Real-time map updating method and system based on multi-source geographic information data fusion

Abstract

The invention provides a real-time map updating method and system based on multi-source geographic information data fusion. Wherein a three-dimensional model is constructed by integrating traffic flow data, remote sensing images and road network topology, the road traffic pressure is quantified, and a feature correlation index is established. And obtaining a vehicle displacement vector and a speed gradient in combination with laser point cloud and video monitoring data, generating a dynamic road feature data set, and establishing a mapping relationship with the index table. On the basis, a road state prediction model is constructed, and through correlation analysis of traffic pressure data and a real-time feature data set, deep correlation between vehicle motion features and road states is established. And dynamically adjusting a road network connection structure, synchronously updating mapping parameters and predicting model precision, and realizing real-time adaptive adjustment of the road network topology. According to the technical scheme provided by the invention, the real-time performance of map updating and the road condition prediction accuracy are remarkably improved, and the urban traffic congestion index can be reduced.

Description

Technical Field

[0001] The present application relates to the technical field of intelligent transportation geographic information systems, and in particular to a real-time map updating method and system based on multi-source geographic information data fusion. Background Art

[0002] In areas such as autonomous driving, smart cities, and logistics route planning, the real-time and accuracy of geographic information data directly impacts the efficiency and safety of system decisions. With the explosive growth of multi-source data, there is an urgent need for a technology that can integrate heterogeneous data from multiple sources, achieve dynamic updates of map elements within seconds, and meet the requirements of high accuracy, low latency, and strong robustness. Traditional solutions based on manual collection or driven by a single data source are no longer adaptable to the high-frequency changes in real-world scenarios.

[0003] One currently targeted technical solution is an automated update method based on the collaborative analysis of crowdsourced trajectories and satellite imagery. This approach uses real-time trajectory data uploaded by devices such as vehicle sensors and mobile phone GPS, combined with change detection algorithms from satellite imagery, to identify changes in road morphology, obstacles, and so on. This solution has been partially automated and is highly effective in addressing changes in road topology.

[0004] However, while these solutions reduce manual intervention, their core challenges lie in the reliability and computational efficiency of multi-source data fusion. First, the quality of crowdsourced data is highly dependent on device accuracy and user distribution. Low-precision GPS signals or sparse trajectories can easily lead to false detections. Second, satellite imagery is subject to cloud cover and revisit cycles, making it difficult to capture sudden changes. Furthermore, the rule engine relies on preset logic and cannot adapt to complex scenarios. This means that updated results still require manual secondary verification, making it difficult to achieve a fully real-time closed-loop system. Summary of the Invention

[0005] The present application provides a real-time map updating method and system based on multi-source geographic information data fusion, which is used to solve the problems of low reliability and computational efficiency of multi-source data fusion in the prior art.

[0006] In a first aspect, the present application provides a real-time map updating method based on multi-source geographic information data fusion, comprising: Collecting traffic flow data, combining remote sensing image data with road network topology data, and constructing a three-dimensional model. Using the three-dimensional model, quantify traffic pressure data of road areas in different time periods to establish a road feature association index table. Obtaining vehicle displacement vectors and velocity gradient changes at key road nodes, synchronously fusing laser point cloud data with video surveillance data to generate a road feature dataset, and establishing a first mapping relationship between the road feature dataset and the road feature association index table; Based on the first mapping relationship, a road state prediction model is constructed to generate prediction data of a road traffic state change trend through the road state prediction model; Performing correlation analysis on the traffic pressure data in the three-dimensional model and the road feature dataset, and establishing a second mapping relationship between the road traffic state and the vehicle motion characteristics based on the analysis results and the predicted data of the road traffic state change trend; Based on the second mapping relationship, the road network connection structure in the road network topology data is adjusted, the associated parameters of the first mapping relationship are iteratively updated through the adjusted road network connection structure, and the updated associated parameters are simultaneously used to optimize the prediction accuracy of the road state prediction model to achieve dynamic optimization of real-time map data.

[0007] Optionally, constructing an input feature vector of a road state prediction model based on a displacement vector change rate in a road feature dataset extracted from the first mapping relationship and a traffic pressure threshold in a road feature association index table; By dividing the input feature vector into time windows, a dynamic weight matching rule between the displacement vector change rate and the traffic pressure threshold in different time periods is established to generate a feature weight allocation table for a road state prediction model; Dynamically binding the feature weight allocation table to the real-time update frequency of the road feature dataset, and incrementally correcting the dynamic matching coefficient in the feature weight allocation table based on the real-time update frequency of the road feature dataset; The revised feature weight distribution table is input into the road state prediction model, and the model is used to perform spatiotemporal coupling calculations on the associated features of the real-time update frequency and the dynamic matching coefficient, and outputs prediction data including the traffic pressure fluctuation range and the displacement vector change trend.

[0008] Optionally, the traffic pressure data of the three-dimensional model is correlated with the road feature dataset in a temporal and spatial dimension, and a real-time correlation between the traffic pressure data and the road feature dataset is calculated to generate an analysis result reflecting a dynamic correlation between pressure intensity and motion parameters; Based on the real-time correlation between the traffic pressure data and the road characteristic data set in the analysis results, the road traffic status level is divided according to the preset pressure threshold range; Combined with the predicted data of the road traffic status level change trend, the vehicle displacement vector distribution characteristics and speed gradient change range corresponding to each state level are dynamically associated to generate a vehicle motion feature mapping rule based on the traffic pressure data to establish a second mapping relationship between the road traffic status and the vehicle motion feature.

[0009] Optionally, dividing the traffic pressure intervals based on the second mapping relationship, and extracting the pressure adaptation scale values of the road nodes within the traffic pressure intervals; converting the real-time traffic pressure data into a real-time pressure parameter having the same dimension as the pressure adaptation scale value using the parameter mapping rule in the second mapping relationship, and adjusting the road network connection structure in the road network topology data according to a deviation between the pressure adaptation scale value and the real-time pressure parameter; Inputting the fluctuation frequency of the real-time pressure parameter of the key road nodes in the adjusted road network connection structure into the parameter correction module of the first mapping relationship, and iteratively updating the associated parameters of the first mapping relationship based on the weight ratio of the displacement vector change rate and the traffic pressure threshold through a dynamic balance mechanism; Based on the updated associated parameters, scale deviation data of the adjusted road network is generated, and the input feature weights of the road state prediction model are adjusted according to the weight difference between the scale deviation data and the feature weight allocation table to optimize the prediction accuracy of the road state prediction model.

[0010] Optionally, based on the displacement vector change rate weight ratio and the traffic pressure threshold weight ratio in the updated associated parameters, extract the real-time pressure parameters of each node in the adjusted road network; Comparing the real-time pressure parameter with the pressure adaptation scale value of the corresponding node using the deviation calculation rule in the second mapping relationship to generate scale deviation data reflecting the degree of dynamic matching between the traffic pressure of the road network node and the vehicle displacement vector; Extracting the weight distribution ratio associated with the adjusted road network node in the feature weight distribution table, and calculating the dynamic weight difference between the weight distribution ratio and the scale deviation data according to the distribution range of the scale deviation data; Based on the positive and negative directions of the dynamic weight difference, the increase or decrease ratio of the input feature weight in the road state prediction model is adjusted to optimize the prediction accuracy of the road state prediction model.

[0011] Optionally, based on the time series distribution of the displacement vector change rate in the input feature vector, the time series distribution is divided into multiple consecutive time periods according to a preset time window division rule, and the fluctuation amplitude of the displacement vector change rate and the fluctuation range of the traffic pressure threshold in each time period are extracted; According to the correlation between the fluctuation amplitude and the fluctuation range, the dynamic impact intensity of the displacement vector change rate on the traffic pressure threshold in different time periods is analyzed to generate a correlation mapping table of time periods, fluctuation amplitudes and impact intensity; Traversing the time periods in the association mapping table, dynamically analyzing the displacement vector change rate and the traffic pressure threshold within each time period to identify abnormal time periods in which the displacement vector change rate exceeds a preset fluctuation threshold; During the abnormal time period, dynamically adjusting the weight parameter of the displacement vector change rate according to the fluctuation range of the traffic pressure threshold, and generating a weight allocation rule set indexed by the time period; Each time parameter in the weight allocation rule set is combined with the real-time update frequency of the corresponding time period to generate a feature weight allocation table.

[0012] Optionally, the dynamic vehicle distribution information in the traffic flow data is fused with the remote sensing image data, and the initial outline of the road geometry is extracted from the fused data; According to the connection relationship and hierarchical structure of road nodes in the road network topology data, the initial outline is spatially reconstructed to generate a three-dimensional model framework with multi-dimensional connection attributes; Mapping the dynamic vehicle distribution information into the three-dimensional model framework to form a three-dimensional model that includes road geometry and real-time traffic flow changes; Dividing the road areas in the three-dimensional model according to time series, and calculating the traffic pressure data of the corresponding time period by the dynamic distribution parameters of the traffic flow data in each road area; The traffic pressure data is bound to the road area features in the road network topology data to form an associated index table with spatiotemporal coordinates and road attributes as indexes.

[0013] In a second aspect, the present application provides a real-time map update system based on multi-source geographic information data fusion, comprising: An acquisition module collects traffic flow data, combines remote sensing image data with road network topology data, and constructs a three-dimensional model. The three-dimensional model is used to quantify traffic pressure data of road areas in different time periods to establish a road feature association index table. a mapping module that obtains vehicle displacement vectors and velocity gradient changes at key road nodes, synchronously fuses laser point cloud data with video surveillance data, generates a road feature dataset, and establishes a first mapping relationship between the road feature dataset and the road feature association index table; A construction module, based on the first mapping relationship, constructs a road state prediction model to generate prediction data of a road traffic state change trend through the road state prediction model; an analysis module that performs correlation analysis on the traffic pressure data in the three-dimensional model and the road characteristic dataset, and establishes a second mapping relationship between the road traffic state and the vehicle motion characteristics based on the analysis result and the predicted data of the road traffic state change trend; An updating module adjusts the road network connection structure in the road network topology data based on the second mapping relationship, iteratively updates the associated parameters of the first mapping relationship through the adjusted road network connection structure, and simultaneously optimizes the prediction accuracy of the road state prediction model using the updated associated parameters to achieve dynamic optimization of real-time map data.

[0014] In a third aspect, an embodiment of the present application provides a computing device comprising a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a real-time map update method based on multi-source geographic information data fusion as described in the first aspect above.

[0015] In a fourth aspect, an embodiment of the present application provides a computer storage medium storing a computer program. When the computer program is executed by a computer, it implements a real-time map update method based on multi-source geographic information data fusion as described in the first aspect.

[0016] In an embodiment of the present application, traffic flow data is collected, and remote sensing image data and road network topology data are combined to construct a three-dimensional model. The three-dimensional model is used to quantify the traffic pressure data of road areas in different time periods to establish a road feature association index table; vehicle displacement vectors and velocity gradient changes at key road nodes are obtained, laser point cloud data and video surveillance data are synchronously integrated to generate a road feature data set, and a first mapping relationship between the road feature data set and the road feature association index table is established; based on the first mapping relationship, a road state prediction model is constructed to generate prediction data of road traffic state change trends through the road state prediction model; the traffic pressure data in the three-dimensional model is associated with the road feature data set, and based on the analysis results and the prediction data of the road traffic state change trends, a second mapping relationship between road traffic state and vehicle motion characteristics is established; based on the second mapping relationship, the road network connection structure in the road network topology data is adjusted, and the association parameters of the first mapping relationship are iteratively updated, and the prediction accuracy of the road state prediction model is optimized to achieve dynamic optimization of real-time map data.

[0017] The embodiments of the present application have the following beneficial effects: The spatiotemporal distribution of road traffic pressure is quantified through three-dimensional models to support dynamic load assessment and visual decision-making of road networks; multi-source data are integrated to construct a high-precision road feature dataset to improve the real-time perception of dynamic features of key nodes and the ability to mutually verify multi-source data; short-term prediction trends of road conditions are generated based on historical and real-time data to enhance the timeliness of prediction of emergencies such as abnormal congestion and accidents; a coupled logical relationship between vehicle motion characteristics and traffic status is established to provide a causal reasoning basis for dynamic path planning; through data closed-loop driven dynamic optimization of road network topology and model iteration, the bottleneck of static map update delay is broken through to achieve multi-dimensional and high-frequency real-time mapping of road traffic status and map self-optimization.

