A method for generating and braking intrusion of light bands along railway hump routes

By constructing a directed track topology map and video mapping of railway hump yards, a unique route status token data object is generated to identify train crossing events. This solves the problem of binding route light strips with trains, realizes the extension of train-specific route light strips and direct association with braking control, and improves the reliability of monitoring and braking in railway hump yard operations.

CN123104734APending Publication Date: 2026-11-17WUHAN ZHIHUIDA TECH CO LTD
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202611275604.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-21
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

In existing technologies, the railway hump route light strip cannot be continuously bound to a specific train, and identity mismatch is prone to occur when multiple trains run in parallel. The light strip is extended without actual operation verification, and there is no direct correlation between personnel intrusion alarm and target braking equipment.

Method used

By collecting video images of railway hump yards, a directed track topology map is constructed and a mapping relationship is established. A route status token data object with a unique token is generated, the virtual event gates at the front and rear ends of the train are identified, a track topology event fingerprint is generated, and the route ownership is verified segment by segment based on the binding result. When the detection personnel intrude, the braking control authority is determined, so as to realize the unique binding and braking linkage between the train and the route.

Benefits of technology

It achieves a unique binding between train and route, reduces identity mismatch when multiple trains operate in parallel, verifies the actual running path of the train section by section, and improves the pertinence and reliability of hump yard operation monitoring and braking control.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123104734A_ABST
    Figure CN123104734A_ABST
Patent Text Reader

Abstract

This invention discloses a method for generating and braking intrusions into railway hump yard routes. It constructs a directed track topology map based on station video, establishing a mapping between interlocking equipment, topology nodes, and video pixel areas. For each hump yard operation, a route status token is generated. Virtual event gates are set at track section entrances / exits and turnout branches. Train front and rear crossing events are associated with section occupancy and clearance events to form a track topology event fingerprint. This invention relates to the field of hump yard operation monitoring technology in railway marshalling yards. This method generates a unique route status token data object for each hump yard operation and uses track topology event fingerprints to uniquely bind trains to routes. This transforms the route light strip from a simple section status display into a train-specific route light strip continuously associated with a specific train, reducing identity mismatches and light strip crosstalk during parallel operations of multiple trains.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of monitoring hump yard operations at railway marshalling yards, and more specifically, to a method for generating and braking intrusions along railway hump route light strips. Background Technology

[0002] Railway hump yards typically display routes based on interlocking status, turnout indication, and track section occupancy information. CN204399207U discloses a white light strip on the control panel to indicate an open route, changing the light strip display status when a train occupies a section. However, this light strip mainly reflects the route or section status and cannot continuously characterize the specific train identity. CN115984200A uses video analysis to obtain track occupancy and clearance status and compares it with computer interlocking status to identify route anomalies. However, it has not yet organized the events before and after the train, turnout branches, and section timing into an event fingerprint that can be used for unique route matching. CN112918515B can determine wrong track and suspend hump pushing based on the location of the shunting train and the accessibility of the target track, but it does not solve the problem of identity inheritance and dedicated light strip transmission when multiple trains are running parallel and video is obstructed. CN113320571B discloses emergency braking when unauthorized personnel approach a train, but it does not uniformly bind the personnel intrusion area, specific route, target train, and braking authority. Therefore, existing technologies still lack a method that can achieve unique binding between train and route, controlled extension of light strips segment by segment, abnormal freezing, and directional braking linkage. Summary of the Invention

[0003] The purpose of this invention is to provide a method for generating and braking intrusions along railway hump routes. This method solves the problems of existing route light strips not being able to be continuously bound to specific trains, identity mismatches easily occurring when multiple trains run parallel or when video is obstructed, light strips extending without actual operational verification, and the lack of direct correlation between personnel intrusion alarms and target braking devices.

[0004] This invention achieves the above objective through the following technical solution: a method for generating and braking intrusions along a railway hump route, the method comprising: Collect video images of railway hump yards, identify rail centerlines, track section boundaries, turnout bifurcation nodes, track merging nodes and marshalling track entrances, construct a directed track topology map according to the train's running direction, and establish a mapping relationship between interlocking equipment numbers, track topology nodes and video pixel areas; The system obtains the peak-pushing operation command, route number, target track, required turnout position, actual turnout status, track section occupancy status, and route locking status. For each peak-pushing operation, a route status token data object with a unique token identifier is generated. The route status token data object records the predetermined track section sequence, predetermined turnout sequence, token permission transmission relationship, current token position, valid light strip section, token running status, and braking control authority. Virtual event gates are set at the entrances and exits of track sections and at turnout branches. Video topology events formed by the front and rear ends of trains passing through the virtual event gates are acquired. The video topology events are associated with track section occupancy events and track section clearance events according to a unified timestamp to generate track topology event fingerprints. The track topology event fingerprint is uniquely matched with the route status token data object that is in the matching allowed state to complete the binding of the train and the route status token data object. Based on the binding results, the predetermined route ownership, actual turnout status, track section occupancy status, and track topology event continuity of the next track section are verified segment by segment. When the verification is successful, the route status token data object is transmitted and the dedicated route light band of the train is extended. When the verification fails, the route status token data object is frozen. When the inspection personnel enter the track area corresponding to the dedicated route light strip of the train or the braking protection area determined by the track area, they determine the target train, braking control authority and target braking execution device based on the route status token data object, and send a braking command to the target braking execution device.

[0005] Furthermore, the construction of the directed orbit topology graph and the establishment of mapping relationships include: Distortion correction is performed on the station video images, and the track pixel area and rail centerline are extracted; Track sections are divided according to the location of track insulation joints, track circuit boundaries, or axle counting section boundaries. The rail bifurcation points, track merging points, and marshalling track entrances are marked as turnout bifurcation nodes, track merging nodes, and track entrance nodes, respectively. Construct a directed track topology graph with track topology nodes as vertices and track segments as directed edges, and record the start node, end node, and upstream and downstream directed edges for each directed edge; Bind the track section number to the corresponding directed edge and track pixel area, bind the turnout number to the corresponding turnout branch node, and record the downstream directed edge connected to the turnout positioning state and reverse position state.

[0006] Further, the generation of the route status token data object includes: When the route locking state is valid, read the sequence of predetermined track sections and the sequence of predetermined turnouts that are connected sequentially from the hump to the target track according to the route number and the target track. Record the positioning or reversing state of each predetermined turnout as the required position of the turnout, and form a token permission transmission relationship between adjacent predetermined track sections; The unique token identifier, route number, target track, predetermined track section sequence, predetermined turnout sequence, turnout required position, token allowed transmission relationship, current token position, valid light strip section, token running status and braking control authority are encapsulated into a route status token data object. The token's running status includes pending binding, allowed matching, already bound, and frozen.

