Multi-source traffic operation data analysis method and system based on AI sensor fusion

Abstract

The present invention provides a multi-source traffic operation data analysis method and system based on AI sensor fusion, which can achieve accurate reconstruction and dynamic monitoring of the spatial scenes of interchanges on highway sections. Traffic operation data streams are generated through AI sensor fusion acquisition technology, and then a detailed traffic dynamic change map is parsed. At the same time, through block mapping and fine extraction of road domain perception data, driving behavior data blocks and non-driving behavior data blocks are effectively separated, providing a data basis for in-depth analysis of factors affecting traffic status. The further optimized traffic dynamic change map, combined with the spatial scenes of interchanges, can accurately estimate traffic safety risk trends and generate forward-looking traffic safety risk trend data. This can thus generate targeted safety tips and suggestions in a timely manner, significantly improving the traffic safety management level and emergency response capabilities of highway sections, effectively preventing and reducing the occurrence of traffic accidents, and ensuring public travel safety.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a multi-source traffic operation data analysis method and system based on AI sensor fusion. Background Art

[0002] With the acceleration of urbanization and the continuous expansion of transportation networks, highways, as crucial links connecting cities, are experiencing increasing traffic volumes and increasingly complex traffic conditions. Interchanges within highways, in particular, often become traffic bottlenecks and accident-prone areas due to their complex structures and frequent traffic flow. Therefore, in-depth analysis of traffic operation data within highway interchanges to understand traffic conditions in real time and predict potential risks is crucial for improving traffic management efficiency and ensuring driving safety.

[0003] Related technologies often rely on single data sources, such as traffic flow monitoring data and vehicle speed data. While these data can provide a certain level of traffic status information, they fail to fully reflect the complexity and dynamic nature of traffic operations. Furthermore, these technologies are inefficient when processing large amounts of data, making it difficult to achieve real-time analysis of traffic conditions and risk prediction. Summary of the Invention

[0004] In view of the above-mentioned problems, in combination with the first aspect of the present invention, an embodiment of the present invention provides a multi-source traffic operation data analysis method based on AI sensor fusion, the method comprising:

[0005] Reconstruct the interchange space scene of the highway section to generate the interchange space scene. Perform AI sensor fusion collection on the highway section based on the interchange space scene to generate a traffic operation data stream. Then, analyze the traffic dynamic change map of the interchange space scene based on the traffic operation data stream to generate a traffic dynamic change map.

[0006] Mapping road domain perception data blocks on the highway section to generate road domain block perception data, extracting driving behavior data blocks from the road domain block perception data to generate driving behavior data blocks, and extracting non-driving behavior data blocks from the road domain block perception data to generate non-driving behavior data blocks;

[0007] Analyze the traffic status influencing factors of the highway section based on the driving behavior data block and the non-driving behavior data block to generate the traffic status influencing factors;

[0008] The traffic dynamic change map is optimized based on the road block perception data and traffic status influencing factors to generate an optimized traffic dynamic change map. The traffic safety risk trend of the expressway section is estimated based on the optimized traffic dynamic change map and the interchange space scenario to generate traffic safety risk trend data. Based on the traffic safety risk trend data, safety tips and suggestions for the interchange space scenario are generated.

[0009] On the other hand, an embodiment of the present invention also provides a multi-source traffic operation data analysis system based on AI sensor fusion, including a processor and a machine-readable storage medium, wherein the machine-readable storage medium is connected to the processor, the machine-readable storage medium is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the machine-readable storage medium to implement the above method.

[0010] Based on the above aspects, the embodiment of the present application can achieve accurate reconstruction and dynamic monitoring of the interchange space scene of the highway section. Through AI sensor fusion acquisition technology, the traffic operation data stream is generated, and then a detailed traffic dynamic change map is parsed. At the same time, through block mapping and fine extraction of road domain perception data, the driving behavior data block and non-driving behavior data block are effectively separated, providing a data basis for in-depth analysis of traffic status influencing factors. The further optimized traffic dynamic change map, combined with the interchange space scene, can accurately estimate the traffic safety risk trend and generate forward-looking traffic safety risk trend data. This can generate targeted safety tips and suggestions in a timely manner, significantly improving the traffic safety management level and emergency response capabilities of the highway section, effectively preventing and reducing the occurrence of traffic accidents, and ensuring the safety of public travel. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 It is a schematic diagram of the execution flow of the multi-source traffic operation data analysis method based on AI sensor fusion provided in an embodiment of the present invention.

[0012] Figure 2 This is a schematic diagram of the hardware architecture of a multi-source traffic operation data analysis system based on AI sensor fusion provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0013] The present invention will be described in detail below with reference to the accompanying drawings. Figure 1 This is a flow chart of a multi-source traffic operation data analysis method based on AI sensor fusion provided by an embodiment of the present invention. The multi-source traffic operation data analysis method based on AI sensor fusion is introduced in detail below.

[0014] Step S110: Reconstruct the interchange space scene of the highway section to generate the interchange space scene, perform AI sensor fusion collection on the highway section based on the interchange space scene to generate a traffic operation data stream, and analyze the traffic dynamic change map of the interchange space scene based on the traffic operation data stream to generate a traffic dynamic change map.

[0015] In this embodiment, the server starts the operation after receiving a task instruction such as reconstructing the spatial scene of the interchange on the highway section. For a specific highway section, for example, a section of the G1 highway connecting multiple cities includes an interchange. The server first performs a partitioned drone aerial scan of the highway section. The server dispatches multiple drones equipped with high-definition cameras to scan according to a pre-set partitioning scheme, such as one partition per square kilometer. During the flight, the drone continuously captures information such as the terrain, road structure, layout of the interchange, and surrounding landforms of the highway section to generate partitioned drone aerial scanning data. For example, the data of a certain partition contains information such as the number of interchange ramps in the area, the angle of the curve, the connection method with the adjacent road sections, and whether there are mountains, rivers, and other landform features that affect the traffic layout in the surrounding area.

[0016] Next, the server fuses the numerous drone aerial scans from these zones. This process involves data calibration and splicing. The server analyzes key information from each zone, such as geographic coordinates and road connectivity, and integrates them into the target drone aerial scan data. For example, by identifying the continuity of roads in adjacent zones and the connection between interchange ramps, the server ensures that the fused data accurately reflects the complete interchange structure of the entire highway section.

[0017] Then, the interchange spatial scene of the highway section is reconstructed based on the target drone aerial scan data. Using specialized modeling software, the server constructs the interchange spatial scene based on parameters such as road dimensions, interchange height, and slope from the aerial scan data. This scene is a 3D model that includes detailed information such as the road entity, spatial coordinates, and traffic sign locations. For example, the model clearly shows how lanes of different directions intersect and diverge at the interchange, as well as the location and direction of traffic lights.

[0018] Afterwards, AI sensor fusion data is collected on the highway section based on the interchange space scenario. The server first configures traffic observation points on the highway section. For example, sensor observation points are set at key locations such as the entrance, exit, curve, and lane merging and separation points of the interchange. These sensors include video surveillance sensors, vehicle speed sensors, traffic flow counters, etc., and the data collected by each of them constitutes the traffic observation point data. Then, the longitudinal traffic section of the section is analyzed for the interchange space scenario based on the traffic observation point data. The server analyzes the vehicle's driving direction, lane distribution, vehicle spacing and other information on the longitudinal section where each observation point is located to generate traffic observation point section data. At the same time, based on the traffic observation point data, the highway section is monitored for traffic flow at the observation point, for example, the number of vehicles passing through each observation point per hour is counted by the traffic flow counter to generate the observation point traffic flow data.

[0019] Next, the interchange traffic flow is calculated based on the cross-sectional data and traffic flow data of the traffic observation points. The server uses an algorithm to take into account the distribution and flow speed of vehicles in different cross-sectional areas, calculates the traffic flow situation within the entire interchange, and generates the observation point traffic flow data. For example, the server calculates the volume of vehicles entering the interchange from a certain entrance ramp and heading to a specific exit, as well as the number of vehicles transferring between different lanes. Finally, based on the observation point traffic flow data and the observation point traffic flow data, the observation point data of the highway section is fused to generate a traffic operation data stream. This traffic operation data stream contains comprehensive traffic information from each observation point along the entire highway section, such as traffic volume, speed distribution, and vehicle type ratio in different time periods.

[0020] Then, based on the traffic operation data stream, the traffic dynamic change map of the interchange space scene is analyzed. First, the traffic operation data stream is analyzed and analyzed for traffic flow and speed. By analyzing data from different time periods, the server determines the difference in traffic flow during peak and off-peak periods, as well as the speed change patterns of different road sections, and generates traffic flow characteristic data. For example, it was found that the traffic flow from the city to the suburbs is peak between 8 and 10 am every day, and the speed drops significantly at certain bends in the interchange. Then, based on the traffic flow characteristic data, the speed impact characteristics of the interchange space scene are simulated. The server simulates the speed changes of vehicles when passing through the interchange under different traffic flows, taking into account the impact of factors such as road slope and bend radius on vehicle speed, and generates speed impact characteristic data.

[0021] Next, a map is constructed of the traffic flow characteristic data and the vehicle speed impact characteristic data to generate a traffic flow change characteristic map. This traffic flow change characteristic map visually displays the changing relationship between traffic flow and vehicle speed in the interchange space scenario. For example, different colored lines represent the speed change trends in different lanes, and the thickness of the lines represents the amount of traffic flow. The traffic operation data stream is then analyzed for traffic congestion and unobstructed conditions. The server determines areas and time periods of traffic congestion and unobstructed traffic based on indicators such as traffic flow, speed, and vehicle queue length, and generates traffic status data. For example, when the traffic flow at a certain exit ramp is excessive and the speed is below a certain threshold, it is determined to be congested.

