doors and train doors according to an embodiment of the present invention. DETAILED DESCRIPTION

[0018] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0019] The following specific embodiments are used to describe the technical solution of the present invention in detail. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.

[0020] Figure 1 FIG. 1 is a flow chart of a method for coordinating alignment, calibration, and parking of platform safety doors and train doors according to an embodiment of the present invention. Figure 1 As shown, the method includes: Determine the location of the platform safety door and the real-time deceleration of the train based on the acquired train head position information and train running speed information; Collecting first image information on both sides of the platform safety door frame and second image information on both sides of the train door frame, and obtaining a relative position deviation value between the platform safety door and the train door through image registration of the first image information and the second image information; generating a train position adjustment instruction including a position compensation parameter according to the relative position deviation value and the real-time deceleration, and sending the train position adjustment instruction to the train automatic driving system; receiving position adjustment feedback information returned by the automatic train driving system based on a preset adjustment step size, calculating a real-time alignment error between the platform safety door and the train door according to the position adjustment feedback information, and sending a stop command to the automatic train driving system when the real-time alignment error is less than a preset error threshold; In response to the stop instruction, the train is smoothly stopped at a preset stop position through closed-loop feedback control of the train braking system.

[0021] In an optional embodiment, determining the position information of the platform safety door and the real-time deceleration of the train based on the acquired train head position information and train running speed information includes: Collect the position signal output by the Doppler radar sensor at the front end of the train to obtain the train head position information; Collect the speed signal output by the wheel speed sensor to obtain the train running speed information; Establishing a mapping relationship between a station coordinate system and a train coordinate system based on the train locomotive position information, converting the train locomotive position information into the station coordinate system, and obtaining the absolute position coordinates of the train locomotive in the station coordinate system; Calculate the spatial geometric relationship between the train head position and the platform safety door position based on the pre-calibrated platform safety door fixed coordinate information; Correcting and compensating the train running speed information based on the spatial geometric relationship to obtain corrected position information of the platform safety door relative to the train head, and determining the position information of the platform safety door according to the corrected position information; Based on the train running speed information, a sliding time window method is used to calculate the real-time deceleration of the train.

[0022] This invention provides a method for determining the position of platform safety gates and the real-time train deceleration based on locomotive position and speed information. This method, through multi-sensor fusion technology, enables precise positioning of platform safety gates and accurate calculation of train deceleration, improving the safety and stability of urban rail transit systems.

[0023] In this embodiment, a Doppler radar sensor is installed at the front of the train to collect locomotive position signals. Operating in the 76 GHz frequency band, the Doppler radar sensor offers high ranging accuracy, reaching ±0.01 meters. Wheel speed sensors are also installed on the train body to collect train speed signals. With a sampling frequency of 100 Hz, the wheel speed sensors offer speed measurement accuracy of ±0.1 km / h.

[0024] When collecting the position signal from the Doppler radar sensor at the front of the train, the system uses a dual filtering algorithm, including median filtering and Kalman filtering, to process the raw signal. The median filter removes burst noise, with a filter window size of 5. The Kalman filter smoothes the position signal, with a process noise covariance of 0.01 and an observation noise covariance of 0.1. After filtering, the train's front position is obtained, with a data update frequency of 20 Hz. For example, at a certain moment, the filtered train's front position is 120.45 meters (relative to the start of the line).

[0025] When collecting speed signals from the wheel speed sensors, the system averages the speed signals from all four wheels and smoothes them through a low-pass filter with a cutoff frequency of 5 Hz. The filtered speed signal is updated at a 50 Hz frequency and represents the train's real-time speed. For example, at a certain moment, the train's speed is 36.8 km / h.

[0026] The system uses a coordinate transformation method to establish the mapping relationship between the station coordinate system and the train coordinate system. The station coordinate system uses the platform starting point as its origin, with the platform direction as its positive direction; the train coordinate system uses the train head as its origin, with the train's travel direction as its positive direction. The transformation relationship between the two coordinate systems is determined by pre-calibrated reference points. The transformation matrix contains a translation vector and a rotation angle. The translation vector is dynamically updated based on the train head position, and the rotation angle is calculated based on the track curvature. For straight track segments, the rotation angle is 0; for curved track segments, the system calculates the rotation angle based on the line geometry. For example, when the train is at milepost K15+360, the corresponding position in the station coordinate system is (25.6, 0.02) meters.

[0027] The fixed coordinates of platform safety gates are obtained through pre-calibration. This calibration process uses a high-precision total station, achieving a measurement accuracy of better than ±0.005 meters. The coordinates of each platform safety gate within the station are recorded in the system database. For example, a platform may have 12 sets of safety gates. The coordinates of the first set of safety gates are (10.00, 0.00) meters, and the distance between adjacent safety gates is 5.00 meters.

[0028] When calculating the spatial geometric relationship between the train's front position and the platform safety gates, the system calculates the Euclidean distance between them in the station coordinate system. Considering the standard distance between the train's front and the first set of platform doors during a stop is 3.00 meters, the system sets a target position deviation threshold of ±0.05 meters. For example, when the train's front is at the (7.02, 0.01) meter position in the station coordinate system, the distance from the first set of platform doors is 2.98 meters, meeting the stop accuracy requirement.

[0029] When compensating for train speed information, the system accounts for Doppler radar measurement errors and wheel speed sensor slip errors. The compensation algorithm, based on a statistical model of historical data, uses different compensation coefficients for different speed ranges. When the train speed is below 5 km / h, the compensation coefficient is 1.02; when the speed is between 5 and 30 km / h, the compensation coefficient is 1.01; and when the speed is above 30 km / h, the compensation coefficient is 1.00. This correction improves position information accuracy by 15%.

[0030] When determining the position information of the platform safety door, the system compares the corrected train head position with the preset fixed coordinates of the platform safety door, and calculates the relative distance. The system takes into account the train length and the target stopping point to ensure that the train door and the platform safety door are accurately matched. For example, a train is 120 meters long, has 6 carriages, and each carriage has 2 doors on each side, with a door spacing of 5 meters. The system calculates the position of the platform safety door corresponding to each door according to these parameters.

