Regional people flow analysis method and system based on agent graph data

Abstract

The invention provides a regional people flow analysis method and system based on agent graph data, and belongs to the technical field of artificial intelligence and regional management. The method comprises the following steps: acquiring a personnel movement track, an area node state and environment perception multi-source data in a preset area, and processing to obtain standardized data; defining a personnel agent, a regional node agent and an association relationship type based on the data, and constructing an initial agent graph; adjusting the agent attribute information and the association relationship weight of the initial atlas through an incremental updating mechanism to obtain an optimized agent atlas; mapping the optimized atlas to a low-dimensional feature space by adopting an atlas embedding algorithm, and constructing a space-time attention model; and capturing a people flow time sequence evolution rule, mining the influence of topological association between the intelligent agents on flow distribution, and outputting a regional people flow analysis report. According to the method, the real-time performance and the accuracy of human traffic analysis are improved through combination of agent map dynamic updating and a space-time attention model, and a scientific decision basis is provided for regional management.

Description

Technical Field

[0001] This application relates to the fields of artificial intelligence and regional management technology, and more specifically, to a method and system for regional pedestrian flow analysis based on intelligent agent map data. Background Technology

[0002] With the acceleration of urbanization, the management needs of densely populated areas such as commercial districts, transportation hubs, and large public areas are becoming increasingly urgent. Real-time monitoring, accurate prediction, and early warning of anomalies in regional pedestrian flow have become crucial for improving management efficiency and safety. Existing methods for analyzing regional pedestrian flow often rely on single data sources (such as surveillance videos and card swipe data), resulting in limited data dimensions and insufficient correlation analysis. Some methods employ static analysis models, which cannot adapt to the dynamic changes in pedestrian flow, leading to delayed analysis results. Furthermore, traditional models do not adequately integrate the temporal evolution patterns and spatial correlation characteristics of pedestrian flow, resulting in low prediction accuracy and delayed anomaly identification. In addition, existing technologies lack dynamic modeling methods based on intelligent agent graphs, making it difficult to comprehensively depict the complex relationships between people and regional nodes, and failing to meet the needs of regional management for refined and real-time pedestrian flow analysis. Therefore, there is an urgent need for a regional pedestrian flow analysis technology that can integrate multi-source data, dynamically adapt to changes in pedestrian flow, and accurately integrate spatiotemporal characteristics to overcome the limitations of existing methods. Summary of the Invention

[0003] This invention provides a method and system for regional pedestrian flow analysis based on agent graph data. It collects pedestrian movement trajectory data, regional node status data, and environmental perception data within a preset area. After data cleaning, format standardization, and noise filtering, standardized multi-source data is obtained. Based on this standardized data, human agents, regional node agents, and association types including human-node access relationships and node-node spatial relationships are defined. An initial agent graph containing agent attribute information and topological associations is constructed. Subsequently, based on the real-time data stream, an incremental update mechanism is used, incorporating a sliding time window triggering strategy, agent addition or removal rules, an association strength decay model, and a topology optimization strategy, to dynamically adjust... The initial agent graph is obtained by integrating agent attribute information and relational weights to obtain an optimized agent graph. Then, a graph embedding algorithm is used to map the optimized agent graph to a low-dimensional feature space, generating low-dimensional feature vectors for agents and constructing a spatiotemporal attention model containing temporal and spatial attention layers. Finally, this model is used to capture the temporal dependencies of pedestrian traffic, mine the topological associations between agents, and integrate environmental perception data features to obtain a regional pedestrian traffic analysis report that includes real-time traffic distribution data, traffic prediction curves, abnormal traffic warning information, traffic hotspot identification, supply and demand balance assessment, and guidance optimization suggestions. This achieves deep fusion and dynamic modeling of multi-source data, improving the comprehensiveness, real-time performance, and accuracy of regional pedestrian traffic analysis.

[0004] This application provides a method for regional pedestrian flow analysis based on intelligent agent map data, including the following steps: Collect personnel movement trajectory data, regional node status data, and environmental perception data within a preset area. Through data cleaning, format standardization, and noise filtering, standardized multi-source data is obtained. Based on the standardized multi-source data, human intelligent agents, regional node intelligent agents and their relationship types are defined, and an initial intelligent agent graph containing intelligent agent attribute information and topological relationship is constructed. Based on the real-time data stream, an incremental update mechanism is used to dynamically adjust the agent attribute information and association weights in the initial agent graph to obtain an optimized agent graph. The optimized agent graph is mapped to a low-dimensional feature space using a graph embedding algorithm to obtain the agent's low-dimensional feature vector and construct a spatiotemporal attention model. Based on the spatiotemporal attention model, a regional pedestrian flow analysis report is obtained by capturing the temporal evolution of pedestrian flow and mining the influence of topological relationships between agents on pedestrian flow distribution.

[0005] In the regional pedestrian flow analysis method based on intelligent agent map data described in this application, the collection of pedestrian movement trajectory data, regional node status data, and environmental perception data within a preset area includes: Collect personnel movement trajectory data, regional node status data, and environmental perception data within the preset area; The personnel movement trajectory data includes terminal positioning data, public transportation travel data, and travel vehicle trajectory data; The regional node status data includes regional node capacity data and facility operation status data; The environmental perception data includes meteorological data, public event data, and traffic control data.