[0018] Furthermore, the method constructs input feature vectors by extracting the displacement vector change rate and traffic pressure threshold from the road feature dataset. Combined with time window partitioning, a dynamic weight matching rule is established to generate a feature weight allocation table. This is then dynamically bound to the real-time update frequency and the matching coefficients are incrementally corrected. Finally, through spatiotemporal coupling, the model outputs forecast data containing the traffic pressure fluctuation range and displacement vector change trends. This dynamic weight allocation and real-time data binding mechanism enables the model to adapt to the spatiotemporal correlation characteristics of road conditions at different time periods, significantly improving the accuracy of the forecast data in capturing sudden changes in traffic pressure (such as congestion formation) and displacement anomalies (such as vehicle detours). Furthermore, incremental corrections are used to reduce the lag of historical data, achieving high-response and low-error short-term road condition trend forecasts.

[0019] These and other aspects of the present application will become more readily apparent from the description of the following embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0021] Figure 1 A flowchart of a real-time map updating method based on multi-source geographic information data fusion provided by the present application is shown; Figure 2 The present invention provides a structural diagram of a real-time map updating system based on multi-source geographic information data fusion; Figure 3 A schematic structural diagram of a computing device provided by the present application is shown. DETAILED DESCRIPTION

[0022] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0023] In some of the processes described in the specification and claims of this application and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this document or may be executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish between different operations, and the serial numbers themselves do not represent any order of execution. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this document are used to distinguish different messages, devices, modules, etc., and do not represent a sequential order, nor do they limit "first" and "second" to being different types.

[0024] Researchers have found that existing real-time map update methods have problems with low spatiotemporal alignment accuracy and insufficient dynamic feature coupling modeling when fusing multi-source heterogeneous data. This results in delayed road traffic status prediction and reliance on manual experience for road network topology adjustments, making it difficult to meet the needs of high-real-time scenarios such as autonomous driving. Based on this, this application provides a real-time map update method based on the fusion of multi-source geographic information data. This method can quantify traffic pressure distribution through a three-dimensional model, construct a dual mapping relationship to achieve dynamic association between vehicle motion characteristics and road network status, and drive self-optimization of road network connection structure based on a data closed loop, ultimately achieving second-level updates of map data and reducing prediction errors.

[0025] The technical solution of this application can be applied to dynamic path planning for autonomous vehicles, smart city logistics scheduling, and emergency traffic management scenarios. For example, in logistics distribution, real-time updates of traffic pressure and predicted trends can be used to dynamically avoid unexpected congested sections of road. In traffic accident handling, the fusion analysis of laser point cloud and video data can quickly reconstruct the road network topology in the accident area and optimize the detour routes for emergency vehicles.

[0026] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making creative efforts are within the scope of protection of this application.

[0027] Figure 1 The present invention provides a flowchart of a real-time map updating method based on multi-source geographic information data fusion, as shown in FIG. Figure 1 As shown, the method includes: 101. Collect traffic flow data, combine remote sensing image data with road network topology data, build a three-dimensional model, and quantify traffic pressure data of road areas in different time periods using the three-dimensional model to establish a road feature association index table; In this step, traffic flow data refers to real-time traffic parameters such as traffic volume, average speed, lane occupancy, etc. collected by sensors.

[0028] Remote sensing image data is high-resolution surface images (such as 0.5-meter resolution) taken by satellites or aerial vehicles, which are used to extract road morphology and surrounding environment.

[0029] Road network topology data describes road connection relationships and attributes (such as the number of lanes and speed limits) in a graph structure (nodes represent intersections and edges represent road sections).

[0030] The three-dimensional model is a three-dimensional road network model built on the GIS platform, superimposing elevation data and traffic flow dynamic parameters.

[0031] Traffic pressure data is a congestion index (0-1 range) calculated by traffic density (number of vehicles / km) and speed decay rate.

[0032] The road feature association index table is a relational database table that establishes a relationship between the road ID and its geometric features (curvature, slope) and traffic pressure value. In the embodiment of the present application, traffic flow data (including traffic volume, average speed, and lane occupancy) is first collected through traffic monitoring sensors (such as geomagnetic coils and cameras), and high-resolution remote sensing image data (such as satellite or drone images) and road network topology data (including road node coordinates and road section connection relationships) are simultaneously obtained. Secondly, three-dimensional modeling software (such as Blender or GIS platform) is used to fuse remote sensing images with road network topology data to construct a three-dimensional model with elevation information (such as intersection structure and bridge slope). Then, based on the traffic flow data, the traffic pressure index of each road area is calculated. The formula is: Traffic pressure index = real-time traffic volume / road design capacity × speed attenuation coefficient (For example, if the design capacity of a road section is 1,000 vehicles per hour, the real-time traffic volume is 800 vehicles, and the speed drops from 60 km / h to 40 km / h, the attenuation coefficient = 40 / 60 ≈ 0.67, and the traffic pressure index = 800 / 1,000 × 0.67 ≈ 0.536.) Finally, the 3D coordinates, traffic pressure index, and time tags of each road section are integrated into a road feature association index table (e.g., fields: section ID, 3D coordinates, pressure index, time window).

[0033] In a dynamic optimization system for intelligent urban transportation, urban traffic management departments deploy multi-source sensing equipment in core urban areas. Using geomagnetic vehicle detectors, they collect real-time, multi-dimensional traffic flow data on major arterial roads, integrating high-resolution satellite remote sensing imagery with open-source road network topology data. Using BIM and 3D GIS technology, they construct a dynamic traffic model, integrating traffic characteristic data from different time periods. The model quantitatively calculates a peak traffic pressure index of 85.3 (normalized on a 0-100 scale) on roads surrounding commercial areas during morning and evening rush hours. The system automatically generates a road characteristic correlation index table containing 20 characteristic parameters, including spatiotemporal pressure distribution and lane-level load factor, to form a traffic status analysis framework.

[0034] 102. Obtain vehicle displacement vectors and velocity gradient changes at key road nodes, synchronously fuse laser point cloud data with video surveillance data to generate a road feature dataset, and establish a first mapping relationship between the road feature dataset and the road feature association index table; In this step, key road nodes are hubs where traffic flows converge or diverge (such as overpass entrances and main road intersections).

[0035] The vehicle displacement vector is the vector data consisting of the vehicle's moving direction (azimuth) and displacement (meters) within the time Δt.

[0036] The speed gradient is the derivative of the vehicle speed change per unit time (e.g. the gradient from 60 km / h to 30 km / h is -30 km / h²).

[0037] Laser point cloud data is a set of three-dimensional coordinates of the road surface obtained by LiDAR equipment (accuracy ±2 cm), which is used to extract lane lines and obstacle information.

[0038] The video surveillance data is real-time traffic flow video (H.264 encoding) captured by the camera, and the vehicle trajectories are extracted through the target detection algorithm.

[0039] The road feature dataset is a structured collection that integrates the above data (including fields such as vectors, point clouds, and video features). The first mapping relationship is a hash mapping structure that associates the data set with the index table through the road ID and the timestamp.

[0040] In this embodiment, laser point cloud data (containing vehicle 3D coordinates and profile information) is first collected at key road nodes (such as intersections and ramps) using LiDAR. Simultaneously, vehicle displacement vectors (e.g., vehicle A moves from coordinates [x1, y1] to [x2, y2] within time Δt) and velocity gradient changes (e.g., acceleration drops from 2 m / s² to -1 m / s²) are extracted through video surveillance. Next, a point cloud registration algorithm (e.g., ICP) is used to spatially align the laser point cloud with the video surveillance data, and a target tracking algorithm (e.g., YOLO+DeepSORT) is used to correlate the displacement and velocity data of the same vehicle. Next, vehicle motion characteristics (e.g., average speed, acceleration standard deviation) and road structural characteristics (e.g., lane width, curvature radius) are extracted to generate a road feature dataset (e.g., fields: node ID, displacement vector, velocity gradient, lane parameters). Finally, a hash table is used to bind the node IDs in the road feature dataset to the road segment IDs in the road feature association index table, establishing a first mapping relationship (e.g., node N1 corresponds to segment RD-001).

[0041] Millimeter-wave radar arrays were deployed at key transportation hubs to capture real-time three-dimensional vehicle trajectory data (with positioning accuracy better than 3 meters) and speed fluctuations. High-precision LiDAR point cloud data (with a density of 280 points / m2) was simultaneously integrated with real-time video streams to construct a road feature dataset encompassing 40 dynamic features, including the frequency of abnormal lane changes and emergency braking hotspots. Using a spatiotemporal feature matching engine, a primary mapping relationship was established between the real-time dataset and an index table. Analysis revealed that the speed gradient decay rate in ramp areas during the weekday evening rush hour was 55% greater than during off-peak hours, revealing the formation patterns of periodic traffic bottlenecks.

[0042] 103. Construct a road state prediction model based on the first mapping relationship, so as to generate prediction data of a road traffic state change trend through the road state prediction model; In this step, the road state prediction model is a time series prediction model built using LSTM (long short-term memory network). The input is the historical traffic pressure and feature data set, and the output is the traffic state (smooth / slow / congested) in the next hour.

[0043] The prediction data is the probability distribution of traffic status with confidence (such as 85%) and the turning point time of change (such as congestion will escalate at 14:30) generated by the model.

[0044] In this embodiment, the road feature dataset (e.g., the velocity gradient at node N1) and the index table traffic pressure data (e.g., the RD-001 pressure index of 0.72) in the first mapping relationship are first input into a spatiotemporal graph convolutional network (ST-GCN) to construct a road state prediction model. The model learns the spatial dependencies between nodes through the graph structure (e.g., congestion at N1 affects downstream N2) and utilizes an LSTM module to capture temporal patterns (e.g., the periodic increase in the pressure index during the morning rush hour). After training, the model outputs the road state change trend for the next hour (e.g., the predicted RD-001 pressure index value of 0.68 → 0.75). The predicted data format is: {road segment ID: RD-001, predicted pressure index: 0.75, time: 09:00-10:00}.

[0045] A deep learning-based spatiotemporal prediction model, incorporating first-order mapping relationship data, accurately predicted during holiday traffic simulations that the scenic area's loop road would experience a critical 58% capacity drop during specific periods. Using a multimodal attention mechanism, the model discovered a strong correlation of 0.84 between parking lot occupancy and arterial road efficiency. The model generated 15-minute granularity forecasts for the next three hours, predicting a 15% increase in traffic queues per period on the westbound arterial road in the core business district, providing decision support for dynamic traffic control.

[0046] 104. Perform correlation analysis on the traffic pressure data in the three-dimensional model and the road feature dataset, and establish a second mapping relationship between the road traffic state and the vehicle motion characteristics based on the analysis results and the predicted data of the road traffic state change trend; In this step, correlation analysis is performed by calculating the linear correlation between traffic pressure and vehicle motion characteristics (such as average acceleration and lane change frequency) through the Pearson correlation coefficient.

[0047] The second mapping relationship is a multidimensional regression equation (such as traffic status = 0.7×traffic density + 0.3×speed gradient) that is established to quantify the impact weight of vehicle behavior on road conditions.

[0048] In this embodiment of the present application, the correlation between traffic pressure data and vehicle motion characteristics is first analyzed using the Pearson correlation coefficient (e.g., when the pressure index is > 0.7, the average vehicle speed decreases by 30%). Next, the predicted data (e.g., RD-001 pressure index 0.75) and real-time vehicle motion characteristics (e.g., speed gradient -4 km / h / s) are input into the random forest model to establish a second mapping relationship (rule example: if the pressure index is > 0.7 and the speed gradient is < -3 km / h / s, then trigger "Forced Lane Change Recommendation"). The resulting mapping relationship table is: {Pressure Interval: [0.7, 1.0], Speed Gradient: <-3, Action: Lane Change Recommendation}.

[0049] By coupling traffic pressure data calculated from the real-time 3D model (current regional peak pressure of 71.6) with a dynamic feature set, an analysis revealed that temporary stops reduce traffic efficiency on specific road sections. Combined with the 45-minute trend output from the prediction model, a second mapping relationship between vehicle movement patterns and road network status was established. The results revealed that when a road section experiences more than six sudden braking events per minute, the probability of a traffic jam in the subsequent 10 minutes exceeds 78%. This pattern has direct application in optimizing signal control system strategies.

[0050] 105. Adjust the road network connection structure in the road network topology data based on the second mapping relationship, iteratively update the associated parameters of the first mapping relationship through the adjusted road network connection structure, and simultaneously optimize the prediction accuracy of the road state prediction model using the updated associated parameters to achieve dynamic optimization of real-time map data.