[0031] When calculating the real-time deceleration of the train, the system uses a sliding time window method. The window size is set to 1 second, and the sliding step is 0.1 second. The least squares method is used to fit the velocity-time curve in the window, and the deceleration is calculated as the rate of change of velocity. To improve the stability of the calculation, the system sets the upper limit of the deceleration rate to 0.2 m / s³. For example, in a 1-second window, the train speed decreases from 28.5 km / h to 25.2 km / h, and the calculated average deceleration is 0.92 m / s², which meets the comfort requirement (less than 1.2 m / s²).

[0032] Through the above technical means, the present application realizes accurate determination of the position of the platform safety door and accurate calculation of the real-time deceleration of the train, which can effectively improve the safety, accuracy and comfort of the urban rail transit system. The method has been tested and verified on multiple subway lines, with a positioning accuracy of ±0.03 meters, meeting the requirements of the automatic driving system for accurate parking.

[0033] In an optional embodiment, first image information on both sides of the door frame of the platform safety door and second image information on both sides of the door frame of the train door are collected, and the relative position deviation value between the platform safety door and the train door is obtained through image registration of the first image information and the second image information, including: Collecting first image information on both sides of the door frame of the platform safety door and second image information on both sides of the door frame of the train door; Motion compensation is performed on the first image information according to the train running speed, the image contrast parameter is adjusted adaptively according to the environmental light intensity, the edge profile features of the door frame of the platform safety door are extracted, and the feature point set of the platform safety door is obtained; The second image information is converted to the frequency domain space, environmental interference is suppressed through adaptive band-pass filtering, the edge profile features of the door frame of the train door are extracted, and the feature point set of the train door is obtained; Based on the feature point set of the platform safety door and the feature point set of the train door, the spatial re-projection of the feature points is performed through the homography matrix transformation, and the correspondence relationship between the two sets of feature points is established; According to the correspondence relationship of the feature point set, a three-dimensional coordinate transformation equation set is established in combination with the platform safety door reference coordinates, and the relative position deviation value between the platform safety door and the train door is obtained by solving the three-dimensional coordinate transformation equation set.

[0034] The present invention provides a method for collecting first image information on both sides of a platform safety door frame and second image information on both sides of a train door frame, and obtains the relative position deviation value between the platform safety door and the train door through image registration.

[0035] In this embodiment of the present invention, at least two high-definition cameras are first installed on the platform to capture image information from both sides of the platform safety door frame. The cameras can be installed on the platform ceiling or above the platform safety door to ensure that both sides of the platform safety door frame are fully captured. Simultaneously, at least two high-definition cameras are installed on the train to capture image information from both sides of the train door frame. The cameras can be installed inside the train car or above the door to ensure that both sides of the train door frame are fully captured. The image resolution captured by the two sets of cameras is set to 1920×1080 pixels, and the frame rate is set to 30 frames per second to ensure image clarity and real-time performance.

[0036] The first image information captured from both sides of the platform safety door frame requires motion compensation. Motion compensation is performed on the first image information based on the train's running speed to eliminate image blur caused by train motion. Specifically, real-time train speed data is obtained from the train control system. When the train speed is 5 km / h, the motion compensation parameter can be set to 0.5; when the train speed is 10 km / h, the motion compensation parameter can be set to 1.0; and when the train speed is 15 km / h, the motion compensation parameter can be set to 1.5. Based on the different motion compensation parameters, the image is translated and rotated accordingly to eliminate motion blur.

[0037] Light sensors are installed at the stations to collect real-time ambient light intensity. When the ambient light intensity is below 100 lux, the image contrast parameter is set to 1.5; when the ambient light intensity is between 100 and 500 lux, the image contrast parameter is set to 1.2; and when the ambient light intensity is above 500 lux, the image contrast parameter is set to 0.9. This adaptive adjustment ensures clear images under varying lighting conditions.

[0038] For the processed platform security door image, the Canny edge detection algorithm is used to extract the edge contour features of the door frame. The specific parameters are set as follows: a high threshold of 150, a low threshold of 50, and a Gaussian filter kernel size of 5×5. These parameters effectively extract the edge contours of the platform security door frame and obtain a set of platform security door feature points. In practice, approximately 200 to 300 feature points can be extracted, primarily distributed along the four edges of the door frame.

[0039] The spatial domain image is converted into a frequency domain representation by performing a two-dimensional transformation on the image. In the frequency domain space, an adaptive band-pass filter is used to suppress environmental interference. The low frequency cutoff frequency of the adaptive band-pass filter is set to 5% of the maximum frequency of the image frequency domain, and the high frequency cutoff frequency is set to 75% of the maximum frequency of the image frequency domain. Through this filtering method, environmental light changes, reflections, shadows and other interference factors can be effectively removed, and the signal-to-noise ratio of the image can be improved.

[0040] After frequency domain processing, the image is converted back to the spatial domain, and the Roberts operator is used to extract the edge profile features of the train door frame. The Roberts operator parameters are set as follows: threshold value is 40. Through this processing, a set of train door feature points can be obtained, usually containing about 150 to 250 feature points, mainly distributed in the four edges of the door frame.

[0041] Based on the extracted platform safety door feature point set and train door feature point set, a homography matrix transformation is used to perform spatial re-projection on the feature points, and a corresponding relationship between the two sets of feature points is established. In specific implementation, 4 pairs of corresponding points are selected from the platform safety door feature point set and the train door feature point set, which are usually selected at the four corner positions of the door frame. The homography matrix is calculated through the 4 pairs of corresponding points, and then all the points in the platform safety door feature point set are mapped into the train door coordinate system using the matrix, thereby establishing a complete corresponding relationship between the two sets of feature points.

[0042] After establishing the corresponding relationship of the feature points, a three-dimensional coordinate transformation equation set is established in combination with the platform safety door reference coordinates. The origin of the platform safety door reference coordinate system is set at the center point of the safety door, the X-axis points to the right along the horizontal direction, the Y-axis points upward along the vertical direction, and the Z-axis points to the train direction perpendicular to the platform plane. By solving the three-dimensional coordinate transformation equation set, the relative position deviation value between the platform safety door and the train door is obtained, including the X-axis direction deviation (horizontal deviation), the Y-axis direction deviation (vertical deviation), and the Z-axis direction deviation (front-back deviation).