[0007] Furthermore, the acquisition of video topology events includes: Virtual event gates intersecting with the rail centerline are set in the video pixel areas corresponding to the hump, track section entrances and exits, turnout branch entrances, and marshalling track entrances; The train train in the video image is continuously tracked and the front and rear ends of the train train are determined. When the train train passes through the virtual event gate at the end of the train and enters the corresponding downstream track pixel area, the crossing event is confirmed. The crossing direction is determined based on the track pixel region before and after the crossing, and the actual turnout branch is determined based on the downstream directed edge after the turnout fork node. The temporary train identifier, virtual event gate identifier, train end identifier, crossing time, crossing direction, actual turnout branch, and unified timestamp are encapsulated into a video topology event.

[0008] Further, the generation of orbital topology event fingerprints includes: By sorting the video topology events of the same train according to a unified timestamp, the order of track topology nodes and the actual turnout branch sequence are obtained. Pair the video topology event of the train's front end passing through the entrance of the track section with the event of the track section changing from idle to occupied, and pair the video topology event of the train's rear end passing through the exit of the track section with the event of the track section changing from occupied to idle. The 95th percentile of the absolute value of the time difference of corresponding events in the normal operation sample, and the sum of the maximum time error of video acquisition, interlocking acquisition and unified time synchronization, are used as the event pairing time window; The sequence of track topology nodes, the actual turnout branch sequence, the track section occupancy order, the track section clearing order, the time interval between adjacent events, and the event pairing results are combined to form a track topology event fingerprint.

[0009] Furthermore, the unique match includes: Extract the confirmed track segment sequence, actual turnout branch sequence, and adjacent event time interval from the track topology event fingerprint; extract the predetermined track segment sequence and turnout requirement position from the route status token data object of each allowed matching state. Retain candidate route status token data objects whose confirmed track segment order is the prefix of the predetermined track segment sequence, whose actual turnout branches are consistent with the required positions of the corresponding turnouts, and whose adjacent track segments are directly connected in the directed track topology graph; Based on the segment length between adjacent track topology nodes, the upper and lower limits of normal train operation speed, and the event pairing time window, the continuity of the time interval between adjacent events is checked. When there is only one candidate route status token data object remaining and the continuity check passes, write the train identifier and update the token running status to the bound status.

[0010] Furthermore, the transmission of the route status token data object and the extension of the train-specific route light strip includes: Read the subsequent reserved track segment of the current token position from the token pass-through relationship; After the front end of the train passes through the entrance of the subsequent predetermined track section, it verifies whether the track section belongs to the predetermined track section sequence, whether the actual representation of the connecting turnout is consistent with the required position of the turnout, whether the track section is not occupied by other trains, and whether the new event is continuous with the track topology event fingerprint. When all verifications pass, update the current token position, add the track segment to the valid light strip segment, and extend the train-specific route light strip according to the mapping relationship; When the rear end of the train crosses the exit of an occupied track section and a clearing event occurs in that track section, the track section is removed from the valid light strip section; if any verification fails, the token running status is updated to frozen.

[0011] Furthermore, when video obstruction occurs during train operation: Record the last confirmed track section, actual turnout branch, direction of operation, last event time, and unique token identifier before obstruction; Starting from the last confirmed track segment, a candidate track segment sequence is generated along the directed track topology graph and the token allowance passing relationship, and candidate results with inconsistent occupancy order are deleted based on the track segment occupancy events and clearing events during the occupancy period. After the obstruction is removed, the topological and temporal continuity of the track section where the train is re-detected, the actual turnout branch, and the time of re-appearance will be checked against the retained candidate track section sequence. If only one candidate result passes the verification, the re-detected train will inherit the unique token identifier; if no unique candidate result is generated, the corresponding token's running status will be updated to frozen.

[0012] Furthermore, the frozen route status token data object includes: When any of the following situations occur: the actual turnout branch of the train does not match the required turnout position; the track section indicated by the video topology event does not match the occupancy status output by the interlocking equipment; the same train matches two or more route status token data objects; or the same route status token data object is bound to two or more trains, a freeze record containing the abnormal type, abnormal track section, abnormal time and unique token identifier is generated. Update the token's running status to frozen, revoke the permission to pass it to subsequent predetermined track sections, maintain the valid light strip sections confirmed before freezing, and prohibit the train's dedicated route light strip from extending to unconfirmed track sections. After regaining continuous video topology events, track segment occupancy events, and track segment clearing events, a unique match is re-executed from the abnormal track segment.

[0013] Furthermore, the inspector sends a braking command, including: Based on the video pixel area corresponding to the effective section of the light strip, a train-specific personnel detection area is formed, and a braking protection area is determined from the current token position along a predetermined track section sequence. The termination position of the braking protection area is determined by the current speed of the train, the command response time of the target braking execution device, the braking force establishment time, and the braking distance. The foot contact point of the target personnel is detected, and the intruded track section is determined when the target personnel continuously enter the train's dedicated personnel detection area or braking protection area. The system queries the route status token data object in the valid section of the light strip that contains the intruded track section, and generates a braking control message based on the train identification, braking control authority and target braking execution device number in the data. Send the braking control message and receive braking execution feedback, and write the sending time, feedback time and execution result into the route status token data object; if braking execution feedback is not received within the feedback period specified by the target braking execution device, output a braking unconfirmed alarm and remain in a frozen state.

[0014] The beneficial effects of this invention are as follows: 1. This invention uses a directed track topology map as a unified spatial reference for video images and interlocking equipment status, and establishes a correspondence between track sections, turnout nodes, track entrances, interlocking equipment numbers and video pixel areas, so that video detection results can be cross-checked with interlocking status at the track topology level.

[0015] 2. This invention generates a unique route status token data object for each peak-pushing operation and uses track topology event fingerprints to complete the unique binding between the train and the route, so that the route light strip changes from a simple section status display to a train-specific route light strip that is continuously associated with a specific train, reducing identity mismatch and light strip crosstalk when multiple trains operate in parallel.

[0016] 3. This invention allows token transmission and light band extension only when the subsequent track section belongs to the predetermined route, the actual status of the turnout is correct, the section is not occupied by other trains, and the new event is continuous with the track topology event fingerprint. This enables the verification of the actual train running path section by section.

[0017] 4. The present invention restores the train identity based on the last confirmed position before obstruction, the token permission transmission relationship, the segment occupancy and clearing order during obstruction, and the continuity of event time. Even when the train is temporarily obstructed or video targets overlap, the continuity between the train, the token, and the dedicated approach light strip can still be maintained.