[0006] In the regional pedestrian flow analysis method based on intelligent agent map data described in this application, the step of obtaining standardized multi-source data through data cleaning, format standardization, and noise filtering specifically involves: Based on the collected personnel movement trajectory data, regional node status data, and environmental perception data, data cleaning is performed using missing value filling algorithm and outlier removal algorithm; The unified data protocol is used to convert and process standardized execution formats. Noise filtering is performed using a sliding window smoothing filter algorithm or a wavelet filter algorithm to obtain standardized multi-source data.

[0007] In the regional pedestrian flow analysis method based on agent graph data described in this application, the step of defining personnel agents, regional node agents, and association relationship types based on the standardized multi-source data includes: Based on the personnel feature data, regional node attribute data and spatial association data in the standardized multi-source data, define personnel intelligent agents and regional node intelligent agents, and define the association relationship types. The types of association relationships include personnel-node access relationships and node-node spatial association relationships; The personnel and node access relationships include temporary access relationships and permanent residency relationships; The spatial relationships between nodes include adjacency relationships, path relationships, and functional relationships.

[0008] In the regional pedestrian flow analysis method based on agent graph data described in this application, the step of dynamically adjusting the agent attribute information and correlation weights in the initial agent graph using an incremental update mechanism based on real-time collected data streams to obtain an optimized agent graph includes: Based on the real-time data stream, an incremental update mechanism with a sliding time window triggering strategy is adopted to add or remove intelligent agents according to the preset dwell time threshold of personnel in the area node. The strength of the correlation is adjusted by an exponential decay function based on the time decay coefficient; The topology is optimized based on a dynamic threshold of association density, and the agent attribute information and association weights are dynamically adjusted to obtain an optimized agent graph.

[0009] In the regional pedestrian flow analysis method based on agent graph data described in this application, the step of using a graph embedding algorithm to map the optimized agent graph to a low-dimensional feature space, obtaining a low-dimensional feature vector of the agent, and constructing a spatiotemporal attention model specifically involves: Based on the optimized agent graph, a preset algorithm is used as the graph embedding algorithm; The optimized agent graph is mapped to a low-dimensional feature space to obtain the agent's low-dimensional feature vector; A spatiotemporal attention model is constructed based on the low-dimensional feature vectors of the agent.

[0010] In the regional pedestrian flow analysis method based on agent graph data described in this application, the step of obtaining a regional pedestrian flow analysis report by capturing the temporal evolution pattern of pedestrian flow and mining the influence of topological associations between agents on the flow distribution according to the spatiotemporal attention model is as follows: The spatiotemporal attention model includes a temporal attention layer and a spatial attention layer; Based on the low-dimensional feature vector of the agent, the temporal dependence of pedestrian flow at different time steps is captured by a time attention layer based on a self-attention mechanism. By mining the influence of topological correlations through a spatial attention layer that dynamically adjusts weights based on topological distance between agents, and by fusing feature vectors corresponding to environmental perception data, a regional pedestrian flow analysis report is obtained.

[0011] In the regional pedestrian flow analysis method based on intelligent agent map data described in this application, the regional pedestrian flow analysis report further includes: Based on the results of traffic hotspot identification, regional traffic supply and demand balance assessment, and traffic diversion optimization suggestions output by the spatiotemporal attention model.

[0012] Secondly, this application provides a regional pedestrian flow analysis system based on intelligent agent map data, characterized in that it includes: The data acquisition and preprocessing module is used to collect personnel movement trajectory data, regional node status data, and environmental perception data within a preset area. Through data cleaning, format standardization, and noise filtering, it outputs standardized multi-source data. The initial agent graph construction module is used to define personnel agents, regional node agents and association types based on the standardized multi-source data, and construct an initial agent graph containing agent attribute information and topological associations. The agent graph incremental update module is used to dynamically adjust the agent attribute information and association weights in the initial agent graph based on the real-time collected data stream using an incremental update mechanism, and output an optimized agent graph. The graph embedding and model building module is used to map the optimized agent graph to a low-dimensional feature space using a graph embedding algorithm, to obtain the agent's low-dimensional feature vector and build a spatiotemporal attention model. The analysis report generation module is used to output a regional pedestrian flow analysis report based on the spatiotemporal attention model by capturing the temporal evolution pattern of pedestrian flow and mining the impact of topological associations between agents on the flow distribution.

[0013] The system also includes a memory and a processor. The memory contains a program for a regional pedestrian flow analysis method based on intelligent agent map data. When the processor executes the program for the regional pedestrian flow analysis method based on intelligent agent map data, it performs the following steps: Collect personnel movement trajectory data, regional node status data, and environmental perception data within a preset area. Through data cleaning, format standardization, and noise filtering, standardized multi-source data is obtained. Based on the standardized multi-source data, human intelligent agents, regional node intelligent agents and their relationship types are defined, and an initial intelligent agent graph containing intelligent agent attribute information and topological relationship is constructed. Based on the real-time data stream, an incremental update mechanism is used to dynamically adjust the agent attribute information and association weights in the initial agent graph to obtain an optimized agent graph. The optimized agent graph is mapped to a low-dimensional feature space using a graph embedding algorithm to obtain the agent's low-dimensional feature vector and construct a spatiotemporal attention model. Based on the spatiotemporal attention model, a regional pedestrian flow analysis report is obtained by capturing the temporal evolution of pedestrian flow and mining the influence of topological relationships between agents on pedestrian flow distribution.

