Traffic network situation data display method and system

Abstract

The invention relates to the technical field of data processing, and provides a traffic road network situation data display method and system, which comprehensively improves the traffic situation perception and presentation capability through multi-step cooperative processing of traffic road network situation data, and finally dynamically adjusts the grid granularity according to deduction errors from multi-source heterogeneous data acquisition. A closed-loop optimization system is formed, accurate, real-time and personalized traffic network situation information can be provided for traffic management departments, travelers and the like, decision making is assisted, the travel efficiency is improved, and the method has remarkable innovativeness and practicability in the field of intelligent traffic.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a method and system for displaying traffic network situation data. Background Art

[0002] In the existing traffic network situation data display technology, the common practice is to use a single or a few data sources to analyze traffic conditions, such as relying only on traffic flow data to judge traffic congestion and perform simple visualization, without considering the complex relationship between different sections and regions and real-time feedback adjustment. Most of them do not solve how to achieve efficient optimization of traffic data processing and dynamic display of traffic situation. Summary of the invention

[0003] In view of the deficiencies of the prior art, the present application provides a method and system for displaying traffic network situation data.

[0004] In a first aspect, the present application provides a method for displaying traffic network situation data, the method comprising: obtaining multi-source heterogeneous data including traffic flow data, radar trajectory data, and event propagation data in a traffic network, interpolating missing multi-source heterogeneous data in real time through a generative adversarial network, and performing spatiotemporal synchronization on the interpolated multi-source heterogeneous data to generate a traffic network topology;

[0005] Based on the grid granularity, the traffic network topology is decomposed into grid units, the spatiotemporal conduction entropy between adjacent grid units is calculated, and the basic field strength of each grid unit is determined according to the spatiotemporal conduction entropy between adjacent grid units to generate the situation field strength map of the traffic network;

[0006] Extracting the situation feature tensor of the situation field strength map, the situation feature tensor includes situation layer features, event layer features and deduction layer features, and visually encoding the situation feature tensor according to the basic field strength of each grid unit to determine the layered rendering instructions;

[0007] Eye tracking data is obtained to dynamically adjust the basic field strength of each grid unit, and the layered rendering instructions are semantically described and labeled with positions through natural language processing and annotation positioning algorithms. Multi-layer situation deduction of the traffic network is performed according to the layered rendering instructions, and the grid granularity is dynamically adjusted according to the error of the multi-layer situation deduction.

[0008] As an optional implementation, the generation logic of the traffic network topology includes:

[0009] Acquire multi-source heterogeneous data including traffic flow data, radar trajectory data and event propagation data from the traffic network;

[0010] Real-time interpolation of missing multi-source heterogeneous data through generative adversarial networks;

[0011] The interpolated multi-source heterogeneous data are synchronized in time and space according to the constructed spatial reference points and time reference points;

[0012] The multi-source heterogeneous data that have been synchronized in time and space are used to determine the connection relationship of nodes and the weight of edges in the transportation network through a topological algorithm to generate the transportation network topology.

[0013] As an optional implementation, the generation logic of the situation field strength map includes:

[0014] Based on the grid granularity, the traffic network topology is decomposed into grid units, and the spatiotemporal conduction entropy between adjacent grid units is calculated;

[0015] The space-time conduction entropy between adjacent grid cells is used as a weight to determine the basic field strength of each grid cell by weighted average.

[0016] The basic field strength and space-time conduction entropy of each grid unit are integrated to generate a situation field strength map, which is then visualized.

[0017] As an optional implementation manner, the decomposition sub-logic of the grid unit includes:

[0018] The convolutional neural network is used to extract the features of the traffic network topology and identify the traffic network type;

[0019] Determine the grid granularity according to the type of traffic network;

[0020] The traffic network topology is decomposed into grid cells based on the grid granularity, and the boundary of each grid cell is determined according to its position in the traffic network.

[0021] As an optional implementation, the calculation sub-logic of the spatiotemporal conduction entropy between adjacent grid cells includes:

[0022] Extract multi-source heterogeneous data for each grid cell from the traffic network topology;

[0023] Analyze the causal relationship and spatial correlation of multi-source heterogeneous data between adjacent grid cells to obtain spatiotemporal correlation;

[0024] The spatiotemporal conduction entropy between adjacent grid cells is calculated based on the spatiotemporal correlation relationship.

[0025] As an optional implementation, the extraction logic of the situation feature tensor includes:

[0026] The situation field strength map is divided into situation layer, event layer and deduction layer;

[0027] A multi-layer extraction unit is configured to extract situation layer features, event layer features and deduction layer features of the situation field strength map in parallel through the multi-layer extraction unit;

[0028] The situation layer features, event layer features and inference layer features are fused through the attention mechanism to obtain the situation feature tensor.

[0029] As an optional implementation, the determination logic of the layered rendering instruction includes:

[0030] The mapping relationship between the basic field strength and situational characteristic tensor of each grid unit is analyzed through a neural network;

[0031] Visually encode the situation layer features, event layer features and deduction layer features respectively, and adjust the parameters of the visual encoding in combination with the basic field strength of each grid unit;

[0032] According to the results of visual encoding and the mapping relationship between the basic field strength of each grid unit and the situation feature tensor, rendering instructions for the situation layer, event layer and deduction layer are generated to determine the hierarchical rendering instructions.

[0033] As an optional implementation manner, the adjustment logic of the basic field strength of each grid unit includes:

[0034] Obtain the user's eye tracking data in real time, including gaze point location, gaze time, and scan path, and map the eye tracking data to grid cells;

[0035] The grid units that users pay attention to are identified through clustering algorithms, and the degree of attention of each grid unit is obtained;

[0036] Determine the adjustment direction and adjustment amplitude of the basic field strength of each grid unit according to the attention degree of each grid unit;

[0037] According to the adjustment direction and adjustment amplitude of the basic field strength of each grid unit and the current basic field strength of each grid unit, the adjusted basic field strength of each grid unit is obtained.

[0038] As an optional implementation, the grid granularity adjustment logic includes:

[0039] In the process of multi-layer situation deduction of the traffic network according to the layered rendering instructions, the error size and error change trend between the deduction results of the situation layer, event layer and deduction layer and the actual traffic conditions are recorded in real time through the mean square error;

[0040] Through fuzzy logic reasoning method, according to the error size and error change trend and actual traffic conditions, determine whether the grid size needs to be adjusted and the direction of adjustment of the grid size;

[0041] The grid size is adjusted according to the adjustment direction of the grid size, and multi-layer situation deduction is performed again to compare the error size before and after the adjustment of the grid size to determine whether the grid size should be readjusted.

[0042] In a second aspect, the present application provides a traffic network situation data display system, the system comprising: a topology generation module, a graph generation module, a layered rendering module and a feedback adjustment module;

[0043] The topology generation module is used to obtain multi-source heterogeneous data in the traffic network, including traffic flow data, radar trajectory data, and event propagation data. The missing multi-source heterogeneous data is interpolated in real time through the adversarial generative network, and the interpolated multi-source heterogeneous data is synchronized in time and space to generate the traffic network topology.

[0044] The map generation module is used to decompose the traffic network topology into grid units based on the grid granularity, calculate the spatiotemporal conduction entropy between adjacent grid units, and determine the basic field strength of each grid unit according to the spatiotemporal conduction entropy between adjacent grid units to generate a situational field strength map of the traffic network;

[0045] The hierarchical rendering module is used to extract the situation feature tensor of the situation field strength map. The situation feature tensor includes situation layer features, event layer features and deduction layer features. The situation feature tensor is visually encoded according to the basic field strength of each grid unit to determine the hierarchical rendering instructions.