[0022] Based on the traffic status data, a traffic situation simulation is then performed for the interchange scenario. The server simulates the spread and impact of traffic congestion or traffic flow within the interchange, generating traffic situation data. For example, the server simulates the impact of congestion on a particular on-ramp on the entire interchange and adjacent road sections. Finally, the traffic flow change characteristic map and traffic situation map are combined to generate a traffic dynamics map. This map comprehensively reflects the dynamic changes in traffic flow, vehicle speed, congestion status, and other information within the interchange scenario.

[0023] Step S120, mapping the road domain perception data blocks of the highway section to generate road domain block perception data, extracting the driving behavior data blocks from the road domain block perception data to generate driving behavior data blocks, and extracting the non-driving behavior data blocks from the road domain block perception data to generate non-driving behavior data blocks.

[0024] In this embodiment, the server performs road-domain awareness data block mapping operations on expressway sections. For example, for the G1 expressway section mentioned earlier, the server utilizes multiple video surveillance devices deployed along the expressway to conduct road-domain detection video surveillance of the expressway section. These video surveillance devices cover different road sections, including straight sections, curves, and interchanges, continuously capturing video streams to generate expressway section road-domain detection video streams. The video streams contain information such as vehicle trajectories, lane occupancy, and traffic sign visibility.

[0025] The highway section road detection video stream is then mapped to road area blocks. The server divides the entire highway section into multiple blocks, for example, one block every 100 meters. By analyzing the geographic location information in the video stream (such as by comparing it with a known road coordinate system), the vehicles and road elements in the video stream are mapped to the corresponding road area blocks, generating road area block perception data. For example, the data of a certain road area block includes information such as the number of vehicles in the area, the direction of vehicle travel, and whether there is any abnormal parking.

[0026] Next, the driving behavior data blocks are extracted from the road block perception data. First, driving operation features are extracted from the road block perception data. The server first extracts block dynamic features from the road block perception data. For example, it analyzes the dynamic behaviors of vehicles in each block, such as acceleration, deceleration, and lane changes, to generate block perception dynamic feature data. Based on this block perception dynamic feature data, the road block perception data is then analyzed for traffic area dynamic features. The dynamic behavior characteristics of vehicles in different traffic areas (such as acceleration lanes, main lanes, and overtaking lanes) are determined to generate traffic area dynamic features.

[0027] Afterwards, the road block perception data is subjected to non-traffic area feature extraction based on the dynamic characteristics of the traffic area. For example, the behavioral characteristics of vehicles in non-traffic areas such as service area entrances and emergency parking strips are analyzed to generate non-traffic area feature data. Regular driving pattern features are then extracted from the non-traffic area feature data. For example, the deceleration and parking patterns of vehicles at the service area entrances are analyzed to generate regular driving pattern feature data. Based on the regular driving pattern feature data, the road block perception data is subjected to regular driving area feature analysis. For example, regular features such as the common speed range and following distance of vehicles in normal driving areas (such as the main lane) are determined to generate regular driving area features. Finally, the dynamic characteristics of the traffic area and the regular characteristics of the normal driving area are fused to generate driving operation feature data.

[0028] Based on the driving operation characteristic data, driving behavior data blocks are extracted from the road segment perception data to generate driving behavior data blocks. For example, vehicle data with similar driving operation characteristics (such as frequent lane changes and stable following) are grouped together as a driving behavior data block. Simultaneously, based on the driving operation characteristic data, non-driving behavior data blocks are extracted from the road segment perception data to generate non-driving behavior data blocks. For example, data on road facility status unrelated to vehicle travel (such as damaged traffic signs and road surface conditions), abnormal parking, and pedestrian intrusions are grouped together as non-driving behavior data blocks.

[0029] Step S130 , analyzing traffic state influencing factors of the highway section based on the driving behavior data block and the non-driving behavior data block to generate traffic state influencing factors.

[0030] In this embodiment, the server analyzes traffic state influencing factors for highway sections based on driving behavior data blocks and non-driving behavior data blocks. First, the driving behavior data blocks are analyzed for normal driving feature vectors. For example, for the normal driving data in the driving behavior data blocks, the server analyzes parameters such as the vehicle's average speed, speed fluctuation range, and average following distance to generate a normal driving feature vector.

[0031] Based on the normal driving behavior feature vector, the server analyzes factors influencing traffic flow on highway sections. For example, if the average speed of vehicles in a certain area is high with minimal speed fluctuations and reasonable following distance, the server will determine that driving behavior in this area has a positive impact on traffic flow on the section and generate a driving behavior impact vector. This vector may include factors affecting traffic speed, traffic stability, and other aspects.

[0032] The non-driving behavior data block is then analyzed for abnormal parking ratios. The server counts the number of abnormal parking events in the non-driving behavior data block and compares it with the total number of vehicles to generate an abnormal parking ratio. For example, if there are 5 abnormal parking events on a certain road section within a certain time period and the total number of vehicles is 1,000, the abnormal parking ratio is 0.5%.

[0033] The server analyzes the congestion impact vector for each highway section based on the abnormal parking ratio. If the abnormal parking ratio is high, the server determines that the section is more likely to experience congestion and generates a congestion impact vector. This vector includes factors that influence the likelihood of congestion and the extent of congestion.

[0034] Next, factors influencing traffic flow on the highway sections during normal driving are analyzed based on the non-driving zones. For example, the server identifies the distribution of traffic facilities within the non-driving zones. Suppose, for example, a section of road is found to have inappropriate traffic signs or excessively long signal cycles. The server calculates the signal cycle for the section based on the traffic facility distribution and generates a signal cycle. Based on the signal cycle, the server then analyzes the extended travel time for the highway section. Excessively long signal cycles increase vehicle wait times and extend travel time.

[0035] The server further analyzes the incremental lane change rate for each highway section based on the extended traffic time data. Due to the extended traffic time, vehicles may frequently change lanes to find faster routes, generating incremental lane change data. Based on this incremental lane change data, the server calculates the reduction in the number of vehicles on the highway that are normally traveling. Because increased lane changes may result in a decrease in the number of vehicles on the highway, this reduction in the number of vehicles on the highway is also generated. Finally, based on this reduction in the number of vehicles on the highway, the server analyzes factors influencing traffic flow on the highway section, generating a vector for the impact of normal traffic flow.

[0036] Finally, the congestion impact vector and the normal driving impact vector are weighted and fused to generate the non-driving behavior impact vector. Based on actual traffic conditions and empirical data, the server assigns different weights to the congestion impact vector and the normal driving impact vector. For example, during peak traffic hours, the congestion impact vector may be weighted higher. The two are then weighted and summed to generate the non-driving behavior impact vector. The driving behavior impact vector and the non-driving behavior impact vector are then fused to generate the traffic state impact factor. This traffic state impact factor comprehensively considers the impact of both driving and non-driving behaviors on the traffic state of the highway section.

[0037] Step S140: Optimize the traffic dynamic change map based on the road block perception data and traffic status influencing factors to generate an optimized traffic dynamic change map, and estimate the traffic safety risk trend of the expressway section based on the optimized traffic dynamic change map and the interchange space scenario to generate traffic safety risk trend data, and generate safety tips and suggestions for the interchange space scenario based on the traffic safety risk trend data.

[0038] In this embodiment, the server optimizes the traffic dynamic change map based on road segment perception data and traffic state influencing factors. First, road infrastructure is extracted from the road segment perception data. For example, various road infrastructure features such as curve type (e.g., sharp bend, gentle bend), road width, and traffic signs (e.g., speed limit signs, lane indicators) are identified from the road segment perception data.

[0039] The traffic dynamics map is optimized based on the characteristics of various road facilities and traffic status influencing factors. For example, if a road section has sharp curves and the traffic status influencing factors indicate that it is prone to congestion, the server will enhance the identification of this section in the traffic dynamics map, such as using more eye-catching colors to indicate areas with higher congestion risk. At the same time, detailed information on the impact of curves on vehicle speeds will be added to the map to generate an optimized traffic dynamics map.

[0040] Next, based on the optimized traffic dynamics map, the server simulates the impact of traffic flow trends on the interchange space. Based on the traffic information in the optimized map, the server simulates the flow of vehicles in the interchange space under different time periods and traffic conditions, generating a trend model of traffic flow impacting highway sections. For example, during peak holiday periods, the server predicts the diversion of vehicles from a specific city at the interchange, as well as potential congestion points.

[0041] Based on the traffic flow impact model for expressway sections, traffic safety risk trends are estimated for each expressway section, generating traffic safety risk trend data. For example, if the model indicates that a certain exit ramp is prone to vehicle collisions due to excessive traffic volume and limited road infrastructure, the server will generate corresponding traffic safety risk trend data, including the probability of an accident and the number of vehicles potentially affected.

[0042] Finally, based on the traffic safety risk trend data, the server generates safety tips and suggestions for the interchange space scenario. For example, if the traffic safety risk trend data for a certain area indicates a high accident risk, the server will generate safety tips and suggestions such as "There is heavy traffic and sharp curves at Interchange Exit A. Please slow down and pay attention to the distance between vehicles." These suggestions can be communicated to drivers through traffic broadcasts, electronic display screens, and other means.