[0018] 5. This invention unifies the personnel detection area, braking protection area, train-specific route light strip, route status token, and braking control authority. When personnel intrusion occurs, the affected train and target braking execution equipment can be directly identified, and the abnormal freezing mechanism can prevent the erroneous light strip from continuing to spread, thereby improving the pertinence and reliability of hump yard monitoring and braking control. Attached Figure Description

[0019] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a flowchart illustrating the overall method of the present invention; Figure 2 This is a schematic diagram of the directed orbit topology and mapping relationship of the present invention; Figure 3 This is a schematic diagram of track topology event fingerprint matching and route status token transmission according to the present invention; Figure 4 This is a schematic diagram of the personnel intrusion detection and target braking control of the present invention. Detailed Implementation

[0020] The technical solution of the present invention will be further described below with reference to the accompanying drawings and specific embodiments. It should be noted that the following embodiments are used to illustrate the technical principles, data processing procedures, and control logic of the present invention, and are not intended to limit the scope of protection of the present invention. Adaptive adjustments made by those skilled in the art to the number of video acquisition devices, device installation locations, data interface forms, number of station tracks, and parameter values ​​without departing from the concept of the present invention can all be considered as implementations of the present invention.

[0021] This invention is applicable to monitoring of peak-pushing operations, train identification tracking, route light strip display, personnel intrusion detection, and braking linkage control in railway hump yards. The equipment executing the method includes a station video acquisition device, a video analysis device, an interlocking status acquisition interface, a track topology data storage device, a route status token management device, a light strip display terminal, a personnel detection device, and a braking control interface.

[0022] The station's video acquisition devices are located at the hump top, turnout bifurcation area, track section intersection, and marshalling track entrance, covering the track area from the hump top to the marshalling track. The video analysis device receives station video images, identifies track areas, train targets, and personnel targets, and generates video topology events of trains passing through virtual event doors. The interlocking status acquisition interface obtains peak-pushing operation commands, route numbers, target tracks, required turnout positions, actual turnout status, track section occupancy status, track section clearance status, and route locking status.

[0023] The video analysis device, interlocking status acquisition interface, and braking control interface all use the same timing source. Video topology events, track section occupancy events, track section clearing events, route status token status changes, and braking control commands are all recorded with a unified timestamp.

[0024] In this embodiment, distance is uniformly expressed in meters, time in seconds, speed in meters per second, and acceleration in meters per second squared. The video image resolution is set to 1920×1080 pixels, and the video sampling frequency is set to 25 frames per second. These parameters can be adjusted according to the station area and camera equipment performance, but the physical quantities used in the same data processing process maintain consistent dimensions.

[0025] Example 1: Please see Figure 1-4 This embodiment takes a railway hump yard with 8 marshalling tracks as the object, and uses 6 video acquisition devices to cover the hump top, the first-level turnout area, the second-level turnout area and the entrance area of ​​the marshalling tracks. The specific implementation process is as follows.

[0026] S1. Construct a directed orbital topology graph and establish mapping relationships.

[0027] Video images of railway hump yards under conditions of no trains and large-scale obstruction are acquired. Lens distortion correction and station area calibration are performed on the video images. During station area calibration, no fewer than 12 control points with fixed locations and known actual coordinates are selected within the station area. Based on the actual station coordinates and video pixel coordinates of each control point, a transformation relationship between the actual station coordinate system and the video pixel coordinate system is established.

[0028] After establishing the transformation relationship, four additional control points not involved in the transformation calculation are selected for verification. The video pixel coordinates of the verification control points are converted to actual station coordinates, and the distance error between the conversion result and the actual location is calculated. When the maximum distance error corresponding to the four verification control points is no greater than 0.30m, the station area calibration is deemed valid; when the maximum distance error is greater than 0.30m, control points are reselected and calibration is performed again.

[0029] The value of 0.30m is determined based on the width of the track area, the positioning error of the video target, and the vibration test results of the camera equipment installation. The determination rule is as follows: under the condition that the calibration error will not cause the same detection position to cross into the adjacent track pixel area, take the sum of the maximum value of the on-site calibration error over 7 consecutive days and the 0.05m installation vibration margin, and round it up to an integer multiple of 0.05m.

[0030] The track pixel region is extracted from the calibrated video image, and the rail centerline is extracted along the track extension direction. The track pixel region can be obtained from a manually annotated station base map or from a segmentation model trained on track image samples. When using a segmentation model, the track bifurcation and merging regions are manually verified to ensure that the extracted rail centerline is consistent with the actual track extension direction.

[0031] Based on the location of the track insulation joint, the boundary of the track circuit, or the boundary of the axle counting section, the rail centerline is divided into multiple track sections. The location where the rail centerline bifurcates is determined as the turnout bifurcation node, the location where the two track paths merge is determined as the track merging node, and the location where the train enters the marshalling track is determined as the track entrance node.

[0032] The vertices of the directed track topology are the boundary nodes of the track sections, the branching nodes of the turnouts, the merging nodes of the tracks, and the track sections between adjacent vertices are the directed edges. The direction of the directed edges is determined according to the direction in which the train runs from the hump to the marshalling track.

[0033] Each directed edge records the corresponding track segment number, start node, end node, track segment length, upstream directed edge, and downstream directed edge. The track segment length is primarily based on the data recorded in the station's as-built drawings, and is verified using laser ranging equipment or trackside mileage markers. When the difference between the as-built drawing data and the verified data exceeds 0.20m, the verified data is used.

[0034] Each turnout branch node records the corresponding turnout number and the downstream directed edges connected when the turnout is in the locating and reversing states, respectively. Each track entry node records the corresponding train group number.

[0035] Establish a mapping relationship between interlocking equipment numbers, track topology nodes, and video pixel areas. Specifically, bind track section numbers to corresponding directed edges and track polygon pixel areas, bind turnout numbers to corresponding turnout branch nodes, and bind train group track numbers to corresponding track entrance nodes.

[0036] If the camera's position, focal length, or installation orientation changes, the mapping calibration must be re-performed. After the mapping calibration is completed, field trials are conducted at the entrances, exits, and turnout branches of each track section to verify the correspondence between the video pixel areas and the directed track topology map.

[0037] S2. Generate a route status token data object.

[0038] The video analysis device receives peak-pushing operation commands through the interlocking status acquisition interface and obtains the corresponding route number, target track, required turnout position, actual turnout status, track section occupancy status, and route locking status.

[0039] When the route locking state is valid, the predetermined track segment sequence connecting the hump to the target track is read from the directed track topology map according to the route number and the target track.

[0040] In this embodiment, the target track for the first peak-pushing operation is G3, and the corresponding predetermined track section sequence is Q1, Q2, Q5, Q8 and G3 in sequence, and the predetermined turnout sequence is D1, D3 and D7 in sequence.

[0041] Based on the connection relationships between adjacent predetermined track sections, the required turnout positions for each predetermined turnout are determined. Each pair of adjacent predetermined track sections forms a token permission transmission relationship. The token permission transmission relationships corresponding to the first peak-pushing operation include Q1 to Q2, Q2 to Q5, Q5 to Q8, and Q8 to G3.