[0046] The feedback adjustment module is used to obtain eye tracking data to dynamically adjust the basic field strength of each grid unit, and to semantically describe and position the layered rendering instructions through natural language processing and annotation positioning algorithm, to perform multi-layer situation deduction of the traffic network according to the layered rendering instructions, and to dynamically adjust the grid granularity according to the error of the multi-layer situation deduction.

[0047] Compared with the existing technology, the beneficial effects of the present application are: through multi-step collaborative processing of traffic network situation data, the traffic situation perception and presentation capabilities are comprehensively improved, from multi-source heterogeneous data acquisition to the final dynamic adjustment of grid granularity according to the deduction error, a closed-loop optimization system is formed, which can provide accurate, real-time and personalized traffic network situation information to traffic management departments and travelers, assist decision-making and improve travel efficiency, and has significant innovation and practicality in the field of intelligent transportation.

[0048] Comprehensive traffic flow data, radar trajectory data and event propagation data can comprehensively reflect the actual situation of the traffic network, interpolate missing data in real time through adversarial generation network, ensure data integrity, synchronize data in time and space to unify the time and space benchmark, and provide a reliable basis for subsequent analysis. The generated traffic network topology accurately presents the node connection relationship and edge weight, laying the foundation for in-depth analysis of traffic situation.

[0049] Based on the grid granularity decomposition of the traffic network topology, the space-time conduction entropy is calculated to determine the basic field strength. The generated situation field strength map can intuitively display the situation strength of each grid unit and their mutual relationship. Through differentiated grid division to adapt to different traffic network types, it helps to more accurately analyze the details of the traffic situation and provide key data support for subsequent feature extraction and visualization.

[0050] The multi-level features of the situation field strength map are extracted to form a situation feature tensor, and the visual encoding is performed according to the basic field strength to determine the layered rendering instructions. This allows the traffic situation data to be presented in a layered visual form, with information at different levels clearly distinguished. Users can focus on specific levels as needed, thereby improving the efficiency of information acquisition. The basic field strength affects the visual encoding, highlighting the situation in key areas.

[0051] Obtain eye tracking data to adjust the basic field strength, realize personalized situation display, highlight the user's focus area, process layered rendering instructions through natural language processing and annotation positioning algorithm, which is easy to understand and apply. Dynamically adjust the grid granularity according to the multi-layer situation deduction error, optimize the situation deduction accuracy, enable the system to adapt to the ever-changing traffic conditions, and continuously provide high-quality situation information. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. Among them:

[0053] Figure 1 A method flow chart of a method for displaying traffic network situation data provided by an embodiment of the present application;

[0054] Figure 2 A grid unit decomposition logic diagram of a traffic network situation data display method provided in an embodiment of the present application;

[0055] Figure 3 A system module diagram of a traffic network situation data display system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0056] In order to make the objectives, technical solutions and advantages of the embodiments of the present application more obvious and easy to understand, the technical solutions in the embodiments of the present application are clearly and completely described below in conjunction with the drawings in the specification. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments.

[0057] Example 1 like Figure 1As shown, a method flow chart of a method for displaying traffic network situation data is provided in an embodiment of the present application, and the method includes:

[0058] S1. Acquire multi-source heterogeneous data including traffic flow data, radar trajectory data and event propagation data in the traffic network, perform real-time interpolation of missing multi-source heterogeneous data through a generative adversarial network, and perform spatiotemporal synchronization of the interpolated multi-source heterogeneous data to generate a traffic network topology.

[0059] Specifically, the generation logic of the traffic network topology includes:

[0060] Acquire multi-source heterogeneous data including traffic flow data, radar trajectory data and event propagation data from the traffic network;

[0061] Real-time interpolation of missing multi-source heterogeneous data through generative adversarial networks;

[0062] The interpolated multi-source heterogeneous data are synchronized in time and space according to the constructed spatial reference points and time reference points;

[0063] The multi-source heterogeneous data that have been synchronized in time and space are used to determine the connection relationship of nodes and the weight of edges in the transportation network through a topological algorithm to generate the transportation network topology.

[0064] Traffic flow data can reflect the traffic conditions of vehicles on the road, such as traffic volume and speed. Radar trajectory data can accurately track the real-time position and movement trajectory of vehicles, while event propagation data contains relevant information on special events such as traffic accidents and road construction. Only by integrating these multi-source heterogeneous data can we comprehensively and accurately describe the actual situation of the traffic network and provide a rich data foundation for subsequent analysis and processing. Induction coils and cameras are installed at key locations on the road (such as intersections and the middle of road sections). The induction coils use the principle of electromagnetic induction to detect the changes in the magnetic field generated when a vehicle passes by, thereby obtaining data such as traffic flow and vehicle speed. The camera uses image recognition technology to analyze the captured video stream, identify vehicles and count relevant information.

[0065] Millimeter-wave radar emits millimeter waves and receives signals reflected by vehicles. The speed and position of the vehicle are calculated by analyzing the frequency changes of the signal (Doppler effect), thereby obtaining radar trajectory data. Millimeter-wave radar is usually installed at a high place to obtain a wider monitoring range. The acquisition of event propagation data is achieved by obtaining official information on traffic accidents and road construction or collecting traffic-related text information from platforms such as traffic information websites using natural language processing technology, and extracting event information. For example, text classification algorithms are used to identify text containing accident information, and further extract key information such as the location, time and type of the accident. Multi-source data fusion can provide more comprehensive traffic network situation awareness. Traffic flow data reflects traffic operation conditions at a macro level, radar trajectory data provides the precise location of vehicles at a micro level, and event propagation data can reveal special events that affect traffic. The three complement each other to make the data more complete and accurate.

[0066] In the actual data acquisition process, due to equipment failure and communication interruption and other reasons, data missing will inevitably occur. Missing data will affect the accuracy and completeness of subsequent data analysis, leading to misjudgment of the traffic network situation. The missing multi-source heterogeneous data is interpolated in real time through the adversarial generative network. The generator of the adversarial generative network adopts a fully connected neural network structure, inputs a random noise vector and some known multi-source heterogeneous data, and outputs an estimated value of the missing data through nonlinear transformation of multiple layers of neurons. The discriminator is also a fully connected neural network, and its input is complete multi-source heterogeneous data, including real data and interpolated data generated by the generator. It outputs a probability value, indicating the possibility that the input data is real data. A large amount of historical multi-source heterogeneous data is used to train the adversarial generative network. During the training process, the generator attempts to generate The generator generates realistic missing data to deceive the discriminator, while the discriminator strives to distinguish between real data and generated data. Through continuous iterative training, the generator can eventually generate missing data interpolation values ​​that are close to the real data distribution. For example, during the training process, the stochastic gradient descent algorithm is used to update the parameters of the generator and the discriminator. The loss function uses the cross entropy loss function to measure the difference between the generated data and the real data. In the real-time data acquisition process, once data missing is detected, the known part of the data is immediately input into the trained generator, and the generator quickly outputs the interpolation value of the missing data to achieve real-time interpolation; thereby improving the integrity and accuracy of the data and reducing the analysis errors caused by missing data. The interpolated data generated by the adversarial generative network can better retain the characteristics and distribution patterns of the original data, making the subsequent analysis results based on these data more reliable.

[0067] Multi-source heterogeneous data may be inconsistent in space and time. Spatiotemporal synchronization can unify the spatiotemporal benchmarks of multi-source heterogeneous data so that they can be analyzed and processed in the same spatiotemporal framework, thereby accurately reflecting the true situation of the transportation network. Based on the map in the geographic information system, key landmarks in the transportation network (such as large buildings and important intersections) are selected as spatial benchmarks. By measuring the coordinates of these key landmarks in different data acquisition methods and using coordinate conversion algorithms, such as the seven-parameter conversion model, the spatial coordinates of different data are uniformly converted to a coordinate system with these spatial benchmarks as references. Atomic clocks are used as time synchronization sources to provide a unified time benchmark to ensure the accuracy and consistency of data acquisition time. At the same time, during the data acquisition process, the acquisition timestamp of each data point is recorded.