[0043] In actual application examples, when the train stops at the platform, the system automatically starts the image acquisition and processing process. The relative position deviation values calculated by the above method may be: the X-axis direction deviation is +15 mm (indicating that the train door deviates 15 mm to the right relative to the platform safety door), the Y-axis direction deviation is -8 mm (indicating that the train door deviates 8 mm downward relative to the platform safety door), and the Z-axis direction deviation is +20 mm (indicating that the train door deviates 20 mm forward relative to the platform safety door). According to these deviation data, the train control system can accurately adjust the stopping position of the train to ensure accurate alignment of the platform safety door and the train door, and improve the safety and convenience of passengers getting on and off the train.

[0044] In an optional embodiment, based on the platform safety door feature point set and the train door feature point set, a homography matrix transformation is used to spatially reproject the feature points to establish a correspondence between the two sets of feature point sets, including: Feature classification is performed on the platform safety door feature point set and the train door feature point set respectively, and the neighborhood grayscale distribution information of each feature point is encoded into a feature description vector to obtain the platform safety door feature description vector set and the train door feature description vector set; Calculating a Euclidean distance matrix between feature points based on the platform safety door feature description vector set and the train door feature description vector set; Performing initial matching of feature points according to the Euclidean distance matrix, selecting the feature point pair with the smallest distance in the Euclidean distance matrix, and generating an initial feature point correspondence set; Constructing a homography matrix based on the corresponding set of the initial feature points, taking the set of feature description vectors as prior knowledge, iteratively optimizing the transformation parameters of the homography matrix through maximum a posteriori probability estimation, and performing spatial reprojection on the feature points; The reprojection error is calculated based on the coordinates of the feature points after spatial reprojection and the coordinates of the original feature points before spatial reprojection. The feature point pairs with a reprojection error greater than a preset threshold are removed from the initial feature point correspondence set, and the correspondence between the platform safety door feature point set and the train door feature point set is established.

[0045] This embodiment provides a method for aligning platform safety gates and train doors based on feature point correspondence. The method uses homography matrix transformation to spatially reproject feature points to establish accurate correspondence.

[0046] In this embodiment, image data of the platform safety door and train door are first collected. A feature extraction algorithm is then used to obtain the platform safety door feature point set and the train door feature point set. For each feature point in the platform safety door feature point set, a 16×16 pixel neighborhood is extracted, and the grayscale gradient directional histogram within this area is calculated and encoded into a 128-dimensional feature description vector. Similarly, the same processing is performed for each feature point in the train door feature point set to obtain the corresponding feature description vector. This results in the platform safety door feature description vector set and the train door feature description vector set.

[0047] Specifically, for each feature point, the SIFT algorithm is used to extract a feature description vector. Assume that 150 feature points are extracted from the platform security door image and 160 feature points are extracted from the train door image. The position of each feature point is represented by two-dimensional coordinates (x, y). For example, the coordinates of a feature point on the platform security door are (125, 345), and the coordinates of a feature point on the train door are (128, 342). The feature description vector of each feature point contains 128 elements. For example, the feature description vector of the platform security door feature point is [0.12, 0.08, 0.25, ..., 0.15], and the feature description vector of the train door feature point is [0.13, 0.07, 0.26, ..., 0.14].

[0048] After obtaining the two sets of feature description vectors, the Euclidean distance matrix between the platform safety door feature points and the train door feature points is calculated. Assuming the platform safety door has m feature points and the train door has n feature points, the Euclidean distance matrix is ​​an m×n matrix, where the element D(i,j) in the matrix represents the Euclidean distance between the i-th feature point of the platform safety door and the j-th feature point of the train door. The Euclidean distance is calculated by subtracting the corresponding elements in the two feature description vectors, finding the sum of the squares, and then taking the square root.

[0049] Taking the actual data as an example, for the platform security door feature point P1 (125, 345) and the train door feature point Q1 (128, 342), their feature description vectors are VP1 = [0.12, 0.08, 0.25, ..., 0.15] and VQ1 = [0.13, 0.07, 0.26, ..., 0.14], respectively. The calculated Euclidean distance is 0.089. Similarly, the Euclidean distances between all feature point pairs are calculated to obtain the complete Euclidean distance matrix.

[0050] Initial matching of feature points is performed based on the Euclidean distance matrix. For each feature point in the platform safety gate, the point with the smallest Euclidean distance among the train door feature points is found as the matching point. To improve the reliability of the matching, a ratio test strategy is also adopted, which requires that the ratio of the minimum distance to the second smallest distance be less than 0.8. In this way, the selected feature point pairs form the initial feature point corresponding set. For example, the Euclidean distance between the platform safety gate feature point P1 and the train door feature point Q1 is 0.089, and the Euclidean distance between the platform safety gate feature point P1 and the train door feature point Q1 is 0.112. Since 0.089 / 0.112=0.79<0.8, P1 and Q1 are added as a pair of matching points to the initial feature point corresponding set.

[0051] After forming the initial set of feature point correspondences, a homography matrix is ​​constructed based on this set. The homography matrix describes the projective transformation relationship between two planes. It is a 3×3 matrix containing eight independent parameters. Initially, the homography matrix can be set to the identity matrix, that is, H = [1, 0, 0; 0, 1, 0; 0, 0, 1].

[0052] Next, the maximum a posteriori probability estimation method is used to iteratively optimize the transformation parameters of the homography matrix. In the specific implementation, the feature description vector is used as a priori knowledge, and the homography matrix parameters are updated by minimizing the reprojection error. In each iteration, the platform safety door feature points are projected into the train door image coordinate system using the current homography matrix, and the distance between the projected point and the actual train door feature point is calculated as the reprojection error. For example, using the initial homography matrix to project the platform safety door feature point P1(125, 345) to P1'(127, 343), while the actual train door feature point Q1 has coordinates of (128, 342), the reprojection error is the Euclidean distance between the two points, which is 1.414.