[0042] Each peak-pushing operation is assigned a unique token identifier, and the unique token identifier, route number, target track, predetermined track section sequence, predetermined turnout sequence, turnout required position, token allowed transmission relationship, current token position, valid light strip section, token running status, train identification and braking control authority are encapsulated into a route status token data object.

[0043] The token's running status includes pending binding, allowed matching, bound, and frozen states. After the route status token data object is generated, the token's running status is initialized to pending binding. When the route locking state is valid, and the actual representation state of each turnout in the predetermined turnout sequence is consistent with the required position of the corresponding turnout, the token's running status is updated to allowed matching.

[0044] The current token position is initialized to the starting track segment in the predetermined track segment sequence, and the valid light strip segment is initialized to an empty set. Braking control permissions are determined based on the braking execution equipment corresponding to the target track, the communication status of the braking control interface, and the current operation control mode.

[0045] Whether a route status token enters the matching-allowed state is not determined by a weighted scoring method. A valid route locking status, correct actual turnout status, and a usable braking control interface are each independent necessary conditions. If any necessary condition is not met, the route status token remains in the pending binding state or enters a frozen state.

[0046] S3. Set up virtual event gates and generate video topology events.

[0047] Virtual event gates are set up in the video pixel areas corresponding to the hump, the entrance of each track section, the exit of each track section, the entrance of each turnout branch, and the entrance of each marshalling track.

[0048] The virtual event gate is set as a detection zone intersecting the centerline of the rail. The actual width of the detection zone along the rail extension direction is set to 0.60m. This width is determined according to twice the maximum permissible error of 0.30m for the station area calibration, so that the crossing process can still be stably identified when the positioning results of the train end fluctuate within the permissible error range.

[0049] The system identifies train targets in video images using a target detection model and generates temporary train markers using a target tracking algorithm. The front and rear ends of the train are determined based on the train's direction of travel.

[0050] A valid crossing event is confirmed when the front or rear of the train enters the downstream area from the upstream area of ​​the virtual event gate, and three consecutive frames are located in the downstream track pixel area. At a video sampling frequency of 25 frames per second, three consecutive frames correspond to a continuous confirmation time of 0.12 seconds.

[0051] The number of consecutive confirmation frames was determined using normal operation video samples. Verification was performed using 2 consecutive frames, 3 consecutive frames, and 4 consecutive frames. When using 2 consecutive frames, the repetition rate due to video jitter exceeded 1%; when using 3 consecutive frames, the repetition rate was less than 0.5%; and when using 4 consecutive frames, the repetition rate decreased by less than 0.1%, but the detection latency increased by 0.04 seconds. Therefore, this embodiment uses 3 consecutive frames as the train crossing confirmation threshold.

[0052] The crossing direction is determined based on the track pixel regions before and after the train's end passes through the virtual event gate. For turnout branch nodes, the actual turnout branch is determined based on the downstream directed edge into which the train enters.

[0053] The temporary train identifier, virtual event gate identifier, train end identifier, crossing time, crossing direction, actual turnout branch, and unified timestamp are encapsulated into a video topology event. The temporary train identifier is only used for continuous target tracking within the same video field of view and is not directly used as the basis for binding the train to the route. The binding between the train and the route is completed through a unique match between the track topology event fingerprint and the route status token data object.

[0054] S4. Generate track topology event fingerprints.

[0055] The interlocking status acquisition interface reads the occupancy status of each track section according to a predetermined acquisition cycle. When a track section changes from an idle state to an occupied state, a track section occupancy event is generated; when a track section changes from an occupied state to an idle state, a track section clearing event is generated.

[0056] Pair the video topology event of the train's front end crossing the entrance of the track section with the occupancy event of that track section, and pair the video topology event of the train's rear end crossing the exit of the track section with the clearing event of that track section.

[0057] The event pairing time window between video topology events and track segment occupancy events or track segment clearing events is determined according to formula (1): In the formula, Pair events with time windows, in units of ; The time difference between video topology events and corresponding track segment status events in a normal operation sample, in units of 1. ; The 95th percentile of the absolute value of the time difference, in units of ; This represents the video capture time error, in units of... ; This represents the time error for interlocking status acquisition, in units of... ; To standardize time synchronization errors, the unit is 1. .

[0058] The 95th percentile of the absolute value of the time difference was determined based on a sample of no less than 100 normal train sets. Using the 95th percentile can cover most normal communication and sampling fluctuations, while avoiding excessive expansion of the event pairing time window by a small number of abnormal delay samples.

[0059] In this embodiment, 100 sets of normal train passing samples were collected, and the 95th percentile of the absolute value of the time difference was 0.18s. The video sampling frequency was 25 frames per second, so one video sampling period of 0.04s was taken as the video acquisition time error; the maximum acquisition time error obtained from the interlocking status acquisition interface test was 0.02s; the maximum time error obtained from the unified time synchronization device test was 0.01s. Therefore, the event pairing time window was determined to be 0.25s.

[0060] When the absolute value of the time difference between a video topology event and its corresponding track segment status event is no greater than 0.25s, the event is considered successfully paired; when the absolute value of the time difference is greater than 0.25s, the video topology event or track segment status event is marked as an unpaired event.

[0061] The video topology events corresponding to the same train's temporary identifier are sorted according to a unified timestamp to obtain the track topology node sequence and the actual turnout branch sequence. The track topology node sequence, the actual turnout branch sequence, the track section occupancy order, the track section clearing order, the time interval between adjacent events, and the event pairing results are combined to form a track topology event fingerprint.

[0062] The track topology event fingerprint is continuously updated as the train runs. Each valid video topology event, track segment occupancy event, or track segment clearing event is added to the track topology event fingerprint of the corresponding train according to a unified timestamp.

[0063] S5. Perform a unique match between the train and the route status token data object.

[0064] Extract the confirmed track segment sequence, actual turnout branch sequence, and adjacent event time interval from the track topology event fingerprint. Extract the predetermined track segment sequence, predetermined turnout sequence, and turnout required position from each route status token data object in the allowed matching state.

[0065] First, a track segment order check is performed. It determines whether the confirmed track segment order is a consecutive prefix of the predetermined track segment sequence. Route status token data objects that are not consecutive prefixes are no longer considered as candidate results.

[0066] Then, turnout branch verification is performed. The actual turnout branch in the track topology event fingerprint is compared one by one with the turnout required position in the candidate route status token data object. If the actual turnout branch does not match the turnout required position, the corresponding candidate result is deleted.

[0067] Then perform a topology continuity check. Two adjacent confirmed track segments should be directly connected in the directed track topology graph, and the connection direction between adjacent track segments should be consistent with the train's running direction.