[0043] Based on the above steps, the embodiment of the present application can achieve accurate reconstruction and dynamic monitoring of the interchange space scene of the highway section. Through AI sensor fusion acquisition technology, the traffic operation data stream is generated, and then a detailed traffic dynamic change map is parsed. At the same time, through block mapping and fine extraction of road domain perception data, the driving behavior data block and the non-driving behavior data block are effectively separated, providing a data basis for in-depth analysis of traffic status influencing factors. The further optimized traffic dynamic change map, combined with the interchange space scene, can accurately estimate the traffic safety risk trend and generate forward-looking traffic safety risk trend data. This can generate targeted safety tips and suggestions in a timely manner, significantly improving the traffic safety management level and emergency response capabilities of the highway section, effectively preventing and reducing the occurrence of traffic accidents, and ensuring the safety of public travel.

[0044] In a possible implementation, step S110 includes:

[0045] Step S111: perform a partitioned drone aerial photography scan on the highway section to generate partitioned drone aerial photography scan data.

[0046] Step S112: fusing the partitioned UAV aerial scanning data to generate target UAV aerial scanning data.

[0047] Step S113 , reconstructing the interchange space scene of the highway section based on the target UAV aerial photography scanning data to generate the interchange space scene.

[0048] Step S114: Perform AI sensor fusion data collection on the highway section based on the interchange space scenario to generate a traffic operation data stream.

[0049] Step S115 , performing traffic dynamic change graph analysis on the interchange space scene based on the traffic operation data stream to generate a traffic dynamic change graph.

[0050] In a possible implementation, step S114 includes:

[0051] Step S1141: configure traffic observation points on the expressway section to generate traffic observation point data, perform longitudinal traffic section analysis on the interchange space scene based on the traffic observation point data to generate traffic observation point section data, and monitor the traffic flow at the observation points on the expressway section based on the traffic observation point data to generate traffic flow data at the observation points.

[0052] Step S1142 , performing interchange traffic flow calculation on the traffic observation point data based on the traffic observation point section data and the observation point vehicle flow data, and generating the observation point traffic flow data.

[0053] Step S1143 , performing observation point data fusion on the highway section based on the observation point vehicle flow data and the observation point traffic flow data to generate a traffic operation data stream.

[0054] In a possible implementation, step S115 includes:

[0055] Step S1151: perform traffic flow and speed analysis on the traffic operation data stream to generate traffic flow characteristic data; simulate the vehicle speed impact characteristics of the interchange space scene based on the traffic flow characteristic data to generate vehicle speed impact characteristic data; construct a map of the traffic flow characteristic data and the vehicle speed impact characteristic data to generate a traffic flow change characteristic map.

[0056] Step S1152: Analyze the traffic congestion and smooth traffic conditions of the traffic operation data stream to generate traffic status data.

[0057] Step S1153: Perform traffic situation simulation on the interchange space scene based on the traffic status data to generate traffic situation data, construct a map of the traffic status data and the traffic situation data to generate a traffic situation map.

[0058] Step S1154: The traffic flow change characteristic map and the traffic situation map are integrated to generate a traffic dynamic change map.

[0059] In this embodiment, for a specific highway section (such as a section of the G2 Expressway connecting multiple cities with high traffic volume and containing complex interchanges), the server first performs a drone aerial scan of the highway section in sections, generating segmented drone aerial scan data. The server pre-plans the scanning plan and divides the section into sections based on the length of the highway section, the complexity of the terrain, and the distribution of interchanges. For example, the entire highway section is divided into 500-meter sections. Multiple drones equipped with high-resolution cameras, lidar, and other equipment fly along predetermined routes, performing detailed scans of each section. The cameras can capture information such as the road's surface condition (such as cracks and potholes), the clarity of lane markings, and the location and status of traffic signs. The lidar can obtain three-dimensional spatial information of the road and surrounding terrain, including the height, slope, and curvature of the interchanges. The scan data for each section will contain this multi-dimensional information, forming segmented drone aerial scan data.

[0060] Next, the server fuses the partitioned drone aerial scanning data to generate target drone aerial scanning data. Since there may be some differences in the data of different partitions (such as overlapping data at the partition boundaries), the server needs to calibrate and integrate these data. It will align the data of each partition by identifying key features in the data (such as specific marking points on the road, iconic features of the terrain, etc.). At the same time, the data of the overlapping parts is optimized and the most accurate and complete data is selected for retention. For example, in the overlapping area of ​​adjacent partitions, if the data of one partition shows that the road width is 10 meters and the other partition shows 10.1 meters, the server will combine other relevant data (such as the continuity of the lane lines, etc.) to determine the final road width value. After such a series of processing, the data of all partitions are fused into a complete target drone aerial scanning data. This data can accurately and comprehensively reflect the condition of the entire highway section, especially the overall structure and detailed information of the interchange.

[0061] Then, the interchange space scene of the highway section is reconstructed based on the target drone aerial scanning data to generate an interchange space scene. The server uses professional 3D modeling software to construct the interchange space scene based on the various parameters in the target drone aerial scanning data. For each lane in the interchange, the lane width, length, slope and other information in the scan data are accurately constructed in the model; the connection relationship of the ramp, the turning radius and the connection point with the main road are also accurately modeled based on the data. At the same time, the location and orientation of traffic signs (such as speed limit signs, signs, etc.), the distribution of street lights and other surrounding facilities are also reflected in the model. This interchange space scene is a highly accurate 3D digital model, which provides a basic spatial framework for subsequent traffic analysis.

[0062] Next, AI sensor fusion is performed on the highway section based on the interchange spatial scenario to generate a traffic operation data stream. This process involves multiple sub-steps. First, the server configures traffic observation points on the highway section and generates traffic observation point data. The server sets up observation points at key locations based on the characteristics of the interchange spatial scenario. For example, observation points are set up at the interchange's entrance and exit ramps, converging points, diverging points, the start and end points of curves, and specific intervals on the main road. Each observation point is equipped with multiple types of sensors, such as video sensors, radar sensors, and geomagnetic sensors. Video sensors can capture information such as the vehicle's appearance, license plate number, and driving trajectory; radar sensors can accurately measure parameters such as vehicle speed and distance; and geomagnetic sensors can detect vehicle passage. The data collected by these sensors collectively constitute the traffic observation point data.

[0063] Based on traffic observation point data, the server analyzes longitudinal traffic sections of the interchange space to generate traffic observation point section data. The server analyzes the distribution and flow of vehicles in the longitudinal direction, centered around each observation point. For example, at an observation point located on an entrance ramp, the server analyzes information such as the queue length, inter-vehicle spacing, and vehicle acceleration for vehicles entering the main road from the ramp over a specific period of time. Furthermore, the server considers the impact of factors such as road slope and number of lanes on vehicle movement, integrating this information to generate traffic observation point section data.

[0064] Based on traffic observation point data, traffic flow at observation points on highway sections is monitored to generate observation point traffic flow data. Sensors at each observation point continuously collect vehicle flow information. For example, geomagnetic sensors can accurately count the number of vehicles passing through an observation point per unit time, and radar sensors can assist in verifying vehicle passage to avoid misjudgments. The server organizes this data to generate observation point traffic flow data, which includes the number of vehicles passing each observation point at different time periods (e.g., every 5 minutes, every 10 minutes, etc.), as well as the proportion of vehicle types (e.g., small cars, large trucks, etc.).

[0065] Based on the cross-sectional data and traffic flow data of the observation points, the server calculates the traffic flow of the interchange and generates the traffic flow data for the observation points. Using complex algorithms, the server combines the vehicle driving status in the cross-sectional data of the observation points with the traffic volume information in the traffic flow data of the observation points to calculate the traffic flow conditions within the entire interchange. For example, when calculating the traffic flow from an entrance ramp to an exit ramp, the merging, exiting, and lane changes of vehicles between the various observation points are taken into account, as well as the distribution of traffic flow between different lanes. By comprehensively analyzing these factors, the observation point traffic flow data is generated, which can reflect the flow direction, flow rate, and other information of traffic flow in each area of ​​the interchange.

[0066] Based on the observation point vehicle flow data and observation point traffic flow data, the observation point data of the expressway section is fused to generate a traffic operation data stream. The server fuses the previously generated observation point vehicle flow data and observation point traffic flow data. For example, the traffic flow information of each observation point is combined with information such as the traffic flow direction and speed in the area where the observation point is located, taking into account the connections and influences between different observation points. During the fusion process, the data is also quality checked and corrected to remove any erroneous data or outliers. The resulting traffic operation data stream contains comprehensive traffic information from each observation point along the entire expressway section, such as detailed information such as traffic flow, speed, and vehicle type distribution in different lanes and time periods.

[0067] Then, based on the traffic operation data stream, the traffic dynamic change map of the interchange space scene is analyzed to generate a traffic dynamic change map. First, the traffic flow and speed analysis of the traffic operation data stream is performed to generate traffic flow characteristic data. The server will conduct an in-depth analysis of the traffic flow and speed data in the traffic operation data stream. For example, analyze the traffic flow change pattern in different time periods within 24 hours a day to find out the peak and trough periods of traffic; at the same time, analyze the speed distribution of different road sections (such as different ramps of the interchange, different sections of the main road, etc.) to determine which areas are prone to slower speeds (such as curves, merging points, etc.). These analysis results are integrated to generate traffic flow characteristic data, which can clearly reflect the distribution characteristics of traffic flow and speed in time and space.