[0053] The optimization process uses the Levenberg-Marquardt algorithm to iteratively update the homography matrix parameters until the maximum number of iterations reaches 50 or the parameter change is less than a threshold of 0.001. After iterative optimization, the final homography matrix is, for example, H = [0.995, -0.018, 2.5; 0.012, 0.992, -1.2; 0.00001, 0.00002, 1].

[0054] All platform safety door feature points are spatially reprojected using the optimized homography matrix to obtain their projected positions in the train door coordinate system. For each pair of feature points, the reprojection error is calculated, which is the distance between the projected point and the actual train door feature point. A reprojection error threshold is set to 3 pixels, and feature point pairs with a reprojection error greater than the threshold are removed from the initial feature point correspondence set. For example, the reprojection error between platform safety door feature point P2 and train door feature point Q2 after projection is 3.5 pixels, exceeding the threshold, so (P2, Q2) is removed from the feature point correspondence set.

[0055] Through the above processing, reliable correspondences between the feature point sets of the platform safety gate and the train door were ultimately established. For example, out of the 100 pairs of feature points initially matched, 85 pairs of reliable correspondences were retained after reprojection error screening. These correspondences can be used for subsequent precise alignment control of the platform safety gate and train door, ensuring accurate alignment when the train stops, improving the safety and efficiency of passenger boarding and alighting.

[0056] In an optional embodiment, generating a train position adjustment instruction including a position compensation parameter based on the relative position deviation value and the real-time deceleration, and sending the train position adjustment instruction to the train automatic driving system includes: Decomposing the relative position deviation value into a lateral deviation component and a longitudinal deviation component, and calculating a train position adjustment compensation coefficient based on the lateral deviation component and the longitudinal deviation component; collecting a train wheel-rail adhesion coefficient, determining a train braking parameter according to the real-time deceleration and the train wheel-rail adhesion coefficient, and combining the train braking parameter with the train position adjustment compensation coefficient to generate a position compensation parameter; Building a position adjustment priority queue based on the position compensation parameters, dynamically adjusting the execution order of instructions in the position adjustment priority queue according to the numerical values ​​of the lateral deviation component and the longitudinal deviation component, and generating a train position adjustment instruction containing priority information; The train position adjustment instruction is sent to the train automatic driving system through the on-board communication network.

[0057] In one specific implementation, the train position adjustment system generates train position adjustment instructions containing position compensation parameters based on relative position deviation and real-time deceleration. During train operation, if the system detects a deviation from the planned trajectory, it performs a series of calculations and processing to send precise position adjustment instructions to the train's automated driving system, ensuring safe and accurate train operation.

[0058] In actual application scenarios, the system first obtains the train's current and scheduled positions and calculates the relative position deviation. For example, when a train arrives at platform A, the train's actual position, measured by onboard sensors and a ground beacon system, is (x1, y1, z1), while the scheduled stop position is (x0, y0, z0). The relative position deviation is the position vector difference between the two points. Simultaneously, the system also obtains the train's real-time deceleration information, such as the current deceleration of 2.3 m / s².

[0059] The system decomposes the relative position deviation value into a lateral deviation component and a longitudinal deviation component. In the railway coordinate system, the lateral deviation refers to the distance deviation between the train and the centerline of the track, and the longitudinal deviation refers to the front-to-back deviation of the train in the direction of the track. For example, the calculated lateral deviation is 0.05 meters and the longitudinal deviation is 0.15 meters. The system calculates the train position adjustment compensation coefficient based on these two components. Specifically, when the lateral deviation is less than 0.08 meters and the longitudinal deviation is less than 0.2 meters, the lateral adjustment compensation coefficient is set to 0.85 and the longitudinal adjustment compensation coefficient is set to 0.92; when the lateral deviation is greater than 0.08 meters or the longitudinal deviation is greater than 0.2 meters, the lateral adjustment compensation coefficient is set to 0.75 and the longitudinal adjustment compensation coefficient is set to 0.85. In this example, the lateral adjustment compensation coefficient is 0.85 and the longitudinal adjustment compensation coefficient is 0.92.

[0060] The system collects the train's wheel-rail adhesion coefficient using wheel speed sensors and track condition monitoring equipment. In dry weather, the coefficient can reach 0.25; in wet conditions, it may drop to 0.15; and in rainy or snowy weather, it may even drop below 0.10. The system determines the train's braking parameters based on the real-time deceleration and the train's wheel-rail adhesion coefficient. For example, when the wheel-rail adhesion coefficient is 0.2 and the real-time deceleration is 2.3 m / s², the system sets the braking pressure parameter to 350 kPa and the braking distance parameter to 120 meters.

[0061] The system combines the train's braking parameters with the train's position adjustment compensation coefficient to generate the position compensation parameter. This combination is a weighted average: the product of the braking pressure parameter and the lateral adjustment compensation coefficient is added to the product of the braking distance parameter and the longitudinal adjustment compensation coefficient, and then divided by the sum of the weights of the two sets of parameters. For example, if the braking pressure parameter is assigned a weight of 0.6 and the braking distance parameter is assigned a weight of 0.4, the position compensation parameter is calculated as (350 × 0.85 × 0.6 + 120 × 0.92 × 0.4) / (0.6 + 0.4) = 192.83.

[0062] Based on the position compensation parameters, the system constructs a position adjustment priority queue. The priority queue contains multiple adjustment instructions, such as "lateral position fine-tuning," "longitudinal position fine-tuning," "braking force adjustment," and "traction adjustment." The system dynamically adjusts the execution order of instructions in the position adjustment priority queue based on the numerical values ​​of the lateral and longitudinal deviation components. When the lateral deviation is greater than the longitudinal deviation, the lateral adjustment instruction takes precedence over the longitudinal adjustment instruction, and vice versa. In this example, because the longitudinal deviation (0.15 meters) is greater than the lateral deviation (0.05 meters), the system sets the priority of the longitudinal position fine-tuning instruction to 1, the priority of the lateral position fine-tuning instruction to 2, the priority of the braking force adjustment instruction to 3, and the priority of the traction adjustment instruction to 4.