[0051] In this step, the road network connectivity structure is the adjacency matrix of the road segment connectivity in the topology data.

[0052] Dynamic optimization is to adjust the traffic weight of road sections through reinforcement learning (Q-Learning) (for example, the weight of congested sections is increased by 20%), and feed it back to the index table and prediction model.

[0053] Iterative updating involves retraining model parameters in 5-minute cycles (e.g., the learning rate decays to 0.001) and updating the dataset using a sliding window mechanism.

[0054] The prediction accuracy was optimized by calculating the model F1-score (increased from 0.82 to 0.89) through the confusion matrix and adaptively adjusting the feature weights.

[0055] In an embodiment of the present application, the road network connection structure is first dynamically adjusted according to the second mapping relationship (such as reducing the traffic priority of the high-pressure road section RD-001, and adjusting the weight from 0.5 to 0.3), and a temporary virtual lane is added (such as the weight of the auxiliary road RD-002 is increased from 0.2 to 0.5). Secondly, through online learning technology, real-time data is fed back to the hash table of the first mapping relationship, and the association parameters of the node ID and the road section ID are updated (such as N1 newly associated with the auxiliary road RD-002). At the same time, incremental training is used to optimize the ST-GCN model to improve prediction accuracy (such as the mean square error is reduced from 0.05 to 0.03). Finally, the dynamically optimized real-time map data is output (such as RD-001 in the AutoNavi Map API is displayed as red congestion and RD-002 is green unobstructed).

[0056] Based on dynamic mapping relationships, the system optimizes road network topology and automatically activates the intelligent lane allocation system during peak hours, improving capacity in key corridors. The system updates feature mapping parameters every two minutes through an online learning mechanism and employs a generative adversarial network to optimize the prediction model, minimizing the error rate in 15-minute short-term predictions. A two-way data channel is established with the navigation platform to deliver optimal route solutions in real time, improving the travel time stability index of the core road network and implementing a closed-loop optimization system from data perception to road network regulation.

[0057] In summary, steps 101 to 105 have achieved a three-dimensional dynamic road network modeling and intelligent optimization system driven by multi-source traffic data. Through the deep integration of remote sensing imagery, laser point clouds, video surveillance, and real-time traffic flow, a bidirectional mapping relationship between the spatiotemporal quantitative model of road traffic pressure and vehicle motion characteristics has been established. This system transcends the limitations of traditional static road network modeling, implementing a closed-loop, coordinated optimization mechanism between the road state prediction model and the road network topology. It can dynamically analyze the correlation between traffic flow evolution and road network connectivity efficiency, continuously improve prediction accuracy through real-time data feedback, and ultimately form a dynamic map generation system with self-learning capabilities. This provides collaborative decision-making support for intelligent traffic management, including road state prediction, bottleneck node prediction, and flexible network structure optimization.

[0058] To address the issue of inaccurate feature weight allocation in road traffic state prediction, a road state prediction system with dynamic weight optimization has been developed. This system implements incremental correction of feature weights and spatiotemporal coupling calculations, accurately outputting prediction data on traffic pressure fluctuation ranges and displacement change trends, and significantly improving the accuracy and timeliness of road state predictions. In some embodiments, step 103 constructs a road state prediction model based on the first mapping relationship to generate prediction data on road traffic state change trends using the road state prediction model, including: 201. Construct an input feature vector of a road state prediction model based on the displacement vector change rate in the road feature data set extracted from the first mapping relationship and the traffic pressure threshold in the road feature association index table; In step 201, the first mapping relationship is a hash association structure between the road feature dataset established in step 102 and the road feature association index table; the displacement vector change rate is the rate of change of the direction and modulus of the vehicle displacement vector at the key road node per unit time (such as a change of 15 degrees per second or a speed increase or decrease of 3 m / s²); the traffic pressure threshold is the road area traffic pressure critical value preset in the index table (such as a congestion state when the congestion index is ≥0.7); the input feature vector is a multidimensional array (example dimension: [ΔV, P_th, D]) spliced together by normalizing the displacement vector change rate, the traffic pressure threshold, and other features (such as traffic density), which serves as the input data structure of the road state prediction model.

[0059] In the embodiment of the present application, the displacement vector change rate (e.g., the displacement change rate of the vehicle at node N1 = Δ displacement / Δ t, in units of m / s²) is first extracted from the first mapping relationship (i.e., the node-road segment association table of the road feature dataset and the index table), and the preset traffic pressure threshold value in the road feature association index table is read (e.g., the threshold value of road segment RD-001 = 0.7). Secondly, the displacement vector change rate is normalized (Z-Score normalization), and the traffic pressure threshold value is converted into a Boolean flag (if the real-time pressure index ≥ the threshold value, it is marked as 1, otherwise it is 0). Then, the normalized displacement change rate, Boolean threshold mark, and time window label (e.g., 08:00-09:00) are concatenated into a multi-dimensional input feature vector. The format example is: [displacement change rate: 0.82, threshold mark: 1, time window: 08:00-09:00]. Finally, a structured feature vector set is generated as the input of the road state prediction model.

[0060] 202. By dividing the input feature vector into time windows, a dynamic weight matching rule between the displacement vector change rate and the traffic pressure threshold in different time periods is established to generate a feature weight allocation table for a road state prediction model; In step 202, time window division is to divide the input feature vector into data blocks of continuous time periods according to a fixed time length (such as 15 minutes); the dynamic weight matching rule is to dynamically assign weights based on the correlation (Pearson coefficient) between the displacement vector change rate within the time window and the traffic pressure threshold (for example, the weight is increased to 0.8 when the correlation is high); the feature weight allocation table is a matrix that stores the weight values of each feature in different time periods (such as {timestamp: {feature name: weight}}), which is used to adjust the model's attention to different features.

[0061] In this embodiment of the present application, the input feature vector is first divided into preset time windows (e.g., hourly), and the distribution patterns of the displacement vector change rate (e.g., the displacement change rate of a vehicle at node N1 = Δdisplacement / Δt) and the traffic pressure threshold mark (e.g., whether the real-time pressure index exceeds a threshold of 0.7) within each window are statistically analyzed. The dynamic time warping (DTW) algorithm is used to calculate the correlation between the two in the time series. For example, the DTW correlation coefficient between the displacement change rate and the threshold mark during the morning rush hour (08:00-10:00) is 0.75, which is determined to be highly correlated. This generates a dynamic weight matching rule: the displacement change rate weight is set to 0.6, and the threshold mark weight is set to 0.4. Finally, a feature weight distribution table is generated, recording the weight distribution relationship between different time windows (e.g., the time window 08:00-09:00 corresponds to a displacement weight of 0.6 and a threshold weight of 0.4), providing a basis for differentiated feature importance for the model.

[0062] 203. Dynamically bind the feature weight allocation table to the real-time update frequency of the road feature dataset, and incrementally modify the dynamic matching coefficient in the feature weight allocation table based on the real-time update frequency of the road feature dataset; In step 203, the real-time update frequency is the data refresh rate of the road feature dataset per second or per minute (e.g., 10 new data items per second); dynamic binding is to align the update frequency with the timestamp of the weight distribution table through a message queue (e.g., Kafka) to ensure that the weight distribution takes effect synchronously with the data update; incremental correction is to use an online learning algorithm (e.g., stochastic gradient descent) to perform small iterative optimizations on the dynamic matching coefficients in the weight distribution table (e.g., each weight value is updated by ±0.05) to avoid model mutations.

[0063] In an embodiment of the present application, by monitoring the update frequency of the road feature data set in real time (such as 10 items per second for high-frequency updates and 1 item per second for low-frequency updates), it is mapped to a weight correction coefficient (high-frequency updates +0.1, low-frequency updates -0.1), and the matching coefficient in the feature weight allocation table is dynamically adjusted. For example, in the high-frequency update scenario during the morning rush hour, the original displacement change rate weight of 0.6 is adjusted to 0.7 based on the correction coefficient +0.1, and the threshold mark weight is adjusted from 0.4 to 0.3. The real-time data stream is continuously received through an online learning mechanism (such as random forest incremental training), and the weights are iteratively optimized to ensure that the weight allocation table is adaptively updated as the traffic status changes. For example, when the data update frequency of a certain road section drops sharply due to an accident, the system automatically reduces the displacement weight to 0.5 and increases the threshold weight to 0.5, so as to prioritize relying on static threshold marks for prediction.

[0064] 204. Input the revised feature weight distribution table into the road state prediction model. The model then performs a spatiotemporal coupling calculation on the correlation between the real-time update frequency and the dynamic matching coefficient, outputting predicted data including the pressure fluctuation range and displacement vector trend. In step 204, the spatiotemporal coupling calculation is a fusion of spatial features (such as road morphology) extracted by the model using a convolutional neural network (CNN) and temporal features extracted by a recurrent neural network (RNN). The pressure fluctuation range is the upper and lower limits of the predicted pressure over the future time period (e.g., [0.65, 0.82]), calculated using quantile regression. The displacement vector trend is a curve showing the predicted change in vehicle displacement direction and velocity (e.g., a 20° deflection and a 2 m / s² decrease in velocity over the next 5 minutes), used to assess path conflict risk.

[0065] In this embodiment, the incrementally corrected feature weight distribution table (e.g., displacement weight 0.7, threshold weight 0.3) is input into a spatiotemporal coupled prediction model (ST-GCN+LSTM hybrid model). The model uses a graph convolutional network (GCN) to extract spatial dependencies between nodes (e.g., the impact of congestion on the adjacent road section RD-002 on road section RD-001) and uses LSTM to capture the temporal regularity of the displacement change rate (e.g., an increase of 0.05 in the displacement change rate per hour). By weightedly fusing the spatial weight (displacement weight 0.7) and the temporal weight (threshold weight 0.3), the model outputs the traffic pressure fluctuation range for the future time period (e.g., the pressure index range [0.68, 0.78] for road section RD-001 between 09:00 and 10:00) and the displacement vector change trend (e.g., the displacement change rate increases from 0.85 to 0.90). For example, during the evening rush hour in the urban area, the model predicts that the pressure index of a main road will exceed the threshold of 0.7, and the displacement change rate will decrease rapidly, triggering the navigation system to push detour suggestions in real time, reducing the traffic volume on this section by 15% and bringing the pressure index back to a safe range.

[0066] Here's a specific example: When deploying a road state prediction model on a city's intelligent transportation system on a commercial district ring road (CBD East Ring Road → Convention and Exhibition Center Ramp → Financial City Auxiliary Road → CBD East Ring Road), the system first extracts the displacement vector change rate (e.g., the average speed fluctuation rate ΔV from north to south on the Convention and Exhibition Center ramp is 42%) based on the vehicle GPS trajectory during the morning rush hour (07:00-09:00). Combined with the historical traffic pressure threshold (maximum traffic flow of 1200 vehicles / hour), the system constructs the input feature vector [ΔV, Pressure threshold ratio] = [0.42, 0.85]; a dynamic weight rule is established by dividing the time window into 15 minutes (the displacement vector weight is 0.6 and the pressure threshold weight is 0.4 during the 07:00-07:15 period, and the ratio is adjusted to 0.45:0.55 during the 07:15-07:30 period due to frequent accidents), and a feature weight distribution table is generated; when the real-time traffic data update frequency is accelerated from 5 minutes / time to 1 minute / time (a sudden pile-up collision on the convention center ramp), the system uses incremental learning to modify the displacement vector weight coefficient from 0.45 to 0.68; the model outputs the prediction result through spatiotemporal coupling calculation: the ramp traffic pressure fluctuation range in the next 30 minutes is [135%, 148%], the displacement vector change trend continued to be negative (speed attenuation rate ≥ 15% / 10 minutes), triggering dynamic lane control (tidal lanes extended the traffic period by 40%) and real-time detour recommendations on the navigation platform, which reduced the congestion index in the area during the morning rush hour, improved the match between the model prediction results and the actual road conditions, and reduced traffic delays compared to before optimization.

[0067] In summary, steps 201 to 204 implement an adaptive feature optimization and dynamic weight allocation mechanism for the road state prediction model. By integrating the multidimensional features of displacement vector change rate and traffic pressure threshold, a dynamic weight matching rule with time window sensitivity is constructed. The system triggers incremental coefficient correction based on the real-time data update frequency, forming a dynamic coupling relationship between feature weights and data streams. Relying on a spatiotemporal correlation calculation engine, the system outputs prediction parameters for traffic pressure fluctuation ranges and displacement vector evolution trends. This capability overcomes the limitations of static weight models, enabling dynamic calibration of prediction features with the spatiotemporal evolution of road conditions, providing a quantitative decision-making basis that can dynamically track changes for road network traffic efficiency assessment.