[0068] Perform a time continuity check on the candidate route status token data objects that have passed the above verification. When the actual time interval between two adjacent video topology events satisfies formula (2), it is determined that the corresponding events belong to the continuous operation process of the same train: In the formula, For the first The actual track distance between adjacent virtual event gates in a group, in units of ; This represents the upper limit of normal operating speed for trains on the corresponding track section, in units of... ; This represents the lower speed limit required for a continuous operation of the corresponding track section, in units of... ; The actual time interval between the front end of the train passing through adjacent virtual event gates, in units of ; Pair events with time windows, in units of .

[0069] Speed ​​limit The lower value between the upper speed limit specified in the station operation rules and the 99th percentile of the normal operation sample speed is used to prevent abnormally high-speed samples from expanding the allowable time range. Lower speed limit The larger of the minimum speed that the video system can reliably measure and the first percentile of the normal operating sample speed is taken to prevent false detection targets that are stationary or moving at too low a speed from being judged as continuously running trains.

[0070] In this embodiment, the maximum speed limit for the station is 5.0. The 99th percentile of the normal operating sample speed is 4.8. ,therefore Take 4.8 The minimum speed that the video system can reliably measure is 1.0. The first percentile of the normal operating sample speed is 1.2. ,therefore Take 1.2 .

[0071] Taking the virtual event gates at entrances Q5 and Q8 as an example, the actual track distance between the two virtual event gates is 36m, and the event pairing time window is 0.25s. According to formula (2), the allowable range of the time interval between adjacent events is 7.25s to 30.25s. When the detected actual time interval is 8.40s, the time continuity check is deemed to have passed.

[0072] The sequence of track sections, actual turnout branches, topological connection relationships, and time intervals between adjacent events are all necessary conditions for a unique match. No matching weight is set, and no weighted summation method is used to calculate the matching score.

[0073] If all necessary conditions are met and only one candidate route status token data object remains, the train identifier is written to that route status token data object, and the token's running status is updated to the bound state. If no candidate route status token data object exists, or if two or more candidate route status token data objects exist, train binding is not performed, and a matching exception record is generated.

[0074] S6. Transmit the route status token data object and extend the train-specific route light strip.

[0075] After the train and route status token data objects are uniquely bound, the subsequent predetermined track section corresponding to the current token position is read from the token pass-through relationship.

[0076] When the front end of a train passes through the virtual event gate at the entrance of a subsequent predetermined track section, the following checks are performed in sequence: determine whether the subsequent track section belongs to the predetermined track section sequence in the route status token data object; determine whether the actual status of the turnout connecting the current token position and the subsequent track section is consistent with the turnout's required position; determine whether the subsequent track section is not occupied by other trains; and determine whether the newly generated video topology event and track section occupancy event are continuous with the current track topology event fingerprint.

[0077] All of the above verification conditions are necessary and no verification weight is set. When all verification conditions are met, the current token position is updated to the subsequent predetermined track segment, and that track segment is added to the valid optical band segment.

[0078] Based on the mapping relationship between track sections and video pixel areas, dedicated light strips for each train are displayed on the station monitoring interface. These dedicated light strips can be displayed on the station track topology map or overlaid on the station video image. Each dedicated light strip is associated with a unique token identifier and a train identifier.

[0079] When the rear of the train crosses the exit of an occupied track section, and the corresponding track section generates a clearing event within a 0.25s event pairing time window, that track section is removed from the valid segments of the light strip. Through the above processing, the dedicated route light strip for the train extends segment by segment according to the actual running position of the train, and fades segment by segment as the rear of the train clears the track section.

[0080] If any verification condition is not met, the route status token data object is stopped from being passed to the subsequent track segment, and the token running status is updated to frozen, while the valid optical band segment that was confirmed before freezing is maintained.

[0081] S7. Handle video obstruction of train tracks.

[0082] When a train is obstructed by station facilities, enters the blind spot of the camera equipment, or overlaps with another train in the video target, record the last confirmed track section before obstruction, the actual turnout branch, the direction of operation, the last event time, and the unique token identifier.

[0083] Starting with the last confirmed track segment before occlusion, a sequence of candidate track segments is generated along the directed track topology graph and the token-allowed passing relationship.

[0084] During the obstruction period, occupancy and clearing events for each candidate track segment are continuously acquired. If the occupancy order of the candidate track segment sequence is inconsistent with the actual occupancy order of the track segments, the candidate result is deleted; if the actual representation state of the turnout corresponding to the candidate track segment is inconsistent with the required position of the turnout, the candidate result is deleted.

[0085] After the obstruction is removed, the track section where the re-detected train is located, the actual turnout branch, and the time of re-appearance are obtained, and topological continuity and time continuity checks are performed with the retained candidate track section sequence, respectively.

[0086] If only one candidate result passes the verification, the re-detected train is identified as the train before the occlusion, and it inherits the original unique token identifier and the train's dedicated path light strip. To reduce the possibility of incorrect train identity inheritance due to a single false detection, after the occlusion is removed, three consecutive sets of mutually paired and topologically contiguous events are required before token inheritance can be restored.

[0087] The number of consecutive events is determined using occluded video samples. When using two consecutive events as the recovery condition, the error inheritance rate is 1.6%; when using three consecutive events, the error inheritance rate decreases to 0.3%; when using four consecutive events, the error inheritance rate decreases by less than 0.1%, but the recovery delay increases significantly. Therefore, this embodiment uses three consecutive events as the token inheritance threshold.

[0088] If no candidate result passes the verification, or if two or more candidate results pass the verification, token inheritance is not performed, and the corresponding path status token data object is updated to a frozen state.

[0089] S8. Perform abnormal freeze and rematch.

[0090] During train operation, the actual turnout branch of the train is continuously compared with the required turnout position recorded in the route status token data object. When the actual turnout branch and the required turnout position are inconsistent, a route deviation anomaly is generated.

[0091] The system continuously compares the track section where the train is located, as indicated by the video topology events, with the track section occupancy status output by the interlocking equipment. If the video shows that the train has entered a certain track section, but no occupancy event is generated for that track section within the 0.25s event pairing time window, an inconsistency in the generated status is detected. Similarly, if an occupancy event is generated for a track section, but no train is detected in the corresponding video pixel area, an inconsistency in the generated status is also detected.

[0092] This function counts the number of route status token data objects that match the same train, as well as the number of trains bound to the same route status token data object. A non-unique match exception is generated when the same train matches more than two route status token data objects, or when the same route status token data object is bound to more than two trains.

[0093] Route deviation anomalies, state inconsistency anomalies, and non-unique match anomalies are all triggered by independent conditions, without using anomaly scoring or weighting methods. When any anomaly occurs, a freeze record is generated.

[0094] The freeze record must include at least the anomaly type, the abnormal track section, the anomaly time, the train identifier, and the unique token identifier. The token running status of the corresponding route status token data object is updated to frozen, its permission to pass on to subsequent predetermined track sections is revoked, and the valid optical band sections confirmed before the freeze are retained.