[0068] Based on traffic flow characteristic data, the server simulates the speed impact characteristics of the interchange space scenario to generate speed impact characteristic data. The server simulates the speed impact characteristics based on information such as traffic volume and lane distribution in the traffic flow characteristic data, combined with factors such as road slope and curve radius in the interchange space scenario. For example, on a ramp with a small curve radius, when traffic volume is high, the server simulates the reasonable speed range for vehicles and the speed change trend affected by traffic volume and road conditions. These simulation results form speed impact characteristic data, which accurately reflects the impact of various factors on vehicle speed.

[0069] Traffic flow characteristic data and vehicle speed impact characteristic data are mapped to generate a traffic flow change characteristic map. The server utilizes specialized map-building tools to visualize these data. In the traffic flow change characteristic map, different colors and lines are used to represent changes in traffic flow and speed in different lanes and areas. For example, red lines represent areas with heavy traffic and slow speeds, while blue lines represent areas with light traffic and fast speeds. The thickness of the lines indicates the volume of traffic. This map can intuitively demonstrate the changing trends of traffic flow in interchange spaces.

[0070] Traffic operation data streams are analyzed for traffic congestion and traffic flow to generate traffic status data. The server determines traffic congestion and traffic flow by analyzing multiple indicators in the traffic operation data stream, such as vehicle volume, vehicle speed, and vehicle queue length. For example, when the traffic volume in a certain area exceeds a certain threshold, the vehicle speed is lower than a set value, and the vehicle queue length reaches a certain standard, the area is judged to be congested; conversely, when the traffic volume is within a normal range and the vehicle speed is stable and high, the area is judged to be unobstructed. These judgment results and related quantitative indicators (such as the level of congestion and the duration of unobstructed periods) are compiled into traffic status data.

[0071] Based on traffic status data, traffic situation simulation is performed on the interchange space scenario to generate traffic situation data. The server simulates traffic situation in the interchange space scenario based on the congestion and smooth traffic information in the traffic status data. For example, when simulating congestion, the server displays the queueing pattern of vehicles in the congested area and the slow movement of vehicles. When simulating smooth traffic, the server displays the normal driving trajectories and lane utilization of vehicles. Through these simulations, traffic situation data is generated that reflects the propagation and evolution of traffic conditions in the interchange space scenario.

[0072] Traffic status data and traffic situation data are mapped to generate a traffic situation map. Similar to constructing a traffic flow change feature map, specialized tools are used to visualize traffic status data and traffic situation data into a traffic situation map. This map uses different colors and patterns to represent different traffic conditions (e.g., red for congestion, green for unimpeded traffic), and can show how traffic conditions change over time and across different regions.

[0073] Finally, the traffic flow change characteristic map and the traffic situation map are integrated to generate a traffic dynamic change map. The server integrates and fuses the information from the traffic flow change characteristic map and the traffic situation map. For example, by combining the vehicle speed and volume information from the traffic flow change characteristic map with the traffic congestion and unblocked status information from the traffic situation map, the traffic dynamic change map can comprehensively and accurately reflect the dynamic changes in traffic flow, speed, congestion, and other information within the interchange space. This traffic dynamic change map provides an important basis for subsequent traffic analysis, management, and decision-making.

[0074] In a possible implementation, step S120 includes:

[0075] Step S121: Perform road area detection video monitoring on the expressway section to generate a road area detection video stream for the expressway section.

[0076] Step S122 , performing road block mapping on the highway section road area detection video stream to generate road area block perception data.

[0077] Step S123 , extracting driving operation features from the road area block perception data to generate driving operation feature data.

[0078] Step S124 , extracting driving behavior data blocks from the road area block perception data based on the driving operation characteristic data to generate driving behavior data blocks.

[0079] In step S125 , non-driving behavior data blocks are extracted from the road area block perception data according to the driving operation characteristic data to generate non-driving behavior data blocks.

[0080] In a possible implementation, step S123 includes:

[0081] Step S1231 , extracting block dynamic features from the road block perception data to generate block perception dynamic feature data.

[0082] Step S1232 , performing traffic area dynamic feature analysis on the road area block perception data based on the block perception dynamic feature data to generate traffic area dynamic features.

[0083] Step S1233 , extracting non-traffic area features from the road block perception data based on the dynamic features of the traffic area to generate non-traffic area feature data.

[0084] Step S1234: extract regular driving pattern features from the non-traffic area feature data to generate regular driving pattern feature data.

[0085] Step S1235 , analyzing the regular characteristics of the normal driving area of ​​the road block perception data based on the regular driving pattern characteristic data to generate regular characteristics of the normal driving area.

[0086] Step S1236: Fusion the dynamic characteristics of the traffic area and the regular characteristics of the normal driving area to generate driving operation characteristic data.

[0087] In this embodiment, for a busy highway section (e.g., the G3 Expressway, which connects multiple major cities and has high traffic volume and complex road conditions), the highway section is first monitored using road detection video surveillance to generate a highway section detection video stream. Along the highway section, a server controls multiple high-definition video surveillance devices, which are carefully positioned to cover key areas of the highway, including interchanges, long straights, curves, and tunnel entrances and exits. Each video surveillance device features high resolution, a wide viewing angle, and the ability to adapt to varying lighting conditions. They continuously capture the highway section at a fixed frame rate (e.g., 30 frames per second), generating a highway section detection video stream. This video stream contains a wealth of information, such as each vehicle's trajectory, its position within the lane, its outline (which can be used to roughly determine vehicle type, such as sedan, truck, or bus), the distance between vehicles, the position of vehicles relative to traffic signs (e.g., lane markings and speed limit signs), and any visible objects on the road (e.g., scattered debris).

[0088] Next, the server maps the highway section's road detection video stream into road blocks, generating block-level perception data. The server divides the entire highway section into multiple blocks. This division can be based on geographic coordinates, fixed distance intervals, or structural features of the road (e.g., by each interchange). For example, each kilometer section is divided into a block. The server then analyzes the image information in the video stream, using image recognition technology and geographic positioning information to accurately map vehicles, traffic signs, and other road elements in the video stream to corresponding blocks. During this process, the server identifies the vehicle's coordinate location and, based on pre-defined block division rules, determines the block in which the vehicle is located. For traffic signs, the server also determines the block in which they are located and records the sign type (e.g., a 60 km / h speed limit sign is located in a specific block). Special road conditions (e.g., road surface damage within a block) are also marked. Thus, each block has corresponding perception data containing information about various road elements, generating block-level perception data.

[0089] Then, the server extracts the driving operation features of the road block perception data and generates driving operation feature data. This process includes multiple sub-steps. First, the server extracts the block dynamic features of the road block perception data and generates block perception dynamic feature data. The server analyzes the dynamic behavior of vehicles in each road block, such as the vehicle's acceleration, deceleration, and constant speed driving state changes, the vehicle's lane change operation (including the starting position of the lane change, the direction of the lane change, and the duration of the lane change), the vehicle's steering operation (such as the steering angle and steering speed at the curve), etc. By monitoring and analyzing these dynamic behaviors, the server can obtain the dynamic characteristics of vehicles in each road block, such as the average acceleration amplitude and lane change frequency of vehicles in a certain road block, thereby generating block perception dynamic feature data.

[0090] Based on the block-perception dynamic feature data, the server analyzes the dynamic characteristics of traffic areas and generates dynamic characteristics of traffic areas. Based on the previously obtained block-perception dynamic feature data, the server further analyzes the dynamic behavior characteristics of vehicles in different traffic areas (such as acceleration lanes, normal driving lanes, overtaking lanes, exit ramps, and entrance ramps). In acceleration lanes, the server focuses on vehicle acceleration, such as calculating the average time it takes for vehicles to reach the specified speed on the highway section. In normal driving lanes, the server analyzes vehicle speed stability and changes in following distance. In overtaking lanes, the server focuses on the frequency and safety of vehicle overtaking maneuvers (such as whether the distance to the preceding and following vehicles during overtaking meets safety standards). On exit and entrance ramps, the server analyzes the compliance of vehicle deceleration, merging, and exiting maneuvers. Through this analysis, the server can obtain the dynamic behavior patterns of vehicles in different traffic areas and generate dynamic characteristics of traffic areas.

[0091] Based on the dynamic characteristics of traffic areas, the server extracts features from non-traffic areas in the road segment perception data to generate non-traffic area feature data. Taking into account the dynamic characteristics of traffic areas, the server extracts features from non-traffic areas (such as service areas, emergency lanes, and road construction areas) within the road segment perception data. For service areas, the server analyzes the time distribution of vehicles entering and leaving the service area, the length of time they stay in the service area, and the vehicle's driving operations at the service area entrance and exit (such as deceleration, parking, and turning). For emergency lanes, the server monitors the reasons for vehicles stopping in emergency lanes (by analyzing the vehicle's previous driving status and surrounding environment, such as whether the stop was due to a sudden breakdown) and the duration of the stop. For road construction areas, the server monitors speed adjustments and lane changes when approaching and passing the construction area. This information on non-traffic areas is integrated to generate non-traffic area feature data.

[0092] Regular driving pattern features are extracted from the non-traffic area feature data to generate regular driving pattern feature data. The server conducts in-depth analysis of the non-traffic area feature data to identify regular driving patterns. For example, in service areas, the server may find that most vehicles will enter the service area to rest after driving a certain mileage (such as every 2-3 hours of driving mileage), and the stay time in the service area shows a certain distribution pattern (such as most vehicles stay for 15-30 minutes); in emergency parking strips, it is found that the proportion of vehicles stopping due to vehicle breakdowns and the waiting time for rescue for these breakdown vehicles also have certain patterns; for road construction areas, there are also certain common patterns in the speed adjustment amplitude and lane change method of vehicles when passing through the construction area. These patterns are summarized to generate regular driving pattern feature data.