[0068] In order to solve the problem of inaccurate modeling of the association between road traffic status and vehicle motion characteristics, a multi-dimensional dynamically associated road status-vehicle motion mapping system has been developed. This system implements dynamic mapping between traffic status classification and vehicle motion characteristics, providing accurate vehicle behavior prediction capabilities for intelligent traffic control. In some embodiments, as described in step 103, the traffic pressure data in the three-dimensional model is associated with the road feature dataset and analyzed, and based on the analysis results combined with the predicted data of the road traffic status change trend, a second mapping relationship between the road traffic status and vehicle motion characteristics is established, including: 301. Perform correlation analysis on the traffic pressure data of the three-dimensional model and the road characteristic dataset in a temporal and spatial dimension, calculate the real-time correlation between the traffic pressure data and the road characteristic dataset, and generate an analysis result reflecting the dynamic correlation between pressure intensity and motion parameters; In step 301, the traffic pressure data of the three-dimensional model is a dynamic congestion index (e.g., a value in the range of 0-1) calculated based on the three-dimensional road network model constructed in step 101, with superimposed timestamps and spatial coordinates (latitude, longitude, and elevation). The spatiotemporal dimension is a joint analysis dimension that divides the data into time windows (e.g., every 5 minutes) and spatial grids (e.g., 100 meters × 100 meters). The road feature dataset includes displacement vectors, velocity gradients, and lane geometry extracted from laser point clouds. Real-time correlation is the instantaneous correlation between traffic pressure and vehicle motion parameters (e.g., average speed, lane change frequency) calculated using the Pearson correlation coefficient and covariance matrix (e.g., a correlation coefficient ≥ 0.6 indicates a strong correlation). The dynamic correlation between pressure intensity and motion parameters is a quantitative result (e.g., "a 10% decrease in speed corresponds to a 0.2 increase in pressure index") expressed through a multivariate linear regression equation. The analysis results are structured tables that record the grid ID, time window, correlation coefficient, and significance level (p-value).

[0069] In this embodiment, the 3D coordinates and pressure index of each road segment in the 3D model (e.g., a pressure index of 0.72 for section RD-001) are first aligned with the vehicle displacement vectors (e.g., a displacement change rate of 0.85 m / s²) and velocity gradients (e.g., -4 km / h / s) in the road feature dataset according to spatiotemporal labels (e.g., the time window 08:00-09:00, spatial grid G01). The Pearson correlation coefficient algorithm is used to calculate the real-time correlation between the pressure data and the displacement vectors (e.g., the correlation coefficient between the pressure index and the displacement change rate is R = 0.65). Simultaneously, a spatiotemporal graph convolutional network (ST-GCN) is used to explore the dynamic spatial diffusion of pressure intensity and velocity gradients (e.g., a high-pressure section RD-001 causes a 20% decrease in the velocity gradient of the adjacent section RD-002). The resulting analysis results include the pressure-displacement correlation strength (e.g., R = 0.65), the pressure-velocity gradient impact factor (e.g., -0.2), and a spatiotemporal distribution heat map (e.g., grid G01 has the highest correlation).

[0070] 302. Based on the real-time correlation between the traffic pressure data and the road characteristic data set in the analysis result, classify the road traffic status level according to the preset pressure threshold range; In step 302, the preset pressure threshold interval is the traffic pressure level boundary divided according to historical data and expert experience (such as low pressure [0, 0.4], medium pressure (0.4, 0.7], and high pressure (0.7, 1]); the road traffic status level is mapped to a discrete label by the pressure index (such as level 1 unimpeded, level 2 slow traffic, and level 3 congested) through K-means clustering or decision tree classification; the division method adopts a dynamic threshold adjustment mechanism. For example, the high pressure threshold during peak hours in the morning and evening is increased by 0.1 to adapt to traffic changes.

[0071] In the embodiment of the present application, the road traffic status level is divided according to the pressure-displacement correlation coefficient (R value) and the pressure-velocity gradient impact factor in the analysis results, combined with the preset pressure threshold range (e.g., low pressure: R < 0.3, impact factor > -0.1; medium pressure: 0.3 ≤ R < 0.6, -0.3 ≤ impact factor ≤ -0.1; high pressure: R ≥ 0.6, impact factor < -0.3). For example, if a road section has R = 0.65 and impact factor = -0.2, it is determined to be at the "high pressure-moderate speed suppression" level and marked as status level L3. At the same time, the spatiotemporal distribution heat map is regionally aggregated using the K-means clustering algorithm to generate a spatial distribution map of the status level (e.g., the urban core area is L3, and the suburbs are L1).

[0072] 303. Combined with the predicted data of the changing trend of the road traffic status level, the vehicle displacement vector distribution characteristics and the speed gradient change range corresponding to each status level are dynamically associated to generate a vehicle motion feature mapping rule based on the traffic pressure data to establish a second mapping relationship between the road traffic status and the vehicle motion feature.

[0073] In step 303, the trend of road traffic status level change is the future time period state transition probability output by the prediction model in step 103 (e.g., the probability of level 1 → level 2 is 60%). The vehicle displacement vector distribution characteristics include directional concentration (standard deviation ≤ 15° indicates high consistency) and modulus distribution (e.g., mean 50 m / s, variance 10). The speed gradient variation range is the statistically derived acceleration distribution interval (e.g., [-3 m / s², +2 m / s²]). Dynamic association establishes a mapping between status level and motion parameters through a random forest model (e.g., level 3 congestion corresponds to a speed gradient ≤ -1 m / s²). The vehicle motion feature mapping rule is conditional logic encapsulated in an IF-THEN format (e.g., "IF pressure level = 3 THEN allow lane change frequency ≤ 2 times / minute"). The second mapping relationship is a regression equation containing a weight matrix (e.g., status level weight 0.6, motion feature weight 0.4), which is used for real-time scheduling decisions.

[0074] In the embodiment of the present application, first, based on the state level change trend (such as an 80% probability of upgrading from L2 to L3), the vehicle displacement vector distribution of the corresponding level (such as a mean displacement change rate of 0.5 m / s² and a standard deviation of 0.1 at the L3 level) and the speed gradient range (such as -5±1 km / h / s) are extracted. Secondly, the dynamic time warping (DTW) algorithm is used to match the vehicle motion characteristics at similar pressure levels in historical data to generate mapping rules: Rule 1: If the road section is at L3 level, the vehicle displacement change rate threshold is ≤0.6m / s², and the speed gradient threshold is ≥-6 km / h / s; Rule 2: If the road section pressure index increases by 10%, the displacement change rate weight is increased to 0.7, and the speed gradient weight is reduced to 0.3. Finally, a second mapping relationship table is established to dynamically bind the road state (such as L3) with the vehicle motion parameter threshold for the real-time scheduling system to call.

[0075] Here's a specific example: When implementing dynamic traffic flow control on a certain smart highway expansion and reconstruction section (main road G25 → interchange H7 → ramp connection area J3 → main road G25), the system generated a three-dimensional road model through the fusion modeling of lidar and camera, and collected real-time traffic pressure data of interchange H7 (traffic volume 1420 vehicles / hour, average speed 48km / h) and road feature data sets (lane number 4→3 gradient section, curvature radius R=300m). Spatiotemporal correlation analysis showed that the Pearson correlation coefficient between curvature radius and speed gradient reached -0.83, and the traffic pressure intensity was strongly positively correlated with the dispersion of displacement vector (R²=0.91). Based on the preset threshold, the traffic status was divided into low pressure (<60% threshold), medium pressure (60%-85%), and high pressure (>85%). The current west-to-east pressure value of interchange H7 reached 92%, triggering the high pressure level. Combined with the pressure fluctuation range [88%, 96%], dynamically associates the lateral offset threshold of the vehicle displacement vector (±1.2m→±0.8m) and the speed gradient control range (45-60km / h→40-50km / h), and generates a second mapping rule: under high-pressure conditions, the lane keeping assist system is forcibly activated and the lane change frequency is limited to ≤1 time / km, and the electronic speed limit sign is simultaneously adjusted to 50km / h. After execution, the discreteness of the displacement vector in the area decreases, and the congestion index drops from red warning (peak 2.1) to yellow (1.3), traffic efficiency is improved, and the matching degree between the mapping rule and the actual traffic movement characteristics is improved.

[0076] In summary, steps 301 to 303 implement a dynamic coupling analysis system for road traffic status and vehicle motion characteristics. Through spatiotemporal correlation modeling and a state classification mechanism, precise mapping rules for traffic pressure and vehicle behavior are constructed. Based on the multi-dimensional correlation calculation between traffic pressure data and road characteristics, the system converts continuous pressure values into discrete traffic status levels, and dynamically matches the vehicle displacement vector distribution and velocity gradient variation range at each status level in combination with the predicted trend, forming a bidirectional mapping relationship between pressure intensity and motion parameters. This capability opens up the analytical channel between macro-road network status assessment and micro-vehicle behavior patterns, providing a quantitative decision-making basis based on state evolution for traffic flow prediction, congestion cause tracing, and dynamic control strategy generation, ultimately achieving coordinated optimization and regulation of road network traffic efficiency and vehicle motion characteristics.

[0077] To address the issue of insufficient coordination between parameter adjustment and model prediction during dynamic road network optimization, a real-time map dynamic optimization system based on closed-loop feedback has been developed to achieve intelligent iterative updates of map data. In some embodiments, step 105 involves adjusting the road network connection structure in the road network topology data based on the second mapping relationship, iteratively updating the associated parameters of the first mapping relationship using the adjusted road network connection structure, and simultaneously optimizing the prediction accuracy of the road state prediction model using the updated associated parameters to achieve dynamic optimization of real-time map data, including: 401. Divide the traffic pressure intervals based on the second mapping relationship, and extract the pressure adaptation scale values of the road nodes within the traffic pressure intervals; In step 401, the second mapping relationship refers to a quantitative correlation model between road traffic status and vehicle motion characteristics (such as the regression equation traffic pressure = 0.6×traffic density + 0.4×speed gradient); the traffic pressure interval divides the traffic pressure into three levels: low (0-0.4), medium (0.4-0.7), and high (0.7-1), and each interval corresponds to a different traffic management strategy; the pressure adaptation scale value is a normalized parameter (0-1) of the road node's carrying capacity, which is obtained through historical data training (such as a node A scale value of 0.8 indicates that it can carry 80% of the maximum theoretical pressure), and is used to evaluate the real-time load status of the node.

[0078] In the embodiment of the present application, historical traffic pressure data and road node attributes (such as the number of lanes, slope, and design capacity) are first obtained from the second mapping relationship, and the K-means clustering algorithm is used to divide the road state into three traffic pressure intervals: low, medium, and high (for example, the low pressure interval is [0, 0.4], the medium pressure interval is [0.4, 0.7], and the high pressure interval is [0.7, 1.0]). Secondly, for each node in each pressure interval, the pressure adaptation scale value is calculated using a multivariate linear regression model, using the formula: The scaling value is calculated as α × number of lanes + β × slope + γ × design capacity (where α = 0.2, β = -0.1, and γ = 0.3 are regression coefficients obtained through least squares fitting). For example, for a three-lane node (number of lanes = 3), with a 2% slope and a design capacity of 1000 vehicles / hour, the scaling value is 0.2 × 3 + (-0.1) × 2 + 0.3 × 1000 / 1000 = 0.6 - 0.2 + 0.3 = 0.7. Finally, a scaling value table indexed by node ID is generated for subsequent real-time data comparison.

[0079] 402. Convert the real-time traffic pressure data into a real-time pressure parameter having the same dimension as the pressure adaptation scale value using the parameter mapping rule in the second mapping relationship, and adjust the road network connection structure in the road network topology data based on a deviation between the pressure adaptation scale value and the real-time pressure parameter. In step 402, the parameter mapping rule is to convert the real-time traffic pressure data into a parameter with the same dimension as the scale value according to a linear formula (e.g., real-time pressure parameter = original pressure × 1.2); the real-time pressure parameter is the normalized value after mapping and can be directly compared with the scale value; the deviation value is the absolute difference between the real-time pressure parameter and the scale value (e.g., |0.75-0.8|=0.05). If the deviation exceeds a threshold (e.g., 0.1), the road network connection structure adjustment is triggered - dynamically modifying the road section connectivity (e.g., closing secondary sections or adding virtual lanes) through graph theory algorithms (e.g., Dijkstra path weight optimization).