[0068] For the interpolated multi-source heterogeneous data, according to their timestamps and spatial coordinates, they are uniformly mapped to the space-time coordinate system based on the constructed spatial reference points and time reference points. For example, for traffic flow data and radar trajectory data, according to their acquisition time and spatial position, they are matched and aligned in a unified space-time coordinate system to ensure that the data at the same time and the same position can accurately correspond; thereby achieving the consistency of multi-source heterogeneous data in the space-time dimension, eliminating data confusion and errors caused by space-time inconsistency, and the unified space-time benchmark makes the subsequent analysis and understanding of the traffic network situation more accurate and intuitive, and can better grasp the changing laws of traffic flow in time and space.

[0069] Traffic network topology is the key to understanding traffic flow operation and situation analysis. By determining the connection relationship of nodes and the weight of edges, the complex traffic network can be abstracted into an analyzable graph structure, which is convenient for subsequent in-depth research and visualization of traffic situation. According to the actual situation of the traffic network, intersections and road section endpoints are taken as nodes. By analyzing the multi-source heterogeneous data after spatiotemporal synchronization, the location information of these nodes is identified through geographic information data, such as extracting the coordinates of intersections from GIS map data and judging the starting and ending positions of sections from traffic flow data, thereby determining the specific location of the nodes. Synchronously based on the actual layout of the traffic network and geographic information, the connection relationship between nodes is determined. For example, adjacent intersections are connected by sections, and the two ends of the section correspond to two nodes respectively. The shortest path algorithm is used in combination with traffic rules, such as one-way street restrictions, to construct the connection relationship between nodes.

[0070] The weight of the edge is calculated according to different traffic indicators. For example, when the traffic flow is used as the weight, the number of vehicles passing through the section (edge) in unit time is counted. When the speed is used as the weight, the average speed of the section is calculated through radar trajectory data and traffic flow data. Multiple indicators can also be considered comprehensively, such as weighted combination of traffic flow and speed to obtain a comprehensive weight value. The Dijkstra algorithm is used to generate the traffic network topology according to the connection relationship of the determined nodes and the weight of the edge. Thus, the complex traffic network is converted into a graph structure that is easy to understand and analyze, providing an intuitive and effective tool for traffic situation analysis. By reasonably setting the weight of the edge, the importance and traffic operation status of different sections in the traffic network can be reflected, which helps traffic managers to quickly understand traffic bottlenecks and congested sections. The generated traffic network topology provides a basic framework for the subsequent steps of decomposing it into grid units and calculating the spatiotemporal conduction entropy between adjacent grid units. The accurate topological structure can make the subsequent grid division and analysis more in line with the actual traffic situation, and improve the accuracy and effectiveness of the entire traffic network situation data display method.

[0071] S2. Based on the grid granularity, the traffic network topology is decomposed into grid units, the spatiotemporal conduction entropy between adjacent grid units is calculated, and the basic field strength of each grid unit is determined according to the spatiotemporal conduction entropy between adjacent grid units to generate a situational field strength map of the traffic network.

[0072] Specifically, the generation logic of the situation field strength map includes:

[0073] Based on the grid granularity, the traffic network topology is decomposed into grid units, and the spatiotemporal conduction entropy between adjacent grid units is calculated;

[0074] The space-time conduction entropy between adjacent grid cells is used as a weight to determine the basic field strength of each grid cell by weighted average.

[0075] The basic field strength and space-time conduction entropy of each grid unit are integrated to generate a situation field strength map, which is then visualized.

[0076] By decomposing the traffic network topology into grid units and calculating the spatiotemporal conduction entropy between adjacent units, the relative situational influence information of each grid unit in the traffic network can be obtained, providing a data basis for the subsequent determination of the basic field strength and the generation of the situational field strength map; according to the determined grid granularity, the traffic network topology is gridded in the GIS environment to determine the boundary of each grid unit, and then the multi-source heterogeneous data of each grid unit is extracted from the traffic network topology, the causal relationship and spatial correlation between adjacent grid units are analyzed, and the spatiotemporal conduction entropy between adjacent grid units is calculated based on the spatiotemporal correlation relationship. In actual operation, parallel computing technology is used to calculate the spatiotemporal conduction entropy of multiple adjacent grid unit pairs at the same time to improve the computing efficiency; thereby, the spatiotemporal conduction relationship between each grid unit and the adjacent units in the traffic network is fully obtained, providing rich and accurate data support for the subsequent generation of the situational field strength map, which helps to more accurately reflect the situational characteristics of the traffic network.

[0077] The basic field strength of each grid unit reflects the relative importance of the grid unit in the traffic network and the intensity of the traffic situation. By taking the spatiotemporal conduction entropy between adjacent grid units as the weight for weighted average, the traffic situation transmission relationship between the grid unit and its adjacent units can be comprehensively considered, so as to more reasonably determine its basic field strength. For each grid unit, the spatiotemporal conduction entropy between it and all adjacent grid units is collected. The grid unit is assumed to be have adjacent grid cells, adjacent grid cells With grid cells The space-time conduction entropy is , then the grid cell Basic field strength The calculation formula is as follows:

[0078] ;

[0079] In the formula, Represents a grid cell The basic field strength, represents the number of adjacent grid cells, For adjacent grid cells With grid cells The space-time conduction entropy between Represents adjacent grid cells A characteristic value related to the traffic situation, such as traffic flow and average vehicle speed.

[0080] In actual calculations, the characteristic values ​​related to the traffic situation used for calculation are first determined, and then the basic field strength of each grid unit is calculated according to the above formula; this makes the determination of the basic field strength more scientific and reasonable, fully considering the temporal and spatial correlation between adjacent grid units, and can more accurately reflect the actual situation strength of each grid unit in the traffic network, providing key support for generating accurate situation field strength maps.

[0081] Integrating basic field strength and spatiotemporal conduction entropy to generate a situation field strength map can present the situation information of each grid unit in the traffic network in an intuitive and visual way; based on the geographic information system, each grid unit is used as an element in the situation field strength map. In the situation field strength map, the grid units are laid out according to their positions. The color, size or other visual attributes of each grid unit can be encoded according to the values ​​of its basic field strength and spatiotemporal conduction entropy. For example, grid units with strong basic field strength (obtained by comparing with the field strength threshold) are represented by darker colors, and grid units with large spatiotemporal conduction entropy (obtained by comparing with the entropy value threshold) are represented by larger sizes. In this way, the basic field strength and spatiotemporal conduction entropy information of each grid unit can be intuitively displayed in the map.

[0082] The generated situation field strength map is displayed through visualization software, providing a variety of display methods, such as static map display and dynamic animation display (displaying the evolution of traffic situation over time), etc. At the same time, interactive functions are added. For example, when the mouse hovers over a grid unit, the detailed basic field strength, space-time conduction entropy and related traffic data information of the unit are displayed, which is convenient for users to have an in-depth understanding of the situation of each grid unit; thus, the situation information of the traffic network is presented in an intuitive and visual way, which greatly improves the readability and comprehensibility of traffic situation data, and helps traffic managers make decisions quickly and take corresponding traffic management measures.

[0083] Furthermore, if Figure 2 As shown, the decomposition sub-logic of the grid unit includes:

[0084] The convolutional neural network is used to extract the features of the traffic network topology and identify the traffic network type;

[0085] Determine the grid granularity according to the type of traffic network;

[0086] The traffic network topology is decomposed into grid cells based on the grid granularity, and the boundary of each grid cell is determined according to its position in the traffic network.