[0014] As can be seen from the above, this invention provides a method and system for regional pedestrian flow analysis based on agent graph data. It collects pedestrian movement trajectory data, regional node status data, and environmental perception data within a preset area. After data cleaning, format standardization, and noise filtering, standardized multi-source data is obtained. Based on this standardized data, human agents, regional node agents, and association types including human-node access relationships and node-node spatial relationships are defined. An initial agent graph containing agent attribute information and topological associations is constructed. Subsequently, based on the real-time data stream, an incremental update mechanism is adopted, including a sliding time window triggering strategy, agent addition or deletion rules, an association strength decay model, and a topology optimization strategy. The algorithm dynamically adjusts the agent attribute information and relational weights of the initial agent graph to obtain an optimized agent graph. Then, it uses a graph embedding algorithm to map the optimized agent graph to a low-dimensional feature space, generating low-dimensional feature vectors for agents and constructing a spatiotemporal attention model with temporal and spatial attention layers. Finally, it uses this model to capture the temporal dependencies of pedestrian traffic, mine the topological associations between agents, and integrate environmental perception data features to obtain a regional pedestrian traffic analysis report that includes real-time traffic distribution data, traffic prediction curves, abnormal traffic warning information, traffic hotspot identification, supply and demand balance assessment, and guidance optimization suggestions. This achieves deep fusion and dynamic modeling of multi-source data, improving the comprehensiveness, real-time performance, and accuracy of regional pedestrian traffic analysis.

[0015] Other features and advantages of this application will be set forth in the following description and will be apparent in part from the description or may be learned by practicing embodiments of this application. The objectives and other advantages of this application may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a high-level flowchart of a regional pedestrian flow analysis method based on intelligent agent graph data provided in an embodiment of this application, used for regional pedestrian flow management.

[0018] Figure 2 A flowchart of a regional pedestrian flow analysis method based on intelligent agent map data provided in this application embodiment; Figure 3 This is a structural block diagram of a regional pedestrian flow analysis system based on intelligent agent map data, provided in an embodiment of this application. Detailed Implementation

[0019] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0020] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first" and "second" are used only to distinguish descriptions and should not be construed as indicating or implying relative importance. It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use, and processing of related data must comply with the relevant laws, regulations, and standards of the relevant countries and regions.

[0021] Please refer to Figure 1 , Figure 1This is a high-level flowchart of a regional pedestrian flow analysis method based on agent graph data in some embodiments of this application. The high-level flowchart can be summarized as follows: First, multi-source data from a preset region is collected and processed through cleaning, standardization, and noise filtering to obtain standardized data. Then, based on this data, two types of agents and their relationship types are defined to construct an initial agent graph. Subsequently, the graph is dynamically adjusted through an incremental update mechanism to obtain an optimized graph. Next, a graph embedding algorithm is used to generate low-dimensional feature vectors and construct a spatiotemporal attention model. Finally, this model is used to capture temporal patterns, mine topological relationships, and integrate environmental features to output a regional pedestrian flow analysis report, achieving dynamic and accurate pedestrian flow analysis.

[0022] Please refer to Figure 2 , Figure 2 This is a flowchart of a regional pedestrian flow analysis method based on intelligent agent map data in some embodiments of this application.

[0023] The regional pedestrian flow analysis method based on intelligent agent map data disclosed in the first aspect of this invention is used in terminal devices, such as computers and mobile terminals. This regional pedestrian flow analysis method based on intelligent agent map data includes the following steps: S201. Collect personnel movement trajectory data, regional node status data, and environmental perception data within the preset area. Through data cleaning, format standardization, and noise filtering, obtain standardized multi-source data. S202. Based on the standardized multi-source data, define personnel intelligent agents, regional node intelligent agents and their relationship types, and construct an initial intelligent agent graph containing intelligent agent attribute information and topological relationship. S203. Based on the real-time data stream, the agent attribute information and association weights in the initial agent graph are dynamically adjusted using an incremental update mechanism to obtain an optimized agent graph. S204. The optimized agent graph is mapped to a low-dimensional feature space using a graph embedding algorithm to obtain the agent's low-dimensional feature vector and a spatiotemporal attention model is constructed. S205. Based on the spatiotemporal attention model, by capturing the temporal evolution of pedestrian flow and mining the influence of topological associations between agents on the flow distribution, a regional pedestrian flow analysis report is obtained.

[0024] First, personnel movement trajectory data, including terminal positioning data and public transportation travel data, are collected within a preset area. Then, regional node status data, including regional node capacity data and facility operation status data, as well as environmental perception data, including meteorological data and public event data, are collected. Next, missing value imputation and outlier removal algorithms are used for data cleaning. A unified data protocol conversion process is used to standardize the format, and a sliding window smoothing filter or wavelet filter algorithm is used to filter noise, resulting in standardized multi-source data. Based on personnel feature data, regional node attribute data, and spatial association data in this standardized multi-source data, personnel agents, regional node agents, and personnel-node access relationships (including temporary and fixed residency relationships) and node-node spatial association relationships (including adjacency, path, and functional relationships) are defined. An initial agent graph containing agent attribute information and topological association relationships is constructed. Based on the real-time collected data stream, an incremental update mechanism with a sliding time window trigger strategy is used to add or remove agents according to a preset dwell time threshold for personnel at regional nodes. This method adjusts the strength of associations using an exponential decay function based on the time decay coefficient, optimizes the topology based on a dynamic threshold of association density, and dynamically adjusts the agent attribute information and association weights in the initial agent graph to obtain an optimized agent graph. A graph embedding algorithm, such as graph convolutional network algorithm or graph attention network algorithm, is used to map the optimized agent graph to a low-dimensional feature space, obtaining low-dimensional feature vectors for the agents. Based on these vectors, a spatiotemporal attention model containing a temporal attention layer and a spatial attention layer is constructed. The model uses a self-attention mechanism-based temporal attention layer to capture the temporal dependencies of pedestrian traffic at different time steps, and a spatial attention layer that dynamically adjusts weights based on topological distance between agents to mine the impact of topological associations on traffic distribution. Furthermore, feature vectors corresponding to environmental perception data are fused to obtain a regional pedestrian traffic analysis report containing real-time traffic distribution data, traffic prediction curves, abnormal traffic warning information, traffic hotspot area identification results, regional traffic supply and demand balance assessment results, and traffic diversion optimization suggestions. This comprehensively achieves dynamic and accurate analysis of regional pedestrian traffic.

