A method and system for intelligent display of subway vehicle information

Abstract

This invention relates to the field of transportation management technology, and in particular to a method and system for intelligent display of subway in-vehicle information. The method includes the following steps: acquiring train operation status data, carriage environment perception data, and train event flow data; analyzing passenger behavior based on the carriage environment perception data to obtain passenger behavior data; obtaining focus status data based on the passenger behavior data; performing status perception scheduling based on the train operation status data, train event flow data, and focus status data to obtain status perception data; obtaining passenger attention screen data based on the status perception data; acquiring ground coordination data, and coordinating the passenger attention screen data based on the ground coordination data to obtain subway passenger coordination data, for auxiliary operation of intelligent display of subway in-vehicle information. This invention realizes the evolution of subway carriage information display from static and passive to dynamic and intelligent coordination, improving the overall quality of rail transit services and passenger travel experience.

Description

Technical Field

[0001] The present invention relates to the technical field of transportation management, and particularly to an intelligent display method and system for subway on-vehicle information. Background Art

[0002] With the continuous expansion of the rail transit network and the increasing density of subway passenger flow, the traditional subway on-vehicle information display system has been difficult to meet the complex application requirements such as multi-source situation awareness, event-driven response, and passenger personalized prompts. Currently, the on-vehicle display systems of mainstream subway trains mostly adopt the method of timed carousel or static information push triggered by stations, lacking the ability to perceive the behavior states of passengers in the carriage, and it is difficult to achieve dynamic adaptation between the displayed content and information such as the areas of interest of passengers, congestion situations, and transfer intentions. In the face of emergencies (such as skip-stop scheduling, temporary transfer, operation delay, etc.) or large passenger flow transfer scenarios, the traditional display system cannot achieve the linkage update between multiple screens in the carriage, and more cannot form event consistency synchronization with the ground dispatching center, resulting in information delay, inconsistency, or failure to reach key groups, reducing the efficiency of passengers obtaining effective information and the safety response ability Summary of the Invention

[0003] In order to solve the above technical problems, the present invention proposes an intelligent display method and system for subway on-vehicle information to solve at least one of the above technical problems.

[0004] The present application provides an intelligent display method for subway on-vehicle information, including the following steps:

[0005] Step S1: Obtain train operation status data, carriage environment perception data, and train event stream data; perform passenger behavior analysis based on the carriage environment perception data to obtain passenger behavior data; perform focused situation analysis based on the passenger behavior data to obtain focused situation data;

[0006] Step S2: Perform status perception scheduling based on the train operation status data, train event stream data, and focused situation data to obtain status perception data;

[0007] Step S3: Perform passenger attention screen driving based on the status perception data to obtain passenger attention screen data;

[0008] Step S4: Obtain ground coordination data, and perform coordinated linkage on the passenger attention screen data according to the ground coordination data to obtain subway passenger coordination data for intelligent display auxiliary operation of subway on-vehicle information.

[0009] This invention analyzes passenger behavior and identifies focus patterns by sensing data from the carriage environment, enabling dynamic perception of passenger attention areas and behavioral trends. Combined with train operation status and event information, it performs status-aware scheduling to prioritize and optimize displayed content. Simultaneously, a passenger attention-driven mechanism controls the screen's content, refresh rate, and visual style, improving information delivery efficiency and audience adaptability. Furthermore, by integrating ground-based collaborative data, it achieves event synchronization between the train and the dispatch center, and cross-carriage display linkage, providing consistent guidance and precise information delivery in critical scenarios such as emergencies and transfer pressures. This effectively enhances the system's intelligent linkage capabilities and the overall optimization of display strategies.

[0010] Preferably, step S1 specifically includes:

[0011] Acquire train operation status data, carriage environment perception data, and train event stream data;

[0012] Spatial behavior maps are constructed based on the passenger compartment environment perception data to obtain passenger compartment behavior map data;

[0013] Behavioral features are extracted from the carriage behavior graph data to obtain behavioral feature data;

[0014] Behavioral generalization enhancement is performed on behavioral feature data to obtain passenger behavior data.

[0015] In this invention, by acquiring train operation status, carriage environment perception, and event stream data, the system can achieve spatiotemporal synchronized data fusion. Subsequently, based on the environmental perception results, a spatial behavior map is constructed, which not only clarifies the passenger's movement paths, dwelling preferences, and behavior density in different areas, but also provides a structured foundation for subsequent feature extraction. Through in-depth analysis of the behavior map data, multi-dimensional behavioral features such as boarding and alighting patterns, dwell time distribution, and passage crossing frequency can be extracted. Then, through a behavior generalization enhancement mechanism, abnormal behaviors, marginal behaviors, or weak signal behaviors are modeled and reinforced, effectively improving the generalization ability and robustness of passenger status recognition.

[0016] Preferably, the focus of situational analysis specifically includes:

[0017] Regional attention data is obtained by calculating regional attention based on passenger behavior data;

[0018] The focus status data of the region is classified into focus status data.

[0019] By performing local congestion evolution on the focused state data, local congestion data is obtained;

[0020] Focus situation data is generated based on local congestion data and focus status data.

[0021] This invention calculates passenger attention levels in different areas, quantifying passenger gaze lingering, station density, and interaction frequency to establish a regional saliency model. It categorizes regional attention into focus states, such as transfer focus, waiting focus, and avoidance-based dispersal, enhancing the semantic recognition of passenger intentions. By modeling local congestion evolution in focus state data, it captures dynamic changes such as sudden density increases, shifting trends, and chain reactions, enabling early detection of sudden gathering situations. Based on focus state and local congestion data, it generates focus situation data, which not only possesses regional positioning accuracy but also temporal trends and event triggering, providing quantifiable input for display scheduling, content priority adjustment, and screen layout optimization, thereby improving the system's understanding of passenger behavior and its proactive response.

[0022] Preferably, step S2 specifically includes:

[0023] Train status feature data is obtained by extracting train status features from train operation status data and train event stream data.

[0024] Train status influence matrix data is obtained by mapping train status characteristic data and focused situation data;

[0025] Key response regions are identified from the train situation matrix data to obtain key response region data.

[0026] The scheduling window is generated based on the key response area data, resulting in scheduling window data.

[0027] State-aware scheduling is generated based on scheduling window data and a preset state-aware strategy library to obtain state-aware data.

[0028] This invention extracts train status feature data to comprehensively obtain core operational indicators such as the current train's operating stage, delay deviation, station proximity, and event sensitivity. Then, combined with focused situational data, a train situational influence matrix is ​​constructed to quantify the impact intensity of different carriage areas on information dispatching strategies under specific operating conditions. This matrix further identifies the key response areas requiring intervention in the current operating situation, providing a spatial range reference for display control. The generated dispatch window, based on this matrix, incorporates parameters such as operating rhythm, arrival countdown, and display duration to ensure a high degree of matching between displayed content and changes in passenger behavior. By calling the status-aware strategy library based on the dispatch window data to generate display strategies, intelligent control over screen content type, priority, and display rhythm can be achieved, significantly improving the timeliness, relevance, and dispatching response capabilities of the display, meeting the information interaction needs under different operating scenarios.

[0029] Preferably, the train situation influence matrix mapping is specifically as follows:

[0030] Initialize the track network topology diagram by analyzing the train operation status data and train event flow data to obtain the track network diagram data;

[0031] Based on the train status characteristic data, the track network diagram data is mapped to the running trajectory to obtain the running trajectory diagram data;

[0032] The trajectory graph data is weighted using a tensor structure to obtain weighted graph data;

[0033] Spatiotemporal coupling is performed on the weighted graph data to obtain coupled graph data;

[0034] Propagation link impact calculations are performed on the coupling graph data to obtain propagation graph data;

[0035] Event-driven structural mutations are performed on the propagation graph data to obtain train situation matrix data.

[0036] This invention initializes the subway line structure as a graph through a track network topology graph, providing a foundation for graph computation. By mapping train trajectory based on train state characteristics, the path distribution and state evolution path of the train in the network at the current moment can be constructed. Through a tensor structure weighting mechanism, multiple indicators such as train speed, station delay, and event intensity are embedded into the weight structure of nodes and edges in the graph in a multi-channel manner, forming high-dimensional weighted graph data. By performing spatiotemporal coupling processing on the weighted graph, the interaction relationships between state changes in different time slices and adjacent station segments can be revealed. Using propagation link influence calculation, the diffusion process of scheduling events or abnormal passenger behavior in the track graph is simulated, forming the event scope and response range. Through an event-driven structural mutation mechanism, the graph structure is dynamically adjusted to generate train status matrix data that can be used for scheduling decisions, thereby achieving precise control over the information display rhythm, response range, and priority strategy, significantly improving the global linkage and predictive scheduling capabilities of the display system.

[0037] Preferably, step S3 specifically includes:

[0038] Acquire data from the carriage screens;

[0039] Spatial modeling is performed on the screen data in the carriage to obtain the carriage screen model;

[0040] Attention probability field data is obtained by calculating the attention probability field based on the carriage screen model and focusing situation data;

[0041] Passenger attention screen data is obtained by multi-frequency driven generation based on attention probability field data and state perception data.

[0042] In this invention, the system spatially models the data from train carriage screens, accurately acquiring the geometric position, orientation, viewing angle, and coverage area of ​​each screen to construct a carriage screen model, providing a physical basis for display strategies. Subsequently, by combining focus situation data, it calculates the attention probability field, quantifying the distribution of attention across different screens among passengers in different areas, thereby capturing passenger gaze concentration trends and regional attention dynamics. By integrating state-aware data, it differentiates the display rhythm, refresh rate, and content type parameters of each screen, forming a multi-frequency driven strategy tailored to different user groups, and outputting passenger attention screen data. This process enables regional adaptation and temporal control of screen content based on real-time passenger behavior and train status, significantly improving the efficiency and accuracy of information acquisition for passengers, while also increasing the display density and energy efficiency of screen resources, and enhancing the system's intelligent adaptability and perception response capabilities in different scenarios such as peak congestion and transfer nodes.