[0087] Different types of traffic networks, such as highways, urban main roads and branch roads, have different traffic characteristics and structural features. Accurately identifying the type of traffic network will help to adopt appropriate grid division strategies according to its characteristics and improve the accuracy of traffic situation analysis. The traffic network topology is converted into image form through convolutional neural networks. For example, the road network nodes are used as pixels, and different colors or grayscale values ​​are assigned to the pixels according to the connection relationship between the nodes and the weight of the edges. Then, these image data are input into the convolutional neural network. The convolutional neural network contains multiple convolutional layers, pooling layers and fully connected layers. The convolutional layer extracts local features in the image through convolution kernels of different sizes. The pooling layer downsamples the output of the convolution layer to reduce the amount of data and retain important features. The fully connected layer integrates the features after multiple convolutions and pooling, and outputs the probability distribution of the traffic network type through the softmax function to determine the type of traffic network. During the training process, a large number of known types of traffic network topologies are used for training, and the model parameters are continuously adjusted through the back propagation algorithm to improve the classification accuracy; thus, the traffic network type can be identified quickly and accurately, providing a basis for the subsequent determination of the grid granularity, making the grid division more in line with the actual situation of different types of traffic networks, thereby more effectively capturing traffic situation information.

[0088] Different types of traffic networks have differences in traffic flow, speed changes, and spatial scales. For example, highways have large traffic flow and fast speed, so larger grid granularity is needed to summarize the overall traffic situation, while branch roads have complex traffic conditions and small traffic flow, so smaller grid granularity is needed to describe local traffic conditions in detail. Therefore, determining the grid granularity according to the type of traffic network can more effectively utilize computing resources and accurately reflect the traffic characteristics of different networks. For highways, a larger grid granularity is set according to the number of lanes, design speed, and historical traffic flow data, such as a square grid with a side length of 5 kilometers or 10 kilometers. For urban main roads, a moderate grid granularity is set considering the changes in traffic flow and road length, such as a grid with a side length of 1 kilometer or 2 kilometers. For branch roads, due to their narrow roads and complex traffic conditions, a smaller grid granularity is used, such as a grid with a side length of 200 meters or 500 meters. This makes the grid division more scientific and reasonable, and can reduce unnecessary calculations and improve data processing efficiency while ensuring accurate description of the traffic situation. At the same time, setting the grid granularity according to the characteristics of different types of road networks can more accurately reflect traffic changes.

[0089] Decomposing the traffic network topology into grid units can divide the complex traffic network into relatively independent and interrelated small areas, which is convenient for detailed analysis of the traffic situation in each area. The purpose of determining the boundary of each grid unit is to clarify the scope of the traffic network covered by each grid unit, so as to accurately extract the traffic data of the area. Based on the determined grid granularity, the traffic network topology is grid-divided in the geographic information system environment. For example, if the grid granularity is a square grid with a side length of 1 km, grid lines are drawn in the horizontal and vertical directions at intervals of 1 km on the GIS map to cover the traffic network topology under the grid. For each grid unit, the spatial analysis function of GIS is used to determine the intersection of its boundary with the traffic network, so as to clarify the road network sections and nodes contained in the grid unit, so as to determine the boundary of the grid unit. The effective segmentation of the traffic network is achieved, so that the analysis of the traffic situation can be refined to each grid unit, which improves the accuracy of the analysis. At the same time, the clear grid unit boundary helps to accurately extract the traffic data of each unit, providing an accurate data source for the subsequent calculation of spatiotemporal conduction entropy and determination of basic field strength.

[0090] Furthermore, the calculation sub-logic of the spatiotemporal conduction entropy between adjacent grid cells includes:

[0091] Extract multi-source heterogeneous data for each grid cell from the traffic network topology;

[0092] Analyze the causal relationship and spatial correlation of multi-source heterogeneous data between adjacent grid cells to obtain spatiotemporal correlation;

[0093] The spatiotemporal conduction entropy between adjacent grid cells is calculated based on the spatiotemporal correlation relationship.

[0094] The spatiotemporal conduction entropy is used to measure the mutual influence of traffic situations between adjacent grid cells in the spatiotemporal dimension. Based on the previously determined grid cell boundaries, multi-source heterogeneous data within each grid cell are extracted from the traffic network topology. Thus, the multi-source data of each grid cell is fully acquired, which provides rich data support for the subsequent analysis of the spatiotemporal correlation between adjacent grid cells and can more accurately reflect the interaction of traffic situations between grid cells.

[0095] Understanding the causal relationship and spatial correlation of multi-source heterogeneous data between adjacent grid cells can provide a deep understanding of the propagation and influence mechanism of traffic situations between different grid cells. The causal relationship can reveal whether the change of traffic status of a grid cell will trigger the corresponding change of adjacent grid cells, while the spatial correlation reflects the similarity of traffic status of adjacent grid cells in space. The spatiotemporal correlation relationship obtained by combining the two is the key factor in calculating spatiotemporal conduction entropy. Through the Granger causality test method, for the time series data of adjacent grid cells, such as the time series of traffic flow, a vector autoregression model is constructed to test whether the data of one grid cell can predict the data of another grid cell. If the data of one grid cell can significantly improve the prediction accuracy of the adjacent grid cell data after adding the vector autoregression model, it is considered that there is a causal relationship. Through spatial autocorrelation analysis methods, such as the Moran index, the multi-source heterogeneous data of each grid cell is first used as a spatial variable, and the spatial autocorrelation coefficient of these variables between adjacent grid cells is calculated. By calculating the Moran index, the degree of spatial correlation of the data between adjacent grid cells is judged.

[0096] The results of causal relationship analysis and spatial correlation analysis are integrated to form the spatiotemporal correlation relationship between adjacent grid units. For example, the strength of causal relationship (measured by the coefficient size of Granger causality test) and spatial autocorrelation coefficient are comprehensively considered to construct a spatiotemporal correlation matrix. The matrix elements represent the degree of spatiotemporal correlation in different aspects between adjacent grid units. This deeply reveals the interaction mechanism of traffic situation between adjacent grid units, provides strong support for understanding the propagation law of traffic flow in the road network, and helps to more accurately grasp the overall operation status of the traffic road network.

[0097] As a quantitative indicator, the space-time conduction entropy can comprehensively reflect the uncertainty and mutual influence of the traffic situation between adjacent grid units in the space-time dimension. Based on the concept of entropy in information theory, the space-time conduction entropy is defined and calculated in combination with the previously obtained space-time correlation relationship. For example, the maximum entropy principle is used to solve an optimization problem under the condition of satisfying the known space-time correlation constraint to obtain the value of the space-time conduction entropy. Specifically, the space-time correlation between adjacent grid units is represented as a set of constraints. By maximizing the entropy function and satisfying the constraints at the same time, the value obtained is the space-time conduction entropy. In actual calculations, the Lagrange multiplier method combined with the gradient descent method can be used to solve the optimization problem. In this way, the complex space-time correlation relationship is quantified into a specific value, which is convenient for subsequent numerical calculation and analysis, and provides an objective and comparable basis for determining the basic field strength of each grid unit, which helps to generate a situation field strength map more accurately, thereby intuitively displaying the situation distribution of the traffic network.

[0098] S3. Extract the situation feature tensor of the situation field strength map. The situation feature tensor includes situation layer features, event layer features and deduction layer features. Visually encode the situation feature tensor according to the basic field strength of each grid unit to determine the layered rendering instructions.

[0099] Specifically, the extraction logic of the situation feature tensor includes:

[0100] The situation field strength map is divided into situation layer, event layer and deduction layer;

[0101] A multi-layer extraction unit is configured to extract situation layer features, event layer features and deduction layer features of the situation field strength map in parallel through the multi-layer extraction unit;

[0102] The situation layer features, event layer features and inference layer features are fused through the attention mechanism to obtain the situation feature tensor.