[0093] Based on the regular driving pattern feature data, the road block perception data is analyzed for regular features of normal driving areas to generate regular features of normal driving areas. Based on the regular driving pattern feature data, the server further analyzes the regular features of vehicles in normal driving areas (such as the main lanes of highway sections). For example, by analyzing the speed distribution of vehicles in normal driving areas, it is found that most vehicles will maintain a certain speed range (such as 90-110 kilometers per hour) in the absence of special circumstances (such as traffic congestion, weather changes, etc.); in terms of following distance, the average following distance between vehicles also exists in a relatively stable range; at the same time, the lane keeping situation of vehicles (such as rarely changing lanes frequently during normal driving) also shows certain regularities. These regular features of normal driving areas are extracted to generate regular features of normal driving areas.

[0094] Finally, the dynamic characteristics of the traffic area and the regular characteristics of the normal driving area are fused to generate driving operation characteristic data. The server fuses the previously obtained dynamic characteristics of the traffic area with the regular characteristics of the normal driving area. For example, the fusion process considers the correlation between the vehicle's acceleration in different traffic areas (such as acceleration in the acceleration lane) and its speed stability in the normal driving area (such as speed stability in the main lane). If a vehicle accelerates too slowly or too quickly in the acceleration lane, it may affect its speed stability after entering the main lane. By comprehensively analyzing these factors, the relevant information from the dynamic characteristics of the traffic area and the regular characteristics of the normal driving area is integrated to generate driving operation characteristic data. This driving operation characteristic data comprehensively reflects the vehicle's driving operation characteristics in various areas of the entire highway section, including the comprehensive characteristics of the vehicle's acceleration, deceleration, lane changing, following, lane keeping, and other operations, providing an important data foundation for subsequent driving behavior analysis.

[0095] Based on the driving operation characteristic data, the server extracts driving behavior data blocks from the road segment perception data to generate driving behavior data blocks. The server classifies the road segment perception data based on similarities in the driving operation characteristic data, grouping vehicle data with similar driving operation characteristics into a single driving behavior data block. For example, vehicles with frequent lane changes, large speed fluctuations, and close following distances may be grouped into one driving behavior data block, potentially indicating aggressive driving behavior. Vehicles with stable speeds, few lane changes, and a moderate following distance may be grouped into another driving behavior data block, potentially indicating more stable driving behavior.

[0096] Based on the driving operation characteristic data, the non-driving behavior data blocks are extracted from the road area block perception data to generate non-driving behavior data blocks. The server also extracts the road area block perception data that is not related to the vehicle driving operation based on the driving operation characteristic data to form non-driving behavior data blocks. For example, information such as the damage of traffic signs on the road, the accumulation of water or snow on the road surface, and temporary traffic control due to non-vehicle reasons (such as the intrusion of wild animals) are classified as non-driving behavior data blocks. These data blocks help to conduct a comprehensive analysis of the traffic conditions on the highway section, taking into account not only the driver's driving behavior, but also non-driving related factors such as the road environment.

[0097] In a possible implementation, step S130 includes:

[0098] Step S131 , performing normal driving feature vector analysis on the driving behavior data block to generate a normal driving feature vector.

[0099] Step S132 , analyzing the traffic flow influencing factors of the highway section based on the normal driving characteristic vector of the driving behavior, and generating a driving behavior influencing traffic flow vector.

[0100] Step S133 , performing abnormal parking ratio analysis on the non-driving behavior data block to generate an abnormal parking ratio.

[0101] Step S134 , performing a road congestion impact vector analysis on the expressway section based on the abnormal parking ratio to generate a road congestion impact vector.

[0102] Step S135 , analyzing the traffic flow impact factors of the normal driving section on the highway section based on the non-driving behavior area, and generating a normal driving impact traffic flow vector.

[0103] Step S136 , weighting the road congestion impact vector and the normal driving impact traffic flow vector to generate a non-driving behavior impact traffic flow vector.

[0104] Step S137 , fusing the traffic flow vectors affected by driving behavior and the traffic flow vectors affected by non-driving behavior to generate a traffic state impact factor.

[0105] In this embodiment, for the aforementioned G3 expressway section, the server performs traffic state influencing factor analysis on the expressway section based on the driving behavior data block and the non-driving behavior data block to generate the traffic state influencing factor.

[0106] First, the server analyzes the driving behavior data block for normal driving feature vectors to generate a normal driving feature vector. The server then conducts a thorough analysis of the data within the driving behavior data block, focusing on various characteristics of the vehicle during normal driving. For example, for each driving behavior data block, the server calculates the average vehicle speed during normal driving. This speed is calculated by averaging the speed data of a large number of vehicles within the data block on normal driving sections (such as straight sections without traffic control, road construction, or other abnormal conditions). The server also calculates the standard deviation of the speed to measure speed fluctuation. A small standard deviation indicates relatively stable speed; a small standard deviation indicates significant speed fluctuation. The server also analyzes the following distance of the vehicle, calculating the average following distance and its distribution. For example, this analysis can reveal the proportion of vehicles whose following distances fall within a certain safe range. Lane keeping is also an important factor. The server calculates the proportion of time a vehicle remains in the same lane during normal driving. A high standard deviation indicates good lane keeping. These normal driving characteristics, such as average speed, speed fluctuation, following distance, and lane keeping, are combined to form a multidimensional vector, the normal driving behavior feature vector. This vector comprehensively describes the comprehensive characteristics of normal vehicle driving under a specific driving behavior data block.

[0107] Next, the server analyzes factors influencing traffic flow on the highway section based on the normal driving behavior feature vector, generating a driving behavior impact traffic flow vector. The server uses the previously generated normal driving behavior feature vector to analyze its impact on traffic flow on the highway section. For example, if the average speed of vehicles in a particular driving behavior data block is high with minimal speed fluctuation, this may indicate high vehicle efficiency and a faster overall traffic flow on that section. In this case, the server calculates the positive impact of this factor on traffic flow on the section based on a pre-defined algorithm and traffic flow model. For example, this may improve the section's capacity and reduce mutual interference between vehicles, thus reflecting a positive contribution to traffic speed and smoothness in the driving behavior impact traffic flow vector. Regarding following distance and lane keeping, if vehicles maintain a reasonable following distance and good lane keeping, this helps maintain traffic stability, reduces the risk of traffic accidents, and ultimately increases the section's traffic capacity. The server generates a driving behavior impact traffic flow vector by comprehensively considering the performance of these factors in the normal driving feature vector of driving behavior. This vector reflects the impact of driving behavior on the speed, stability and capacity of traffic flow on the road section.

[0108] The server then analyzes the non-driving behavior data block for abnormal parking ratios to generate an abnormal parking ratio. The server carefully combs through the data in the non-driving behavior data block, focusing on information related to abnormal parking. For example, the non-driving behavior data block records all parking events on the highway. The server determines which stops are abnormal based on information such as the reason, location, and time of the stop. Irregular stops may include parking in non-emergency lanes or parking without a reasonable reason (such as non-malfunction or non-emergency). The server counts these abnormal parking events and divides them by the total number of parking events (including normal and abnormal stops) to calculate the abnormal parking ratio. Suppose that within a certain time period, a total of 100 parking events occurred on the highway, and 10 of them were determined to be abnormal stops after analysis. Therefore, the abnormal parking ratio is 10%. This abnormal parking ratio is an important indicator that reflects the degree to which non-driving behavior factors disrupt traffic order on the highway.

[0109] Based on the abnormal parking ratio, the server analyzes the congestion impact vector for each highway section and generates a congestion impact vector. Once the server determines the abnormal parking ratio, it analyzes its impact on congestion on the highway section. A higher abnormal parking ratio generally increases the risk of congestion on the section. For example, if the abnormal parking ratio reaches 10%, the server will assess the impact on congestion on the section based on traffic flow theory and historical data. Abnormal parking may cause following vehicles to slow down or change lanes, disrupting normal traffic flow. The server considers the impact of this disruption on congestion-related factors such as traffic speed, traffic density, and queue length to construct the congestion impact vector. This vector may include components such as the increase in the probability of congestion formation, the impact on the spread of congestion, and the extent of the congestion impact on the section. For example, a 10% abnormal parking ratio may increase the probability of congestion formation by 20%, accelerate the spread of congestion by 10%, and potentially affect the 500-meter section surrounding the abnormal parking point.

[0110] Based on the non-driving zones, the server analyzes factors influencing traffic flow on normal driving sections of highways and generates a vector for traffic flow influencing normal driving. The server then conducts a thorough analysis of the impact of data related to the non-driving zones on traffic flow on normal driving sections. For example, the non-driving zones include the distribution of traffic facilities. The server first identifies the distribution of traffic facilities. For example, on a particular road section, the server identifies the distribution of traffic signs (such as speed limit signs and lane indicators) and discovers that speed limit signs in some sections are improperly set, causing vehicles to frequently decelerate or accelerate. The server also calculates the signal cycle duration of the road section based on the distribution of traffic facilities. For example, at the entrance or exit of a road section controlled by traffic lights, the server calculates the cycle duration of the traffic lights. If the signal cycle is long, vehicle waiting time increases, which affects traffic flow on normal driving sections. The server then analyzes traffic duration extension on highway sections based on the signal cycle. When the signal cycle is long, vehicles stay longer in the area, resulting in extended traffic duration. This extended traffic duration changes the spacing between vehicles, affecting the continuity of traffic flow.