[0080] In this embodiment, real-time traffic pressure data (e.g., a node pressure index of 0.9) is first converted to a parameter with the same dimension as the scale value (e.g., if the scale value is 0.8, the real-time parameter is 0.72) based on the parameter mapping rules preset in the second mapping relationship (e.g., the linear conversion formula: real-time pressure parameter = real-time pressure index × pressure adaptation scale value). Secondly, a deviation calculation module monitors the difference between the scale value and the real-time parameter in real time (deviation value = |0.8-0.72| = 0.08). If the deviation exceeds a threshold (e.g., 0.1), a dynamic adjustment process is triggered: a gradient descent algorithm is used to reduce the node connection weight (e.g., from 0.8 to 0.75, with the step size controlled by a preset learning rate of 0.05), or a Dijkstra algorithm is used to search for the shortest alternative path in the road network topology, inserting virtual lane nodes (e.g., adding a bypass lane) to divert traffic. Finally, the weight table and topological relationship data of the road network connection structure are updated, and the message queue (e.g., Kafka) notifies downstream systems to synchronize and take effect.

[0081] 403. Input the fluctuation frequency of the real-time pressure parameter of the key road node in the adjusted road network connection structure into the parameter correction module of the first mapping relationship, and iteratively update the associated parameters of the first mapping relationship based on the weight ratio of the displacement vector change rate and the traffic pressure threshold through a dynamic balance mechanism; In step 403, the fluctuation frequency is the number of times the node pressure parameter exceeds the scale value in unit time (such as 5 times / hour), reflecting the stability of the node; the parameter correction module is an adaptive unit integrated in the first mapping relationship, and uses a PID controller to adjust the parameters; the dynamic balancing mechanism automatically adjusts the weight ratio of the displacement vector change rate and the traffic pressure threshold according to the fluctuation frequency (such as from 6:4 to 5:5), and the balance factor α=0.01; the iterative update of the associated parameters is to roll over the coefficient matrix of the first mapping relationship according to the time window (such as 10 minutes), and the weights are optimized using the stochastic gradient descent method with momentum (β=0.9).

[0082] In this embodiment, the real-time pressure parameter fluctuation frequency of key nodes in the adjusted road network (e.g., a 15Hz fluctuation frequency at node N1) is first input into a PID controller. This controller then calculates the weighted proportional adjustment between the displacement vector change rate and the pressure threshold. This involves using a proportional term (Kp=0.1) to respond to the current frequency deviation (target frequency 10Hz → 5Hz deviation), an integral term (Ki=0.01) to accumulate historical deviations (e.g., a cumulative deviation of 50Hz·s over 10 seconds), and a differential term (Kd=0.05) to predict trend changes. Ultimately, a dynamic weight adjustment instruction is output (e.g., increasing the displacement weight from 0.6 to 0.65 and decreasing the pressure weight from 0.4 to 0.35). Secondly, based on the updated weight ratios, an incremental online learning algorithm (e.g., a partial fit of a random forest) is used to retrain the associated parameter model of the first mapping relationship. The latest window data (e.g., the latest 1000 feature vectors) is used to adjust the decision tree's splitting rules and node matching coefficients (e.g., adjusting the coefficient for node N1 from 0.8 to 0.82). Finally, the updated parameter table is written to a real-time database (such as Redis) for the next round of prediction calls.

[0083] 404. Generate scale deviation data of the adjusted road network based on the updated associated parameters, and adjust the input feature weights of the road state prediction model by the scale deviation data and the weight difference in the feature weight allocation table to optimize the prediction accuracy of the road state prediction model.

[0084] In step 404, the scale deviation data is an N-dimensional vector (N = number of nodes) that records the difference between the real-time pressure of all nodes in the entire network and the scale value, and the element value is the standardized residual (such as the residual of node B -0.12); the weight difference is the gap between the predicted weight in the feature weight allocation table and the actual demand (such as displacement weight 0.7 → target value 0.6, difference -0.1); the input feature weight adjustment updates the model input layer weight through the back propagation algorithm, and the adjustment amplitude Δw = η × scale deviation × weight difference (η = 0.05); the prediction accuracy optimization requires that the mean square error (MSE) of the pressure prediction value output by the model and the measured value is reduced by at least 15% (such as from 0.25 to 0.18).

[0085] In the embodiment of the present application, the scale deviation data of the adjusted road network is first calculated based on the updated associated parameters (such as the node N1 matching coefficient of 0.82): combining the real-time pressure index (such as 0.75) and the theoretical expected value (scale value × matching coefficient = 0.8 × 0.82 = 0.656), the deviation value Δ = 0.75-0.656 = + 0.094 is obtained. Secondly, the weight difference in the feature weight distribution table is extracted (such as displacement weight 0.65 vs initial weight 0.6, difference + 0.05), and the model input feature weight is adjusted through the back propagation algorithm: the mean square error loss is calculated (such as Loss = (predicted value 0.656-actual value 0.75)² = 0.0084), and the weight gradient ( Loss / The weight distribution is updated using a learning rate of η = 0.01 (e.g., the displacement weight variance is adjusted from 0.1 to 0.08). Finally, if the scaling deviation continues to exceed the threshold (e.g., Δ > 0.1), the network adjustment and parameter optimization cycle of steps 402-403 is retriggered; otherwise, the trained weights are solidified into the model parameter file, completing this round of closed-loop optimization.

[0086] Here's a specific example: During the evening rush hour on a city's expressway network, dynamic closed-loop optimization was triggered: real-time traffic pressure in the west-to-east direction at the N12 interchange hub exceeded the threshold by 21% (1,450 vehicles / hour), and the standard deviation of the lateral displacement vector of the vehicle reached 2.3 meters. The system dynamically adjusted the direct dual lanes of the main ring road M5 to a single lane and activated the auxiliary road R3 for diversion, reducing the pressure to 1,120 vehicles / hour (deviation +6.7%). Simultaneously, the data update frequency was increased to 1 minute / time, and the first mapping relationship weights were iteratively optimized (displacement vector weight 0.55→0.68, pressure threshold weight 0.45→0.32). ), a new dynamic matching coefficient of 0.81 was generated, and the displacement vector standard deviation threshold was inversely compressed to 1.8m based on the pressure deviation matrix (N12 residual -14.3%, Q8 residual +8.2%). After parameter convergence, the road network was adjusted twice, the two-lane system was restored, a dynamic speed limit (60→50km / h) was implemented, and the Q8 tidal lane was expanded. Finally, the N12 pressure was stabilized at 1180 vehicles / hour (3.5% above the threshold), the displacement vector standard deviation was reduced to 1.1m, the model error rate was reduced, the evening peak efficiency was improved, the congestion duration was shortened, and three rounds of closed-loop self-iteration reduced manual intervention.

[0087] In summary, the intelligent road network optimization system implemented through steps 401 to 404 significantly improves traffic management efficiency. The system accurately predicts road congestion and proactively identifies potential bottlenecks, significantly increasing road network efficiency. Through real-time dynamic optimization, average vehicle travel times are shortened, fuel consumption is reduced, and traffic incident response speeds are increased. The system also improves the accuracy of urban traffic control decisions, significantly alleviating peak-hour traffic pressures. This solution provides strong technical support for the development of modern smart cities, enabling a shift in traffic management from reactive response to proactive prevention.

[0088] To address the issue of inaccurate matching between pressure deviation and weight parameters during dynamic road network optimization, an intelligent prediction system based on pressure-weight collaborative optimization has been developed to achieve adaptive optimization of the road state prediction model. In some embodiments, in step 403, the process of generating scale deviation data for the adjusted road network based on the updated associated parameters and adjusting the input feature weights of the road state prediction model based on the scale deviation data and the weight difference in the feature weight allocation table to optimize the prediction accuracy of the road state prediction model includes: 501. Based on the displacement vector change rate weight ratio and the traffic pressure threshold weight ratio in the updated associated parameters, extract the real-time pressure parameters of each node in the adjusted road network; In step 501, the updated correlation parameters refer to the displacement vector change rate weight ratio (e.g., 0.55) and the traffic pressure threshold weight ratio (e.g., 0.45), adjusted by the dynamic balancing mechanism in step 403. Together, these two parameters constitute the core parameters for road network analysis. The real-time pressure parameter is the standardized pressure data collected by sensors in real time (e.g., the pressure value at node C is 0.68). Its dimension is consistent with the pressure adaptation scale value and can be directly used for comparative analysis. Based on the weight ratios of the correlation parameters, the system performs a weighted fusion of the pressure data from each node in the adjusted road network (e.g., displacement weight 0.55 × change rate + pressure weight 0.45 × threshold), generating a node pressure parameter set with spatiotemporal labels (e.g., {node A: 0.72, time: 08:15}).

[0089] In this embodiment of the present application, the displacement vector change rate weight ratio (e.g., 0.65) and the traffic pressure threshold weight ratio (e.g., 0.35) are first obtained from the updated associated parameters. Combined with the real-time collected vehicle displacement change rate (e.g., the average displacement change rate of a node is 0.8 m / s²) and the traffic pressure index (e.g., 0.75), the real-time pressure parameter of each node is calculated using the weighted fusion formula: real-time pressure parameter = displacement change rate × displacement weight + pressure index × pressure weight. For example, the real-time parameter of node N1 = 0.8 × 0.65 + 0.75 × 0.35 = 0.78. The calculation results are stored in a distributed database (e.g., HBase) using spatiotemporal tags (e.g., timestamp 08:00, grid coordinate G01) and pushed to a message queue (Kafka) for subscription by downstream modules. 502. Compare the real-time pressure parameter with the pressure adaptation scale value of the corresponding node using the deviation calculation rule in the second mapping relationship to generate scale deviation data reflecting the degree of dynamic matching between the traffic pressure of the road network node and the vehicle displacement vector; In step 502, the deviation calculation rule is the comparison method defined by the second mapping relationship (such as the absolute deviation formula | real-time pressure parameter - scale value | or percentage deviation (difference / scale value) × 100%); the scale deviation data is a quantitative indicator reflecting the degree of matching between the node pressure and the theoretical bearing capacity (such as node B scale value 0.8, real-time parameter 0.9, absolute deviation 0.1), and the deviation concentration area can be identified through spatial clustering (such as three consecutive nodes on a certain road section with deviations > 0.15); the dynamic matching degree is evaluated by the deviation level classification (low matching [0, 0.1), medium matching [0.1, 0.2), high deviation ≥ 0.2) to provide a basis for weight adjustment.

[0090] In an embodiment of the present application, the pressure adaptation scale value of the node (such as the scale value of node N1 of 0.8) is first loaded from the second mapping relationship, and the deviation value is calculated by the normalized difference formula scale deviation = (real-time pressure parameter - scale value) / scale value. For example, the deviation between the real-time parameter 0.78 of node N1 and the scale value 0.8 = (0.78-0.8) / 0.8=-0.025. Secondly, the matching level is divided based on the preset threshold: the absolute value of the deviation ≤5% is marked as a high match (green), 5%~10% is a medium match (yellow), and >10% is a low match (red). Finally, a scale deviation matrix is generated and written into a time series database (such as InfluxDB) for call by the visualization platform and decision module.

[0091] 503. Extract the weight distribution ratio associated with the adjusted road network node from the feature weight distribution table, and calculate the dynamic weight difference between the weight distribution ratio and the scale deviation data based on the distribution range of the scale deviation data; In step 503, the weight distribution ratio is the displacement and pressure weight values associated with the node in the feature weight allocation table (for example, node D has a displacement weight of 0.6 and a pressure weight of 0.4). The dynamic weight difference is calculated using linear algebra to determine the degree of match between the weight distribution and the scale deviation (formula: difference = weight ratio × scale deviation direction coefficient). The direction coefficient is determined by the sign of the deviation (+1 for positive deviations and -1 for negative deviations). For example, if node E has a scale deviation of +0.12 and a displacement weight of 0.6, the difference is 0.6 × 1 × 0.12 = 0.072, indicating that the displacement weight ratio needs to be reduced.