[0103] The traffic network situation contains many different types of information. Through layering, features of different natures can be processed separately, which facilitates more targeted extraction of information specific to each layer and improves the accuracy and efficiency of feature extraction. The situation layer mainly reflects the regular operation situation of the traffic network, such as traffic flow and speed distribution, and the event layer focuses on the impact of special events on traffic, such as traffic accidents and road construction, while the deduction layer predicts future traffic situations based on existing data. This layering method can clearly sort out complex traffic situation information and provide a basis for subsequent accurate feature extraction; in the data structure of the situation field strength map, different storage areas or data labels are defined to distinguish each layer of data, such as using a three-dimensional array to store the situation field strength map data, and the first dimension of the array represents the grid The unit number, the second dimension represents the feature dimension, such as basic field strength and space-time conduction entropy, etc. The third dimension is used to distinguish the situation layer, event layer and deduction layer. For the situation layer, the data related to the normal state of traffic flow is stored in the corresponding dimension position, the event layer stores the attribute information related to the event, such as event type, occurrence time and impact range, etc., and the deduction layer stores the future traffic situation related data predicted according to historical data, such as the predicted traffic flow change trend and congestion probability, etc. Through this data structure design, a clear stratification of the situation field strength map is achieved; thus, the complex traffic situation data is structured, and the information of each layer is independent and clear, which is helpful in the subsequent feature extraction process. According to the characteristics of different layers, more suitable algorithms are used to improve the accuracy and efficiency of feature extraction.

[0104] Features at different layers have different properties and feature extraction requirements. Parallel extraction can make full use of distributed computing resources, speed up feature extraction, and improve processing efficiency. The spatial features in the situation layer data are extracted through convolutional neural networks. For example, the situation layer data is scanned by convolution kernels of different sizes to capture the spatial distribution pattern and change trend of traffic flow. The features extracted by the convolutional neural network are then reduced in dimension through principal component analysis to remove redundant information and retain the most representative features. For example, the high-dimensional feature vectors output by CNN are used with the help of principal component analysis to calculate the eigenvalues ​​and eigenvectors, and the principal components whose cumulative contribution rate reaches a certain threshold (such as 95%) are selected as the situation layer features.

[0105] The text information in the event layer (such as accident descriptions and construction notices) is preprocessed through natural language processing technology, including word segmentation, part-of-speech tagging and removal of stop words. The preprocessed text data is then converted into word vector representation and input into the long short-term memory network. The long short-term memory network is used to learn a series of event-related text data to extract event layer features, such as the type of event, degree of impact and duration. The historical traffic situation data is modeled through the autoregressive integral moving average model to predict future trends. Through continuous iterative training, features that can accurately reflect future traffic situation changes are extracted. For example, historical traffic flow data is input into the autoregressive integral moving average model to obtain prediction results to extract deduction layer features. This greatly improves the efficiency of feature extraction and shortens the processing time, making it possible to analyze a large amount of traffic situation data in a shorter time. At the same time, customized extraction methods for different layers of features improve the accuracy and pertinence of feature extraction, and can better mine key information in each layer of data.

[0106] Features at different layers have different importance under different traffic scenarios and decision-making requirements. The attention mechanism can automatically learn the weights of features at each layer and dynamically adjust the contribution of features at each layer to the final result according to the actual situation, so as to more accurately reflect the comprehensive situation of the traffic network. For example, in traffic congestion, event layer features (such as traffic accident information) are more critical to decision-making, while under normal traffic conditions, situation layer features (such as traffic flow distribution) are more important. The query vector, key vector and value vector corresponding to each layer are obtained through the attention mechanism, and then the similarity score between the query vector and the key vector is calculated by dot product operation. After normalization, the attention weights of features at each layer are obtained, and finally the value vectors are weighted and summed according to the attention weights to obtain the fused situation feature tensor; thus, the weights of features at each layer can be adaptively adjusted, key features can be highlighted, and the ability to accurately describe and analyze traffic situations can be improved. Compared with simple feature splicing or average fusion methods, the attention mechanism can better capture the complex relationship between features at different layers, generate more representative situation feature tensors, and provide more reliable data support for subsequent visual encoding and layered rendering instruction determination.

[0107] Specifically, the determination logic of the layered rendering instructions includes:

[0108] The mapping relationship between the basic field strength of each grid unit and the situation characteristic tensor is analyzed through a neural network;

[0109] Visually encode the situation layer features, event layer features and deduction layer features respectively, and adjust the parameters of the visual encoding in combination with the basic field strength of each grid unit;

[0110] According to the results of visual encoding and the mapping relationship between the basic field strength of each grid unit and the situation feature tensor, the rendering instructions of the situation layer, event layer and deduction layer are generated to determine the hierarchical rendering instructions. (The rendering instructions include the type of visual element, attribute value and position in the traffic network)

[0111] The basic field strength reflects the relative importance of each grid unit in the traffic network and the intensity of the traffic situation, while the situation feature tensor contains the comprehensive situation information of the traffic network. By analyzing the mapping relationship between them, we can understand how the basic field strength affects and is related to the overall traffic situation characteristics, and provide a basis for the subsequent visual encoding of the situation feature tensor according to the basic field strength. The basic field strength of each grid unit is combined with the multi-layer perceptron neural network to obtain the prediction value corresponding to the dimension of the situation feature tensor. During the training process, a large number of grid unit basic field strengths and their corresponding real situation feature tensor data are used as training samples. Through the back propagation algorithm, the weights and biases of the multi-layer perceptron neural network are continuously adjusted, so that the multi-layer perceptron neural network can learn the mapping relationship between the basic field strength and the situation feature tensor. Thus, a quantitative relationship between the basic field strength and the situation feature tensor is established, which provides a scientific basis for the subsequent visual encoding and layered rendering instruction determination according to the basic field strength. Through this mapping relationship, the key information of the basic field strength can be effectively integrated into the visualization process of the traffic situation, improving the accuracy of the visualization results and the guiding value to users.

[0112] Visual coding can transform abstract traffic situation features into intuitive visual elements, which are easy for users to understand and observe. By adjusting the visual coding parameters in combination with the basic field strength, it can highlight the areas with higher basic field strength, making the display of traffic situation more focused and layered. For situation layer features, such as traffic flow, color mapping is used for visual coding. A color mapping table is defined, and different flow values ​​are mapped to different colors according to the size range of traffic flow. For example, green is set to represent low flow, yellow is set to represent medium flow, and red is set to represent high flow. Then, according to the situation layer features (i.e., traffic flow) of each grid unit, the corresponding color is selected from the color mapping table, and the color brightness is adjusted in combination with the basic field strength. The color brightness of the grid unit with a strong basic field strength is appropriately increased to highlight it.

[0113] For the event layer features, an icon library is used for encoding. Different icons are defined for different types of events (such as traffic accidents, road construction, and traffic control, etc.). According to the event layer feature information of each grid unit, the corresponding icon is selected from the icon library and displayed at the grid unit position. The size of the icon is adjusted in combination with the basic field strength. The icon size of the grid unit with a strong basic field strength is appropriately increased. For the deduction layer features, dynamic graphics are used for visual encoding. For example, the thickness and color changes of the lines are used to represent the changing trend of future traffic flow. Thicker lines indicate an increase in flow, and darker colors indicate a larger increase. The display speed of the dynamic graphics is adjusted in combination with the basic field strength. The display speed of the dynamic graphics of the grid unit with a strong basic field strength is appropriately accelerated to highlight the future situation changes in the area. The speed adjustment is achieved by setting the frame rate of the animation. In this way, the abstract traffic situation features are transformed into intuitive and easy-to-understand visual elements, and the key areas can be highlighted according to the basic field strength, which improves the visualization effect of the traffic situation and the efficiency of information communication. Users can quickly understand the overall situation of the traffic network and the situation of key areas by observing the visualization results, providing intuitive support for traffic decision-making.