[0095] After freezing, continue collecting video topology events, track section occupancy events, and track section clearing events. When three consecutive sets of events are successfully matched, the actual turnout branch matches the required turnout position, the track section occupancy order matches the predetermined track section sequence, and the unique matching result remains unchanged, re-execute matching from the abnormal track section.

[0096] The unfreezing threshold for three consecutive sets of events uses the same determination rule as the occlusion inheritance threshold to reduce the possibility of erroneous unfreezing due to a single state restoration. If the rematch result is unique, and the route locking state, turnout actual display state, and track section occupancy state are restored to the same state, the token running state is restored to the bound state. If the rematch result is still not unique, the frozen state is maintained, and a manual verification prompt is output.

[0097] S9. Determine the current speed of the train and the braking protection zone.

[0098] The current speed of the train is calculated based on the time difference between the front end of the train passing through two adjacent virtual event gates, according to formula (3): In the formula, The current speed of the train, in units of ; For the first The actual track distance between adjacent virtual event gates in a group, in units of ; The time difference between the front end of the train passing through adjacent virtual event gates, in units of .

[0099] The actual track distance between adjacent virtual event gates is determined by the segment length recorded in the directed track topology graph. Train speed is calculated only using cross-traffic events that are successfully paired and topologically continuous.

[0100] In this embodiment, the actual track distance between a group of adjacent virtual event gates is 16.8m, the time difference between the front end of the train passing through two virtual event gates is 4.0s, and the calculated train speed is 4.2m / s.

[0101] To reduce the impact of a single outlier timestamp on the speed calculation results, the median of the three most recent valid speed calculation results is used as the current train speed for braking control. Using the median eliminates a speed value that deviates significantly from the rest of the results, eliminating the need to set weights.

[0102] Based on the current train speed, braking control message transmission and equipment response time, braking force establishment time, effective deceleration, and safety margin distance, the longitudinal length of the braking protection zone is determined and calculated according to formula (4): In the formula, The longitudinal length of the braking protection zone, in units of ; The current speed of the train, in units of ; The time for brake control message transmission and target brake execution device response is expressed in units of 1. ; The time for braking force to build up, in units of ; The effective deceleration during the braking phase, in units of ; For safety margin distance, the unit is .

[0103] Braking control message transmission and target braking actuator response time The result is determined through at least 50 command response tests, and the 95th percentile of the test results is taken. In this embodiment, the 95th percentile of the test results is 0.15. ,therefore Take 0.15 .

[0104] Braking force build-up time The braking force build-up time is determined based on the test time taken for the target braking actuator to reach 90% of the rated braking force from receiving a valid braking command. In this embodiment, the braking force build-up time is set to 0.40. .

[0105] Effective deceleration The effective deceleration is determined based on the braking test results of the target braking device under maximum permissible train load and unfavorable track surface conditions, taking the 5th percentile of the test results. In this embodiment, the effective deceleration is taken as 0.45. This is to avoid underestimating the braking distance by using average deceleration.

[0106] Safety margin distance The distance margin is composed of the distance margin corresponding to the video position error, the distance margin corresponding to the velocity measurement error, and the station safety margin. In this embodiment, the distance margin corresponding to the maximum video position error is 0.8. The distance margin corresponding to the speed measurement error is 0.7. The station's safety margin is 1.5. Therefore, the safety margin distance is taken as 3.0. .

[0107] When the train's current speed is 4.2 Substituting the above parameters into formula (4), the longitudinal length of the braking protection zone is obtained as 24.91. When actually generating the braking protection zone, the value is increased to 25.0. This avoids reducing the protected area due to numerical truncation.

[0108] Starting from the current token position, set a length of 25.0 along the predetermined track segment sequence in the route status token data object forward of the train. The braking protection zone. When the protection zone spans two track segments, it is continuously mapped to the corresponding video pixel area according to the track path in the directed track topology map.

[0109] S10, Personnel intrusion detection and braking control.

[0110] Based on the valid light band segment in the route status token data object, the corresponding video pixel area is read and combined to form a train-specific personnel detection area.

[0111] The braking protection area determined according to formula (4) is mapped to the video image, so that the train-specific personnel detection area and the braking protection area are dynamically updated according to the current token position, train running direction and train current speed.

[0112] The system identifies human targets in video images using a human detection model. The center of the lower edge of the human target detection bounding box is used as the grounding point for the human's feet, and the video pixel region where the human is located is determined based on this grounding point.

[0113] An intrusion event is confirmed when a person's foot touchpoint is located within the train's dedicated personnel detection area or braking protection area for five consecutive frames. At a video sampling frequency of 25 frames per second, five consecutive frames correspond to a 0.20s confirmation time.

[0114] The number of frames required to confirm personnel intrusion is determined using a labeled video verification set. When using 3 consecutive frames as the confirmation condition, the false alarm rate caused by trackside tools, shadows, and localized reflections is 1.4%. With 5 consecutive frames, the false alarm rate decreases to 0.2%, and the confirmation delay does not exceed 0.20 seconds. When using more than 5 frames as the confirmation condition, the false alarm rate decreases less, but the braking trigger delay continues to increase. Therefore, this embodiment uses 5 consecutive frames as the personnel intrusion confirmation threshold.

[0115] After confirming the personnel intrusion event, the intruded track segment is determined based on the video pixel area where the personnel's feet are grounded, and the path status token data object containing the intruded track segment is queried in the valid light band segment.

[0116] Read the train identification, braking control authority, and target braking execution device number from the retrieved route status token data object.

[0117] When the route status token data object has braking control authority, a braking control message is generated. The braking control message includes a unique token identifier, train identifier, intruded track section, target braking actuator number, braking level, and sending timestamp.

[0118] The system sends braking control messages to the target braking actuator via the braking control interface and receives braking execution feedback from the target braking actuator. The timing of the braking control message transmission, the timing of the braking execution feedback, and the braking execution result are written into the route status token data object.

[0119] The braking execution feedback timeout is determined based on the maximum feedback time specified by the target braking execution device. In this embodiment, the maximum feedback time specified by the target braking execution device is 0.25s. Adding a 20% communication margin to this maximum feedback time, the braking execution feedback timeout is set to 0.30s.

[0120] If no braking execution feedback is received within 0.30s, an alarm for unconfirmed braking is output, and the corresponding route status token data object remains frozen.

[0121] After sending the braking control message, the corresponding route status token data object is updated to a frozen state, stopping the train's dedicated route light strip from extending further into unconfirmed track sections. After personnel leave the personnel detection area and braking protection area, the freeze can only be lifted after confirming that the target braking execution equipment status, the actual turnout indication status, the track section occupancy status, and the route locking status all meet the recovery conditions.