[0025] According to an embodiment of the present invention, the collection of personnel movement trajectory data, area node status data, and environmental perception data within a preset area includes: Collect personnel movement trajectory data, regional node status data, and environmental perception data within the preset area; The personnel movement trajectory data includes terminal positioning data, public transportation travel data, and travel vehicle trajectory data; The regional node status data includes regional node capacity data and facility operation status data; The environmental perception data includes meteorological data, public event data, and traffic control data.

[0026] The personnel movement trajectory data includes terminal positioning data (such as mobile phone GPS positioning data and wearable device positioning data), public transportation travel data (such as subway card swiping data, bus QR code payment data, and high-speed rail ticket purchase data), and travel tool trajectory data (such as ride-hailing vehicle travel trajectory data and shared bicycle riding trajectory data), used to accurately depict the movement paths and travel characteristics of personnel within a preset area; the regional node status data includes regional node capacity data (such as the maximum capacity of shops in a business district, the capacity of waiting areas in transportation hubs, and the capacity of public squares) and facility operation status data (such as elevator operation status, escalator start / stop status, parking lot space occupancy status, and public restroom usage status). The data is used to reflect the carrying capacity and service guarantee status of key nodes in the region; environmental perception data includes meteorological data (such as real-time meteorological parameters such as temperature, precipitation, and wind force), public event data (such as information on the holding of large-scale exhibitions, performances, and promotional activities in the region), and traffic control data (such as information on temporary traffic restrictions, road construction closures, and traffic control periods), which are used to capture external environmental factors that may affect the flow of people. Through the comprehensive collection of the above multi-dimensional data, a complete and comprehensive data source is provided for subsequent standardized processing, intelligent agent map construction, and people flow analysis, ensuring the accuracy and reliability of the analysis results.

[0027] According to an embodiment of the present invention, the process of obtaining standardized multi-source data through data cleaning, format standardization, and noise filtering specifically involves: Based on the collected personnel movement trajectory data, regional node status data, and environmental perception data, data cleaning is performed using missing value filling algorithm and outlier removal algorithm; The unified data protocol is used to convert and process standardized execution formats. Noise filtering is performed using a sliding window smoothing filter algorithm or a wavelet filter algorithm to obtain standardized multi-source data.

[0028] The process of cleaning, standardizing, and filtering the collected personnel movement trajectory data, regional node status data, and environmental perception data to obtain standardized multi-source data is a crucial preprocessing step to ensure the accuracy of subsequent intelligent agent map construction and pedestrian flow analysis. First, to address potential issues such as missing information and logical contradictions in the three types of data, missing value imputation algorithms (such as imputation based on statistical mean or median, or predictive imputation based on machine learning) are used to supplement missing data according to data characteristics. Outlier removal algorithms (such as removal based on the 3σ criterion or box plot analysis) are used to remove outlier data that exceeds reasonable limits, achieving initial optimization of data quality. Next, to solve the problem of heterogeneous multi-source data formats, a unified data protocol conversion process is adopted (such as converting different formats of positioning data, status data, and perception data into JSON or XML format). The data is formatted to ensure that all types of data follow consistent syntax rules and data structures, guaranteeing data compatibility and reusability. Finally, to address noise interference caused by equipment errors and signal interference during data acquisition, a sliding window smoothing filter or wavelet filter algorithm is used to reduce noise. The sliding window smoothing filter reduces the impact of instantaneous noise by performing a moving average calculation on the data within a fixed-length time window, while the wavelet filter separates noise signals from valid signals through multi-scale decomposition and reconstruction. Through these three steps, standardized multi-source data with reliable quality, unified format, and noise elimination is finally obtained, providing a high-quality data foundation for the accurate construction of the initial intelligent agent map, incremental updates, and effective training of the spatiotemporal attention model.

[0029] According to an embodiment of the present invention, defining personnel intelligent agents, regional node intelligent agents, and association relationship types based on the standardized multi-source data includes: Based on the personnel feature data, regional node attribute data and spatial association data in the standardized multi-source data, define personnel intelligent agents and regional node intelligent agents, and define the association relationship types. The types of association relationships include personnel-node access relationships and node-node spatial association relationships; The personnel and node access relationships include temporary access relationships and permanent residency relationships; The spatial relationships between nodes include adjacency relationships, path relationships, and functional relationships.