[0043] Preferably, the spatial modeling specifically includes:

[0044] Based on the data from the carriage screens, a carriage screen space model is obtained by creating a carriage screen space model.

[0045] A visual cone model is constructed from the screen space model of the carriage to obtain the screen visual cone model;

[0046] Crowd attention heatfield is constructed on the screen visual cone model to obtain crowd attention heatfield data;

[0047] The actual screen visibility probability is calculated from the crowd attention heat field data to obtain the actual screen visibility probability data.

[0048] The actual visibility probability of the screen and the crowd attention heat field data are used to annotate the spatial model of the carriage screen to obtain the carriage screen model.

[0049] In this invention, by modeling the space of the train carriage screens, the system can obtain the specific position and orientation of each screen within the three-dimensional carriage structure; construct a visual cone model to clarify the theoretical visible area of ​​each screen, and spatially restrict this area based on passenger distribution. The system performs crowd attention heat field modeling, dynamically generating a spatial heat map of passenger gaze focus based on information such as passenger position, gaze direction, and dwell time; calculates the actual visibility probability of the screens by the overlap area between the heat field and the visual cone, quantifying the effective information coverage and accessibility of each screen in the current carriage state. The attention heat field and visibility probability results are then labeled back into the carriage screen model, forming a structured and resolvable display visibility model.

[0050] Preferably, step S4 specifically includes:

[0051] Acquire ground-based collaborative data;

[0052] A full-line operational status map is generated based on ground collaborative data, resulting in full-line operational status map data.

[0053] Based on the data from the overall line operation status map, cross-carriage passenger distribution trend processing is performed to obtain cross-carriage passenger distribution trend data.

[0054] Event synchronization is performed based on cross-carriage passenger distribution trend data and full-line operation status map data to obtain full-line event synchronization data;

[0055] By cross-domain display and linkage of passenger attention screen data based on the synchronous data of events across the entire line, subway passenger collaboration data is obtained to assist in the intelligent display of subway in-vehicle information.

[0056] In this invention, the system acquires ground-based collaborative data and generates a full-line operational status map based on the line structure and train operation status, forming a spatiotemporal status expression model for the entire network. Subsequently, the status map undergoes cross-carriage passenger distribution trend processing to identify passenger flow trends and spatial load distribution between different trains or carriages, achieving a precise depiction of passenger flow dynamics across the entire line. Combined with event information, spatiotemporal synchronization is performed to construct a unified full-line event synchronization model, ensuring that dispatching events are synchronously reached and responded to consistently across multiple trains. Based on the event synchronization results, the system drives cross-domain display linkage on in-train passenger attention screens, enabling content broadcasting, screen synchronization, or zoned adjustment at key transfer nodes, during large passenger flow warnings, or in abnormal operation scenarios. It outputs subway passenger collaborative data, significantly improving the global coordination, dynamic adaptability, and dispatching intervention efficiency of information display, effectively ensuring the consistency of information acquisition and guidance stability for passengers in different operating scenarios.

[0057] Preferably, the generation of the overall operational status map specifically involves:

[0058] The track alignment structure diagram is initialized based on ground collaborative data to obtain track alignment diagram data;

[0059] By injecting operational status attributes into the track alignment map data, attribute map data is obtained.

[0060] Schedule event-affected domain subgraph mapping to attribute graph data to obtain graph mapping data;

[0061] Graph situational analysis is performed on the graph mapping data to obtain situational map data;

[0062] Based on the situation map data, a time-series snapshot sequence is constructed to obtain the overall operational situation map data.

[0063] This invention initializes the track network structure diagram, modeling the relationships between stations and sections as a graph structure, providing a topological foundation for concurrent state representation and propagation calculations across multiple nodes. Through operational state attribute injection, dynamic attributes such as current train density, delay status, and capacity load for each line are mapped to graph nodes and edges, forming a high-dimensional attribute graph. The system performs subgraph mapping of scheduling event impact domains, allowing events such as construction skips and transfer congestion to be locally labeled in the track network as propagable subgraphs, accurately modeling the spatial impact boundaries of these events. Through graph situational awareness construction, structural information and attribute changes are fused to generate node / edge embeddings for situational awareness and intervention decisions. The constructed temporal snapshot sequence preserves the state evolution of the graph at continuous moments, providing continuous contextual support for train-ground coordinated scheduling and interactive display, significantly enhancing the system's perception completeness and response foresight in situations involving multiple trains, multiple events, and multiple regions.

[0064] Preferably, this application also provides a subway vehicle-mounted information intelligent display system for executing the subway vehicle-mounted information intelligent display method described above, the subway vehicle-mounted information intelligent display system comprising:

[0065] The passenger behavior perception and focus situation extraction module is used to acquire train operation status data, carriage environment perception data, and train event flow data; to analyze passenger behavior based on carriage environment perception data to obtain passenger behavior data; and to analyze focus situation based on passenger behavior data to obtain focus situation data.

[0066] The state-aware intelligent scheduling decision-making module is used to perform state-aware scheduling based on train operation status data, train event flow data, and focused situation data to obtain state-aware data.

[0067] The passenger attention perception-driven screen control module is used to drive the passenger attention screen based on the state perception data and obtain the passenger attention screen data.

[0068] The ground-based collaborative display optimization module is used to acquire ground-based collaborative data and, based on this data, to coordinate and link passenger attention screen data to obtain subway passenger collaborative data, thereby assisting in the intelligent display of subway onboard information.

[0069] The beneficial effects of this invention are as follows: By acquiring and analyzing data on train operation status, carriage environment perception, and event flow, it can comprehensively identify passenger behavior patterns and focus states within the carriage, achieving dynamic modeling of passenger attention areas and behavioral trends; combining train status and emergency event information for status-aware scheduling allows for precise control over information display content, priority, and presentation methods; driving carriage screen display strategies based on status perception results enables adaptive refresh of screen content, rhythm control, and visual parameter optimization, improving the efficiency and accuracy of passengers obtaining key information; acquiring ground-based collaborative data allows for the construction of a full-line operation status map and the realization of cross-carriage and cross-train synchronous linkage of displayed events, ensuring consistency in information transmission and stability of guidance during peak transfers, scheduling adjustments, or abnormal event scenarios. This invention enhances the timeliness, local adaptability, and global coordination of the subway information display system, providing passengers with more intelligent and reliable travel information services. Attached Figure Description

[0070] Other features, objects, and advantages of this application will become more apparent from the following detailed description of the non-limiting embodiments, taken with reference to the accompanying drawings:

[0071] Figure 1 A flowchart illustrating the steps of an embodiment of a method for intelligent display of subway vehicle information is shown.

[0072] Figure 2 A flowchart illustrating the steps of a passenger behavior analysis method according to one embodiment is shown.

[0073] Figure 3 A flowchart illustrating the steps of a state-aware intelligent scheduling decision-making method according to one embodiment is shown.

[0074] Figure 4 A flowchart illustrating the steps of a passenger attention perception-driven screen control method according to an embodiment is shown.

[0075] Figure 5 A flowchart illustrating the steps of a ground-based collaborative linkage demonstration optimization method according to an embodiment is shown. Detailed Implementation

[0076] The technical method of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0077] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. Functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0078] It should be understood that although the terms "first," "second," etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are used merely to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0079] Please see Figures 1 to 5 This application provides a method for intelligent display of subway vehicle information, including the following steps:

[0080] Step S1: Acquire train operation status data, carriage environment perception data, and train event flow data; perform passenger behavior analysis based on carriage environment perception data to obtain passenger behavior data; perform focused situation analysis based on passenger behavior data to obtain focused situation data.

[0081] In one embodiment, the system achieves perception of the current train operating environment and passenger behavior by fusing train operation status data, carriage environment perception data, and train event stream data. Train operation status data includes the current station number, train arrival timestamp, next station identifier, and current operating speed (in words). The system includes indicators such as spatial density distribution, spatial density distribution, and a flag indicating whether the train is currently in operation ("0" indicates operation, and "1" indicates a stop). The train's environmental perception data is collected from cameras, infrared sensors, or thermal imaging equipment deployed inside the train and mainly includes the following three categories of indicators: : Used to represent the passenger density at time t at spatial location (x, y) in the carriage; thermal trajectories for boarding and alighting. This indicates the thermal distribution of passenger entry and exit frequencies in a specific area; passenger dwell time statistics. : indicates that the number is The cumulative time passengers spend in each area; train event stream data includes information on construction-related station skips, temporary transfer announcements, and line-wide delay broadcasts, with the record format including event type, trigger time, and the geographical or line-wide affected area. The carriage space is divided into... There are 1 regular grid cells, each denoted as _____. Each cell is considered a node in a graph structure. Based on the passenger's movement trajectories and dwell transitions between grid cells, an edge weight function is constructed: the weight of adjacent edges is calculated by the spatial overlap ratio of movement frequency and dwell trajectory; the result is a graph structure. ,in Represents the set of all spatial unit nodes. This represents the set of movement-related edges between nodes. Each node has the following attributes, including passenger density, average dwell time, and number of boarding and alighting times within that cell. The constructed carriage behavior graph is clustered, node attributes are standardized, and clustering algorithms (such as K-Means or DBSCAN) are applied to classify the nodes. Nodes are labeled based on the clustering results (mapped based on cluster centers and a pre-defined label parameter library), for example, "dense dwelling area," "transfer gathering area," or "passage crossing area." If a node's location is a passenger travel path and there are high-attention areas (i.e., high density and high dwell time) in its neighborhood, the node is labeled as a "secondary attention area." The output is a structured passenger behavior feature dataset, recording the behavior category and spatial attributes of each node. Combining multiple passenger behavior features, regional attention is calculated, and focus situation data is generated, with each node receiving an attention score. Defined by the following weighting formula: ,in This is the weighting factor for the passenger density term, with a value of 0.5. This is the weighting factor for the average length of stay, with a value of 0.3. The weighting coefficient for the gaze direction concentration term is 0.2. The aforementioned weights are obtained based on empirical data fitting. For nodes The standardized value of passenger density, i.e., current passenger density / maximum passenger density; This is the standardized value of the average dwell time in the area, i.e., the average dwell time / maximum dwell time in the area; This is an indicator representing the concentration of passengers' gaze. Areas are categorized based on attention thresholds: areas with high scores and long dwell times are marked as "waiting area"; areas that are entrances / exits with high frequency are marked as "transfer area"; and areas where people significantly avoid the area are marked as "avoidance area". The output is focus trend data, including focus type, corresponding area number, focus weight value, and trend prediction direction.