[0114] Rendering instructions are used to convert visually encoded information into commands that can be understood and executed by computers, and are used to accurately draw traffic situations on a visual interface. Based on the previous visual encoding results and mapping relationships, rendering instructions containing information such as visual element types, attribute values, and positions in the traffic network are generated. For the situation layer, the generated rendering instructions include the color value corresponding to each grid unit (determined according to the visual encoding), the color brightness adjustment value (determined in combination with the basic field strength), and the geographic coordinate position of the grid unit in the traffic network. For the event layer, the rendering instructions include the icon type corresponding to each grid unit (determined according to the event type), the icon size adjustment value (determined in combination with the basic field strength), and the color brightness adjustment value (determined in combination with the basic field strength). Field strength determination) and coordinate position; for the deduction layer, the rendering instructions include parameters of dynamic graphics (such as line thickness, color change and frame rate, etc.) and coordinate position, and the rendering instructions of these situation layers, event layers and deduction layers are integrated to form a complete layered rendering instruction, which is used for layered rendering of traffic network situation on visualization software or platform; the generated layered rendering instructions can accurately guide the computer to draw graphics that conform to the actual traffic situation on the visualization interface, highlight key areas, and make the display of traffic network situation clearer and more intuitive. Users can quickly obtain comprehensive and focused traffic information through the visualization interface, providing strong support for traffic management and decision-making.

[0115] S4. Obtain eye tracking data to dynamically adjust the basic field strength of each grid unit, and use natural language processing and annotation positioning algorithms to semantically describe and position the layered rendering instructions, perform multi-layer situation deduction of the traffic network based on the layered rendering instructions, and dynamically adjust the grid granularity based on the error of the multi-layer situation deduction.

[0116] Specifically, the adjustment logic of the basic field strength of each grid unit includes:

[0117] Obtain the user's eye tracking data in real time, including gaze point location, gaze time, and scan path, and map the eye tracking data to grid cells;

[0118] The grid units that users pay attention to are identified through clustering algorithms, and the degree of attention of each grid unit is obtained;

[0119] Determine the adjustment direction and adjustment amplitude of the basic field strength of each grid unit according to the attention degree of each grid unit;

[0120] According to the adjustment direction and adjustment amplitude of the basic field strength of each grid unit and the current basic field strength of each grid unit, the adjusted basic field strength of each grid unit is obtained.

[0121] The user's eye movement behavior can directly reflect the degree of his or her attention to different areas of the traffic network. By acquiring this data and mapping it to the corresponding grid units, we can understand the user's interest in the traffic situation in each area, thereby providing a basis for adjusting the basic field strength according to the user's attention, so that the displayed traffic situation is more in line with user needs; the user's eye movement is captured in real time by an infrared camera, and the image processing algorithm is used to analyze information such as the pupil position and the eyeball rotation angle to determine the gaze point position. The gaze time is obtained by recording the duration of the user's gaze at a certain point, and the scanning path is determined by tracking the continuous changes of the gaze point. The infrared camera is connected to the display of the traffic network A grid coordinate system corresponding to the traffic network topology is established on the display interface. According to the coordinates of the gaze point on the screen, it is mapped to the grid unit of the traffic network through a coordinate conversion algorithm. For example, the grid unit number corresponding to the gaze point is calculated through a formula when the screen resolution, the position of the display interface on the screen, and the scaling ratio of the traffic network topology on the display interface are known. This realizes the association between the user's subjective attention information and the objective traffic network area, provides a data basis for personalized adjustment of the traffic situation display, can highlight the traffic situation of the relevant area according to the user's attention point, and improves the efficiency of users in obtaining key information.

[0122] The directly acquired eye tracking data is relatively scattered, and it is difficult to intuitively judge the user's attention to each grid unit. The clustering algorithm can classify grid units with similar eye movement characteristics (such as dense gaze points and longer gaze time, etc.) into one category, thereby identifying the areas where the user focuses on and quantifying the attention level of each grid unit; the DBSCAN density clustering algorithm is used to divide the data points into core points, boundary points, and noise points. For each grid unit, the number of other grid units contained in the neighborhood radius area is calculated (i.e., density). If the density of a grid unit is greater than or equal to the minimum number of points, it is marked as a core point; if the density of a grid unit is less than the minimum number of points, but it is within the neighborhood radius area of ​​a core point, it is marked as a boundary point; others The grid cells are marked as noise points, and clusters are formed by connecting the core points and the boundary points in their neighborhood radius. The degree of attention can be expressed by the ratio of the number of core points in the cluster where the grid cell is located to the total number of core points. For example, the total number of core points in all clusters and the number of core points in the cluster where a grid cell is located are calculated. The degree of attention of the grid cell is the ratio of the number of core points in the cluster where the grid cell is located to the total number of core points in all clusters. In this way, the grid cells that the user pays attention to can be automatically and accurately identified from a large amount of scattered eye movement data, and the degree of attention can be quantified, which provides a scientific basis for the subsequent targeted adjustment of the basic field strength. Compared with manual judgment or simple statistical methods, the clustering algorithm can more comprehensively and objectively reflect the user's attention pattern.

[0123] The area of ​​user concern is usually the focus of traffic situation analysis and needs to be highlighted. The adjustment direction and amplitude of the basic field strength are determined according to the degree of concern, which can make the displayed traffic situation more in line with user needs, highlight the area of ​​user concern, and not affect the basic display of other areas. The adjustment direction determines whether to enhance or weaken the basic field strength, and the adjustment amplitude determines the degree of enhancement or weakening. The combination of the two can achieve refined control of traffic situation display; establish an adjustment rule table, divide the degree of concern into different intervals, and each interval corresponds to a different adjustment direction and amplitude. For example, when the degree of concern is greater than or equal to 0.8, the adjustment direction is to enhance the basic field strength, and the adjustment amplitude is to increase the original basic field strength by 30%; when the degree of concern is greater than or equal to 0.5 and is less than 0.8, the adjustment direction is to enhance the basic field strength, and the adjustment amplitude is to increase by 15%; when the attention level is less than 0.5, the adjustment direction is to keep the basic field strength unchanged or appropriately weaken it according to the actual situation, such as weakening it by 5%. In actual applications, these interval ranges and corresponding adjustment parameters can be flexibly adjusted according to user feedback and traffic situation analysis needs. By querying the adjustment rule table, the adjustment direction and amplitude of the basic field strength of each grid unit can be determined according to the attention level of each grid unit; thereby realizing personalized and precise adjustment of the basic field strength according to the user's attention level, making the traffic situation display more targeted and practical, and users can obtain traffic information in their areas of concern more quickly and clearly, improving the user experience and information communication efficiency of traffic situation data display.

[0124] The adjustment direction and amplitude are determined through the previous steps. Combined with the current basic field strength, the final adjusted basic field strength can be obtained for calculation, thereby realizing real-time update of the traffic network situation display, making the display result more in line with user needs and actual traffic conditions; the adjusted basic field strength is obtained according to the determined adjustment direction and amplitude. In actual calculations, each grid unit is traversed to obtain the adjusted basic field strength according to its corresponding adjustment direction and amplitude, and the corresponding value in the basic field strength data storage structure is updated; thereby completing the adjustment process of the basic field strength, so that the traffic network situation display can be optimized in real time according to the user's concerns, highlighting the areas that the user is concerned about, and providing users with more valuable traffic information. At the same time, the adjusted basic field strength data provides input data that better meets user needs for subsequent multi-layer situation deduction and other related analyses.