[0063] The system generates a train position adjustment instruction containing priority information. The instruction content includes: adjustment type (such as "longitudinal position fine-tuning"), adjustment amplitude (such as "adjust forward 0.15 meters"), execution priority (such as "priority 1") and execution time window (such as "complete within 2 seconds").

[0064] The system sends train position adjustment commands to the automatic train driving system via the onboard communication network. The onboard communication network utilizes a redundant design, consisting of a primary and backup communication channel. The primary channel utilizes industrial Ethernet, with a transmission rate of 100 Mbps and a transmission delay of no more than 10 ms. The backup channel utilizes wireless communication technology, with a transmission rate of 50 Mbps and a transmission delay of no more than 20 ms. Upon receiving the position adjustment commands, the automatic train driving system executes the corresponding adjustment actions according to the priority order specified in the commands, achieving precise adjustment of the train's position.

[0065] In actual tests, the system was able to control the docking position error within ±5cm at a train speed of 80km / h. This improved docking accuracy by 60% compared to traditional fixed parameter adjustment methods, while also reducing brake wear by 30%, significantly improving train safety, comfort, and punctuality. In particularly severe weather conditions, the system improved the stability of train docking accuracy by 45% by dynamically adjusting parameters related to the wheel-rail adhesion coefficient.

[0066] In an optional embodiment, receiving position adjustment feedback information returned by the automatic train driving system based on a preset adjustment step size, calculating a real-time alignment error between the platform safety door and the train door based on the position adjustment feedback information, and sending a stop instruction to the automatic train driving system when the real-time alignment error is less than a preset error threshold, includes: receiving position adjustment feedback information returned by the train automatic driving system based on a preset adjustment step size; Establishing a coordinate mapping relationship between the platform coordinate system and the train coordinate system based on the position adjustment feedback information, mapping the train motion parameters to the platform coordinate system, and obtaining the motion state parameters of the train in the platform coordinate system; constructing a train motion compensation function according to the motion state parameters, calculating a compensation gain coefficient of the train motion compensation function according to track curvature information, performing compensation correction on train position information based on the compensation gain coefficient, and calculating a lateral alignment error and a longitudinal alignment error between a platform safety door and a train door in a platform coordinate system; Constructing a position deviation matrix according to the lateral alignment error and the longitudinal alignment error, and calculating a correction parameter of the preset adjustment step size based on the motion state parameter; The position deviation matrix is ​​decomposed into a transverse projection component and a longitudinal projection component in the orthogonal projection space, and a preset error threshold is adjusted according to the ratio of the projection components. When both the transverse projection component and the longitudinal projection component are less than the adjusted preset error threshold, a stop command is sent to the train automatic driving system.

[0067] To ensure precise alignment between platform safety gates and train doors in station automatic train stopping systems, this paper provides an automatic train stopping method based on real-time alignment error calculation. This method establishes a coordinate mapping relationship between the platform and the train, and uses track curvature information to compensate for the train's position, thereby achieving high-precision train stopping control.

[0068] The system first receives position adjustment feedback from the train's automated driving system based on a preset adjustment step size. This feedback typically includes parameters such as the train's current position coordinates, speed, and acceleration. The preset adjustment step size can be set based on the characteristics of different stations and lines, typically ranging from 10 to 50 mm. This step size can be adjusted through the system parameter configuration interface. For example, on straight platform sections, the system can set a larger adjustment step size, such as 50 mm; on curved platform sections, a smaller adjustment step size, such as 20 mm, is used to ensure positioning accuracy.

[0069] After receiving feedback, the system uses this data to establish a coordinate mapping between the platform coordinate system and the train coordinate system. The platform coordinate system is typically based on the platform centerline, with the x-axis running along the length of the platform, the y-axis running along the width of the platform, and the z-axis perpendicular to the platform plane. The train coordinate system, on the other hand, has the center point of the front end of the train as its origin, with the x-axis running in the direction of travel, the y-axis perpendicular to the track plane, and the z-axis perpendicular to the cross-section of the train body. Using a coordinate transformation matrix, the system maps the train's motion parameters from the train coordinate system to the platform coordinate system, obtaining the train's motion parameters in the platform coordinate system, including position, velocity, and acceleration. In practical applications, this mapping can be achieved using rotation and translation matrices, achieving millimeter-level accuracy. For example, if a train is detected to be 50 meters from the platform starting point, 15 mm laterally offset from the centerline, and traveling at a speed of 0.2 m / s, the system can calculate the train's precise position and motion in the platform coordinate system through coordinate mapping.

[0070] The system constructs a train motion compensation function according to the mapped motion state parameters. The compensation function mainly considers the dynamic characteristics of the train during low-speed driving, including inertia, wheel-rail contact characteristics, etc. At the same time, the system calculates the compensation gain coefficient of the compensation function according to the track curvature information. In the curve platform, the track curvature will cause the lateral gap between the train and the platform to change, and additional compensation is needed. The compensation gain coefficient is usually inversely proportional to the track curvature radius. The smaller the curvature radius, the larger the compensation coefficient. For a curve platform with a curvature radius of 300 meters, the compensation coefficient can be set to 1.25; while for a curve platform with a curvature radius of 500 meters, the compensation coefficient can be set to 1.15. The system compensates and corrects the train position information based on the compensation gain coefficient, and finally calculates the lateral alignment error and longitudinal alignment error of the platform safety door and the train door in the platform coordinate system. The lateral error reflects the distance difference between the train door and the platform safety door in the direction perpendicular to the track, and the longitudinal error reflects the distance difference between the two in the direction parallel to the track.

[0071] The system constructs a position deviation matrix according to the calculated lateral alignment error and longitudinal alignment error. This matrix contains the alignment error information of each door and the corresponding safety door. For example, a train has 8 doors, then the position deviation matrix is an 8x2 matrix, each row contains the lateral and longitudinal alignment error of a door. The system calculates the correction parameters of the preset adjustment step based on the previously calculated motion state parameters. When the train speed is high, a larger adjustment step is used; when the train approaches the target stopping point, the adjustment step gradually decreases to improve accuracy. For example, when the train speed is greater than 0.5 meters / second, the adjustment step can be set to 1.2 times the original preset value; when the speed drops below 0.2 meters / second, the adjustment step can be reduced to 0.8 times the original preset value.