[0082] Step S2: Perform state-aware scheduling based on train operation status data, train event flow data, and focused situation data to obtain state-aware data;

[0083] In one embodiment, train status features are extracted from the operating status: ,in For the length of time the train was delayed, This refers to the actual arrival / departure time. The timetable is designed to schedule train departures. Current operational phase identifiers are extracted and coded to represent different phase states: "0" indicates the train is in the middle of its journey; "1" indicates the train is approaching a station; "2" indicates the train is stopped; and "3" indicates the train is about to depart. Event flow data is parsed to identify event types (e.g., construction skips stations, transfer anomalies, overall delays) and categorized using event type codes. The scope of an event is mapped to the train's spatial structure; for example, if the event affects carriages C3 to C5, its corresponding physical location interval is recorded. By integrating train status characteristics and focused situational data, a two-dimensional train situational impact matrix is ​​constructed. To assess the scheduling response priority of each carriage area at the current moment: ,in Assign a priority score to the dispatch response of the carriage area under the current dispatching system. Let be the state fusion function, representing the th The carriage area and the first The response strength between focused situations, such as weighted summation, These are the weighting coefficients for the carriage state factors. For the first The state factors (such as delay, arrival, stop, etc.) corresponding to each carriage area are first encoded based on a preset state parameter table and then standardized. To focus on the weighting coefficients of the situation weights, For the first The focus intensity weights for each key situational region. From the situational matrix... Select the top K region pairs with the highest weight values, such as 5-10, as the priority display or broadcast target regions; generate a time window that can be used for display based on the current train time status and the arrival time of the next station: ,in, This is the current timestamp. The estimated time for the train to arrive at the next station. A buffer time (e.g., 10 seconds) is reserved to prevent sudden information changes from affecting passenger cognition. Based on the attributes of key response areas and the current operational status, a pre-set state-aware display strategy library is invoked to generate a specific display scheme. For example, if an area is identified as a "waiting focus area," a guidance route map is displayed first, and the readability of the text content is enhanced (e.g., larger font size). If an area is in a transfer focus area and there is a transfer anomaly, a scrolling animated graphic is used, accompanied by voice prompts. If there is avoidance of focus behavior in the current carriage (e.g., people are significantly moving away from a certain area), a safety prompt display can be triggered. The output state-aware data format includes fields such as target screen identifier, display content type, priority level, animation style identifier, and whether voice prompts are enabled.

[0084] Step S3: Drive passenger attention screen based on state perception data to obtain passenger attention screen data;

[0085] In one embodiment, the structural parameters of all information display screens within the carriage are obtained to construct a spatial model of the carriage screens. Specifically, this includes the three-dimensional spatial coordinates (x, y, z) of the screens within the carriage; the orientation vector of the screens, i.e., the normal direction perpendicular to the screen surface; and the viewing angle range of the screens, defined as the horizontal and vertical viewing angle boundaries of the viewing cone. These parameters are uniformly encoded into a spatial cone model, where the viewing area of ​​each screen is defined as a cone-shaped region with an included angle range of [missing information]. This indicates whether a passenger's line of sight enters the visible area of ​​the screen. The output is a carriage screen model data with spatial cone structure attributes. Combining passenger focus behavior data, the degree of projection overlap of each focus area within the visible cone of each screen is calculated to quantify the combination of screen visibility and passenger attention. Specifically, it iterates through all focus areas, determining whether their center point or main activity trajectory falls within the visible cone of a screen; if overlap is established, the attention intensity of that focus area is taken as the attention contribution of that screen; the passenger attention coverage probability of that screen is defined as: ,in For the first The probability value of passenger attention coverage per screen. Indicates the first One screen; Indicates the first A focused area; Indicates screen The set of visible cone coverage areas; It is an indicator function, indicating the first... Is the focal area on the screen? Within the visible range, it is 1 if it is there and 0 if it is not. Indicates the focus area The output is the attention intensity score. The output is the passenger attention coverage probability value for each screen. Based on the attention probability values ​​of each screen, combined with the aforementioned state-aware scheduling weights, all screens are sorted to determine the display update priority. For example, to calculate the display priority index: ,in This indicates the policy weight for this screen during the state-aware scheduling phase (e.g., transfer alarms take precedence over station guidance), where For the first The probability value of passenger attention coverage per screen. For the screen Policy scheduling weights during the state-aware scheduling phase; for all screens... The values ​​are sorted from high to low; for screens with high priority, a high-frequency update display strategy is formulated, which includes setting the refresh rate to fast mode, such as updating once every 3-5 seconds; prioritizing graphical full-screen display or scrolling broadcast mode for display format; synchronously setting content category and voice control parameters (if enabled); outputting passenger attention screen data, including fields such as the display content category, update frequency, display mode (full-screen, scrolling, icon), priority label, and voice control switch for each screen.

[0086] Step S4: Obtain ground coordination data and coordinate with passenger attention screen data based on the ground coordination data to obtain subway passenger coordination data for assisting in the intelligent display of subway information.

[0087] In one embodiment, ground-based collaborative data is acquired, including operational status maps generated by the dispatch center, regional abnormal event synchronization plans, and transfer pressure prediction matrices, in the format of {station identifier, event type, event severity level (or priority), target duration}. Based on the aforementioned system, a full-line operational status map is generated, based on the track network topology G=(V,E), where node V is a station and edge E is an interval. The system injects status into each node: attributes such as passenger density, delay, and construction impact; and injects delay propagation probability into each edge. Distribution trend processing and event synchronization are performed by using a prediction model (such as temporal LSTM) to generate cross-carriage trends: which carriage will experience transfer passenger flow; matching the trends to the current train's focus area to generate a synchronization control table; and generating linkage control. If a transfer congestion trend is detected across the entire line, multi-screen synchronous broadcasting is triggered; if the current carriage is a dispatch focus area, the screen highlights the warning information issued from the ground. Output: subway passenger collaborative data (including the screen IDs to be linked, display content, display strategies, and synchronization windows).

[0088] Preferably, step S1 specifically includes:

[0089] Step S11: Acquire train operation status data, carriage environment perception data, and train event stream data;

[0090] In one embodiment, train operation status data is acquired through an interface to the train control system (such as a Train Control and Monitoring System (TCMS)). The system collects the following fields at fixed time intervals: the train's unique identifier, the current station number and platform number, and the train's actual arrival and departure times (denoted as follows). The system collects data including the current operating speed v (in km / h), current acceleration / deceleration status, and operating phase status codes (e.g., running, entering a station, stopping, etc.). These operating phase status codes are used to determine the dynamic operation of the train between stations. The carriage environment perception data is acquired through multimodal edge sensing devices deployed inside the train, including edge computing cameras, thermal infrared imaging arrays, infrared counters, and ultra-wideband (UWB) array nodes. The perception system collects the following spatial data indicators: real-time crowd density distribution. , indicating at time Below, the two-dimensional spatial position of the carriage Passenger density; hotspot trajectory point set for boarding and alighting, used to describe the thermal trajectory area formed by passengers during boarding or alighting, which can be further analyzed to determine the distribution of hotspot boarding / alighting areas; dwell time cluster distribution. , representing the dwell time of the i-th passenger in a specific area, used to calculate area congestion and activity frequency; local channel flow direction vector field Describing at time Below, various locations inside the carriage Information on passenger flow direction and intensity. Train event stream data is provided by the ground dispatching system or train event management middleware. The system receives event data through message queues (such as the MQ protocol) or REST-based asynchronous communication interfaces. Each event message includes at least the following fields: Event type: for example, skip-station scheduling, temporary transfer notice, construction closure notice, specific carriage disabling, etc.; Event effective period: indicating the time range of the event occurrence and failure; Impact range identifier: including the affected station number or carriage section number; Event priority label: used for priority ranking of event handling and response strategies.

[0091] Step S12: Construct a spatial behavior map based on the carriage environment perception data to obtain carriage behavior map data;

[0092] In one embodiment, each carriage is divided into Equal grid cells (e.g., 50cm per cell) 50cm); each grid Corresponding to a graph node Record the location index and attributes. If two adjacent cells are connected by personnel movement trajectories, then establish an undirected edge. The edge weight is defined as the frequency of passenger movement per unit time. Or, the superposition value of thermal trajectory density. Graph structure definition: Each node contains the current population density; average dwell time; whether it belongs to the entrance / exit area (marked by the vehicle door); each edge can contain directional probability, supporting the construction of directed graphs for flow analysis.

[0093] Step S13: Extract behavioral features from the carriage behavior map data to obtain behavioral feature data;

[0094] In one embodiment, the system extracts multi-dimensional behavioral features from each node and its adjacency relationships in the constructed carriage behavior graph. The extraction dimensions include the following four categories, including dwell intensity features, for each graph node... (That is, the average passenger dwell time of a certain grid unit in the corresponding carriage) is calculated and denoted as . The system includes several key features: **Alighting and boarding activity characteristics:** Grid cells within a certain spatial range from the carriage door are selected, and the number of passenger trajectories originating from or arriving at each node per unit time is counted. **Migration kinetic characteristics:** Based on the edge weights in the carriage behavior graph (e.g., passenger movement frequency per unit time), the movement intensity value associated with each node is extracted, representing the degree of passenger mobility or behavioral fluctuation within the area. **Path heat sequence characteristics:** By analyzing the main paths passengers take from the boarding entrance area to the center of the carriage or the exit area, the cumulative heat intensity along the route (estimated by movement frequency or trajectory density) is counted to identify the main passageways and their heat distribution, thereby inferring structured population migration trends.