[0122] Example 2: This embodiment uses the parallel peak pushing of two trains and the video obstruction of the first train as an example to illustrate the unique binding of the train and the route status token data object, the obstruction inheritance, and the personnel intrusion braking process.

[0123] The first train corresponds to the first peak-pushing operation, with its target track being G3. The predetermined track section sequence is Q1, Q2, Q5, Q8 and G3, and the predetermined turnout sequence is D1, D3 and D7.

[0124] The second train corresponds to the second peak-pushing operation, with its target track being G5. The predetermined track section sequence is Q1, Q3, Q6, Q9 and G5, and the predetermined turnout sequence is D1, D4 and D8.

[0125] The route status token management device generates a first route status token data object and a second route status token data object. The two route status token data objects have different unique token identifiers and record their respective predetermined track section sequences, predetermined turnout sequences, and token permission transmission relationships.

[0126] After the first train passed through the virtual event gate at the top of the hump, the video analysis device sequentially detected that the first train had entered tracks Q1 and Q2. The interlocking status acquisition interface sequentially acquired track section occupancy events for tracks Q1 and Q2.

[0127] The time differences between the video topology event corresponding to the first train and the track section occupancy events corresponding to Q1 and Q2 are 0.11s and 0.16s, respectively, which are both no greater than the 0.25s event pairing time window, so the pairing is successful.

[0128] The first train branches off at turnout D1 to Q2, and the actual turnout branch matches the required turnout position recorded in the first route status token data object. The confirmed track section sequence Q1 and Q2 of the first train are consecutive prefixes of the predetermined track section sequence in the first route status token data object, but not consecutive prefixes of the predetermined track section sequence in the second route status token data object. Therefore, the first train is uniquely bound to the first route status token data object.

[0129] The second train then passes through the virtual event gate at the top of the hump and branches off to Q3 at turnout D1. The confirmed track section sequence and actual turnout branch of the second train are matched with the second route status token data object, and the second train is uniquely bound to the second route status token data object.

[0130] After the first train entered Q5, it was obstructed by station facilities. The video analysis device recorded the last confirmed track section Q5, the actual turnout branch, the direction of travel, the last event time, and the unique token identifier of the first route status token data object before the first train was obstructed.

[0131] Based on the token permission relationship in the first route status token data object, the subsequent predetermined track segment for Q5 is Q8. During the occlusion period, the interlocking status acquisition interface detects that Q8 has changed from an idle state to an occupied state.

[0132] The actual track distance between the virtual event gate at entrance Q5 and the virtual event gate at entrance Q8 is 36m. According to formula (2), the allowable range of the time interval between adjacent events is 7.25s to 30.25s.

[0133] After the obstruction was removed, the video analysis device re-detected the train in the video pixel area corresponding to Q8 at 8.40s. Q8 and Q5 are directly connected in the directed track topology diagram. Q8 belongs to the subsequent predetermined track section of the first route status token data object. The actual indicated state of turnout D7 is consistent with the required position of the turnout, and 8.40s is within the allowable time range.

[0134] After obtaining three sets of mutually paired and topologically continuous events in succession, the re-detected train is identified as the first train and inherits the first route status token data object and its train-specific route light strip.

[0135] The current speed of the first train is determined to be 4.2 m / s based on the median of the most recent three effective speed calculations. According to formula (4), the longitudinal length of the braking protection zone of the first train is determined to be 25.0 m.

[0136] When a person's foot touch point enters the train's dedicated personnel detection area corresponding to train Q8 for 5 consecutive frames, the video analysis device confirms that a personnel intrusion event has occurred.

[0137] Based on the intruded track segment Q8, the system queries the valid light band segment containing the route status token data object containing Q8, and the query result is the first route status token data object. According to the train identifier, braking control authority, and target braking actuator number in the first route status token data object, the system sends a braking control message to the target braking actuator corresponding to the first train.

[0138] The valid optical band segment of the second route status token data object does not include Q8, therefore no braking control message is sent to the braking actuator corresponding to the second train.

[0139] Therefore, this invention can maintain the one-to-one correspondence between train identification, route status token data object, train-specific route light strip and target braking execution device when two routes have a common starting track section, multiple trains push peaks in parallel, train video obstruction occurs and personnel enter the local track area, and realize the controlled extension of the route light strip section by section, abnormal freezing and directional braking for personnel intrusion.

[0140] The above embodiments provide a detailed description of the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for generating and braking intrusion of light bands along railway hump routes, characterized in that, The method includes: Collect video images of railway hump yards, identify rail centerlines, track section boundaries, turnout bifurcation nodes, track merging nodes and marshalling track entrances, construct a directed track topology map according to the train's running direction, and establish a mapping relationship between interlocking equipment numbers, track topology nodes and video pixel areas; The system obtains the peak-pushing operation command, route number, target track, required turnout position, actual turnout status, track section occupancy status, and route locking status. For each peak-pushing operation, a route status token data object with a unique token identifier is generated. The route status token data object records the predetermined track section sequence, predetermined turnout sequence, token permission transmission relationship, current token position, valid light strip section, token running status, and braking control authority. Virtual event gates are set at the entrances and exits of track sections and at turnout branches. Video topology events formed by the front and rear ends of trains passing through the virtual event gates are acquired. The video topology events are associated with track section occupancy events and track section clearance events according to a unified timestamp to generate track topology event fingerprints. The track topology event fingerprint is uniquely matched with the route status token data object that is in the matching allowed state to complete the binding of the train and the route status token data object. Based on the binding results, the predetermined route ownership, actual turnout status, track section occupancy status, and track topology event continuity of the next track section are verified segment by segment. When the verification is successful, the route status token data object is transmitted and the dedicated route light band of the train is extended. When the verification fails, the route status token data object is frozen. When the inspection personnel enter the track area corresponding to the dedicated route light strip of the train or the braking protection area determined by the track area, they determine the target train, braking control authority and target braking execution device based on the route status token data object, and send a braking command to the target braking execution device.

2. The method for generating and braking intrusion of railway hump route light strips according to claim 1, characterized in that, The construction of the directed orbit topology graph and the establishment of mapping relationships include: Distortion correction is performed on the station video images, and the track pixel area and rail centerline are extracted; Track sections are divided according to the location of track insulation joints, track circuit boundaries, or axle counting section boundaries. The rail bifurcation points, track merging points, and marshalling track entrances are marked as turnout bifurcation nodes, track merging nodes, and track entrance nodes, respectively. Construct a directed track topology graph with track topology nodes as vertices and track segments as directed edges, and record the start node, end node, and upstream and downstream directed edges for each directed edge; Bind the track section number to the corresponding directed edge and track pixel area, bind the turnout number to the corresponding turnout branch node, and record the downstream directed edge connected to the turnout positioning state and reverse position state.