[0092] In this embodiment, the node weight distribution ratios (e.g., displacement weight 0.65, pressure weight 0.35) are first extracted from the feature weight distribution table. Combined with the scale deviation data (e.g., node N1 deviation -0.025), the dynamic weight difference is calculated using the residual analysis formula: for a positive deviation (Δ>0), weight difference = displacement weight × Δ - pressure weight × Δ; for a negative deviation (Δ<0), weight difference = pressure weight × |Δ| - displacement weight × |Δ|. For example, if node N1 has Δ = -0.025, the weight difference = 0.35 × 0.025 - 0.65 × 0.025 = 0.00875 - 0.01625 = -0.0075. The calculated results are stored in the dynamic difference table, and time series analysis (e.g., sliding window statistics) is used to identify weight adjustment trends.

[0093] 504. Based on the positive and negative directions of the dynamic weight difference, adjust the increase or decrease ratio of the input feature weight in the road state prediction model to optimize the prediction accuracy of the road state prediction model.

[0094] In step 504, the weight adjustment ratio is set according to the positive and negative signs and numerical values of the dynamic weight differences: a positive difference indicates that the current weight needs to be reduced (e.g., a displacement weight difference of +0.1 means a 10% reduction), while a negative difference indicates an increase in the opposite direction (e.g., a pressure weight difference of -0.08 means an 8% increase). The prediction accuracy is optimized by iteratively updating the model parameters using the gradient descent method. For example, the displacement weight is adjusted from 0.55 to 0.53, and the pressure weight is adjusted from 0.45 to 0.47, so that the mean square error (MSE) of the pressure prediction output by the model is reduced from 0.18 to 0.15, achieving the preset optimization target (e.g., error reduction ≥15%).

[0095] In this embodiment, the input feature weights are first adjusted using a gradient descent algorithm based on the sign of the dynamic weight difference (e.g., +0.04 or -0.0075). A positive difference increases the displacement weight ratio, using the formula: new displacement weight = original weight + learning rate × difference (learning rate = 0.01); a negative difference decreases the displacement weight ratio, using the formula: new displacement weight = original weight - learning rate × |difference|. For example, if the difference for node N1 is -0.0075, the new displacement weight is 0.65 - 0.01 × 0.0075 ≈ 0.6499. The updated weights are then fed into a road state prediction model (e.g., a hybrid LSTM + Attention model). Backpropagation is used to minimize the mean squared error (MSE) between the predicted and actual values until the error converges to a threshold (e.g., ≤3%). Here's a specific example: When implementing dynamic parameter optimization on a ring expressway (North Ring S3 → Interchange T9 → Cross-River Tunnel V6 → North Ring S3) in a certain city's smart transportation system, the following steps were taken: First, based on the real-time east-west traffic pressure data of Interchange T9 (peak value 1560 vehicles / hour, dynamic matching coefficient 0.79), the pressure fluctuation range was divided into [1280, 1640] vehicles / hour, and a node-indexed traffic pressure deviation matrix was generated (e.g., T9 node deviation +19.3%, V6 node deviation -8.7%). Spatial projection matching was used to find that the displacement vector weight distribution position (coordinate grid G4) of Interchange T9 in the feature weight distribution table was consistent with the pressure deviation quantization range [+15%, +25%], there was an 8.6% regional mismatch, triggering a reverse correction: the displacement vector weight density within the G4 grid was reduced, and the pressure threshold weight density was increased. After the optimization, the dynamic matching error of the distributed parameter set decreased (below the threshold of 5%), and the system restarted the road network structure adjustment accordingly, adjusting the east-west dual lanes of T9 to tidal lanes (3 to 2 lanes during the morning peak, 2 to 3 lanes during the evening peak). At the same time, a dynamically variable speed limit was implemented in the cross-river tunnel V6 (60 to 50 km / h during peak hours). After two iterations, the matching error stabilized at 3.2%, the pressure fluctuation of the overpass T9 narrowed to [1340, 1480] vehicles / hour, the standard deviation of the lateral offset of the displacement vector was optimized from 2.1m to 1.3m, the evening peak traffic efficiency increased year-on-year, and the model prediction accuracy was improved, achieving a closed-loop automation of the entire link "deviation identification-parameter correction-structure tuning".

[0096] In summary, steps 501 to 504 have implemented a self-optimizing intelligent road network management system. This system, through real-time pressure monitoring and a dynamic matching mechanism, significantly improves urban traffic operational efficiency. The system accurately captures traffic pressure fluctuations at road nodes and establishes a pressure deviation matrix with meter-level accuracy, accelerating traffic anomaly detection and response. The system's "monitor-correct-verify" closed-loop optimization mechanism improves the efficiency of road network adjustment decisions. This solution achieves a qualitative shift from reactive response to proactive prevention, reducing urban traffic management costs, lowering traffic accident rates, and improving citizen travel satisfaction, providing a new generation of intelligent transportation infrastructure for smart city development.

[0097] To address the issue of inaccurate time-dimensional feature weight allocation in road state prediction, a dynamic time window-based feature weight optimization system has been developed to achieve adaptive allocation of prediction model feature weights. In some embodiments, step 202 divides the input feature vector into time windows to establish a dynamic weight matching rule between the displacement vector change rate and the traffic pressure threshold within different time periods to generate a feature weight allocation table for the road state prediction model, including: 601. Based on the time series distribution of the displacement vector change rate in the input feature vector, the time series distribution is divided into multiple consecutive time periods according to a preset time window division rule, and the fluctuation amplitude of the displacement vector change rate and the fluctuation range of the traffic pressure threshold in each time period are extracted; In step 601, the input feature vector is a multidimensional array containing parameters such as the displacement vector change rate and the traffic pressure threshold, which serves as the input of the road state prediction model; the time series distribution of the displacement vector change rate is the vehicle displacement direction and speed change rate data arranged in chronological order (e.g., recorded once per second); the preset time window division rule is to divide the continuous time series into equal-length segments according to a fixed time length (e.g., 15 minutes); the fluctuation amplitude is the range of the displacement vector change rate within the time window (maximum value minus minimum value); and the fluctuation range of the traffic pressure threshold is the variation range of the road traffic pressure critical value within the same time window (e.g., [0.65, 0.75]).

[0098] In this embodiment of the present application, the displacement vector change rate in the input feature vector (e.g., the displacement change rate of the vehicle at node N1 = Δdisplacement / Δt, in m / s²) is first segmented into a time series, and then divided into continuous time periods (e.g., 08:00-08:15 and 08:15-08:30) according to preset time windows (e.g., every 15 minutes). The STL (Seasonal-Trend Decomposition) algorithm is used to decompose the time series, extracting the fluctuation amplitude of the displacement change rate within each time period (e.g., the amplitude of the 08:00-08:15 period = ±0.2 m / s²), and simultaneously calculating the fluctuation range of the traffic pressure threshold (e.g., the pressure index fluctuates between 0.68 and 0.75). For example, during the morning rush hour in a certain city, the displacement change rate amplitude increases from ±0.1 m / s² to ±0.3 m / s², corresponding to an expansion of the pressure fluctuation range from [0.65, 0.70] to [0.60, 0.78].

[0099] 602. Analyze the dynamic impact of the displacement vector change rate on the traffic pressure threshold in different time periods based on the correlation between the fluctuation amplitude and the fluctuation range, so as to generate a correlation mapping table of time periods, fluctuation amplitudes, and impact intensities; In step 602, the association relationship is calculated by the Pearson correlation coefficient to calculate the linear correlation between the fluctuation amplitude and the fluctuation range (for example, r=0.8 indicates a strong positive correlation); the dynamic impact intensity is the contribution of the displacement vector change rate to the change in the traffic pressure threshold (for example, an impact coefficient of 0.6 indicates that the displacement change explains 60% of the pressure fluctuation); the association mapping table is a structured table that records the starting point of the time period, the fluctuation amplitude value, and the impact intensity value (for example: {t1: [0.3, 0.7]} indicates that the fluctuation amplitude in the t1 period is 0.3 and the impact intensity is 0.7).

[0100] In the embodiment of the present application, the correlation between the displacement change rate fluctuation amplitude and the pressure fluctuation range is calculated by the Pearson correlation coefficient (such as R = 0.85 indicates a strong positive correlation), and a linear regression model is used to quantify the dynamic impact intensity (such as for every 0.1 m / s² increase in the displacement change rate, the pressure fluctuation range increases by 0.05). Generate an association mapping table to record the fluctuation parameters and impact intensity of each time period (for example: 08:00-08:15 period, displacement amplitude ±0.2, pressure fluctuation 0.65-0.70, impact intensity = 0.05 / 0.1 = 0.5). For example, in a sorting node in a logistics park during the 09:00-10:00 period, the displacement amplitude of ±0.25 m / s² causes the pressure fluctuation range to expand to 0.58-0.80, and the impact intensity is increased to 0.6.

[0101] 603. Traverse the time periods in the association mapping table and perform dynamic association analysis on the displacement vector change rate and the traffic pressure threshold within each time period to identify abnormal time periods in which the displacement vector change rate exceeds a preset fluctuation threshold; In step 603, the dynamic correlation analysis uses the Granger causality test to determine whether the displacement vector change rate significantly affects the traffic pressure threshold (p value < 0.05); the preset fluctuation threshold is the displacement change rate safety boundary set based on historical data (such as ±0.2 / s²); the abnormal time period is the time window when the displacement vector change rate exceeds the threshold and is significantly correlated with pressure fluctuations (such as the change rate in the t5 period reaches 0.35 / s², p = 0.02).

[0102] In an embodiment of the present application, the association mapping table is traversed, a displacement fluctuation threshold value (such as an amplitude of ±0.25 m / s²) is set, and abnormal time periods exceeding the threshold value are detected (such as a displacement amplitude of ±0.3 m / s² during the 08:30-09:00 period). At the same time, it is checked whether the pressure fluctuation range during the abnormal time period is synchronously exceeded (such as a pressure fluctuation range > a preset 0.6-0.75). If both conditions are met, it is marked as a high-conflict period. For example, the displacement amplitude of a cross-border transport hub reaches ±0.35 m / s² (>threshold 0.25) during the 12:00-13:00 period, and the pressure fluctuation range is expanded to 0.55-0.82. The system determines that this period is an abnormal time period and triggers the subsequent weight adjustment process.

[0103] 604. During the abnormal time period, dynamically adjust the weight parameter of the displacement vector change rate according to the fluctuation range of the traffic pressure threshold, and generate a weight allocation rule set indexed by the time period; In step 604, the weight parameter is dynamically adjusted by scaling the weight of the displacement vector change rate based on the pressure fluctuation range during the abnormal time period (formula: new weight = original weight × (1 + pressure fluctuation ratio)). The weight allocation rule set is a dictionary structure with time periods as keys (for example, {t5: 0.8} indicates that the weight of the displacement change rate during the t5 period is adjusted to 0.8).

[0104] In this embodiment, for abnormal time periods (e.g., 12:00-13:00), the weight parameters of the displacement vector change rate are dynamically adjusted based on the pressure fluctuation range (0.55-0.82): if the upper limit of the pressure fluctuation is greater than 0.8, the displacement weight is increased (e.g., from 0.6 to 0.7); if the lower limit is less than 0.6, the displacement weight is decreased (e.g., from 0.6 to 0.5). A set of weight assignment rules indexed by time period is generated (e.g., {time period: 12:00-13:00, displacement weight: 0.7, pressure weight: 0.3}). For example, during abnormal time periods, the displacement weight of a port yard is adjusted from 0.6 to 0.75, and the pressure weight is reduced from 0.4 to 0.25, allowing the model to focus more on the impact of displacement changes on pressure.

[0105] 605. Collect each time parameter in the weight allocation rule set and the real-time update frequency of the corresponding time period to generate a feature weight allocation table.

[0106] In step 605, the time parameter is the time period identifier in the weight allocation rule set (such as t1 to tn); the real-time update frequency is the number of records pushed by the data source per second (such as 10 records / second); the feature weight allocation table is a two-dimensional matrix that integrates the time parameter and the update frequency. The columns represent features (displacement change rate, pressure threshold, etc.), the rows represent time periods, and the cell values are normalized weights (such as the displacement column weight of 0.8 and the pressure column weight of 0.2 in the t5 period).

[0107] In an embodiment of the present application, the time period parameters (such as 12:00-13:00) in the weight allocation rule set are bound to the real-time data update frequency (such as 20 high-frequency data per second in this period) to generate a dynamic feature weight allocation table. Through the time slice rotation mechanism, a preset weight is assigned to each time period (such as a displacement weight of 0.7 for high-frequency time periods and 0.5 for low-frequency time periods), and real-time updates are supported. For example, a smart highway system has a data update frequency of 30 pieces / second during the evening peak period (17:00-19:00). The weight allocation table dynamically sets the displacement weight to 0.75 and the pressure weight to 0.25, which improves the response speed of the prediction model to sudden traffic.