[0095] Step S14: Perform behavioral generalization enhancement on the behavioral feature data to obtain passenger behavior data.

[0096] In one embodiment, for nodes located at the edge of the carriage or in sparsely populated areas, whose original behavioral features are not significant, the system enhances their features using neighborhood information. Specifically, if a node's neighboring nodes (e.g., 4 or 8 neighbors in the top, bottom, left, and right directions) exhibit high behavioral focus (e.g., high density), then the node's features are enhanced. Taking passenger density as an example, the enhanced density value... Expressed as ,in It is a compensation factor (set by experience or obtained through training). Represents a node The set of adjacent nodes, This represents the average density of adjacent nodes. The system analyzes behavioral graph sequences over several consecutive time periods. A time series graph set is constructed using five frames as the time series: The system performs time-weighted smoothing on the behavioral state of each node in the aforementioned behavioral graph sequence, identifying regions exhibiting consistent behavior across multiple consecutive time points (e.g., continuous high density, stable dwell time). Based on rule templates or lightweight classification models (e.g., rule trees or shallow decision trees), the system categorizes the features of each behavioral region, transforming multi-dimensional behavioral features into unified behavioral semantic labels. The categorization rules are as follows: waiting behavior: long dwell time and high-density clustering; transfer focus behavior: short dwell time and frequent boarding and alighting; avoidance behavior: rapid movement and low density. Each identified region is accompanied by structured attribute information, specifically including a behavioral type label (e.g., "transfer_focus"); a feature confidence score (e.g., the degree of matching between behavioral features and rules, ranging from 0 to 1); and a behavioral evolution trend (e.g., "increasing", "stable", "decreasing"), derived through time series analysis. The output format is a set of structured passenger behavior data items, each describing a passenger behavior region with a semantic label. For example, the area with the area code "Z07" is identified as a "transfer concentration area" with a behavior recognition confidence level of 0.91, and shows an increasing trend in activity.

[0097] Preferably, the focus of situational analysis specifically includes:

[0098] Regional attention data is obtained by calculating regional attention based on passenger behavior data;

[0099] In one embodiment, the entire carriage space is divided into several discrete region units based on a carriage space grid. Each region is [size missing] Alternatively, it can be divided into equal sections based on the layout of the carriages. The level of attention given to each area... It can be calculated using the following formula: ,in For the first The attention score for each region The weighting factor for the population density term. The average population density per unit time. The weighting factor for the length of stay. For the first Average length of stay in each region The weighting coefficients for the frequency of action events. For the first The frequency of action events per unit time (such as looking up at the screen, walking, standing still); the system calculates the frequency of action events per unit time based on the above calculation model for each region unit. The level of attention is evaluated, and the output includes the region number, attention score, and relevant intermediate variables used for attention evaluation.

[0100] The focus status data of the region is classified into focus status data.

[0101] In one embodiment, a fixed threshold is used to segment and judge the attention value. The specific classification criteria are as follows: when the region unit Attention value When the area is in "strong focus" state, mark it as such; when... When the area is in "center focus" state, mark it as such; when... When this happens, the area is marked as "unfocused".

[0102] In one embodiment, the attention value of all regional units is... The system inputs a feature vector; executes a clustering algorithm to automatically divide the region into several clusters with similar levels of attention; each cluster corresponds to a potential focused region, and its focus level can be deduced from the average attention level of the cluster centers. For each region determined to be in a focused state, the system needs to add records such as: spatial coverage (i.e., a set of several region unit numbers, representing the spatial distribution of the focused region within the carriage); historical focus duration (in seconds, used to measure the length of time the focused region has been continuously focused on); the associated in-vehicle display screen number (indicating the passenger information display device that the focused region is mainly facing); and the average gaze angle between the display screen and the crowd (which can be calculated by the eye tracking or camera posture estimation module to evaluate visibility and screen attractiveness).

[0103] By performing local congestion evolution on the focused state data, local congestion data is obtained;

[0104] In one embodiment, the system for each region unit Calculate its congestion value The formula is as follows: ,in For the region The congestion value, This is the weighting factor for the population density term, with a value of 0.5. The population density term, i.e., the region at the current time. Average population density within the area / preset maximum population density This is the weighting factor for the inverse proportional term of movement speed, with a value of 0.3. For movement speed, i.e., minimum movement speed / area Average speed of passengers inside the vehicle (unit: meters per second). This is the weighting factor for the input-output imbalance term, with a value of 0.2. This refers to the imbalance between inbound and outbound traffic within a region; the system measures regional congestion values. Evolutionary trend identification is performed based on historical continuous time windows (length is...). Within a time window, the following two indicators are calculated for the congestion value sequence: a volatility indicator, i.e., the standard deviation of the sliding window, used to determine the magnitude of changes in congestion levels; and a trend indicator, i.e., the mean of the derivatives of the sliding window, used to determine whether the congestion level is continuously increasing, stabilizing, or decreasing. When the congestion value shows an increasing trend at multiple consecutive time points, and the change exceeds a preset threshold, the system marks the area as "evolving congestion". To identify contiguous congested areas composed of multiple adjacent areas, the system performs spatial diffusion analysis, including the following steps: calculating the spatial adjacency relationships and corresponding edge weights between all regional units, where the edge weights can be represented as the probability of passage between regions or the intensity of pedestrian flow coupling; if a pair of adjacent regions... Congestion value If both exceed the set congestion threshold and their edge weights are higher than the preset adjacency threshold, the system merges them into a single connected congested area. This merging operation is repeated to form a cluster of regions with spatial connectivity and consistent congestion characteristics, which is output as a local congestion area. The local congestion data output by the system includes the following structured fields: congestion area number; congestion status level (e.g., rising, stable, easing); average population density; inflow and outflow frequency; and average movement speed.

[0105] Focus situation data is generated based on local congestion data and focus status data.

[0106] In one embodiment, by focusing on the situational scoring function Perform the calculation: ,in To score based on attention, This is the weighting coefficient for the attention score. To score the congestion evolution trend, The weighting coefficient for the congestion trend score. The historical duration factor (with higher credibility assigned to long-term focus areas) is calculated by dividing the historical maximum duration factor by the historical average duration factor. This is a weighting coefficient for the historical duration factor. The concentration of crowd focus is assessed based on the entropy index (higher concentration indicates a stronger trend). Each focus area is overlaid using a train carriage map as the base map. It generates two-dimensional heat maps or region label maps; it supports priority sorting of multiple regions in the same carriage for screen strategy matching.

[0107] Preferably, step S2 specifically includes:

[0108] Step S21: Extract train status features based on train operation status data and train event stream data to obtain train status feature data;

[0109] In one embodiment, the data originates from a train control system (such as TCMS) or train dispatching system, and specific fields include the current station number (for location); current operating speed (unit: km / h); train direction (e.g., up / down); relative distance between trains ahead (for assessing congestion); and planned arrival or entry time (for deviation evaluation). Train event stream data is provided by the dispatching center or event management system and accessed asynchronously, including data such as temporary transfer plans (e.g., changes in transfer stations); skip-station dispatching strategies (e.g., skipping some stations to save time); construction, temporary stops, or station closure events (including duration and affected area); and train fault records or delay status (e.g., power failure, system stagnation). Based on the above data, the system constructs a train state feature vector, denoted as: The meanings of each state dimension are as follows: The time offset represents the difference between the current actual arrival time and the scheduled arrival time (in seconds). The speed deviation value is the difference between the current operating speed and the historical average speed of the train over the entire journey, reflecting the acceleration or deceleration status. The indicator for the passability of the section ahead is a Boolean variable (0 indicates blockage; 1 indicates passability), and is determined based on the threshold of the distance between trains ahead. This is an event intensity coding value used to quantify the severity of the event currently affecting the train. For example, construction impact is set to 0.8, sudden accident is set to 1.0, and temporary stop is set to 0.6. The transfer pressure index is a normalized assessment based on the estimated number of transfer passengers at the transfer stations that the current train is passing through or about to arrive at (the value ranges from 0 to 1).

[0110] Step S22: Map the train situation influence matrix based on the train status characteristic data and the focused situation data to obtain the train situation matrix data;

[0111] In one embodiment, train status characteristic data includes operational indicators extracted in the preceding steps, such as train delay time, speed deviation, forward traffic status, event level, and transfer pressure. Focused situational data includes the results of carriage behavior map analysis, including high-density passenger areas, areas with concentrated dwell time, and areas with active entry and exit. Each focused area includes the following parameters: area location index; real-time passenger density; dwell behavior focus degree; passenger attention level or abnormal warning level. The system constructs a two-dimensional influence matrix. Each matrix element represents a dimension of a train's state features. Focus on the situation area The intensity of the influence. Specifically, it is expressed as: ,in Indicates the first Individual train state characteristic components (such as delay duration, speed deviation, etc.); Indicates the first Behavioral characteristics of passenger focus areas (such as density or dwell time); The influence function is specifically a linear function, such as a weighted summation; nonlinear relationships are learned through deep learning training; and state mapping is performed through logical rules or conditional judgments. If the train's current station is less than one station away from a station in a high-focus area for a certain passenger, and the train delay exceeds 60 seconds, then the train is marked as a "high interference source" for that area, and the corresponding matrix element is assigned a high value. If the train event type is "construction" or "skipping a station," the system maps the event's impact to its influence range, applies a disturbance factor to the relevant focus area, and represents it as a situational interruption or behavioral disturbance. The train situational influence matrix output by the system is a two-dimensional array, where the row dimension represents train state factors (such as delays, accidents, transfer pressure, etc.); the column dimension represents the focus area within the carriage or platform; and the value of each element is the corresponding influence intensity (the value range can be 0~1).