3. The method for generating and braking intrusions along railway hump routes according to claim 1, characterized in that, The generated route status token data object includes: When the route locking state is valid, read the sequence of predetermined track sections and the sequence of predetermined turnouts that are connected sequentially from the hump to the target track according to the route number and the target track. Record the positioning or reversing state of each predetermined turnout as the required position of the turnout, and form a token permission transmission relationship between adjacent predetermined track sections; The unique token identifier, route number, target track, predetermined track section sequence, predetermined turnout sequence, turnout required position, token allowed transmission relationship, current token position, valid light strip section, token running status and braking control authority are encapsulated into a route status token data object. The token's running status includes pending binding, allowed matching, already bound, and frozen.

4. The method for generating and braking intrusion of railway hump route light strips according to claim 2, characterized in that, The video topology acquisition event includes: Virtual event gates intersecting with the rail centerline are set in the video pixel areas corresponding to the hump, track section entrances and exits, turnout branch entrances, and marshalling track entrances; The train train in the video image is continuously tracked and the front and rear ends of the train train are determined. When the train train passes through the virtual event gate at the end of the train and enters the corresponding downstream track pixel area, the crossing event is confirmed. The crossing direction is determined based on the track pixel region before and after the crossing, and the actual turnout branch is determined based on the downstream directed edge after the turnout fork node. The temporary train identifier, virtual event gate identifier, train end identifier, crossing time, crossing direction, actual turnout branch, and unified timestamp are encapsulated into a video topology event.

5. The method for generating and braking intrusions along railway hump routes according to claim 4, characterized in that, The generation of orbital topology event fingerprints includes: By sorting the video topology events of the same train according to a unified timestamp, the order of track topology nodes and the actual turnout branch sequence are obtained. Pair the video topology event of the train's front end passing through the entrance of the track section with the event of the track section changing from idle to occupied, and pair the video topology event of the train's rear end passing through the exit of the track section with the event of the track section changing from occupied to idle. The 95th percentile of the absolute value of the time difference of corresponding events in the normal operation sample, plus the sum of the maximum time errors of video acquisition, interlocking acquisition and unified time synchronization, is used as the event pairing time window; The sequence of track topology nodes, the actual turnout branch sequence, the track section occupancy order, the track section clearing order, the time interval between adjacent events, and the event pairing results are combined to form a track topology event fingerprint.

6. The method for generating and braking intrusions along railway hump routes according to claim 3, characterized in that, The unique match includes: Extract the confirmed track segment sequence, actual turnout branch sequence, and adjacent event time interval from the track topology event fingerprint; extract the predetermined track segment sequence and turnout requirement position from the route status token data object of each allowed matching state. Retain candidate route status token data objects whose confirmed track segment order is the prefix of the predetermined track segment sequence, whose actual turnout branches are consistent with the required positions of the corresponding turnouts, and whose adjacent track segments are directly connected in the directed track topology graph; Based on the segment length between adjacent track topology nodes, the upper and lower limits of normal train operation speed, and the event pairing time window, the continuity of the time interval between adjacent events is checked. When there is only one candidate route status token data object remaining and the continuity check passes, write the train identifier and update the token running status to the bound status.

7. The method for generating and braking intrusions along railway hump routes according to claim 5, characterized in that, The transmission of route status token data objects and the extension of the train-specific route light strip include: Read the subsequent reserved track segment of the current token position from the token pass-through relationship; After the front end of the train passes through the entrance of the subsequent predetermined track section, it verifies whether the track section belongs to the predetermined track section sequence, whether the actual representation of the connecting turnout is consistent with the required position of the turnout, whether the track section is not occupied by other trains, and whether the new event is continuous with the track topology event fingerprint. When all verifications pass, update the current token position, add the track segment to the valid light strip segment, and extend the train-specific route light strip according to the mapping relationship; When the rear end of the train crosses the exit of an occupied track section and a clearing event occurs in that track section, the track section is removed from the valid light strip section; if any verification fails, the token running status is updated to frozen.

8. The method for generating and braking intrusions along railway hump routes according to claim 6, characterized in that, When video is obstructed during train operation: Record the last confirmed track section, actual turnout branch, direction of operation, last event time, and unique token identifier before obstruction; Starting from the last confirmed track segment, a candidate track segment sequence is generated along the directed track topology graph and the token allowance passing relationship, and candidate results with inconsistent occupancy order are deleted based on the track segment occupancy events and clearing events during the occupancy period. After the obstruction is removed, the topological and temporal continuity of the track section where the train is re-detected, the actual turnout branch, and the time of re-appearance will be checked against the retained candidate track section sequence. If only one candidate result passes the verification, the re-detected train will inherit the unique token identifier; If no unique candidate result is generated, the running status of the corresponding token will be updated to frozen.

9. The method for generating and braking intrusion of railway hump route light strips according to claim 7, characterized in that, The frozen route status token data object includes: When any of the following situations occur: the actual turnout branch of the train does not match the required turnout position; the track section indicated by the video topology event does not match the occupancy status output by the interlocking equipment; the same train matches two or more route status token data objects; or the same route status token data object is bound to two or more trains, a freeze record containing the abnormal type, abnormal track section, abnormal time and unique token identifier is generated. Update the token's running status to frozen, revoke the permission to pass it to subsequent predetermined track sections, maintain the valid light strip sections confirmed before freezing, and prohibit the train's dedicated route light strip from extending to unconfirmed track sections. After regaining continuous video topology events, track segment occupancy events, and track segment clearing events, a unique match is re-executed from the abnormal track segment.

10. The method for generating and braking intrusions along railway hump routes according to claim 1, characterized in that, The inspector then sends a braking command, including: Based on the video pixel area corresponding to the effective section of the light strip, a train-specific personnel detection area is formed, and a braking protection area is determined from the current token position along a predetermined track section sequence. The termination position of the braking protection area is determined by the current speed of the train, the command response time of the target braking execution device, the braking force establishment time, and the braking distance. The foot contact point of the target personnel is detected, and the intruded track section is determined when the target personnel continuously enter the train's dedicated personnel detection area or braking protection area. The system queries the route status token data object in the valid section of the light strip that contains the intruded track section, and generates a braking control message based on the train identification, braking control authority and target braking execution device number in the data. Send the braking control message and receive braking execution feedback, and write the sending time, feedback time and execution result into the route status token data object; if braking execution feedback is not received within the feedback period specified by the target braking execution device, output a braking unconfirmed alarm and remain in a frozen state.

Citation Information

Patent Citations

  • An automatic calculation method for the peak pushing speed of a railway hump yard

    CN112918515B

  • A Smart Coupling Control Method for Railway Flatbed Shunting

    CN113320571B

  • Track shunt defect detection method and system based on video analysis

    CN115984200A

  • Microcomputer interlocking centralized control system for critical region of multiple groups of turnouts for railway shunting

    CN204399207U