[0030] First, personnel characteristic data (such as travel frequency, preferred routes, and identity tags), regional node attribute data (such as node function type, geographical location, and carrying capacity), and spatial correlation data (such as distance between nodes and connection paths) are extracted from standardized multi-source data. Based on this, personnel intelligent agents and regional node intelligent agents are defined. The personnel intelligent agent is used to abstractly represent individuals or groups of people within the region, while the regional node intelligent agent is used to abstractly represent key locations within the region (such as shops, platforms, and entrances / exits). Simultaneously, based on data characteristics and regional personnel flow patterns, correlation types are defined. These correlation types specifically include personnel-node access relationships and node-node spatial correlation relationships. The relationships between people and nodes include temporary access relationships (such as the association of people passing through a node for a short time) and fixed residence relationships (such as the association of people working or consuming at a node for a long time). The spatial relationships between nodes include adjacency relationships (such as the association between nodes that are directly adjacent in terms of geographical location), path relationships (such as the association between nodes that have a common path), and functional relationships (such as the association between nodes that have complementary or similar functions). By clearly defining the above types of agents and relationships, a clear modeling basis is provided for the subsequent construction of an initial agent graph that includes agent attribute information and topological relationships, ensuring that the graph can comprehensively and accurately reflect the essential characteristics and association logic of people and nodes in the region.

[0031] According to an embodiment of the present invention, the step of dynamically adjusting the agent attribute information and association weights in the initial agent graph based on the real-time acquired data stream using an incremental update mechanism to obtain an optimized agent graph includes: Based on the real-time data stream, an incremental update mechanism with a sliding time window triggering strategy is adopted to add or remove intelligent agents according to the preset dwell time threshold of personnel in the area node. The strength of the correlation is adjusted by an exponential decay function based on the time decay coefficient; The topology is optimized based on a dynamic threshold of association density, and the agent attribute information and association weights are dynamically adjusted to obtain an optimized agent graph.

[0032] First, an incremental update mechanism based on a sliding time window triggering strategy is adopted. This mechanism captures changes in the real-time data stream within a preset sliding time window. Agents are added or deregistered based on a preset dwell time threshold for personnel at a region node. When a person's dwell time at a region node reaches or exceeds the threshold, a new agent is added and associated with that region node agent. When a person leaves the region and their dwell time is below the threshold or outside the monitoring range, the corresponding agent is deregistered. Second, regarding the relationships between agents, the strength of the relationships is dynamically adjusted using an exponential decay function based on a time decay coefficient. As time progresses, the relationship strength is adjusted accordingly. The correlation strength of new interactive behaviors decays exponentially, ensuring that the correlations reflect recent population flow characteristics. Finally, the topology is optimized based on a dynamic threshold of correlation density. When the correlation density between agents in a certain area is higher than a preset threshold, redundant correlations are simplified; when it is lower than the threshold, critical path correlations are strengthened. At the same time, agent attribute information (such as the real-time location of personnel agents and the current carrying capacity of regional node agents) and correlation weights are dynamically adjusted. Ultimately, an optimized agent graph that accurately reflects the real-time population flow status and correlation characteristics is obtained, providing dynamic and reliable basic data support for subsequent graph embedding and spatiotemporal attention model construction.

[0033] According to an embodiment of the present invention, the step of using a graph embedding algorithm to map the optimized agent graph to a low-dimensional feature space to obtain a low-dimensional feature vector of the agent and construct a spatiotemporal attention model specifically involves: Based on the optimized agent graph, a preset algorithm is used as the graph embedding algorithm; The optimized agent graph is mapped to a low-dimensional feature space to obtain the agent's low-dimensional feature vector; A spatiotemporal attention model is constructed based on the low-dimensional feature vectors of the agent.

[0034] First, for the agent attribute information, topological relationships, and dynamic update features contained in the optimized agent graph, a preset graph embedding algorithm is selected (such as one or more combinations of Graph Convolutional Network (GCN), Graph Attention Network (GAT), Node 2vec, or DeepWalk algorithms). This type of algorithm can effectively mine the topological structure features and agent association semantics of the graph. Then, the high-dimensional and complex optimized agent graph is mapped to a low-dimensional feature space through the selected graph embedding algorithm, and the agent attribute information and relationship features are compressed and encoded into low-dimensional, high-density low-dimensional agent feature vectors, which retain the core feature information and reduce the computational complexity of subsequent models. Finally, based on the obtained low-dimensional agent feature vectors, a spatiotemporal attention model that integrates spatiotemporal dimension association features is constructed. This model uses the low-dimensional feature vectors as input and provides a structured algorithmic model support for subsequent capture of the temporal evolution of pedestrian flow and mining the impact of topological relationships between agents on traffic distribution, ensuring the efficiency and accuracy of pedestrian flow analysis.

[0035] According to an embodiment of the present invention, the step of obtaining a regional pedestrian flow analysis report by capturing the temporal evolution pattern of pedestrian flow and mining the influence of topological associations between agents on the flow distribution based on the spatiotemporal attention model is specifically as follows: The spatiotemporal attention model includes a temporal attention layer and a spatial attention layer; Based on the low-dimensional feature vector of the agent, the temporal dependence of pedestrian flow at different time steps is captured by a time attention layer based on a self-attention mechanism. By mining the influence of topological correlations through a spatial attention layer that dynamically adjusts weights based on topological distance between agents, and by fusing feature vectors corresponding to environmental perception data, a regional pedestrian flow analysis report is obtained.