[0111] The server further analyzes the incremental lane change rate for each highway section based on the extended traffic time data. Due to extended traffic times, vehicles may increase their lane changes due to impatience or the desire to find a faster route. The server generates incremental lane change data by analyzing the change in lane change frequency before and after the extended traffic time. For example, if, under normal traffic times, 10 out of every 100 vehicles on a particular highway section change lanes, but after the extended traffic time, 20 out of every 100 vehicles change lanes, the incremental lane change rate is 10. Based on this incremental lane change data, the reduction in the number of vehicles in normal traffic on the highway section is calculated. This is because increased lane changes may reduce the number of vehicles in normal traffic. For example, interference between vehicles during lane changes may cause some vehicles to slow down or even stop, reducing the number of vehicles in normal traffic. Assuming that the incremental lane change rate reduces the number of vehicles in normal traffic by 5%, this is the reduction in the number of vehicles in normal traffic. Finally, based on the reduction in the number of vehicles in normal traffic, the server analyzes the factors affecting traffic flow on the highway section and generates a vector for the impact of normal traffic flow. This vector reflects the impact of non-driving behavior area-related factors (such as traffic facilities and traffic duration) on the number of vehicles, traffic speed and stability of traffic on normal driving sections.

[0112] The congestion impact vector and the normal driving impact vector are then weighted and fused to generate the non-driving behavior impact vector. The server assigns different weights to the congestion impact vector and the normal driving impact vector based on actual traffic conditions and empirical data. For example, during peak traffic hours or on congested sections, the congestion impact vector may be given a higher weight because congestion has a more significant impact on traffic flow. Conversely, during periods or sections with relatively smooth traffic, the normal driving impact vector may be given a higher weight. Suppose, in a specific scenario, the congestion impact vector is weighted 0.6, and the normal driving impact vector is weighted 0.4. The server multiplies each component of the congestion impact vector (such as the increase in congestion probability and the impact of congestion spread) by 0.6, and each component of the normal driving impact vector (such as the reduction in the number of vehicles and the change in traffic speed) by 0.4. The server then adds the corresponding components together to generate the non-driving behavior impact vector. This vector comprehensively considers the impact of non-driving behavior factors on traffic flow in congested and normal driving conditions.

[0113] Finally, the driving behavior-influenced traffic flow vector and the non-driving behavior-influenced traffic flow vector are integrated to generate a traffic state impact factor. The server integrates the information from the driving behavior-influenced traffic flow vector and the non-driving behavior-influenced traffic flow vector. For example, the speed-related components of the driving behavior-influenced traffic flow vector are combined for consideration. If the normal speed of a vehicle in the driving behavior-influenced traffic flow vector contributes to increased traffic flow speed, while the non-driving behavior-influenced traffic flow vector indicates a decrease in traffic flow speed due to abnormal parking or traffic facility problems, the server will combine the impact of these two factors to determine the ultimate impact on traffic flow speed. The same applies to other aspects of traffic flow, such as stability and capacity. Through this comprehensive integration approach, a traffic state impact factor is generated. This factor comprehensively reflects the impact of driving and non-driving behavior on the traffic state of a highway section in terms of traffic flow speed, stability, capacity, and other aspects, providing an important basis for subsequent traffic management, planning, and risk assessment.

[0114] In a possible implementation, step S135 includes:

[0115] Step S1351: Identify the distribution of traffic facilities in the non-driving behavior area and generate a distribution of traffic facilities.

[0116] Step S1352: Calculate the road section signal period for the distribution of traffic facilities to generate the road section signal period.

[0117] Step S1353 , performing traffic time extension analysis on the expressway section according to the road section signal cycle, and generating expressway section traffic time extension data.

[0118] Step S1354: performing vehicle lane change increment analysis on the highway section based on the traffic duration extension data of the highway section to generate vehicle lane change increment data.

[0119] Step S1355, calculating the reduction amount of vehicles traveling normally on the highway section based on the vehicle lane change incremental data, and generating the reduction amount data of vehicles traveling normally.

[0120] Step S1356: analyzing the traffic flow influencing factors of the expressway section based on the reduced volume data of vehicles in normal driving, and generating a normal driving influencing traffic flow vector.

[0121] In this embodiment, for the aforementioned expressway sections (e.g., the G3 expressway section), the server first identifies the distribution of traffic facilities in the non-driving area and generates a traffic facility distribution map. The non-driving area of ​​the expressway section contains numerous traffic-related facilities. The server utilizes its stored detailed map data for the expressway section and image recognition technology (if supplemented by image information from monitoring equipment) to identify the distribution of these traffic facilities. For example, the server accurately determines the location of speed limit signs, the specific road section coordinates of each speed limit sign, and the road section range corresponding to different speed limit values. For lane indicators, the server identifies the type of sign for each lane (e.g., through lane, left turn lane, right turn lane sign) and its exact location within the road section. Traffic light distribution is also a crucial component, and the server records the specific location of each traffic light, whether it is located on an entrance ramp, exit ramp, or somewhere on the main road. Furthermore, the distribution of traffic facilities such as traffic monitoring equipment, emergency stop signs, and service area signs is also identified and recorded in detail. This information is integrated to generate the traffic facility distribution map. This traffic facility distribution data comprehensively and accurately describes the layout of various types of traffic facilities in the non-driving behavior area of ​​the expressway, providing a basis for subsequent analysis.

[0122] Next, the server calculates the signal cycle for each section of traffic infrastructure, generating a signal cycle. Within the distribution of traffic infrastructure, the signal cycle of traffic lights has a significant impact on traffic flow. The server calculates the signal cycle for each traffic light based on its pre-set control logic and relevant parameters. For example, for a traffic light located at the junction of a highway entrance ramp and the main road, the server determines that its red light duration is 30 seconds, its green light duration is 60 seconds, and its yellow light duration is 3 seconds. Therefore, a complete cycle for this light is 93 seconds. For complex traffic lights with multiple phases (such as those near large interchanges), the server analyzes the duration of each phase in detail and combines all phases to calculate the total signal cycle for the section. This calculation process requires interpreting the logic of the traffic light control program and distinguishing between different time periods (e.g., peak and off-peak hours may have different signal cycle settings). Through this calculation, the server accurately determines the signal cycle for each traffic light. This signal cycle data reflects the time-controlled characteristics of the traffic light.

[0123] Next, the server analyzes the extended travel time for each highway section based on the signal cycle, generating data on the extended travel time for that section. Once the server obtains the signal cycle, it analyzes its impact on the highway section's travel time. For example, if a traffic light has a longer signal cycle, vehicles will have to wait longer to pass through it. Suppose, without the influence of traffic lights, the average travel time for vehicles to pass through a particular section is 1 minute. However, due to the presence of a traffic light with a signal cycle of 93 seconds, the average waiting time at that light is 30 seconds (calculated based on the red light duration and the probability distribution of vehicle arrival at the light). Therefore, the actual travel time for vehicles passing through this section is extended to 1.5 minutes. The server performs this analysis on all highway sections affected by traffic lights, taking into account the randomness of different vehicles' arrival times at the light and the influence of different phases within the light cycle, and calculates the extended travel time for each affected section. Similar analysis is also performed for other traffic features that may affect travel time, such as speed bumps and special road signs that cause vehicle deceleration. By combining the traffic time extension caused by all these factors, we can generate the traffic time extension data of expressway sections, which can accurately reflect the additional increase in traffic time caused by traffic facilities.

[0124] Afterwards, the vehicle lane change increments on the highway section are analyzed based on the traffic duration extension data of the highway section to generate vehicle lane change increment data. The server uses the traffic duration extension data of the highway section to analyze the changes in vehicle lane change behavior. When the traffic duration is extended, the driving state of the vehicle will change. For example, under normal traffic duration, the vehicle can maintain a relatively stable lane, but due to the extension of traffic duration, the vehicle may become impatient or try to change lanes in order to find a faster passage. The server analyzes the driving trajectory data of a large number of vehicles on the highway section (these data can be obtained from monitoring equipment) and compares the differences in vehicle lane change behavior before and after the traffic duration is extended. Assuming that under normal circumstances, the number of lane changes for every 100 vehicles on a certain section is 10 times, and after the traffic duration is extended, this number becomes 20 times, then the vehicle lane change increment is 10 times. The server will perform such an analysis on each section of the highway, taking into account the impact of factors such as traffic flow and road structure in different sections on vehicle lane-changing behavior, and accurately calculate the incremental vehicle lane changes caused by the extension of traffic time on each section. By integrating these data, it generates vehicle lane-changing incremental data, which reflects the degree of impact of the extension of traffic time on vehicle lane-changing behavior.

[0125] Based on the lane change increment data, the server calculates the reduction in the number of vehicles traveling normally on the highway, generating the reduction in the number of vehicles traveling normally. Lane change increments affect the number of vehicles traveling normally. Frequent lane changes can trigger a series of chain reactions. For example, the need to maintain a safe distance between vehicles during lane changes can cause the vehicle behind to slow down or the vehicle in front to accelerate. This speed change can disrupt the original traffic balance. In some cases, interference between vehicles during lane changes can prevent some vehicles from traveling at their normal speeds, forcing them to stop and wait for the right opportunity to change lanes. The server calculates the reduction in the number of vehicles traveling normally by analyzing the lane change increment data and related data such as vehicle speed and spacing (which can be obtained from monitoring equipment or calculated using traffic flow models). Assuming that the lane change increments cause 5 out of every 100 vehicles to slow to a point where they cannot travel normally (for example, by falling below the minimum safe speed or being stopped for an extended period), the reduction in the number of vehicles traveling normally is 5%. The server will perform such calculations on each section of the highway, comprehensively considering the differences in traffic conditions on different sections, and generate accurate data on the reduction in the number of normally traveling vehicles. This data reflects the reduction in the number of normally traveling vehicles due to the increase in vehicle lane changes.