[0072] The system decomposes the position deviation matrix into a lateral projection component and a longitudinal projection component in the orthogonal projection space. This decomposition allows the system to evaluate the alignment error in both directions separately. According to the ratio of the projection components, the system dynamically adjusts the preset error threshold. When the lateral error proportion is large, the longitudinal error threshold is appropriately relaxed; when the longitudinal error proportion is large, the lateral error threshold is appropriately relaxed. For example, the standard error threshold is set to ±15mm, if the ratio of the lateral projection component to the longitudinal projection component is greater than 1.5, the lateral error threshold is adjusted to ±12mm, and the longitudinal error threshold is adjusted to ±18mm. When the calculated lateral projection component and longitudinal projection component are both less than the adjusted preset error threshold, the system sends a stop command to the train automatic driving system, instructing the train to stop completely and lock the vehicle braking system.

[0073] Through this method, the system can achieve precise parking of trains at platforms, ensuring precise alignment of train doors with platform safety gates, improving safety and convenience for passengers boarding and alighting. Practical application tests have shown that this method can maintain parking alignment accuracy within ±10 mm, an improvement of approximately 40% over traditional methods. It also offers greater adaptability and can handle a variety of complex track conditions.

[0074] In an optional embodiment, the position deviation matrix is ​​decomposed into a transverse projection component and a longitudinal projection component in an orthogonal projection space, a preset error threshold is adjusted according to a ratio of the projection components, and when both the transverse projection component and the longitudinal projection component are less than the adjusted preset error threshold, a stop instruction is sent to the automatic train driving system, including: Extract the eigenvectors from the position deviation matrix and construct a biorthogonal basis matrix; Mapping the biorthogonal basis matrix to a Riemannian manifold space to construct a nonlinear orthogonal projection space; Decomposing the position deviation matrix into a transverse position component and a longitudinal position component in the nonlinear orthogonal projection space, and calculating a projection mapping relationship between the transverse position component and the longitudinal position component; Constructing a Riemann curvature tensor based on the projection mapping relationship, and calculating the principal curvatures of the transverse position component and the longitudinal position component using the Riemann curvature tensor; constructing a Riemannian metric tensor according to the principal curvature, solving a geodesic equation in the Riemannian manifold space based on the Riemannian metric tensor, and using the shortest path length of the geodesic equation as a dynamic adjustment coefficient of a preset error threshold; When the lateral position component is less than the preset error threshold after adjustment by the dynamic adjustment coefficient and the longitudinal position component is less than the preset error threshold after adjustment by the dynamic adjustment coefficient, a stop instruction is sent to the train automatic driving system.

[0075] The present invention provides a train automatic parking control system and method. To make the purpose, technical solutions and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be described in detail below in conjunction with the embodiments.

[0076] This embodiment first obtains the deviation information between the train's current position and the target parking position, generating a position deviation matrix containing spatial position deviation data. The position deviation matrix contains the coordinate deviation values ​​of the train relative to the target parking position in three-dimensional space. Each element in the position deviation matrix represents the train's deviation in a specific direction, typically measured in meters. For example, for a high-speed train approaching a station, the position deviation matrix might be [0.05, 0.08, 0.02], representing the deviation values ​​in the x, y, and z directions, respectively.

[0077] Next, the eigenvectors in the position deviation matrix are extracted to construct a biorthogonal basis matrix. The system uses singular value decomposition (SVD) to process the position deviation matrix and extract the primary eigenvectors. Specifically, eigenvalue analysis is performed on the position deviation matrix, and the eigenvector corresponding to the largest eigenvalue is selected as the first basis vector. The remaining orthogonal basis vectors are then constructed through a Gram-Schmidt orthogonalization process. For example, from the aforementioned position deviation matrix, the orthogonal basis vector groups [0.707, 0.707, 0] and [0, 0, 1] can be extracted to form a biorthogonal basis matrix.

[0078] The biorthogonal basis matrices are mapped onto a Riemannian manifold space to construct a nonlinear orthogonal projection space. This mapping process uses an exponential mapping function to transform vectors in Euclidean space onto a Riemannian manifold with nonzero curvature. In practice, the system selects a spherical manifold as the mapping target, with a radius of 10 meters, determined based on the characteristics of the train's braking system. This mapping transforms the original biorthogonal basis matrices into a nonlinear orthogonal basis, establishing a curved coordinate system that more accurately describes the train's motion characteristics.

[0079] In the constructed nonlinear orthogonal projection space, the system decomposes the position error matrix into lateral and longitudinal position components. The lateral position component represents the lateral distance deviation between the train and the track centerline, while the longitudinal position component represents the longitudinal distance deviation between the train and the target stopping point. This decomposition is achieved by calculating the projection coefficients of the position error matrix onto a nonlinear orthogonal basis. For example, for the measured position error matrix, the decomposition yields a lateral position component of 0.035 meters and a longitudinal position component of 0.075 meters.

[0080] The system calculates the projection mapping relationship between the lateral and longitudinal position components, establishing a dependency model between them. This mapping relationship describes the response change in the train's position in one direction when the train adjusts its position in the other direction. In this embodiment, the covariance matrix of the two components is calculated to obtain a projection mapping coefficient matrix. For example, a lateral-to-longitudinal mapping coefficient of 0.28 is obtained, representing the ratio of longitudinal response caused by a unit lateral change.

[0081] Based on the projective mapping relationship, the system constructs the Riemann curvature tensor. The curvature tensor describes the geometric properties of the projected space, particularly the degree of spatial curvature. The construction process involves calculating the second-order derivatives of the projective mapping and combining them to form the curvature tensor based on the principles of manifold geometry. In practice, for the above mapping relationship, the calculated principal element value of the curvature tensor is approximately 0.18.

[0082] The Riemann curvature tensor is used to calculate the principal curvatures of the transverse and longitudinal position components. Principal curvatures represent the curvature of space in each principal direction and are an important basis for adjusting the error threshold. Solving the characteristic equation yields a transverse principal curvature of 0.15 and a longitudinal principal curvature of 0.22.