[0036] First, using the low-dimensional feature vectors of the agents as input, a time attention layer based on a self-attention mechanism adaptively allocates attention weights for different time steps, accurately capturing the temporal dependencies of pedestrian traffic on hourly, daily, and weekly time scales (such as peak-hour traffic fluctuation patterns and traffic differences between weekdays and holidays). Second, a spatial attention layer dynamically adjusts weights based on topological distance between agents, allocating spatial attention weights according to the strength of topological associations between personnel agents and regional node agents, and between regional node agents, to deeply explore the impact of spatial associations on traffic distribution (such as traffic transmission patterns between adjacent nodes and the diversion and convergence effects of critical paths on regional traffic). Simultaneously, feature vectors corresponding to environmental perception data (such as meteorological data features, public event data features, and traffic control data features) are integrated into the model calculation process to achieve collaborative analysis of spatiotemporal features and external environmental features. Finally, through the above multi-level and multi-dimensional feature fusion and pattern mining, a regional pedestrian traffic analysis report is generated, including real-time traffic distribution data, traffic prediction curves, abnormal traffic early warning information, as well as traffic hotspot area identification results, regional traffic supply and demand balance assessment results, and traffic diversion optimization suggestions, providing comprehensive and accurate data analysis support for regional management decisions.

[0037] According to an embodiment of the present invention, the regional pedestrian flow analysis report further includes: Based on the results of traffic hotspot identification, regional traffic supply and demand balance assessment, and traffic diversion optimization suggestions output by the spatiotemporal attention model.

[0038] In this technical solution, the regional traffic flow analysis report, in addition to including real-time traffic distribution data, traffic prediction curves, and abnormal traffic early warning information, further incorporates the traffic hotspot area identification results, regional traffic supply and demand balance assessment results, and traffic diversion optimization suggestions based on the spatiotemporal attention model, so as to achieve the comprehensiveness and decision-making guidance of the analysis report. The results of traffic hotspot area identification are achieved by mining the spatiotemporal correlation features of intelligent agents through models to locate core areas with high concentrations of people within a preset area (such as the area around popular shops in commercial districts, transfer passages of transportation hubs, and the area around venues for public events) and the temporal patterns of traffic aggregation. The results of regional traffic supply and demand balance assessment are achieved by combining regional node capacity data, facility operation status data, and real-time and predicted traffic data. Through models, the degree of matching between the carrying capacity of each regional node and the actual traffic demand is quantitatively analyzed to identify nodes with supply and demand imbalances (such as overloaded waiting areas and idle public spaces). Traffic diversion optimization suggestions are based on the distribution of hotspot areas, the situation of supply and demand imbalances, and spatiotemporal evolution patterns. The model generates targeted adjustment plans (such as optimizing the flow routes of people within the area, adjusting the opening hours of facilities, adding temporary diversion channels, and coordinating public transportation scheduling). By supplementing the above three types of core information, the regional traffic analysis report can not only reflect the current traffic status and predict future trends, but also provide regional managers with accurate and feasible optimization decision-making basis, fully supporting the refined management needs of scenarios such as commercial district operation optimization, transportation hub scheduling, and public area safety management.

[0039] Please refer to Figure 3 , Figure 3 This is a structural block diagram of a regional pedestrian flow analysis system based on intelligent agent map data provided in an embodiment of this application.

[0040] A second aspect of the present invention also discloses a regional pedestrian flow analysis system based on intelligent agent map data, comprising: The data acquisition and preprocessing module 301 is used to collect personnel movement trajectory data, regional node status data and environmental perception data within a preset area, and output standardized multi-source data through data cleaning, format standardization and noise filtering. The initial agent graph construction module 302 is used to define personnel agents, regional node agents and association types based on the standardized multi-source data, and construct an initial agent graph containing agent attribute information and topological associations. The agent graph incremental update module 303 is used to dynamically adjust the agent attribute information and association weights in the initial agent graph based on the real-time collected data stream using an incremental update mechanism, and output an optimized agent graph. The graph embedding and model building module 304 is used to map the optimized agent graph to a low-dimensional feature space using a graph embedding algorithm, to obtain the agent's low-dimensional feature vector and build a spatiotemporal attention model. The analysis report generation module 305 is used to output a regional pedestrian flow analysis report based on the spatiotemporal attention model by capturing the temporal evolution pattern of pedestrian flow and mining the influence of topological associations between intelligent agents on the flow distribution.

[0041] The system also includes a memory and a processor. The memory contains a program for a regional pedestrian flow analysis method based on agent mapping data. When the program for the regional pedestrian flow analysis method based on agent mapping data is executed by the processor, it implements the steps of the regional pedestrian flow analysis method based on agent mapping data as described in any one of the first aspects.