[0112] Step S23: Identify key response regions from the train situation matrix data to obtain key response region data;

[0113] In one embodiment, the system sets an influence threshold. (e.g., 0.8) is used to filter out focused regions that are significantly disturbed or affected. Specifically, the system accumulates the influence score under the corresponding state factor dimension for each focused region j. If the accumulated value exceeds the set threshold If so, the area is identified as a critical response area. This is denoted as: ,in Represents train state factor For the region Its influence; Key response screening thresholds (set empirically or adjusted dynamically); This is the set of focused region numbers identified as critical responses. For each candidate region, the higher the frequency with which it was triggered as a critical response in the past time period and the lower the response delay, the more weight is applied to its impact score at the current moment to reflect its spatiotemporal inertia and response sensitivity. The system limits critical region identification to the following areas: the carriage where the current train is located; the carriage directly adjacent to the current carriage; and excluding structural dead zones within the carriage (such as camera blind spots) and platform end areas (due to their unfavorable spatial structure for intervention). By setting these spatial constraints, the system avoids including difficult-to-intervene or inefficient sections in the response domain, thereby concentrating resources on controllable areas.

[0114] Step S24: Generate a scheduling window based on the key response area data to obtain scheduling window data;

[0115] In one embodiment, for each passenger focus area identified as a critical response, the system generates a corresponding dispatch window, denoted as: ,in The window start time is defined as 15 seconds before the train is expected to enter the critical response zone. The window termination time is defined as 5 seconds before the train is expected to leave the area. This time interval is used to plan the optimal timing for displaying relevant information or providing voice reminders, ensuring that passengers receive guidance or warnings in a timely manner before interacting with the area. If the scheduling windows of multiple critical response areas overlap in time, the system will execute the following merging or optimization strategy: if multiple overlapping windows belong to adjacent areas, the system can merge the time range to form a display window with a larger coverage; if window merging is not feasible, the window corresponding to the area with the higher impact score will be retained first (i.e., the one with stronger focus and higher interference risk will be prioritized). Each scheduling window must be accompanied by a content type recommendation tag to indicate the main type of information to be displayed within that time window. The recommendation tag can be automatically generated based on the dominant state factor that triggers the critical response. These include "Construction Avoidance Reminder": corresponding to areas dominated by construction event interference; "Transfer Guidance": corresponding to areas with high transfer pressure; "Train Front Congestion Warning": corresponding to areas with high density or delays in front of the train; "Station Entry Calm Reminder": corresponding to areas with evacuation needs.

[0116] Step S25: Generate state-aware scheduling based on scheduling window data and preset state-aware strategy library to obtain state-aware data.

[0117] In one embodiment, the system dynamically generates dispatch control instructions for specific carriage situations based on pre-generated dispatch window data and a preset state-aware strategy library. These instructions drive the train's onboard screens or broadcasting equipment to switch displayed content, adjust layouts, and control playback frequency, thereby achieving passenger guidance, behavioral intervention, or risk alerts. The preset state-aware strategy library is a set of rule-based strategies. Each strategy includes trigger conditions, such as event type (e.g., skipping stations, construction, transfer), focus area characteristics (e.g., high density, concentrated in a specific screen area), display template (specifying the text or animation display format to be switched when the trigger conditions are met), animation playback frequency (e.g., high frequency (rotating every 5 seconds), medium frequency, or low frequency (refreshing every 30 seconds), visual cues (e.g., red highlighting indicates dangerous areas, blue guidance indicates transfer directions), and screen layout and priority strategies (determining display priority among multiple screens or the position of content displayed on a single screen, such as left-hand or upper right corner) to avoid overlap with passenger gathering areas. The strategies include Strategy A: When the train event type is "skip station", the content is switched to "skip station explanation diagram + voice warning template"; Strategy B: If the passenger's focus area is located in the upper right corner of the screen, the displayed content will be shifted to the right side of the screen to avoid visual obstruction.

[0118] Preferably, the train situation influence matrix mapping is specifically as follows:

[0119] Initialize the track network topology diagram by analyzing the train operation status data and train event flow data to obtain the track network diagram data;

[0120] In one embodiment, the system abstractly models the subway line as a graph structure G=(V,E) to describe the spatial topology and operational characteristics of the rail network. The set of nodes... Represents each subway station; edge set This indicates the accessible track sections between stations, i.e., the stations. and There are direct operational connections between them. The system further assigns the following attributes to the above nodes and edges, including node attributes such as: Line Number: identifies the line code to which the station belongs, used to support multi-line network fusion modeling; Station Level Label: used to distinguish the network role level of the station, such as "Ordinary Station", "Transfer Station", "Transportation Hub", etc.; Transfer Capability: quantifies the transfer capability of the station, such as the number of lines it supports for transfers, the number of transfer channels, etc. Edge attributes such as: Segment Length: records the physical distance between tracks between stations; Historical Throughtime Statistics: represents the average running time, standard deviation, and other statistical indicators of trains on this segment; Whether Temporary Station Skipping is Supported: a Boolean flag used to indicate whether the segment has the capability to skip stations as needed for scheduling. System Anomaly Flag Field , used to represent edges Is it currently in an invalid state? This indicates that the track section is in a normal, passable condition. This indicates that the section is temporarily closed or its operation is restricted due to construction, emergencies, malfunctions, or other reasons. Through the above topology modeling process, the system has completed the initialization of the graph structure of the track network.

[0121] Based on the train status characteristic data, the track network diagram data is mapped to the running trajectory to obtain the running trajectory diagram data;

[0122] In one embodiment, the system searches for a feasible path of 3 to 5 stations in the constructed track network graph structure, based on the node where the current train is located. This path represents the train's expected operating range within the next time window and is denoted as the path sequence. ,in For the current site node, The system identifies the target station nodes. For each track segment (i.e., the edge in the graph) in the above path, the system injects corresponding train scheduling prediction information, specifically including the expected entry time into the segment; the path stability indicators of the segment (such as historical success rate and event impact frequency); and whether there are passenger focus areas around the segment and their spatial distance. The path with the injected scheduling attributes is mapped onto the track network graph as a "dynamic running trajectory," and each node and edge is assigned a time label (e.g., entry time, expected dwell time, etc.), thus forming running trajectory graph data containing time-series scheduling information.

[0123] The trajectory graph data is weighted using a tensor structure to obtain weighted graph data;

[0124] In one embodiment, for each track segment edge in the trajectory diagram (i.e., the path segment connecting any two station nodes in the diagram), the system constructs a three-dimensional tensor-based edge weight structure. This tensor uses "start node - end node - weight dimension" as coordinates to form a set of multi-dimensional indicators, used to comprehensively represent the actual operational characteristics and environmental influencing factors of the path segment. In this embodiment, the third dimension of the tensor (i.e., the weight dimension) includes, but is not limited to, the following four core weight factors, including a time delay factor, which represents the operational delay of the current path segment and is calculated as the deviation between the actual running time and the planned running time. ,in For the weight of the time delay factor, This refers to the actual running time of the track section. The planned operating time of the track segment; the passenger density disturbance factor, which characterizes the degree of passenger gathering or congestion level in the area connected by the route segment, can be obtained through statistical values ​​from area density sensors. , For passenger density interference factor weights, The average passenger density value within the area connected by the route segment; the transfer interaction intensity factor, which characterizes the frequency of transfer behavior and the activity level of transfer nodes within the time window of the route segment. ,in For the weight of the transfer interaction intensity factor, To represent the transfer frequency of path segment (i,j) within a time window, it is calculated based on historical data or current data. The accident impact factor indicates whether the path segment is currently under the influence of construction, accidents, or special events; it can be a binary state (affected / not affected) or a continuous value model using historical accident frequencies. The tensor weights of all edges are summarized to form a weighted trajectory graph structure with three elements: a node set, an edge set, and weight tensors. The output graph structure is represented as a triple, containing a node set, an edge set, and the tensor weight set corresponding to each edge.

[0125] Spatiotemporal coupling is performed on the weighted graph data to obtain coupled graph data;

[0126] In one embodiment, based on track segments (i.e., edges in the diagram) as basic units, the operating state of each track edge is progressively updated within each time slice at set time intervals (e.g., 1 minute per time slice), forming a sequence structure of path segment evolution over time. For the multidimensional tensor weights carried by each edge, the system pre-defines a set of corresponding time-varying functions. These functions are used to model the dynamic changes of different influencing factors on the time axis. For example, for accident-related factors, an event decay function is used to represent the decrease in their influence intensity over time; for passenger density factors, a peak response function is used to represent their rapid increase during peak boarding and alighting times; for transfer interaction factors, a periodic function or short-term disturbance function is used to reflect their changing rhythm; and time delay factors can be combined with historical regression trends to generate smooth change curves. The state of each edge at each time t is represented as: ,in Let be the time-varying function of the k-th type of feature (e.g., event decay function, peak function); this state characterizes the operational efficiency, stability, and potential risk level of the path segment at the current moment. The state sequences of each edge in different time slices collectively constitute an evolutionary graph with a time dimension. The state scoring structures of all path segments in different time slices are uniformly merged to construct a coupled graph structure with node sets, edge sets, and temporal state sequences. This coupled graph not only preserves the spatial connectivity of the track network but also explicitly describes the dynamic evolution trend of the state of each path segment in the network over time.