[0083] The Riemann metric tensor is constructed based on the principal curvatures, which defines the distance metric on the manifold. The system constructs the metric tensor based on the principal curvature values, with the elements of the tensor proportional to the principal curvatures. For example, based on the principal curvature values ​​above, the diagonal elements of the metric tensor are [1.15, 1.22].

[0084] The system solves the geodesic equation in a Riemannian manifold space based on the Riemannian metric tensor. A geodesic represents the shortest path between two points on a manifold, and its length reflects the actual distance after accounting for spatial curvature. The system uses a variational method to solve the geodesic equation and determine the optimal path from the current location to the target parking point. In this example, the calculated geodesic length is 0.85, which the system uses as a dynamic adjustment factor.

[0085] The preset error threshold is initially set at 0.05 meters. After the dynamic adjustment coefficient is adjusted, the lateral error threshold is increased to 0.0425 meters (0.05 × 0.85), and the longitudinal error threshold is also adjusted to 0.0425 meters. The system continuously monitors the train's position and determines that the train has met the stopping accuracy requirement when it detects a lateral position component of 0.035 meters, which is less than the adjusted lateral error threshold of 0.0425 meters, and a longitudinal position component of 0.04 meters, which is less than the adjusted longitudinal error threshold of 0.0425 meters.

[0086] At this point, the system sends a stop command to the train's automated driving system. This command includes a stop confirmation flag, the train's current location, and a checksum. Upon receiving the stop command, the automated driving system controls the train and executes the parking procedure, including activating the electric and pneumatic brakes to ensure the train stops smoothly at the target location.

[0087] Through the above technical solution, the present invention realizes precise parking control of trains, adapts to parking requirements under different operating conditions, improves parking accuracy and reliability, and optimizes passenger experience.

[0088] The platform safety door and train door coordinated alignment, calibration, and docking positioning system according to an embodiment of the present invention includes: The first unit is used to determine the position information of the platform safety door and the real-time deceleration of the train based on the acquired train head position information and train running speed information; The second unit is configured to collect first image information on both sides of the platform safety door frame and second image information on both sides of the train door frame, and obtain a relative position deviation value between the platform safety door and the train door through image registration of the first image information and the second image information; a third unit, configured to generate a train position adjustment instruction including a position compensation parameter according to the relative position deviation value and the real-time deceleration, and send the train position adjustment instruction to a train automatic driving system; a fourth unit, configured to receive position adjustment feedback information returned by the automatic train driving system based on a preset adjustment step length, calculate a real-time alignment error between the platform safety door and the train door according to the position adjustment feedback information, and send a stop command to the automatic train driving system when the real-time alignment error is less than a preset error threshold; The fifth unit is used to respond to the stop instruction and make the train stop smoothly at a preset stop position through closed-loop feedback control of the train braking system.

[0089] According to a third aspect of an embodiment of the present invention, an electronic device is provided, including: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0090] According to a fourth aspect of an embodiment of the present invention, a computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the method described above is implemented.

[0091] The present invention may be a method, an apparatus, a system and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for executing various aspects of the present invention.

[0092] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for coordinating the alignment, calibration, and parking of platform safety doors and train doors, characterized in that: include: Determine the location of the platform safety door and the real-time deceleration of the train based on the acquired train head position information and train running speed information; Collecting first image information on both sides of the platform safety door frame and second image information on both sides of the train door frame, and obtaining a relative position deviation value between the platform safety door and the train door through image registration of the first image information and the second image information; generating a train position adjustment instruction including a position compensation parameter according to the relative position deviation value and the real-time deceleration, and sending the train position adjustment instruction to the train automatic driving system; receiving position adjustment feedback information returned by the automatic train driving system based on a preset adjustment step size, calculating a real-time alignment error between the platform safety door and the train door according to the position adjustment feedback information, and sending a stop command to the automatic train driving system when the real-time alignment error is less than a preset error threshold; In response to the stop instruction, the train is smoothly stopped at a preset stop position through closed-loop feedback control of the train braking system.

2. The method according to claim 1, characterized in that Based on the obtained train head position information and train running speed information, the platform safety door position information and the real-time deceleration of the train are determined, including: Collect the position signal output by the Doppler radar sensor at the front end of the train to obtain the train head position information; Collect the speed signal output by the wheel speed sensor to obtain the train running speed information; Establishing a mapping relationship between a station coordinate system and a train coordinate system based on the train locomotive position information, converting the train locomotive position information into the station coordinate system, and obtaining the absolute position coordinates of the train locomotive in the station coordinate system; Calculate the spatial geometric relationship between the train head position and the platform safety door position based on the pre-calibrated platform safety door fixed coordinate information; Correcting and compensating the train running speed information based on the spatial geometric relationship to obtain corrected position information of the platform safety door relative to the train head, and determining the position information of the platform safety door according to the corrected position information; Based on the train running speed information, a sliding time window method is used to calculate the real-time deceleration of the train.

3. The method according to claim 1, characterized in that Collecting first image information of both sides of the platform safety door frame and second image information of both sides of the train door frame, and obtaining a relative position deviation value between the platform safety door and the train door through image registration of the first image information and the second image information, including: Collecting first image information on both sides of the platform safety door frame and second image information on both sides of the train door frame; Performing motion compensation on the first image information according to the train running speed, adaptively adjusting the image contrast parameter based on the ambient light intensity, extracting edge contour features of the platform safety door frame, and obtaining a platform safety door feature point set; Converting the second image information into a frequency domain, suppressing environmental interference through adaptive bandpass filtering, extracting edge contour features of the train door frame, and obtaining a train door feature point set; Based on the platform safety door feature point set and the train door feature point set, the feature points are spatially reprojected using a homography matrix transformation to establish a correspondence between the two sets of feature point sets; According to the correspondence between the feature point sets and in combination with the platform safety door reference coordinates, a three-dimensional coordinate transformation equation group is established, and the relative position deviation value between the platform safety door and the train door is obtained by solving the three-dimensional coordinate transformation equation group.