[0042] This invention addresses the technical pain points of existing regional pedestrian flow analysis methods, such as single data source, weak dynamic adaptation capability, and insufficient spatiotemporal feature fusion. It provides a regional pedestrian flow analysis method and corresponding system based on intelligent agent graph data, aiming to achieve comprehensive, real-time, and accurate analysis of regional pedestrian flow and provide scientific decision support for regional management.This method first comprehensively collects personnel movement trajectory data (including terminal positioning data, public transportation travel data, and vehicle trajectory data), regional node status data (including regional node capacity data and facility operation status data), and environmental perception data (including meteorological data, public event data, and traffic control data) within a preset area through multi-source data acquisition terminals and interfaces, constructing a multi-dimensional, full-scenario data source. Subsequently, systematic preprocessing is performed on the collected multi-source data, employing missing value imputation and outlier removal algorithms for data cleaning, standardizing the format through unified data protocol conversion, and filtering noise using sliding window smoothing filtering or wavelet filtering algorithms to obtain reliable data. Standardized multi-source data with a unified format lays the foundation for subsequent modeling. Based on this standardized multi-source data, personnel feature data, regional node attribute data, and spatial correlation data are extracted. Personnel agents (representing individuals or groups of people within a region) and regional node agents (representing key locations within a region) are defined, and the types of correlations are clarified (including personnel and node access relationships such as temporary access relationships and fixed residence relationships, and node-to-node spatial correlation relationships including adjacency relationships, path relationships, and functional relationships). An initial agent graph containing agent attribute information and topological correlations is constructed to achieve abstract modeling of complex correlations between personnel and regional nodes. To adapt to the dynamic changes in personnel flow, the foundation... Based on the real-time data stream, an incremental update mechanism with a sliding time window trigger strategy is adopted. Agents are added or removed according to a preset dwell time threshold at regional nodes. The strength of associations is adjusted using an exponential decay function based on a time decay coefficient. The topology is optimized based on a dynamic threshold of association density, dynamically adjusting agent attribute information and association weights in the initial agent graph to obtain an optimized agent graph that reflects the real-time flow of people. Then, one or more graph embedding algorithms, such as Graph Convolutional Network (GCN) and Graph Attention Network (GAT), are used to map the high-dimensional and complex optimized agent graph to a low-dimensional feature space, generating a graph that retains core features. The system generates low-dimensional feature vectors for intelligent agents and constructs a spatiotemporal attention model based on these vectors, including temporal and spatial attention layers. The model captures the temporal dependencies of pedestrian traffic at different time steps through a self-attention-based temporal attention layer, and mines the influence of topological relationships through a spatial attention layer that dynamically adjusts weights based on topological distance between agents. Simultaneously, it integrates feature vectors corresponding to environmental perception data to achieve deep collaborative analysis of spatiotemporal features and external environmental features. Finally, it generates a regional pedestrian traffic analysis report containing real-time traffic distribution data, traffic prediction curves, abnormal traffic warning information, traffic hotspot area identification results, regional traffic supply and demand balance assessment results, and traffic diversion optimization suggestions.The corresponding system, through the collaborative work of data acquisition and preprocessing modules, initial agent graph construction modules, incremental agent graph update modules, graph embedding and model construction modules, and analysis report generation modules, combined with hardware support from memory and processors, ensures the efficient implementation of the aforementioned methods. This invention, through three core technological innovations—multi-source data fusion, dynamic modeling of agent graphs, and spatiotemporal attention feature fusion—effectively overcomes the limitations of traditional methods, improving the real-time performance, comprehensiveness, and accuracy of regional pedestrian flow analysis. It can be widely applied to scenarios such as business district operation, transportation hub scheduling, and public area management, providing strong support for refined management and scientific decision-making.

[0043] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be electrical, mechanical, or other forms.

[0044] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units. They may be located in one place or distributed across multiple network units. Some or all of the units may be selected to achieve the purpose of this embodiment according to actual needs.

[0045] In addition, in the various embodiments of the present invention, each functional unit can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

[0046] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media that can store program code, such as mobile storage devices, read-only memory, random access memory, magnetic disks, or optical disks.

[0047] Alternatively, if the integrated units of this invention are implemented as software functional modules and sold or used as independent products, they can also be stored in a readable storage medium. Based on this understanding, the technical solutions of the embodiments of this invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROM, RAM, magnetic disks, or optical disks.

Claims

1. A method for analyzing regional pedestrian flow based on intelligent agent map data, characterized in that, Includes the following steps: Collect personnel movement trajectory data, regional node status data, and environmental perception data within a preset area. Through data cleaning, format standardization, and noise filtering, standardized multi-source data is obtained. Based on the standardized multi-source data, human intelligent agents, regional node intelligent agents and their relationship types are defined, and an initial intelligent agent graph containing intelligent agent attribute information and topological relationship is constructed. Based on the real-time data stream, an incremental update mechanism is used to dynamically adjust the agent attribute information and association weights in the initial agent graph to obtain an optimized agent graph. The optimized agent graph is mapped to a low-dimensional feature space using a graph embedding algorithm to obtain the agent's low-dimensional feature vector and construct a spatiotemporal attention model. Based on the spatiotemporal attention model, a regional pedestrian flow analysis report is obtained by capturing the temporal evolution of pedestrian flow and mining the influence of topological relationships between agents on pedestrian flow distribution.

2. The regional pedestrian flow analysis method based on intelligent agent map data according to claim 1, characterized in that, The data collected includes personnel movement trajectory data, area node status data, and environmental perception data within the preset area, including: Collect personnel movement trajectory data, regional node status data, and environmental perception data within the preset area; The personnel movement trajectory data includes terminal positioning data, public transportation travel data, and travel vehicle trajectory data; The regional node status data includes regional node capacity data and facility operation status data; The environmental perception data includes meteorological data, public event data, and traffic control data.