[0125] Specifically, the grid granularity adjustment logic includes:

[0126] In the process of multi-layer situation deduction of the traffic network according to the layered rendering instructions, the error size and error change trend between the deduction results of the situation layer, event layer and deduction layer and the actual traffic conditions are recorded in real time through the mean square error;

[0127] Through fuzzy logic reasoning method, according to the error size and error change trend and actual traffic conditions, determine whether the grid size needs to be adjusted and the direction of adjustment of the grid size;

[0128] The grid size is adjusted according to the adjustment direction of the grid size, and multi-layer situation deduction is performed again to compare the error size before and after the adjustment of the grid size to determine whether the grid size should be readjusted.

[0129] The purpose of multi-layer situation simulation is to predict the future situation of the traffic network, while the actual traffic conditions are constantly changing. By recording the errors between the simulation results and the actual conditions in real time, the accuracy and reliability of the current situation simulation model can be evaluated. The size of the error reflects the degree of deviation between the current simulation prediction value and the true value, and the error change trend can reflect the changes in the prediction performance of the simulation over time. This information is crucial for determining whether the grid granularity needs to be adjusted and how to adjust the grid granularity, because the choice of grid granularity will affect the situation simulation's ability to capture the details of the traffic network and the computational efficiency.

[0130] In the multi-layer situation simulation process, the situation layer obtains actual traffic flow data in real time, such as traffic volume and vehicle speed, and compares them with the corresponding data in the simulation results. Similarly, for the event layer, the actual event information (such as event type, time and place of occurrence, etc.) is compared with the simulation results, and the mean square error is calculated through a certain quantitative method (such as encoding the event type as a numerical value and calculating the error of the time and place of the event). For the simulation layer, the future traffic situation predicted based on the current model is compared with the subsequent traffic situation actually observed to calculate the mean square error. The moving average method is used to smooth these mean square errors to highlight the error change trend. By drawing a curve of the mean square error changing with time, the error size and change trend are intuitively displayed; thus, the performance of the multi-layer situation simulation model can be evaluated in real time and quantitatively, providing an objective basis for the subsequent adjustment of the grid granularity. By analyzing the error size and change trend, inaccurate simulation predictions can be discovered in a timely manner, so that the grid granularity can be adjusted in a targeted manner to improve the accuracy of situation simulation.

[0131] The complexity and dynamics of the traffic network make it difficult to determine whether and how to adjust the grid size through simple rules or fixed thresholds. Fuzzy logic reasoning can handle uncertainty and fuzzy information, comprehensively consider multiple factors such as the error size, error change trend, and actual traffic conditions, and make decisions that are more in line with the actual situation. For example, when the error is large and on an upward trend, and the actual traffic conditions show that the traffic flow changes are complex, it is necessary to reduce the grid size to improve the accuracy of situation deduction. When the error is small and stable, and the actual traffic conditions are relatively stable, there is no need to adjust the grid size or the grid size can be appropriately increased to improve computational efficiency.

[0132] A fuzzy logic reasoning unit is constructed, which includes four parts: fuzzification, fuzzy rule base, fuzzy reasoning and defuzzification. The input variables such as error size, error change trend and actual traffic conditions are fuzzified. For example, the error size is divided into three fuzzy sets of "small", "medium" and "large", and the specific mean square error value is mapped to the membership of the corresponding fuzzy set through the membership function. Similarly, the error change trend is divided into "declining", "stable" and "increasing", and the actual traffic conditions are divided into fuzzy sets such as "simple", "general" and "complex", and the corresponding membership function is established.

[0133] Based on the experience of experts in the traffic field and the actual situation, a series of fuzzy rules are formulated. For example, Rule 1: If the error size is "large" and the error change trend is "increasing" and the actual traffic condition is "complex", then the grid granularity adjustment direction is "decreasing"; Rule 2: If the error size is "small" and the error change trend is "stable" and the actual traffic condition is "simple", then the grid granularity adjustment direction is "unchanged". Through a large number of experiments and data analysis, the fuzzy rule library is continuously optimized and improved.

[0134] Through the Mamdani reasoning method, fuzzy reasoning is performed according to the fuzzification results of the input variables and the fuzzy rule library. For example, for the input at a certain moment, the corresponding fuzzy rules are activated according to the fuzzified membership value, and the fuzzy set of the reasoning result is obtained by taking the smallest operation; the result obtained by fuzzy reasoning is defuzzified to obtain a specific grid granularity adjustment decision. The commonly used defuzzification method is the centroid method, which is to calculate the centroid position of the fuzzy set and convert it into a specific adjustment direction, such as "increase", "decrease", and "unchanged"; thereby, multiple complex factors can be comprehensively considered to make more reasonable and intelligent grid granularity adjustment decisions to adapt to the dynamic changes and uncertainties of the traffic network. Compared with the traditional method based on fixed thresholds or simple rules, fuzzy logic reasoning can more flexibly respond to different traffic scenarios and improve the adaptability and accuracy of the situation deduction model.

[0135] After actual adjustments are made to the grid granularity adjustment direction determined by fuzzy logic reasoning, it is necessary to perform multi-layer situation deduction again. By comparing the error sizes before and after the adjustment, the effect of the adjustment can be evaluated. If the error is significantly reduced after the adjustment, it means that the adjustment direction is correct and the grid granularity is more suitable for the current traffic network situation. If the error does not improve or even increases, it is necessary to reconsider the adjustment strategy, readjust the grid granularity or check other parameter settings of the situation deduction model.

[0136] If the grid granularity adjustment direction is "increase", then in the geographic information system environment, the original grid units are merged, for example, four adjacent small grid units are merged into a large grid unit, and the basic field strength and space-time conduction entropy and other related parameters of the merged grid unit are recalculated; if the adjustment direction is "decrease", the original grid unit is subdivided, for example, a large grid unit is divided into four small grid units, and the related parameters of the subdivided grid units are also recalculated. After completing the grid granularity adjustment and parameter calculation, the multi-layer situation simulation is run again, and the mean square error between the adjusted simulation results and the actual traffic conditions is recorded, and the adjusted mean square error is compared with the mean square error before the adjustment. If the adjusted mean square error is smaller than the mean square error before the adjustment and meets certain threshold conditions, such as the error reduction is greater than 10%, the adjustment is considered to be effective and the current grid granularity is maintained; if the condition is not met, the fuzzy logic reasoning step is returned to re-evaluate whether the grid granularity needs to be adjusted and the adjustment direction, or other parameters of the situation simulation model are optimized.

[0137] Through actual adjustment and effect evaluation, the grid granularity can be continuously optimized to improve the accuracy and adaptability of multi-layer situation deduction. This iterative optimization process can make the traffic network situation data display method better adapt to different traffic scenarios and changes, and provide more reliable support for traffic management and decision-making. After grid granularity adjustment and effect evaluation, the final grid granularity determined will affect the subsequent entire traffic network situation data display process. The appropriate grid granularity can make the situation field strength map generation, situation feature tensor extraction, layered rendering instruction determination and multi-layer situation deduction more accurate and efficient, thereby improving the quality and practicality of traffic network situation data display.

[0138] Example 2 like Figure 3 As shown, a system module diagram of a traffic network situation data display system is provided for an embodiment of the present application, and the system includes a topology generation module, a map generation module, a layered rendering module and a feedback adjustment module.

[0139] The topology generation module is used to obtain multi-source heterogeneous data in the traffic network, including traffic flow data, radar trajectory data, and event propagation data. The missing multi-source heterogeneous data is interpolated in real time through the adversarial generative network, and the interpolated multi-source heterogeneous data is synchronized in time and space to generate the traffic network topology.