[0127] Propagation link impact calculations are performed on the coupling graph data to obtain propagation graph data;

[0128] In one embodiment, the event impact intensity of each site node at the initial moment is defined. This propagation intensity function represents the degree of interference caused by a specific event to the site node at a given point in time. The initial impact intensity can be derived from external input events (such as construction closures or accident reports), and its value is set according to the event level, timeliness, and scope of impact. The system performs event impact modeling. ,in For nodes In time The intensity of the propagation effect at any given moment. For the currently affected target nodes, This represents the current propagation time step (discrete time). For the propagation factor from the weighted edge, For nodes The adjacent nodes, For nodes In time The system simulates the dynamic diffusion of influence between nodes using graph convolutional propagation mechanisms or Bayesian network propagation methods. Within each iteration time step, the propagation intensity of a node is jointly determined by the influence of its neighboring nodes. For any target node, its propagation intensity value at the next time step is obtained by the weighted accumulation of the propagation intensity of its neighboring nodes at the current time step; the weighting factor originates from the propagation factor in the edge weights of the coupled graph, calculated based on the density influence of adjacent edges, spatiotemporal disturbance factors, etc.; the adjacency relationship is based on the boundary definition of the graph in the track network topology, limiting the propagation path. Throughout the iteration process, the system progresses step by step according to the time step, updating the propagation intensity values ​​of all nodes in each iteration until the propagation intensity converges or the propagation range reaches a set threshold. For example, in the initial stage of an accident, only the accident site has a high influence value; as propagation progresses, adjacent path segments and sites affected by related transfers gradually experience varying degrees of diffusion interference; this process can identify the event propagation chain and its evolution rhythm. The system encodes the propagation impact results onto each node of the track graph, constructing an event propagation graph structure. The propagation graph retains the node and edge set structure of the original graph and adds the influence strength value at the current propagation time to each node.

[0129] Event-driven structural mutations are performed on the propagation graph data to obtain train situation matrix data.

[0130] In one embodiment, the system sets a propagation intensity threshold parameter, representing the critical condition for the event's impact to reach a significant level. For each node (i.e., a track station) in the propagation graph, the system monitors its propagation intensity value in real time. If the propagation intensity does not belong to [0,1], it is first normalized. If the propagation intensity of a node exceeds the set threshold at a certain moment, the node is marked as a high-risk node and considered a key area with significant event impact. The system sets a propagation intensity threshold θ (e.g., θ=0.8∈[0,1]), which represents the critical condition for the propagation impact to reach a significant level. This threshold can be set based on the 95th percentile of historical event propagation data or dynamically set through manual optimization. After identifying multiple high-risk nodes, the system checks whether these nodes form a continuous path in the propagation graph. The system determines whether there is a direct or two-hop-less edge connection between high-risk nodes. If several consecutive path segments are in a high propagation intensity state (e.g., the propagation intensity of three consecutive edges all exceeds θ), they are aggregated into a high-risk propagation path. Path aggregation allows one node to be interrupted (e.g., a non-high-risk node), but the overall span does not exceed five nodes. The system categorizes and analyzes event types in the current propagation graph, extracting key situational factors, including but not limited to construction impacts, operational delays, transfer pressure, and passenger congestion density. For each high-situation node or propagation path, the quantitative impact intensity of the aforementioned situational factors is calculated and coded according to type. The system generates a two-dimensional train situational impact matrix. The row dimension of this matrix represents different situational factor categories, such as construction factors, delay factors, and transfer factors; the column dimension corresponds to the node or link index identified in the propagation graph. Each element in the matrix represents the impact intensity of the corresponding factor at that location within that region or link.

[0131] Preferably, step S3 specifically includes:

[0132] Step S31: Obtain the screen data of the carriage;

[0133] In one embodiment, the system collects the following basic information for each carriage screen: a screen identification code to uniquely identify each screen, typically including the carriage number and screen serial number; installation location coordinates, using a three-dimensional coordinate system (X, Y, Z) to represent the screen's physical installation location within the carriage space, in meters; installation orientation angles, using three rotation angles to represent the screen's pitch, yaw, and roll angles, in degrees (°), to describe the screen's orientation; and screen dimensions, recording the length and width of the screen's display surface (in meters). To support content rendering and driver instruction compatibility for multiple screen types, the system also needs to record a "type label" for each screen to distinguish screen types. For example, different types of devices such as LED strip screens, LCD advertising screens, and dynamic interactive touch screens can be labeled with corresponding category labels.

[0134] Step S32: Perform spatial modeling on the carriage screen data to obtain the carriage screen model;

[0135] In one embodiment, a three-dimensional coordinate system is used to construct the carriage space model. Each screen is embedded as a directional field of view cone (visual cone): the center point of the screen is the apex of the cone; the viewing angle range (set to 30° horizontally and 20° vertically) determines the cone's angle; the viewing radius is set according to the width of the carriage (e.g., 3 meters); based on the above parameters, the system constructs the visible area of ​​each screen. Its definition is in the spatial model The set of points that satisfy the following conditions: the points are located within the coverage area of ​​the screen cone; the distance from the point to the screen center does not exceed the set visible radius; and the orientation of the points satisfies the screen's installation direction constraints. This can be formally expressed as: The system needs to synchronously mark the area that each screen "faces," that is, the main direction it faces towards the crowd activity area or structural orientation (such as the boarding door area, the alighting door area, the center of the carriage, etc.). This orientation labeling information will serve as a key reference dimension for "matching the focusing posture with the screen position." Through this spatial modeling process, the system completes the visual space mapping of the screens inside the carriage.

[0136] Step S33: Calculate the attention probability field based on the carriage screen model and focusing situation data to obtain the attention probability field data;

[0137] In one embodiment, the system for each screen Calculate their attention score This is used to measure the degree to which a device can currently attract passenger attention. The calculation is based on an integral weighted average of the focusing behavior within the visible area of ​​the screen, as shown in the following formula: ,in : No. Attention score for each block of screen; : No. The visible cone area of ​​the block screen; Passenger attention at corresponding spatial points; The viewpoint attenuation function is used to weighted suppress the degree of deviation from the center of the screen's line of sight. Its expression is a cosine function judgment formula; points with a cosine value greater than 0 are retained, while those less than 0 are considered 0. This function exhibits a monotonically decreasing characteristic. For example, it can be modeled using a cosine angle function or a Gaussian attenuation form, making the weight of attention points located at the edge of the cone lower than that of attention points directly facing the line of sight. : Spatial voxel integral element. The system performs the above calculation on all screens, and the output is an attention probability field dataset, defined as: ,in Indicates the first The probability or priority of attention for a block of screen at the current moment. This is the screen sequence number, with values ​​1, 2, 3...n.

[0138] Step S34: Generate passenger attention screen data by multi-frequency drive based on attention probability field data and state perception data.

[0139] In one embodiment, the system uses attention scores... Screen hierarchical management is implemented based on set upper and lower thresholds. The specific driving strategy is as follows: if the attention score of a certain screen... The system marks these screens as high-attention screens. These screens are prioritized for pushing important or urgent information, such as emergency transfer notices, accident response updates, or breaking news broadcasts, ensuring timely delivery of information to the target audience; if... Screens with low visibility are classified as low-attention screens. These screens are only used to display background information, such as the current station name, energy-saving reminders, or static advertisements, to avoid wasting resources. For screens falling between the high and low thresholds, a periodic carousel strategy or a switching of prompt content can be implemented, flexibly configured according to actual operational needs. Based on the above classification results, the system assigns a corresponding content refresh rate to each screen. For example, high-attention screens have a refresh cycle of 1 second and are suitable for pushing frequently updated dynamic content; medium-attention screens have a refresh cycle of 5 seconds and are suitable for lightweight notifications or reminders; low-attention screens have a refresh cycle of 10 seconds or more and only maintain basic information. The system has a multi-frequency content generation module that generates content based on the refresh frequency. Together with the state event set E, it calls the backend content scheduling service to generate specific information update packages and generate the drive sequence to control each screen. Its structure includes displaying event information ( ): Corresponds to specific message content in the event set; animation style attribute ( ): Defines the dynamic color scheme for information presentation; layout parameters ( This is used to control the information display structure, such as horizontal scrolling, partitioned display, and side-by-side display of text and images. Through the above mechanism, the system can achieve dynamic screen driving for different attention levels.

[0140] Preferably, the spatial modeling specifically includes:

[0141] Based on the data from the carriage screens, a carriage screen space model is obtained by creating a carriage screen space model.

[0142] In one embodiment, a three-dimensional Cartesian coordinate system is established with the carriage as the reference object. This coordinate system has its origin at the lower left corner of the carriage floor, and three axes represent the horizontal direction (X-axis), vertical direction (Y-axis), and carriage depth direction (Z-axis). This coordinate system serves as a unified geometric reference basis, used to describe the position and orientation of all screens. For each display screen installed in the carriage, its core spatial attributes are extracted, including spatial position coordinates, representing the three-dimensional absolute position of the screen's center point in the carriage coordinate system; installation orientation information, defined in Euler angles, including pitch, yaw, and roll angles, used to represent the orientation vector of the screen surface; and display size parameters, including the screen's width and height. The system treats each screen as a rigid body object and embeds it into the carriage coordinate space according to its spatial position and orientation parameters. All screen models are consistent with the actual physical structure of the carriage in three-dimensional space, ensuring error-free expression of orientation, position, and scale. Based on the spatial parameters of all screens, a spatial model of the train carriage screens is constructed, identifying the precise 3D arrangement of each screen within the carriage. This model serves as the foundational structural data for visual cone construction, passenger attention distribution mapping, and screen content-driven strategies.

[0143] A visual cone model is constructed from the screen space model of the carriage to obtain the screen visual cone model;

[0144] In one embodiment, a cone-shaped field-of-view model is constructed for each display screen in the carriage. This cone has its center as the vertex, and its axis is aligned with the screen's mounting orientation, i.e., the direction vector determined by the screen's pitch, yaw, and roll angles. The cone's angle is set by the viewing angle, with ±30 degrees horizontally and ±20 degrees vertically, simulating the naturally acceptable viewing range of the human eye. A maximum viewing radius of the cone is set. This radius is adjusted based on the carriage's length and the screen's resolution performance; for example, for a clear viewing distance of 3.5 meters or less, the maximum viewing radius is set to 3.5 meters. This parameter is used to shield the screen from the possibility of distant areas drawing attention. The system uses a spatial geometric ray tracing method to emit multiple line-of-view rays from the cone's apex along its direction, simulating the possible paths of the passenger's gaze in three-dimensional space. The ray trajectories are cross-calculated with the carriage's spatial structure to generate a corresponding three-dimensional cone-shaped polygon, accurately representing the screen's visible coverage area. The cone can be approximated as a combination of multiple subdivided triangular pyramids, forming a closed field-of-view volume. Each constructed cone structure is encoded to generate a screen-visible cone model data structure, including screen identifiers, cone vertex coordinates, angular parameters, and the set of boundary points of the visible range. This model will serve as the spatial constraint basis for constructing the passenger attention heatfield and calculating the actual visibility probability.