[0126] Finally, based on the data on the reduction in the number of vehicles in normal operation, the server analyzes factors affecting traffic flow on the expressway section and generates a normal operation impact traffic flow vector. The server uses the data on the reduction in the number of vehicles in normal operation to analyze its impact on traffic flow on the expressway section, thereby generating a normal operation impact traffic flow vector. The reduction in the number of vehicles in normal operation affects multiple aspects of traffic flow. For example, in terms of traffic speed, the reduction in the number of vehicles in normal operation changes the spacing between vehicles in the traffic flow, potentially reducing the overall traffic speed. For example, if the original traffic speed was 100 kilometers per hour, the increase in spacing between vehicles after the reduction in the number of vehicles in normal operation would require more time to accelerate and decelerate, potentially reducing the traffic speed to 80 kilometers per hour. From the perspective of traffic flow stability, the reduction in the number of vehicles in normal operation may disrupt the continuity of the traffic flow, resulting in more gaps and fluctuations in the traffic flow, increasing the risk of traffic accidents. In terms of traffic capacity, the reduction in the number of vehicles in normal operation means that fewer vehicles pass through a particular section per unit time, reducing the traffic capacity of the section. Taking these factors into account, the server uses the data on the reduction in normal vehicle traffic, combined with traffic flow theory and historical data, to calculate the impact of the reduction in normal vehicle traffic on traffic speed, stability, and capacity. These impacts are then integrated into a vector, the normal driving impact traffic flow vector. This vector comprehensively reflects the combined impact of the reduction in normal vehicle traffic on freeway traffic due to factors such as traffic facilities in the non-driving area.

[0127] In a possible implementation, step S140 includes:

[0128] Step S141 : extracting road facilities from the road area block perception data and generating features of various types of road facilities.

[0129] Step S142 , optimizing the traffic dynamic change map based on the characteristics of various types of road facilities and traffic state influencing factors to generate an optimized traffic dynamic change map.

[0130] Step S143 , simulating the traffic flow impact trend on the interchange space scene based on the optimized traffic dynamic change map, and generating a traffic flow impact trend model for the highway section.

[0131] Step S144 , estimating the traffic safety risk trend of the highway section based on the traffic flow impact highway section trend model, and generating traffic safety risk trend data.

[0132] In this embodiment, for the previously mentioned expressway sections (such as the G3 expressway section), the server first extracts road facilities from the road block perception data and generates features for each type of road facility. The road block perception data contains rich information about the expressway section, from which the server extracts data related to road facilities. During this process, the server identifies various road facilities and analyzes their characteristics. For example, for a curve, the server accurately determines the radius of curvature of the curve, which can be obtained by measuring the road shape information in the road block perception data. The server also obtains information such as the length of the curve, the slope at the curve, and the specific location of the curve (expressed in road section coordinates). For road signs, the server identifies the specific value of the speed limit sign, the location of the sign, and the visibility of the sign (by analyzing factors such as the clarity of the sign in the surveillance video). For road dividers, the server determines their width, material (such as concrete, metal guardrail, etc.), and whether they are damaged. Furthermore, for the distribution of street lights, the server records information such as the location, brightness level, and illumination range of each street light. Service area information is also extracted, including the location of entrances and exits, and the size of the service area (e.g., number of parking spaces, type of service facilities, etc.). By integrating this information about various road infrastructure features, such as curves, signs, medians, streetlights, and service areas, we generate a comprehensive and detailed data set for each type of road facility. This data provides a comprehensive and detailed description of the road infrastructure along the expressway, providing a crucial basis for subsequent analysis.

[0133] Next, the traffic dynamic change map is optimized based on the characteristics of various types of road facilities and traffic status influencing factors to generate an optimized traffic dynamic change map. The traffic dynamic change map originally contains information on traffic flow, vehicle speed, congestion conditions, and other aspects. The server integrates the characteristics of various types of road facilities and traffic status influencing factors into this map for optimization. For example, considering the road facility characteristics such as the curvature radius and slope of the curve, if the curvature radius of the curve is small and the slope is large, and the traffic status influencing factors show that the traffic speed in this area is fast and the distance between vehicles is small, then the server will highlight this curve area in the traffic dynamic change map. The color of the area in the map may be adjusted (such as changing the original green representing normal to yellow representing higher risk), and detailed information annotations about this curve will be added to the map, such as "The curvature radius of the curve is small, the slope is large, the traffic speed is fast, please slow down." For speed limit signs, if the speed limit sign value on a certain road section is low, and the traffic status impact factor shows that the average speed of vehicles on this road section is close to or exceeds the speed limit, the server will highlight this speed limit sign and the corresponding road section in the map, reminding relevant personnel to pay attention to the speed management of this road section. Similarly, near the entrance of a service area, if the road facility characteristics show that the entrance is narrow and the traffic status impact factor indicates that vehicles in this area are prone to congestion when merging into the main road, the server will provide a more detailed display of the traffic dynamics around the service area entrance in the map, such as adding a trajectory simulation of vehicles merging into the main road. By combining the road facility characteristics and traffic status impact factors with the traffic dynamics change map in this way, an optimized traffic dynamics change map is generated. This optimized map can more comprehensively and accurately reflect the traffic conditions of the highway section, especially considering the impact of road facilities on traffic.

[0134] Then, based on the optimized traffic dynamics map, the server simulates the impact of traffic flow on the interchange space scenario, generating a traffic flow impact model for the highway section. The server uses information from the optimized traffic dynamics map to simulate traffic flow impact trends based on the interchange space scenario. The interchange space scenario includes detailed road structure information such as individual ramps, main roads, confluence points, and diverging points. The server simulates vehicle flow trends within the interchange based on information such as traffic flow, speed, congestion conditions, and road infrastructure from the optimized traffic dynamics map. For example, if the optimized traffic dynamics map shows that a particular on-ramp has high traffic flow and slow speeds, and the road infrastructure (such as narrow lane widths) on that ramp prevents vehicles from quickly merging onto the main road, the server will simulate the queueing situation at that on-ramp in the traffic flow impact model for the highway section, including queue length and the distribution of queued vehicles (such as the ratio of small to large vehicles). At a diverging point, if the map shows that traffic is heavy in one direction and the number of lanes in that direction is about to decrease, the server will simulate the lane-changing behavior of vehicles at the diverging point, such as how long in advance the vehicle starts changing lanes and how the speed changes during the lane change. At the same time, taking into account the traffic differences in different time periods (such as peak and non-peak hours), the server will simulate the traffic flow impact trend model of the highway section separately to accurately reflect the traffic flow impact trend in different time periods. This traffic flow impact trend model of the highway section can dynamically display the flow trend of vehicles in the interchange space scenario and the impact of various factors (such as traffic volume, road facilities, etc.) on the traffic flow.

[0135] Finally, based on the traffic flow impact highway section trend model, the server estimates traffic safety risk trends for the highway section and generates traffic safety risk trend data. The server estimates traffic safety risk trends based on simulation results from the traffic flow impact highway section trend model. For example, if the model shows frequent close encounters (where the distance between vehicles is less than the safe distance) at a merging point, this indicates a high collision risk in that area. The server calculates the probability of a collision based on factors such as vehicle speed, traffic volume, and the road structure at the merging point. Assume that the model calculates a 0.1% probability of a collision per hour at that merging point. For curved areas, if the model indicates that vehicle speeds are generally high and exceed the safe speed (determined by factors such as the curve radius and slope), the server assesses the risk of vehicle loss of control or skidding, potentially calculating a 0.05% probability of vehicle loss of control in this scenario. Furthermore, for areas with problematic road infrastructure (such as damaged road signs or non-working streetlights), the server considers the impact of these factors on traffic safety and increases the corresponding risk value. For example, a damaged speed limit sign on a particular road section may increase the risk of speeding. Based on historical data and traffic flow theory, the server estimates that the risk of speeding accidents due to damaged speed limit signs increases by 0.03%. By integrating the probabilities of various traffic safety risks, such as collisions, loss of control, and speeding, along with related risk factors, traffic safety risk trend data is generated. This data comprehensively reflects traffic safety risk trends on expressways, providing an important basis for traffic management departments to formulate safety strategies and implement preventive measures.

[0136] Figure 2 The hardware structure of the multi-source traffic operation data analysis system 100 based on AI sensor fusion for implementing the multi-source traffic operation data analysis method based on AI sensor fusion provided by an embodiment of the present invention is shown as follows: Figure 2 As shown, the multi-source traffic operation data analysis system 100 based on AI sensor fusion may include a processor 110 , a machine-readable storage medium 120 , a bus 130 , and a communication unit 140 .

[0137] The machine-readable storage medium 120 may store data and / or instructions. In some embodiments, the machine-readable storage medium 120 may store data acquired from an external terminal. In some embodiments, the machine-readable storage medium 120 may store data and / or instructions used by the AI ​​sensor fusion-based multi-source traffic operation data analysis system 100 to execute or use to implement the exemplary methods described herein.

[0138] During the specific implementation process, one or more processors 110 execute computer executable instructions stored in the machine-readable storage medium 120, so that the processor 110 can execute the multi-source traffic operation data analysis method based on AI sensor fusion in the above method embodiment. The processor 110, the machine-readable storage medium 120 and the communication unit 140 are connected through the bus 130, and the processor 110 can be used to control the sending and receiving actions of the communication unit 140.