4. The method according to claim 3, characterized in that Based on the platform safety door feature point set and the train door feature point set, the feature points are spatially reprojected using a homography matrix transformation to establish a correspondence between the two sets of feature point sets, including: Feature classification is performed on the platform safety door feature point set and the train door feature point set respectively, and the neighborhood grayscale distribution information of each feature point is encoded into a feature description vector to obtain the platform safety door feature description vector set and the train door feature description vector set; Calculating a Euclidean distance matrix between feature points based on the platform safety door feature description vector set and the train door feature description vector set; Performing initial matching of feature points according to the Euclidean distance matrix, selecting the feature point pair with the smallest distance in the Euclidean distance matrix, and generating an initial feature point correspondence set; Constructing a homography matrix based on the corresponding set of the initial feature points, taking the set of feature description vectors as prior knowledge, iteratively optimizing the transformation parameters of the homography matrix through maximum a posteriori probability estimation, and performing spatial reprojection on the feature points; The reprojection error is calculated based on the coordinates of the feature points after spatial reprojection and the coordinates of the original feature points before spatial reprojection. The feature point pairs with a reprojection error greater than a preset threshold are removed from the initial feature point correspondence set, and the correspondence between the platform safety door feature point set and the train door feature point set is established.

5. The method according to claim 1, wherein Generating a train position adjustment instruction including a position compensation parameter according to the relative position deviation value and the real-time deceleration, and sending the train position adjustment instruction to the train automatic driving system, comprising: Decomposing the relative position deviation value into a lateral deviation component and a longitudinal deviation component, and calculating a train position adjustment compensation coefficient based on the lateral deviation component and the longitudinal deviation component; collecting a train wheel-rail adhesion coefficient, determining a train braking parameter according to the real-time deceleration and the train wheel-rail adhesion coefficient, and combining the train braking parameter with the train position adjustment compensation coefficient to generate a position compensation parameter; Building a position adjustment priority queue based on the position compensation parameters, dynamically adjusting the execution order of instructions in the position adjustment priority queue according to the numerical values ​​of the lateral deviation component and the longitudinal deviation component, and generating a train position adjustment instruction containing priority information; The train position adjustment instruction is sent to the train automatic driving system through the on-board communication network.

6. The method according to claim 1, characterized in that Receiving position adjustment feedback information returned by the automatic train driving system based on a preset adjustment step size, calculating a real-time alignment error between the platform safety door and the train door according to the position adjustment feedback information, and sending a stop instruction to the automatic train driving system when the real-time alignment error is less than a preset error threshold, including: receiving position adjustment feedback information returned by the train automatic driving system based on a preset adjustment step size; Establishing a coordinate mapping relationship between the platform coordinate system and the train coordinate system based on the position adjustment feedback information, mapping the train motion parameters to the platform coordinate system, and obtaining the motion state parameters of the train in the platform coordinate system; constructing a train motion compensation function according to the motion state parameters, calculating a compensation gain coefficient of the train motion compensation function according to track curvature information, performing compensation correction on train position information based on the compensation gain coefficient, and calculating a lateral alignment error and a longitudinal alignment error between a platform safety door and a train door in a platform coordinate system; Constructing a position deviation matrix according to the lateral alignment error and the longitudinal alignment error, and calculating a correction parameter of the preset adjustment step size based on the motion state parameter; The position deviation matrix is ​​decomposed into a transverse projection component and a longitudinal projection component in the orthogonal projection space, and a preset error threshold is adjusted according to the ratio of the projection components. When both the transverse projection component and the longitudinal projection component are less than the adjusted preset error threshold, a stop command is sent to the train automatic driving system.

7. The method according to claim 6, characterized in that Decomposing the position deviation matrix into a transverse projection component and a longitudinal projection component in an orthogonal projection space, adjusting a preset error threshold according to a ratio of the projection components, and sending a stop instruction to the train automatic driving system when both the transverse projection component and the longitudinal projection component are less than the adjusted preset error threshold, including: Extract the eigenvectors from the position deviation matrix and construct a biorthogonal basis matrix; Mapping the biorthogonal basis matrix to a Riemannian manifold space to construct a nonlinear orthogonal projection space; Decomposing the position deviation matrix into a transverse position component and a longitudinal position component in the nonlinear orthogonal projection space, and calculating a projection mapping relationship between the transverse position component and the longitudinal position component; Constructing a Riemann curvature tensor based on the projection mapping relationship, and calculating the principal curvatures of the transverse position component and the longitudinal position component using the Riemann curvature tensor; constructing a Riemannian metric tensor according to the principal curvature, solving a geodesic equation in the Riemannian manifold space based on the Riemannian metric tensor, and using the shortest path length of the geodesic equation as a dynamic adjustment coefficient of a preset error threshold; When the lateral position component is less than the preset error threshold after adjustment by the dynamic adjustment coefficient and the longitudinal position component is less than the preset error threshold after adjustment by the dynamic adjustment coefficient, a stop instruction is sent to the train automatic driving system.

8. A system for coordinating alignment, calibration, and docking of platform safety doors and train doors, for implementing the method according to any one of claims 1 to 7, characterized in that: include: The first unit is used to determine the position information of the platform safety door and the real-time deceleration of the train based on the acquired train head position information and train running speed information; The second unit is configured to collect first image information on both sides of the platform safety door frame and second image information on both sides of the train door frame, and obtain a relative position deviation value between the platform safety door and the train door through image registration of the first image information and the second image information; a third unit, configured to generate a train position adjustment instruction including a position compensation parameter according to the relative position deviation value and the real-time deceleration, and send the train position adjustment instruction to a train automatic driving system; a fourth unit, configured to receive position adjustment feedback information returned by the automatic train driving system based on a preset adjustment step length, calculate a real-time alignment error between the platform safety door and the train door according to the position adjustment feedback information, and send a stop command to the automatic train driving system when the real-time alignment error is less than a preset error threshold; The fifth unit is used to respond to the stop instruction and make the train stop smoothly at a preset stop position through closed-loop feedback control of the train braking system.

9. An electronic device, characterized in that: include: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that: When the computer program instructions are executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images