[0145] Crowd attention heatfield is constructed on the screen visual cone model to obtain crowd attention heatfield data;

[0146] In one embodiment, current passenger distribution data within the carriage is collected. Each passenger includes their location coordinates; facing direction (head orientation); dwell time; and a three-dimensional thermal field function is constructed. The calorific value is calculated as follows: ,in For three-dimensional space points The intensity of the crowd's attention heat field at that location assign passenger index numbers, This represents the total number of passengers currently in the carriage. The passenger's attention coefficient (e.g., those who stay longer and those facing the screen have higher weights) is used to measure attention. It is a natural exponential function. Let be any three-dimensional spatial coordinate point in the thermal field. For the first The location coordinates of each passenger To control the degree of diffusion, a value of 0.5 was used, so that passengers' attention was mainly concentrated within a radius of 0.5 meters around them, which is suitable for passengers to be stationary or densely packed; the result is a three-dimensional heat map of attention distribution in the entire carriage.

[0147] The actual screen visibility probability is calculated from the crowd attention heat field data to obtain the actual screen visibility probability data.

[0148] In one embodiment, for each screen visible cone Inner region, integral thermal field intensity calculation: ,in For the first The weighted thermal field integral value (visual intensity score) of the screen. Let be the three-dimensional coordinates of any voxel point in space. For the first The visible cone space area of ​​the block screen. To warm up the crowd's attention at a spatial point The intensity value, Let be the integral volume element, representing the volume of a spatial infinitesimal element. A viewpoint fit weighting function is used (the smaller the angle with the screen centerline, the greater the weight); the actual viewability probability vector is obtained by standardizing the calculation for all screens. Output the visibility score for each screen.

[0149] The actual visibility probability of the screen and the crowd attention heat field data are used to annotate the spatial model of the carriage screen to obtain the carriage screen model.

[0150] In one embodiment, the system uses the actual visibility probability values ​​of each screen obtained in previous steps as a "visibility score" field and labels it in the carriage screen model. This score reflects the relative probability of each screen being actually noticed under the current passenger distribution and their gaze direction, and is one of the key indicators for display priority scheduling. The system extracts the three-dimensional spatial slices that overlap with the visual field of each screen in the attention heat field and stores them in a structured form as "heat slice data". Each slice contains the spatial coordinates of the location point and its attention heat value, which is used for screen content projection simulation or heat distribution visualization analysis. In the formed carriage screen model, each screen not only contains its original spatial location, orientation, and size geometric information, but also its visibility score and corresponding heat field slice. The system uses this structure as the basic unit of the enhanced carriage screen model to realize refined modeling and heat-driven control of the multi-screen system in the carriage. The output of the model is a structure with triple data fusion of spatial geometry, passenger attention probability, and heat field information.

[0151] Preferably, step S4 specifically includes:

[0152] Step S41: Acquire ground collaborative data;

[0153] In one embodiment, the system collects ground-based collaborative data from multiple external platforms or subsystems. For example, traffic management system data includes transfer route information, passenger flow density distribution, and transfer pressure warnings, used to infer the current passenger flow situation and transfer urgency at stations or lines. Operations and maintenance system data includes routine operations and maintenance events, such as construction schedules, temporary scheduling changes, sudden station closures, and temporary stop arrangements, affecting train routes and passenger experience. The city-level big data platform interface accesses external environmental data through the city data platform, such as real-time weather conditions (rain, snow, high temperatures, etc.) and holiday peak passenger flow forecasts, to enhance the system's responsiveness to sudden passenger flow and environmental fluctuations. The system supports two real-time data access methods: streaming data service based on Kafka, used for continuously pushing high-frequency events, such as traffic emergencies and sudden changes in passenger density, achieving millisecond-level transmission; and polling retrieval based on a RESTful interface, suitable for operations and maintenance or forecast events (such as holiday forecasts and construction plans), periodically retrieved in conjunction with scheduling strategies. The system is configured with a refresh window period of 3-5 seconds, meaning data updates or pushes are performed every such time period.

[0154] Step S42: Generate a full-line operational status map based on ground coordination data to obtain full-line operational status map data;

[0155] In one embodiment, a graph structure is constructed. ,in : All stations and nodes along the entire line; : Reachable paths (including transfers and parallel branches); Node attributes include the current pedestrian flow intensity, congestion index, and transfer delay rate at each station node; ground event influencing factors are injected into each station node, such as: Setting thresholds (Influence threshold) is marked as "high pressure point"; the output is a graph structure with event correlation, transfer risk and node dynamic pressure.

[0156] Step S43: Process the passenger distribution trend across carriages based on the overall line operation status map data to obtain passenger distribution trend data across carriages;

[0157] In one embodiment, a cross-carriage population heat map is performed on the currently operating train, collecting data on the number of passengers, stationary rate, and boarding / alighting fluctuation rate for each carriage; the system pre-sets a station stop prediction model (based on a linear function of historical data or a trained deep learning model) to predict future passenger flow changes at each station; and a time series trend function is constructed. This represents the passenger density change in the k-th carriage over the next t seconds; prediction is performed using sliding window regression or LSTM neural network; the output is a passenger flow trend matrix across carriages. .

[0158] Step S44: Synchronize events based on cross-carriage passenger distribution trend data and overall line operation status map data to obtain overall line event synchronization data;

[0159] In one embodiment, an event synchronization table is constructed, such as: horizontal axis: carriage area ID; vertical axis: ground coordination event type (construction, delay, transfer fluctuation); table value: whether there is a significant correlation (based on distance weight + synchronization window matching); the system determines whether an event is related to a carriage based on the following criteria: calculating the stations affected by the current event. With train carriages Distance weighting based on spatial location. Spatial weights can be calculated based on station distances on the route map or the current train running progress. Trend linkage condition detection: If predicting a certain carriage... Passenger density change trend within the future time window (from the aforementioned trend function) The level increased significantly and exceeded the set density threshold. Then it is determined that the carriage is affected by... The event has indirect effects. Check if the event's effective time overlaps with the time window of the train arriving at the target station or the target carriage entering the station's section (i.e., whether the synchronization condition is met). If the above conditions are met, the system activates the event's linkage response for the corresponding carriage and marks it in the event synchronization table.

[0160] Step S45: Based on the synchronized event data across the entire line, cross-domain display linkage of passenger attention screen data is carried out to obtain subway passenger collaboration data for intelligent display assistance of subway onboard information.

[0161] In one embodiment, when a synchronous event is marked as "urgent" or highly relevant to the carriage status (such as high-density overlay, construction closure, etc.), the system inserts or replaces the display content corresponding to that event into the current display plan, with higher-priority content overriding the original content. The system selects the area with the highest score in the screen ID set for display based on the attention score of each screen in the passenger attention data (such as the attention weight evaluation from step S3). This process prioritizes ensuring that the displayed content is played within the passenger's main field of vision, improving information acquisition efficiency. Display frequency parameters are set according to the event level, including the number of loops; the display delay duration; and whether to highlight the original content in overlay mode. High-level events (such as sudden accidents) are pushed at a higher frequency, while ordinary transfer guidance information uses a normal frequency. For example, on screens S1 and S2 in carriage C4, an event reminder with content number X2043 needs to be displayed in overlay mode, with a priority of 2 and a loop playback of 3 times.

[0162] Preferably, the generation of the overall operational status map specifically involves:

[0163] The track alignment structure diagram is initialized based on ground collaborative data to obtain track alignment diagram data;

[0164] In one embodiment, an undirected graph G=(V,E) is initialized, where V represents all subway stations (including main line and branch line nodes); E represents the track connection edges between stations (including transfer passages and transfer cost attributes); edge attributes include, but are not limited to, distance: track distance between two stations; travel_time: theoretical travel time; transfer_cost: cross-line transfer time cost; node attributes include, station_type: ordinary / hub / transfer node; capacity_max: maximum passenger capacity; the structure source includes subway GIS layers, track BIM model interfaces, official station structure diagrams, etc.

[0165] By injecting operational status attributes into the track alignment map data, attribute map data is obtained.

[0166] In one embodiment, state attributes are dynamically injected into each node (station) and edge (track segment) to construct a state attribute graph. Node status fields: crowding_index: congestion index; event_flag: presence of ground events (construction, closure, etc.); delay_time: estimated delay duration; Edge status fields: flow_velocity: passenger flow velocity; congestion_risk: current risk score; Attribute data source: dispatching system (SCADA) of ground traffic control center; Capacity estimation model output (based on flow meter or AI camera recognition); Status is marked using timestamp t.

[0167] Schedule event-affected domain subgraph mapping to attribute graph data to obtain graph mapping data;

[0168] In one embodiment, for each scheduling event, the system records its event identifier information and trigger time. For example, event sample: 14:05, Station A is temporarily closed; event type: station closure event; trigger node: the starting station affected by the event, denoted as... The system uses two propagation control parameters to define the identification boundary and propagation of the influence domain: `radius_threshold`, which defines the maximum distance the event's impact can propagate into the graph (usually measured by the number of adjacent nodes, such as 3 stations); and `transfer_dependency`, which indicates whether transfer corridor nodes or cross-line edges should be included in the influence domain expansion process to cover affected transfer paths. The system uses a graph traversal algorithm to start from the event nodes... Starting from this point, the system performs an influence propagation path search, employing algorithms including Breadth-First Search (BFS) and Depth-First Search (DFS). During the traversal, the system constructs an event influence subgraph, using a propagation radius threshold as a constraint and satisfying the propagation rules. This subgraph contains a graph structure consisting of all nodes and edges affected by the current scheduling event. The system processes the edges of each path within the subgraph. Set the "Propagation Time Decay Weight" field The calculation formula is as follows: ,in Indicates from the event source node to the edge starting point The path length (which can be the number of hops or the weighted distance); The initial influence intensity parameter is set to 1.0; This is the attenuation rate adjustment parameter, with a value ranging from 0.5 to 2.0; This is an exponentially decaying term, simulating the trend of strong impact of event propagation on neighboring regions and weak impact on distant regions. The graph mapping data output by the system includes the subgraph structure of the influence domain. This indicates the area covered by the current event; the propagation attenuation weight field is attached to each edge. Updates on the operational status of the nodes involved (such as delay time, changes in accessibility, and transfer disruptions).