[0139] The specific implementation process of the processor 110 can be found in the various method embodiments executed by the above-mentioned multi-source traffic operation data analysis system 100 based on AI sensor fusion. The implementation principles and technical effects are similar and will not be repeated here in this embodiment.

[0140] In addition, an embodiment of the present invention also provides a readable storage medium, in which computer-executable instructions are preset. When the processor executes the computer-executable instructions, the multi-source traffic operation data analysis method based on AI sensor fusion as described above is implemented.

[0141] It should be noted that in order to simplify the description of the present invention and thus help understand one or more embodiments of the invention, in the foregoing description of the embodiments of the present invention, multiple features are sometimes combined into one embodiment, figure or description thereof.

Claims

1. A multi-source traffic operation data analysis method based on AI sensor fusion, characterized in that: The method comprises: Reconstruct the interchange space scene of the highway section to generate the interchange space scene. Perform AI sensor fusion collection on the highway section based on the interchange space scene to generate a traffic operation data stream. Then, analyze the traffic dynamic change map of the interchange space scene based on the traffic operation data stream to generate a traffic dynamic change map. Mapping road domain perception data blocks on the highway section to generate road domain block perception data, extracting driving behavior data blocks from the road domain block perception data to generate driving behavior data blocks, and extracting non-driving behavior data blocks from the road domain block perception data to generate non-driving behavior data blocks; Analyze the traffic status influencing factors of the highway section based on the driving behavior data block and the non-driving behavior data block to generate the traffic status influencing factors; The traffic dynamic change map is optimized based on the road block perception data and traffic status influencing factors to generate an optimized traffic dynamic change map. The traffic safety risk trend of the expressway section is estimated based on the optimized traffic dynamic change map and the interchange space scenario to generate traffic safety risk trend data. Based on the traffic safety risk trend data, safety tips and suggestions for the interchange space scenario are generated.

2. The multi-source traffic operation data analysis method based on AI sensor fusion according to claim 1 is characterized in that: The steps of reconstructing the interchange space scene of the highway section to generate the interchange space scene, performing AI sensor fusion collection on the highway section based on the interchange space scene to generate a traffic operation data stream, and performing traffic dynamic change map analysis on the interchange space scene based on the traffic operation data stream to generate the traffic dynamic change map include: Conduct drone aerial photography scans of highway sections to generate drone aerial photography data; Fuse the partitioned UAV aerial scanning data to generate target UAV aerial scanning data; Reconstruct the interchange space scene of the highway section based on the target UAV aerial scanning data to generate the interchange space scene; AI sensor fusion is used to collect data on highway sections based on interchange spatial scenarios to generate traffic operation data streams; The traffic dynamic change map of the interchange space scene is analyzed based on the traffic operation data flow to generate a traffic dynamic change map.

3. The multi-source traffic operation data analysis method based on AI sensor fusion according to claim 2 is characterized in that: The steps of performing AI sensor fusion collection on the highway section based on the interchange space scenario to generate a traffic operation data stream include: Configure traffic observation points on expressway sections and generate traffic observation point data; Based on the traffic observation point data, the longitudinal traffic section of the interchange space scene is analyzed to generate traffic observation point section data; Carry out traffic flow monitoring at observation points on highway sections based on traffic observation point data and generate traffic flow data at observation points; Based on the cross-sectional data of traffic observation points and the traffic flow data of observation points, the traffic flow of interchanges is calculated for traffic observation point data to generate traffic flow data of observation points; Based on the observation point vehicle flow data and observation point traffic flow data, the observation point data of the expressway section is fused to generate a traffic operation data stream.

4. The multi-source traffic operation data analysis method based on AI sensor fusion according to claim 2 is characterized in that: The step of analyzing the traffic dynamic change map of the interchange space scene based on the traffic operation data stream to generate the traffic dynamic change map includes: Analyze and resolve traffic flow and speed on traffic operation data streams to generate traffic flow characteristic data; Based on the traffic flow characteristic data, the vehicle speed impact characteristics of the interchange space scene are simulated to generate vehicle speed impact characteristic data; Construct a map of traffic flow characteristic data and vehicle speed impact characteristic data to generate a traffic flow change characteristic map; Analyze traffic congestion and smooth traffic conditions of traffic operation data streams to generate traffic status data; Conduct traffic situation simulation on interchange space scenarios based on traffic status data to generate traffic situation data; Construct a map of traffic status data and traffic situation data to generate a traffic situation map; The traffic flow change characteristic map and the traffic situation map are integrated to generate a traffic dynamic change map.

5. The multi-source traffic operation data analysis method based on AI sensor fusion according to claim 1 is characterized in that: The steps of mapping the road domain perception data blocks on the highway section to generate the road domain block perception data, extracting the driving behavior data blocks from the road domain block perception data to generate the driving behavior data blocks, and extracting the non-driving behavior data blocks from the road domain block perception data to generate the non-driving behavior data blocks include: Conduct road detection video surveillance on highway sections and generate highway section road detection video streams; Perform road block mapping on the highway section road detection video stream to generate road block perception data; Extract driving operation features from road block perception data to generate driving operation feature data; Extracting driving behavior data blocks from the road area block perception data based on the driving operation characteristic data to generate driving behavior data blocks; The non-driving behavior data blocks are extracted from the road area block perception data based on the driving operation feature data to generate non-driving behavior data blocks.

6. The multi-source traffic operation data analysis method based on AI sensor fusion according to claim 5 is characterized in that: The step of extracting driving operation features from the road area block perception data to generate driving operation feature data includes: Extract block dynamic features from road block perception data to generate block perception dynamic feature data; Analyze the traffic area dynamic characteristics of the road block perception data based on the block perception dynamic feature data to generate traffic area dynamic features; Extract non-traffic area features from the road block perception data based on the dynamic characteristics of the traffic area to generate non-traffic area feature data; Extracting regular driving pattern features from the non-traffic area feature data to generate regular driving pattern feature data; Based on the regular driving pattern feature data, the road block perception data is analyzed for regular features of the normal driving area to generate regular features of the normal driving area; The dynamic characteristics of the traffic area and the regular characteristics of the normal driving area are integrated to generate driving operation characteristic data.

7. The multi-source traffic operation data analysis method based on AI sensor fusion according to claim 1 is characterized in that: The step of analyzing the traffic state influencing factors of the highway section based on the driving behavior data block and the non-driving behavior data block to generate the traffic state influencing factors includes: Performing normal driving feature vector analysis on the driving behavior data block to generate a normal driving feature vector of the driving behavior; The traffic flow influencing factors of the highway section are analyzed based on the normal driving characteristic vector of driving behavior, and the driving behavior influencing traffic flow vector is generated; Analyze the abnormal parking ratio of the non-driving behavior data block to generate the abnormal parking ratio; Analyze the congestion impact vector of the highway section based on the abnormal parking ratio and generate the congestion impact vector of the highway section; The factors affecting the traffic flow of the normal driving section of the highway are analyzed based on the non-driving behavior area, and the normal driving impact traffic flow vector is generated; The weighted fusion of the road congestion impact vector and the normal driving impact traffic flow vector is used to generate the non-driving behavior impact traffic flow vector; The traffic flow vectors affected by driving behavior and the traffic flow vectors affected by non-driving behavior are fused to generate traffic status influencing factors.

8. The multi-source traffic operation data analysis method based on AI sensor fusion according to claim 7 is characterized in that: The step of analyzing the traffic flow impact factors of the normal driving section of the highway section based on the non-driving behavior area to generate the normal driving impact traffic flow vector includes: Identify the distribution of traffic facilities in non-driving areas and generate traffic facility distribution; Calculate the road section signal cycle based on the distribution of traffic facilities and generate the road section signal cycle; Analyze the traffic time extension of expressway sections based on the road signal cycle and generate traffic time extension data of expressway sections; Carry out lane change increment analysis on highway sections based on traffic time extension data of highway sections to generate lane change increment data; Calculate the reduction of vehicles traveling normally on the highway section based on the vehicle lane change increment data to generate the reduction data of vehicles traveling normally; The factors affecting traffic flow on expressway sections are analyzed based on the reduced volume data of vehicles in normal driving, and the vector of traffic flow affected by normal driving is generated.

9. The multi-source traffic operation data analysis method based on AI sensor fusion according to claim 1 is characterized in that: The steps of optimizing the traffic dynamic change map based on the road block perception data and the traffic state influencing factors to generate the optimized traffic dynamic change map, estimating the traffic safety risk trend of the highway section based on the optimized traffic dynamic change map and the interchange space scenario to generate traffic safety risk trend data include: Extract road facilities from road area block perception data and generate features of various types of road facilities; Optimize the traffic dynamic change map based on the characteristics of various types of road facilities and traffic status influencing factors to generate an optimized traffic dynamic change map; Based on the optimized traffic dynamic change map, the traffic flow impact trend of the interchange space scenario is simulated to generate a trend model of traffic flow impact on the highway section; Based on the trend model of traffic flow affecting highway sections, the traffic safety risk trend of highway sections is estimated to generate traffic safety risk trend data.

10. A multi-source traffic operation data analysis system based on AI sensor fusion, characterized in that: The multi-source traffic operation data analysis system based on AI sensor fusion includes a processor and a memory, the memory is connected to the processor, the memory is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the memory to implement the multi-source traffic operation data analysis method based on AI sensor fusion according to any one of claims 1 to 9 above.

Images