[0108] Here's a specific example: A certain city's ring expressway (East Ring E2 → Hub Bridge K5 → Tunnel Group L8 → East Ring E2) implements dynamic weight optimization during the morning rush hour: the system divides the 7:00-9:00 period into 15-minute windows, extracts the fluctuation amplitude of the displacement vector change rate (for example, the ΔV of the north-to-south direction of Hub Bridge K5 reaches 38% during the 7:30-7:45 period) and the fluctuation range of the traffic pressure threshold (the carrying capacity threshold fluctuates from 1050 vehicles / hour to 1320 vehicles / hour), and generates an association mapping table through Pearson correlation coefficient analysis (the influence intensity R²=0.91 of the displacement vector change rate on the pressure threshold during the 7:30-7:45 period). It is found that the displacement vector during the 7:45-8:00 period has a significant impact on the traffic pressure threshold. A sudden increase in the volume change rate triggers a dynamic weight adjustment: the displacement vector weight for this period is increased from 0.65 to 0.82, and the pressure threshold weight is reduced from 0.35 to 0.18, generating a period index rule set (7:45-8:00_weight[0.82,0.18]). Combined with the real-time data update frequency of 1 minute, a feature weight allocation table is dynamically generated (the displacement vector weight for the 7:45 period increases by 0.06 for every 10% increase in ΔV). After optimization, the pressure prediction error of the model during the 7:45-8:15 period decreases, the displacement vector offset warning response speed is shortened, the overall traffic efficiency during the morning rush hour is improved, the congestion peak period is compressed, and the accuracy of weight adaptive adjustment is improved.

[0109] In summary, steps 601 to 605 have implemented an intelligent traffic forecasting and optimization system based on spatiotemporal analysis. This system significantly improves the accuracy of road network status predictions through multi-dimensional data fusion. The system utilizes adaptive time window technology to intelligently identify traffic flow characteristics at different time periods and accurately capture the dynamic changes in vehicle displacement vectors. Based on a proprietary correlation analysis engine, the system effectively evaluates the interactions between traffic parameters, accurately identifies various traffic anomaly periods, and significantly improves emergency response time. Through an innovative dynamic weight allocation mechanism, the system automatically optimizes prediction model parameters, significantly improving prediction accuracy in complex scenarios such as peak hours.

[0110] To address the issue of inaccurate dynamic quantification of road pressure, a multi-source data fusion 3D modeling and pressure analysis system has been developed. This system integrates traffic flow, remote sensing imagery, and road network topology data to achieve spatiotemporal dynamic quantification and indexing of road pressure. In some embodiments, the traffic flow data collected in step 101 is combined with remote sensing imagery and road network topology data to construct a 3D model. This 3D model is used to quantify pressure data for road areas over different time periods to establish a road feature association index table, including: 701. Fusing the dynamic vehicle distribution information in the traffic flow data with the remote sensing image data, and extracting the initial outline of the road geometry from the fused data; In step 701, dynamic vehicle distribution information is real-time vehicle position, speed and density data collected by traffic monitoring equipment (such as a GPS trajectory point set updated every 5 seconds); remote sensing image data is high-resolution satellite or aerial imagery (such as 0.5-meter resolution), containing road surface texture and surrounding environmental features; fusion processing uses pixel-level alignment and feature matching technology (such as the SIFT algorithm) to map vehicle coordinates to the image geographic coordinate system; road geometry includes vector features such as lane curvature and intersection turning radius; and the initial contour is the road boundary polygon extracted by edge detection (such as the Canny operator) (with an accuracy of ±0.3 meters).

[0111] In an embodiment of the present application, high-resolution road image data (e.g., 0.5-meter resolution) is first acquired through satellite or UAV remote sensing images, and dynamic vehicle distribution information (e.g., vehicle position, speed, and density) is simultaneously collected from on-board GPS, ground sensing coils, and other equipment. Multimodal data fusion technology (e.g., image registration algorithm based on deep learning) is used to spatially align the vehicle distribution data with the remote sensing image, and an edge detection algorithm (e.g., Canny operator) is used to extract the initial contours of the road geometry (e.g., lane lines, intersection shapes). For example, a remote sensing image of a main road in a city shows that the intersection is a ring structure. After fusing the vehicle trajectory data, the inner and outer radii of the ring lane are automatically identified as 30 meters and 50 meters, respectively, to generate the initial contour line.

[0112] 702. Perform spatial hierarchical reconstruction on the initial outline according to the connection relationship and hierarchical structure of the road nodes in the road network topology data to generate a three-dimensional model framework with multi-dimensional connection attributes; In step 702, the road network topology data is the connection relationship between nodes (intersections) and edges (road sections) stored in a graph structure (such as an adjacency matrix); the hierarchical structure is the level divided according to the function of the road (such as expressway = level 1, main road = level 2); the spatial hierarchy reconstruction is to stratify the initial contour according to the road level through Delaunay triangulation (such as the triangle mesh density of level 1 road is 2 meters, and that of level 2 road is 5 meters); the three-dimensional model framework is a BIM model (IFC format) that includes elevation data (generated by LiDAR point cloud) and topological attributes, and supports LOD (level of detail) switching.

[0113] In an embodiment of the present application, based on the connection relationship (such as N1 connecting N2 and N3) and hierarchical structure (such as elevated layer and ground layer) of nodes (such as intersections and overpasses) in the road network topology data, the initial road contour is reconstructed using a three-dimensional spatial interpolation algorithm (such as Delaunay triangulation). For example, the two-dimensional contour line of the roundabout is stretched into a three-dimensional structure according to the elevation data (such as the height of the overpass is 8 meters), and hierarchical attributes are assigned (such as the elevated layer label L1 and the ground layer label L2). At the same time, multi-dimensional connection attributes are added to each node (such as the travel direction of N1 is east and west, and the speed limit is 60km / h), forming a three-dimensional model framework that contains spatial topology and semantic information.

[0114] 703. Mapping the dynamic vehicle distribution information into the three-dimensional model framework to form a three-dimensional model including road geometry and real-time traffic flow changes; In step 703, dynamic vehicle distribution information mapping is to match the real-time collected vehicle GPS coordinates, speed and other data to the grid cells of the three-dimensional model framework through a spatial interpolation algorithm (such as Kriging interpolation); the road geometry refers to the geometric structure of infrastructure such as lane lines and curbs generated by BIM modeling in the three-dimensional model; real-time traffic flow changes are represented by a dynamically updated vehicle density heat map in the three-dimensional model (red indicates high density, green indicates low density); the three-dimensional model ultimately generates a traffic simulation scene that includes a time dimension, supporting the playback of historical traffic flow status with second-level accuracy.

[0115] In this embodiment of the present application, dynamic vehicle distribution information (e.g., vehicle coordinates and velocity vectors) is matched to the spatial coordinates of the 3D model framework using a spatiotemporal mapping algorithm (e.g., quadtree indexing). For example, a vehicle's real-time position at a roundabout (e.g., [x=121.50°E, y=31.20°N, z=8m]) is mapped to the corresponding lane on the elevated level L1 and displayed as a dynamic particle flow overlay. Using a real-time rendering engine (e.g., Unity3D), a 3D visualization model is generated that combines static road geometry (lanes, markings) with dynamic traffic flow (vehicle movement, density heatmaps), allowing for observation of traffic conditions from any perspective.

[0116] 704. Divide the road areas in the three-dimensional model according to time series, and calculate the traffic pressure data of the corresponding time period using the dynamic distribution parameters of the traffic flow data in each road area; In step 704, time series segmentation refers to dividing continuous traffic flow data into discrete analysis periods at preset time intervals (e.g., 15-minute intervals). This can be processed using either a sliding window (5-minute step) or a fixed window. Road regions are spatial analysis units generated based on a three-dimensional road network model using regular grid division (100m×100m) or a dynamic clustering algorithm. Each region has a unique geocoding identifier (e.g., a GeoHash string). Dynamic distribution parameters include three core indicators: traffic density reflects the number of vehicles per unit area (vehicles / km², with an accuracy of ±5%), average speed represents the median vehicle speed within the area (km / h), and acceleration dispersion quantifies the degree of fluctuation in vehicle acceleration and deceleration behavior (m / s²). Traffic pressure data is a comprehensive assessment value generated through a multi-indicator fusion algorithm.

[0117] In this embodiment of the present application, the road area in the three-dimensional model (e.g., a 50-meter radius around an intersection) is divided into spatiotemporal units according to a preset time window (e.g., every 15 minutes), and traffic flow dynamic parameters (e.g., traffic volume, average speed, and lane occupancy) within each unit are extracted. The pressure calculation formula is: Traffic Pressure Index = (Real-time Traffic Volume / Design Capacity) × (1 - Speed Decay Rate). For example, if a road section has a design capacity of 1,000 vehicles per hour, a real-time traffic volume of 800 vehicles, and a speed drop from 60 km / h to 40 km / h, the decay rate is 0.33, and the pressure index is 0.8 × 0.67 ≈ 0.536. The final output is the traffic pressure data for each time period (e.g., a pressure index of 0.72 for the 08:00-08:15 period).

[0118] 705. Bind the traffic pressure data with the road area features in the road network topology data to form an associated index table indexed by spatiotemporal coordinates and road attributes.

[0119] In step 705, the road area characteristics include static attributes such as the number of lanes, slope, and speed limit; the binding operation uses spatial connection to mount the pressure data to the surface features of the road network topology; the associated index table is a PostGIS database table with a primary key of a spatiotemporal composite index, which supports fast queries by road ID or time range.

[0120] In an embodiment of the present application, traffic pressure data (such as a pressure index of 0.72) is bound to road attributes (such as road section ID, number of lanes, slope) and spatiotemporal coordinates (such as latitude and longitude, timestamp) in the road network topology data. A multi-level index is established through a hash table structure: the first-level index is a spatial grid code (such as the Geohash string wtw3sp), the second-level index is a time window (such as 08:00-08:15), and the third-level index is a road attribute (such as the number of lanes = 3). For example, the pressure index of 0.75 of a logistics park section RD-001 at 08:00 is stored as: {Key: "wtw3sp_0800_RD-001", Value: {Pressure index: 0.75, Number of lanes: 3, Slope: 2%}} This associated index table supports efficient query and spatiotemporal analysis, providing a data basis for subsequent path planning and congestion prediction.

[0121] Here's a specific example: During the construction of a complex road network (overpass G9 → ground auxiliary road H4 → tunnel group P6 → underground roundabout Q2) in a new urban area, the intelligent transportation system extracted the road geometry by fusing drone remote sensing images (spatial resolution 0.2 meters) with real-time GPS data from 100,000 connected vehicles. A point cloud segmentation algorithm was used to identify geometric parameters such as the curvature radius of overpass G9 (R=210 meters) and the length of the gradient section at the entrance of tunnel group P6 (85 meters) from the remote sensing data. These parameters were then matched and calibrated with the on-board GPS trajectory to generate the initial road profile. Spatial reconstruction was performed based on the road network topology hierarchy (overpass-ground layer-underground layer), and a three-dimensional model framework was established (for example, the vertical connection relationship between overpass G9 and underground roundabout Q2 was modeled as Z-axis +18 meters to Z-axis -12 meters). Real-time traffic flow data was mapped to the three-dimensional model to dynamically present the traffic density at different levels (the morning peak density on the elevated layer was 42 vehicles). The model area was divided into 15-minute time periods, and traffic pressure data for the G9 viaduct from west to east between 07:30 and 07:45 was calculated (peak traffic volume of 1,380 vehicles / hour, exceeding the design capacity by 23%). This data was then linked to the road network characteristics (lane number narrowing from 3 to 2, slope of 4%) to generate an associated index table (space-time coordinates [G9, 07:30-07:45]_pressure value 1.23_slope influence coefficient 0.38). Based on this data, the system visually warned in the 3D model and activated tidal lane control on the elevated level (adding one eastbound lane) 15 minutes in advance, bringing traffic pressure during this period back within the design capacity (1,120 vehicles / hour). The 3D model's pressure prediction error rate was reduced, and the average speed of the road network during the morning rush hour increased to 51 km / h (from the original plan's 43 km / h), improving the balance of cross-level traffic flow distribution.