[0169] Graph situational analysis is performed on the graph mapping data to obtain situational graph data;

[0170] In one embodiment, in the graph structure, each node The operational status is represented by a comprehensive indicator called the "pressure value," denoted as... This is used to assess the current operational complexity, load intensity, or scheduling risk faced by the site. The calculation method is as follows: ,in For nodes Pressure value; For nodes The state function represents its operational state characteristics (such as congestion, delays, event flags, etc.). Represents a node For nodes The attention weight or propagation attenuation coefficient is calculated based on the propagation strength of adjacent edges or the event impact coefficient. This represents the aggregation operation that affects the state of all adjacent nodes, and can be one of the following: weighted summation, max pooling, or attention-weighted average. The system calculates the stress value for each node's own state characteristic function, ensuring that local information is included in the stress assessment. The system performs this stress value calculation operation on all nodes in the graph, and the stress values ​​of all nodes are aggregated into a single situational heatmap.

[0171] Based on the situation map data, a time-series snapshot sequence is constructed to obtain the overall operational situation map data.

[0172] In one embodiment, a uniform time window interval is set. For example, sampling once every 1 minute, that is, at time point This indicates that from the first to the nth time point, corresponding snapshots of the graphical situation are saved respectively. Each snapshot map contains the following core components: a site status score set, recording the pressure value or risk score of each site at that time point; an edge flow estimation set, recording the passenger flow estimate or capacity score of the track edges between adjacent sites, used to assist in track segment load assessment or bottleneck identification. The snapshot structure at each moment can be organized in the following format: a timestamp field, recording the current sampling time point (format ISO 8601, e.g., "2025-05-29T14:30:00"); a site status score set, with the site number as the key and the pressure value as the value, representing the status intensity of each site at that moment; and a track connection flow estimation set, with the site pair number as the key, recording the flow estimate or traffic pressure index of the track segment between the site pairs. All snapshots at all time points are organized chronologically to form a time-chained map sequence. , representing a snapshot of the graph at each point in time, and this sequence constitutes the trajectory of the operational status evolution of the entire subway network over a period of time.

[0173] Preferably, this application also provides a subway in-vehicle information intelligent display system for executing the subway in-vehicle information intelligent display method described above, the subway in-vehicle information intelligent display system comprising:

[0174] The passenger behavior perception and focus situation extraction module is used to acquire train operation status data, carriage environment perception data, and train event flow data; to analyze passenger behavior based on carriage environment perception data to obtain passenger behavior data; and to analyze focus situation based on passenger behavior data to obtain focus situation data.

[0175] The state-aware intelligent scheduling decision-making module is used to perform state-aware scheduling based on train operation status data, train event flow data, and focused situation data to obtain state-aware data.

[0176] The passenger attention perception-driven screen control module is used to drive the passenger attention screen based on the state perception data and obtain the passenger attention screen data.

[0177] The ground-based collaborative display optimization module is used to acquire ground-based collaborative data and, based on this data, to coordinate and link passenger attention screen data to obtain subway passenger collaborative data, thereby assisting in the intelligent display of subway onboard information.

[0178] Therefore, the embodiments should be regarded as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended application documents rather than the foregoing description. Thus, it is intended that all variations falling within the meaning and scope of the equivalents of the application documents be incorporated into the invention.

[0179] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A method for intelligent display of information on subway vehicles, characterized in that, Includes the following steps: Step S1: Acquire train operation status data, carriage environment perception data, and train event stream data; Passenger behavior data is obtained by analyzing passenger behavior based on the passenger cabin environment perception data; focus situation analysis is then performed based on the passenger behavior data to obtain focus situation data. Step S2: Perform state-aware scheduling based on train operation status data, train event flow data, and focused situation data to obtain state-aware data; Step S3: Drive passenger attention screen based on state perception data to obtain passenger attention screen data; Step S4: Acquire ground coordination data and coordinate with passenger attention screen data based on ground coordination data to obtain subway passenger coordination data for assisting in the intelligent display of subway information. The focus of situational analysis specifically includes: Regional attention data is obtained by calculating regional attention based on passenger behavior data; The focus status data of the region is classified into focus status data. By performing local congestion evolution on the focused state data, local congestion data is obtained; Focus situation data is generated based on local congestion data and focus status data.

2. The method according to claim 1, characterized in that, Step S1 is as follows: Acquire train operation status data, carriage environment perception data, and train event stream data; Spatial behavior maps are constructed based on the passenger compartment environment perception data to obtain passenger compartment behavior map data; Behavioral features are extracted from the carriage behavior graph data to obtain behavioral feature data; Behavioral generalization enhancement is performed on behavioral feature data to obtain passenger behavior data.

3. The method according to claim 1, characterized in that, Step S2 is as follows: Train status feature data is obtained by extracting train status features from train operation status data and train event stream data. Train status influence matrix data is obtained by mapping train status characteristic data and focused situation data; Key response regions are identified from the train situation matrix data to obtain key response region data. The scheduling window is generated based on the key response area data, and the scheduling window data is obtained. State-aware scheduling is generated based on scheduling window data and a preset state-aware strategy library to obtain state-aware data.

4. The method according to claim 3, characterized in that, The train situation influence matrix mapping is specifically as follows: Initialize the track network topology diagram by analyzing the train operation status data and train event flow data to obtain the track network diagram data; Based on the train status characteristic data, the track network diagram data is mapped to the running trajectory to obtain the running trajectory diagram data; The trajectory graph data is weighted using a tensor structure to obtain weighted graph data; Spatiotemporal coupling is performed on the weighted graph data to obtain coupled graph data; Propagation link impact calculations are performed on the coupling graph data to obtain propagation graph data; Event-driven structural mutations are performed on the propagation graph data to obtain train situation matrix data.

5. The method according to claim 1, characterized in that, Step S3 is as follows: Acquire data from the carriage screens; Spatial modeling is performed on the screen data in the carriage to obtain the carriage screen model; Attention probability field data is obtained by calculating the attention probability field based on the carriage screen model and focusing situation data; Passenger attention screen data is obtained by multi-frequency drive generation based on attention probability field data and state perception data. The multi-frequency drive generation involves screen hierarchical management based on attention probability field data and set upper and lower thresholds, and calling the backend content scheduling service to generate specific information update packages based on the classification results of screen hierarchical management and state perception data, and generating a drive sequence to control each screen.

6. The method according to claim 5, characterized in that, Spatial modeling specifically involves: Based on the data from the carriage screens, a carriage screen space model is obtained by creating a carriage screen space model. A visual cone model is constructed from the screen space model of the carriage to obtain the screen visual cone model; Crowd attention heatfields were constructed using a screen visual cone model to obtain crowd attention heatfield data, where the crowd attention heatfield construction was performed as follows: , For three-dimensional space points The intensity of the crowd's attention heat field at that location assign passenger index numbers, This represents the total number of passengers currently in the carriage. Passenger attention coefficient It is a natural exponential function. Let be any three-dimensional spatial coordinate point in the thermal field. For the first The location coordinates of each passenger To control the extent of diffusion; The actual screen visibility probability is calculated from the crowd attention heat field data to obtain the actual screen visibility probability data. The actual visibility probability of the screen and the crowd attention heat field data are used to annotate the spatial model of the carriage screen to obtain the carriage screen model.

7. The method according to claim 1, characterized in that, Step S4 is as follows: Acquire ground-based collaborative data; A full-line operational status map is generated based on ground collaborative data, resulting in full-line operational status map data. Based on the data from the overall line operation status map, cross-carriage passenger distribution trend processing is performed to obtain cross-carriage passenger distribution trend data. Event synchronization is performed based on cross-carriage passenger distribution trend data and full-line operation status map data to obtain full-line event synchronization data; By cross-domain display and linkage of passenger attention screen data based on the synchronous data of events across the entire line, subway passenger collaboration data is obtained to assist in the intelligent display of subway in-vehicle information.

8. The method according to claim 7, characterized in that, The generation of the overall operational status map is specifically as follows: The track alignment structure diagram is initialized based on ground collaborative data to obtain track alignment diagram data; By injecting operational status attributes into the track alignment map data, attribute map data is obtained. Schedule event-affected domain subgraph mapping to attribute graph data to obtain graph mapping data; Graph situational analysis is performed on the graph mapping data to obtain situational map data; Based on the situation map data, a time-series snapshot sequence is constructed to obtain the overall operational situation map data.

9. A subway vehicle-mounted intelligent information display system, characterized in that, For executing the intelligent display method for subway vehicle information as described in claim 1, the intelligent display system for subway vehicle information includes: The passenger behavior perception and focus situation extraction module is used to acquire train operation status data, carriage environment perception data, and train event flow data; to analyze passenger behavior based on carriage environment perception data to obtain passenger behavior data; and to analyze focus situation based on passenger behavior data to obtain focus situation data. The state-aware intelligent scheduling decision-making module is used to perform state-aware scheduling based on train operation status data, train event flow data, and focused situation data to obtain state-aware data. The passenger attention perception-driven screen control module is used to drive the passenger attention screen based on the state perception data and obtain the passenger attention screen data. The ground-based collaborative display optimization module is used to acquire ground-based collaborative data and, based on this data, to coordinate and link passenger attention screen data to obtain subway passenger collaborative data, thereby assisting in the intelligent display of subway onboard information.

Images