[0140] The map generation module is used to decompose the traffic network topology into grid units based on the grid granularity, calculate the spatiotemporal conduction entropy between adjacent grid units, and determine the basic field strength of each grid unit according to the spatiotemporal conduction entropy between adjacent grid units to generate a situational field strength map of the traffic network;

[0141] The hierarchical rendering module is used to extract the situation feature tensor of the situation field strength map. The situation feature tensor includes situation layer features, event layer features and deduction layer features. The situation feature tensor is visually encoded according to the basic field strength of each grid unit to determine the hierarchical rendering instructions.

[0142] The feedback adjustment module is used to obtain eye tracking data to dynamically adjust the basic field strength of each grid unit, and to semantically describe and position the layered rendering instructions through natural language processing and annotation positioning algorithm, to perform multi-layer situation deduction of the traffic network according to the layered rendering instructions, and to dynamically adjust the grid granularity according to the error of the multi-layer situation deduction.

[0143] The principle of the above system can be found in the implementation steps of the above method, which will not be described in detail here.

Claims

1. A method for displaying traffic network situation data, characterized in that: include: Acquire multi-source heterogeneous data in the traffic network, including traffic flow data, radar trajectory data, and event propagation data, interpolate the missing multi-source heterogeneous data in real time through a generative adversarial network, and synchronize the interpolated multi-source heterogeneous data in time and space to generate the traffic network topology; Based on the grid granularity, the traffic network topology is decomposed into grid units, the spatiotemporal conduction entropy between adjacent grid units is calculated, and the basic field strength of each grid unit is determined according to the spatiotemporal conduction entropy between adjacent grid units to generate the situation field strength map of the traffic network; Extracting the situation feature tensor of the situation field strength map, the situation feature tensor includes situation layer features, event layer features and deduction layer features, and visually encoding the situation feature tensor according to the basic field strength of each grid unit to determine the layered rendering instructions; Eye tracking data is obtained to dynamically adjust the basic field strength of each grid unit, and the layered rendering instructions are semantically described and labeled with positions through natural language processing and annotation positioning algorithms. Multi-layer situation deduction of the traffic network is performed according to the layered rendering instructions, and the grid granularity is dynamically adjusted according to the error of the multi-layer situation deduction.

2. A method for displaying traffic network situation data according to claim 1, characterized in that: The generation logic of the traffic network topology includes: Acquire multi-source heterogeneous data including traffic flow data, radar trajectory data and event propagation data from the traffic network; Real-time interpolation of missing multi-source heterogeneous data through generative adversarial networks; The interpolated multi-source heterogeneous data are synchronized in time and space according to the constructed spatial reference points and time reference points; The multi-source heterogeneous data that have been synchronized in time and space are used to determine the connection relationship of nodes and the weight of edges in the transportation network through a topological algorithm to generate the transportation network topology.

3. A method for displaying traffic network situation data according to claim 2, characterized in that: The generation logic of the situation field strength map includes: Based on the grid granularity, the traffic network topology is decomposed into grid units, and the spatiotemporal conduction entropy between adjacent grid units is calculated; The space-time conduction entropy between adjacent grid cells is used as a weight to determine the basic field strength of each grid cell by weighted average. The basic field strength and space-time conduction entropy of each grid unit are integrated to generate a situation field strength map, which is then visualized.

4. A method for displaying traffic network situation data as claimed in claim 3, characterized in that: The decomposition sub-logic of the grid unit includes: The convolutional neural network is used to extract the features of the traffic network topology and identify the traffic network type; Determine the grid granularity according to the type of traffic network; The traffic network topology is decomposed into grid cells based on the grid granularity, and the boundary of each grid cell is determined according to its position in the traffic network.

5. A method for displaying traffic network situation data as claimed in claim 4, characterized in that: The calculation sub-logic of the spatiotemporal conduction entropy between adjacent grid cells includes: Extract multi-source heterogeneous data for each grid cell from the traffic network topology; Analyze the causal relationship and spatial correlation of multi-source heterogeneous data between adjacent grid cells to obtain spatiotemporal correlation; The spatiotemporal conduction entropy between adjacent grid cells is calculated based on the spatiotemporal correlation relationship.

6. A method for displaying traffic network situation data according to claim 5, characterized in that: The extraction logic of the situation feature tensor includes: The situation field strength map is divided into situation layer, event layer and deduction layer; A multi-layer extraction unit is configured to extract situation layer features, event layer features and deduction layer features of the situation field strength map in parallel through the multi-layer extraction unit; The situation layer features, event layer features and inference layer features are fused through the attention mechanism to obtain the situation feature tensor.

7. A method for displaying traffic network situation data according to claim 6, characterized in that: The determination logic of the layered rendering instruction includes: The mapping relationship between the basic field strength and situational characteristic tensor of each grid unit is analyzed through a neural network; Visually encode the situation layer features, event layer features and deduction layer features respectively, and adjust the parameters of the visual encoding in combination with the basic field strength of each grid unit; According to the results of visual encoding and the mapping relationship between the basic field strength of each grid unit and the situation feature tensor, rendering instructions for the situation layer, event layer and deduction layer are generated to determine the hierarchical rendering instructions.

8. A method for displaying traffic network situation data according to claim 7, characterized in that: The adjustment logic of the basic field strength of each grid unit includes: Obtain the user's eye tracking data in real time, including gaze point location, gaze time, and scan path, and map the eye tracking data to grid cells; The grid units that users pay attention to are identified through clustering algorithms, and the degree of attention of each grid unit is obtained; Determine the adjustment direction and adjustment amplitude of the basic field strength of each grid unit according to the attention degree of each grid unit; According to the adjustment direction and adjustment amplitude of the basic field strength of each grid unit and the current basic field strength of each grid unit, the adjusted basic field strength of each grid unit is obtained.

9. A method for displaying traffic network situation data according to claim 8, characterized in that: The grid granularity adjustment logic includes: In the process of multi-layer situation deduction of the traffic network according to the layered rendering instructions, the error size and error change trend between the deduction results of the situation layer, event layer and deduction layer and the actual traffic conditions are recorded in real time through the mean square error; Through fuzzy logic reasoning method, according to the error size and error change trend and actual traffic conditions, determine whether the grid size needs to be adjusted and the direction of adjustment of the grid size; The grid size is adjusted according to the adjustment direction of the grid size, and multi-layer situation deduction is performed again to compare the error size before and after the adjustment of the grid size to determine whether the grid size should be readjusted.

10. A traffic network situation data display system, characterized in that: include: Topology generation module, graph generation module, layered rendering module and feedback adjustment module; The topology generation module is used to obtain multi-source heterogeneous data in the traffic network, including traffic flow data, radar trajectory data, and event propagation data. The missing multi-source heterogeneous data is interpolated in real time through the adversarial generative network, and the interpolated multi-source heterogeneous data is synchronized in time and space to generate the traffic network topology. The map generation module is used to decompose the traffic network topology into grid units based on the grid granularity, calculate the spatiotemporal conduction entropy between adjacent grid units, and determine the basic field strength of each grid unit according to the spatiotemporal conduction entropy between adjacent grid units to generate a situational field strength map of the traffic network; The hierarchical rendering module is used to extract the situation feature tensor of the situation field strength map. The situation feature tensor includes situation layer features, event layer features and deduction layer features. The situation feature tensor is visually encoded according to the basic field strength of each grid unit to determine the hierarchical rendering instructions. The feedback adjustment module is used to obtain eye tracking data to dynamically adjust the basic field strength of each grid unit, and to semantically describe and position the layered rendering instructions through natural language processing and annotation positioning algorithm, to perform multi-layer situation deduction of the traffic network according to the layered rendering instructions, and to dynamically adjust the grid granularity according to the error of the multi-layer situation deduction.

Images