[0122] In summary, steps 701 to 705 have implemented a 3D traffic modeling system based on multi-source data fusion. This innovative data processing method significantly improves the accuracy and visualization of road network status analysis. This solution transcends the technical limitations of traditional 2D traffic analysis, enabling 3D visualization and spatiotemporal dynamic analysis of road network status. This provides precise decision-making support for traffic planning, intelligent navigation, and emergency management, significantly enhancing the intelligent level of urban traffic management.

[0123] Figure 2 The present invention provides a structural diagram of a real-time map updating system based on multi-source geographic information data fusion, as shown in FIG. Figure 2 As shown, the system includes: The acquisition module 21 collects traffic flow data, combines remote sensing image data with road network topology data, and constructs a three-dimensional model. The three-dimensional model is used to quantify the traffic pressure data of the road area in different time periods to establish a road feature association index table; A mapping module 22 acquires vehicle displacement vectors and velocity gradient changes at key road nodes, synchronously fuses laser point cloud data with video surveillance data, generates a road feature dataset, and establishes a first mapping relationship between the road feature dataset and the road feature association index table; A construction module 23 constructs a road state prediction model based on the first mapping relationship, so as to generate prediction data of a road traffic state change trend through the road state prediction model; An analysis module 24 performs correlation analysis on the traffic pressure data in the three-dimensional model and the road characteristic dataset, and establishes a second mapping relationship between the road traffic state and the vehicle motion characteristics based on the analysis result and the predicted data of the road traffic state change trend; The updating module 25 adjusts the road network connection structure in the road network topology data based on the second mapping relationship, iteratively updates the associated parameters of the first mapping relationship through the adjusted road network connection structure, and simultaneously uses the updated associated parameters to optimize the prediction accuracy of the road state prediction model to achieve dynamic optimization of real-time map data.

[0124] Figure 2 The real-time map update system based on multi-source geographic information data fusion can be executed Figure 1 The implementation principles and technical effects of the real-time map update method based on multi-source geographic information data fusion described in the illustrated embodiment will not be elaborated on here. The specific manner in which the various modules and units perform operations in the real-time map update system based on multi-source geographic information data fusion in the above-mentioned embodiment have been described in detail in the relevant embodiments of the method and will not be elaborated on here.

[0125] In one possible design, Figure 2 The real-time map updating system based on multi-source geographic information data fusion of the embodiment shown can be implemented as a computing device, such as Figure 3 As shown, the computing device may include a storage component 31 and a processing component 32; The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 32 .

[0126] The processing component 32 is used for the above Figure 1 The embodiment provides a real-time map updating method based on multi-source geographic information data fusion.

[0127] The processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented as one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above method.

[0128] The storage component 31 is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile memory device, or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk, or optical disk.

[0129] Of course, a computing device may also include other components, such as input / output interfaces, display components, communication components, etc.

[0130] The input / output interface provides an interface between the processing component and the peripheral interface module, which can be an output device, an input device, etc.

[0131] The communication component is configured to facilitate, among other things, wired or wireless communications between the computing device and other devices.

[0132] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. In this case, the computing device can refer to a cloud server, and the above-mentioned processing components, storage components, etc. can be basic server resources rented or purchased from the cloud computing platform.

[0133] The present application also provides a computer storage medium storing a computer program, wherein the computer program can achieve the above-mentioned Figure 1 The embodiment shown is a real-time map updating method based on multi-source geographic information data fusion.

[0134] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0135] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0136] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer, server, or network device) to execute the methods described in each embodiment or certain portions of the embodiments.

[0137] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A real-time map updating method based on multi-source geographic information data fusion, characterized in that: include: Collecting traffic flow data, combining remote sensing image data with road network topology data, and constructing a three-dimensional model. Using the three-dimensional model, quantify traffic pressure data of road areas in different time periods to establish a road feature association index table. Obtaining vehicle displacement vectors and velocity gradient changes at key road nodes, synchronously fusing laser point cloud data with video surveillance data to generate a road feature dataset, and establishing a first mapping relationship between the road feature dataset and the road feature association index table; Based on the first mapping relationship, a road state prediction model is constructed to generate prediction data of a road traffic state change trend through the road state prediction model; Performing correlation analysis on the traffic pressure data in the three-dimensional model and the road feature dataset, and establishing a second mapping relationship between the road traffic state and the vehicle motion characteristics based on the analysis results and the predicted data of the road traffic state change trend; Based on the second mapping relationship, the road network connection structure in the road network topology data is adjusted, the associated parameters of the first mapping relationship are iteratively updated through the adjusted road network connection structure, and the updated associated parameters are simultaneously used to optimize the prediction accuracy of the road state prediction model to achieve dynamic optimization of real-time map data.

2. The method according to claim 1, characterized in that Building a road state prediction model based on the first mapping relationship to generate prediction data of a road traffic state change trend through the road state prediction model includes: Constructing an input feature vector of a road state prediction model based on the displacement vector change rate in the road feature data set extracted from the first mapping relationship and the traffic pressure threshold in the road feature association index table; By dividing the input feature vector into time windows, a dynamic weight matching rule between the displacement vector change rate and the traffic pressure threshold in different time periods is established to generate a feature weight allocation table for a road state prediction model; Dynamically binding the feature weight allocation table to the real-time update frequency of the road feature dataset, and incrementally correcting the dynamic matching coefficient in the feature weight allocation table based on the real-time update frequency of the road feature dataset; The revised feature weight distribution table is input into the road state prediction model, and the model is used to perform spatiotemporal coupling calculations on the associated features of the real-time update frequency and the dynamic matching coefficient, and outputs prediction data including the traffic pressure fluctuation range and the displacement vector change trend.

3. The method according to claim 1, characterized in that Correlation analysis is performed on the traffic pressure data in the three-dimensional model and the road feature dataset, and a second mapping relationship between the road traffic state and the vehicle motion characteristics is established based on the analysis results and the predicted data of the road traffic state change trend, including: performing correlation analysis on the traffic pressure data of the three-dimensional model and the road characteristic dataset in a temporal and spatial dimension, calculating the real-time correlation between the traffic pressure data and the road characteristic dataset, and generating an analysis result reflecting the dynamic correlation between pressure intensity and motion parameters; Based on the real-time correlation between the traffic pressure data and the road characteristic data set in the analysis results, the road traffic status level is divided according to the preset pressure threshold range; Combined with the predicted data of the road traffic status level change trend, the vehicle displacement vector distribution characteristics and speed gradient change range corresponding to each state level are dynamically associated to generate a vehicle motion feature mapping rule based on the traffic pressure data to establish a second mapping relationship between the road traffic status and the vehicle motion feature.

4. The method according to claim 2, characterized in that Adjusting the road network connection structure in the road network topology data based on the second mapping relationship, iteratively updating the associated parameters of the first mapping relationship using the adjusted road network connection structure, and simultaneously optimizing the prediction accuracy of the road state prediction model using the updated associated parameters to achieve dynamic optimization of real-time map data, including: Dividing the traffic pressure interval based on the second mapping relationship, and extracting the pressure adaptation scale value of the road node within the traffic pressure interval; converting the real-time traffic pressure data into a real-time pressure parameter having the same dimension as the pressure adaptation scale value using the parameter mapping rule in the second mapping relationship, and adjusting the road network connection structure in the road network topology data according to a deviation between the pressure adaptation scale value and the real-time pressure parameter; Inputting the fluctuation frequency of the real-time pressure parameter of the key road nodes in the adjusted road network connection structure into the parameter correction module of the first mapping relationship, and iteratively updating the associated parameters of the first mapping relationship based on the weight ratio of the displacement vector change rate and the traffic pressure threshold through a dynamic balance mechanism; Based on the updated associated parameters, scale deviation data of the adjusted road network is generated, and the input feature weights of the road state prediction model are adjusted according to the weight difference between the scale deviation data and the feature weight allocation table to optimize the prediction accuracy of the road state prediction model.

5. The method according to claim 4, characterized in that Generating scale deviation data of the adjusted road network based on the updated association parameters, and adjusting the input feature weights of the road state prediction model by the scale deviation data and the weight difference of the feature weight distribution table to optimize the prediction accuracy of the road state prediction model, including: Based on the displacement vector change rate weight ratio and the traffic pressure threshold weight ratio in the updated associated parameters, the real-time pressure parameters of each node in the adjusted road network are extracted; Comparing the real-time pressure parameter with the pressure adaptation scale value of the corresponding node using the deviation calculation rule in the second mapping relationship to generate scale deviation data reflecting the degree of dynamic matching between the traffic pressure of the road network node and the vehicle displacement vector; Extracting the weight distribution ratio associated with the adjusted road network node in the feature weight distribution table, and calculating the dynamic weight difference between the weight distribution ratio and the scale deviation data according to the distribution range of the scale deviation data; Based on the positive and negative directions of the dynamic weight difference, the increase or decrease ratio of the input feature weight in the road state prediction model is adjusted to optimize the prediction accuracy of the road state prediction model.

6. The method according to claim 2, characterized in that By dividing the input feature vector into time windows, a dynamic weight matching rule between the displacement vector change rate and the traffic pressure threshold in different time periods is established to generate a feature weight allocation table for the road state prediction model, including: Based on the time series distribution of the displacement vector change rate in the input feature vector, the time series distribution is divided into multiple continuous time periods according to a preset time window division rule, and the fluctuation amplitude of the displacement vector change rate and the fluctuation range of the traffic pressure threshold in each time period are extracted; Analyzing the dynamic impact of the displacement vector change rate on the traffic pressure threshold in different time periods based on the correlation between the fluctuation amplitude and the fluctuation range, so as to generate a correlation mapping table of time periods, fluctuation amplitudes, and impact intensities; Traversing the time periods in the association mapping table, dynamically analyzing the displacement vector change rate and the traffic pressure threshold within each time period to identify abnormal time periods in which the displacement vector change rate exceeds a preset fluctuation threshold; During the abnormal time period, dynamically adjusting the weight parameter of the displacement vector change rate according to the fluctuation range of the traffic pressure threshold, and generating a weight allocation rule set indexed by the time period; Each time parameter in the weight allocation rule set is combined with the real-time update frequency of the corresponding time period to generate a feature weight allocation table.

7. The method according to claim 1, characterized in that Traffic flow data is collected and combined with remote sensing image data and road network topology data to construct a three-dimensional model. The three-dimensional model is used to quantify the traffic pressure data of road areas in different time periods to establish a road feature association index table, including: The dynamic vehicle distribution information in the traffic flow data is fused with the remote sensing image data, and the initial outline of the road geometry is extracted from the fused data; According to the connection relationship and hierarchical structure of road nodes in the road network topology data, the initial outline is spatially reconstructed to generate a three-dimensional model framework with multi-dimensional connection attributes; Mapping the dynamic vehicle distribution information into the three-dimensional model framework to form a three-dimensional model that includes road geometry and real-time traffic flow changes; Dividing the road areas in the three-dimensional model according to time series, and calculating the traffic pressure data of the corresponding time period by the dynamic distribution parameters of the traffic flow data in each road area; The traffic pressure data is bound to the road area features in the road network topology data to form an associated index table with spatiotemporal coordinates and road attributes as indexes.

8. A real-time map updating system based on multi-source geographic information data fusion, characterized in that: include: An acquisition module collects traffic flow data, combines remote sensing image data with road network topology data, and constructs a three-dimensional model. The three-dimensional model is used to quantify traffic pressure data of road areas in different time periods to establish a road feature association index table. a mapping module that obtains vehicle displacement vectors and velocity gradient changes at key road nodes, synchronously fuses laser point cloud data with video surveillance data, generates a road feature dataset, and establishes a first mapping relationship between the road feature dataset and the road feature association index table; A construction module, based on the first mapping relationship, constructs a road state prediction model to generate prediction data of a road traffic state change trend through the road state prediction model; an analysis module that performs correlation analysis on the traffic pressure data in the three-dimensional model and the road characteristic dataset, and establishes a second mapping relationship between the road traffic state and the vehicle motion characteristics based on the analysis result and the predicted data of the road traffic state change trend; An updating module adjusts the road network connection structure in the road network topology data based on the second mapping relationship, iteratively updates the associated parameters of the first mapping relationship through the adjusted road network connection structure, and simultaneously optimizes the prediction accuracy of the road state prediction model using the updated associated parameters to achieve dynamic optimization of real-time map data.

9. A computing device, characterized in that It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a real-time map updating method based on multi-source geographic information data fusion as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that A computer program is stored, and when the computer program is executed by a computer, a real-time map updating method based on multi-source geographic information data fusion as described in any one of claims 1 to 7 is implemented.

Images