3. The regional pedestrian flow analysis method based on intelligent agent map data according to claim 1, characterized in that, The process of data cleaning, format standardization, and noise filtering yields standardized multi-source data, specifically: Based on the collected personnel movement trajectory data, regional node status data, and environmental perception data, data cleaning is performed using missing value filling algorithm and outlier removal algorithm; The unified data protocol is used to convert and process standardized execution formats. Noise filtering is performed using a sliding window smoothing filter algorithm or a wavelet filter algorithm to obtain standardized multi-source data.

4. The method for regional pedestrian flow analysis based on intelligent agent map data according to claim 1, characterized in that, Based on the standardized multi-source data, the definition of personnel intelligent agents, regional node intelligent agents, and association relationship types includes: Based on the personnel feature data, regional node attribute data and spatial association data in the standardized multi-source data, define personnel intelligent agents and regional node intelligent agents, and define the association relationship types. The types of association relationships include personnel-node access relationships and node-node spatial association relationships; The personnel and node access relationships include temporary access relationships and permanent residency relationships; The spatial relationships between nodes include adjacency relationships, path relationships, and functional relationships.

5. The method for regional pedestrian flow analysis based on intelligent agent map data according to claim 1, characterized in that, The data stream, based on real-time acquisition, uses an incremental update mechanism to dynamically adjust the agent attribute information and association weights in the initial agent graph to obtain an optimized agent graph, including: Based on the real-time data stream, an incremental update mechanism with a sliding time window triggering strategy is adopted to add or remove intelligent agents according to the preset dwell time threshold of personnel in the area node. The strength of the correlation is adjusted by an exponential decay function based on the time decay coefficient; The topology is optimized based on a dynamic threshold of association density, and the agent attribute information and association weights are dynamically adjusted to obtain an optimized agent graph.

6. The method for regional pedestrian flow analysis based on intelligent agent map data according to claim 5, characterized in that, The process involves using a graph embedding algorithm to map the optimized agent graph to a low-dimensional feature space, obtaining the agent's low-dimensional feature vector, and constructing a spatiotemporal attention model. Specifically: Based on the optimized agent graph, a preset algorithm is used as the graph embedding algorithm; The optimized agent graph is mapped to a low-dimensional feature space to obtain the agent's low-dimensional feature vector; A spatiotemporal attention model is constructed based on the low-dimensional feature vectors of the agent.

7. The method for regional pedestrian flow analysis based on intelligent agent map data according to claim 1, characterized in that, Based on the spatiotemporal attention model, a regional pedestrian flow analysis report is obtained by capturing the temporal evolution pattern of pedestrian flow and mining the impact of topological relationships between agents on pedestrian flow distribution. Specifically: The spatiotemporal attention model includes a temporal attention layer and a spatial attention layer; Based on the low-dimensional feature vector of the agent, the temporal dependence of pedestrian flow at different time steps is captured by a time attention layer based on a self-attention mechanism. By mining the influence of topological correlations through a spatial attention layer that dynamically adjusts weights based on topological distance between agents, and by fusing feature vectors corresponding to environmental perception data, a regional pedestrian flow analysis report is obtained.

8. The method for regional pedestrian flow analysis based on intelligent agent map data according to claim 7, characterized in that, The regional pedestrian flow analysis report also includes: Based on the results of traffic hotspot identification, regional traffic supply and demand balance assessment, and traffic diversion optimization suggestions output by the spatiotemporal attention model.

9. A regional pedestrian flow analysis system based on intelligent agent map data, characterized in that, include: The data acquisition and preprocessing module is used to collect personnel movement trajectory data, regional node status data, and environmental perception data within a preset area. Through data cleaning, format standardization, and noise filtering, it outputs standardized multi-source data. The initial agent graph construction module is used to define personnel agents, regional node agents and association types based on the standardized multi-source data, and construct an initial agent graph containing agent attribute information and topological associations. The agent graph incremental update module is used to dynamically adjust the agent attribute information and association weights in the initial agent graph based on the real-time collected data stream using an incremental update mechanism, and output an optimized agent graph. The graph embedding and model building module is used to map the optimized agent graph to a low-dimensional feature space using a graph embedding algorithm, to obtain the agent's low-dimensional feature vector and build a spatiotemporal attention model. The analysis report generation module is used to output a regional pedestrian flow analysis report based on the spatiotemporal attention model by capturing the temporal evolution pattern of pedestrian flow and mining the impact of topological associations between agents on the flow distribution.

10. A regional pedestrian flow analysis system based on intelligent agent map data, characterized in that, The system also includes a memory and a processor. The memory contains a program for a regional pedestrian flow analysis method based on intelligent agent map data. When the processor executes the program for the regional pedestrian flow analysis method based on intelligent agent map data, it performs the following steps: Collect personnel movement trajectory data, regional node status data, and environmental perception data within a preset area. Through data cleaning, format standardization, and noise filtering, standardized multi-source data is obtained. Based on the standardized multi-source data, human intelligent agents, regional node intelligent agents and their relationship types are defined, and an initial intelligent agent graph containing intelligent agent attribute information and topological relationship is constructed. Based on the real-time data stream, an incremental update mechanism is used to dynamically adjust the agent attribute information and association weights in the initial agent graph to obtain an optimized agent graph. The optimized agent graph is mapped to a low-dimensional feature space using a graph embedding algorithm to obtain the agent's low-dimensional feature vector and construct a spatiotemporal attention model. Based on the spatiotemporal attention model, a regional pedestrian flow analysis report is obtained by capturing the temporal evolution of pedestrian flow and mining the influence of topological relationships between agents on pedestrian flow distribution.

Images