Traffic resilience dynamic assessment and prediction system based on deep learning

Abstract

The present invention belongs to the technical field of urban traffic resilience analysis, and discloses a dynamic assessment and prediction system for traffic resilience based on deep learning. The present invention breaks through the limitations of traditional single indicators by constructing a traffic resilience assessment matrix that includes the dimensions of absorption capacity, recovery capacity, and maintenance capacity, and realizes a three-dimensional characterization of traffic resilience. It combines the road network structure with the accident parameter correction to improve the assessment accuracy and scenario adaptability, provide a scientific basis for traffic emergency management and road network planning, and help improve the comprehensive performance of the traffic system under emergencies. The present invention integrates the identification of key nodes in the road network and the calculation of accident space parameters to establish a multi-factor weighted accident impact correction model, breaking through the limitations of traditional single accident parameter correction, realizing the coordinated analysis of road network topology and accident scale, and improving the adaptability of the assessment results to complex traffic scenarios.

Description

Technical Field

[0001] The present invention belongs to the technical field of urban traffic resilience analysis and relates to a traffic resilience dynamic assessment and prediction system based on deep learning. Background Art

[0002] Traffic resilience is a measure of the ability of a transportation system to maintain its functions, absorb shocks and recover quickly in the event of an emergency (such as a traffic accident). Its accurate assessment is of key significance to traffic emergency management, road network planning and improvement of traffic efficiency.

[0003] Traffic accidents are a significant factor impacting the normal operation of the transportation system, making traffic resilience analysis based on accident handling capabilities of the system crucial. On the one hand, efficient accident handling can quickly absorb the impact of an accident, reduce the scope and duration of congestion, and maintain the operation of some traffic functions, demonstrating the system's adaptability and robustness. On the other hand, the ability to quickly handle accidents and restore traffic order allows the system to quickly return to a stable state, demonstrating resilience. Furthermore, analyzing each stage of accident handling can identify potential risks, optimize resource redundancy, and provide insights for future accident prevention and handling, comprehensively enhancing the resilience of the transportation system and ensuring smooth traffic flow and socioeconomic stability.

[0004] Traditional methods mostly conduct static assessments based on historical statistical data, focusing only on single indicators such as traffic flow and speed. They fail to build a multi-dimensional assessment system covering absorption capacity, recovery capacity, and maintenance capacity, and are unable to dynamically reflect the real-time resilience changes of the transportation system during the development of an accident.

[0005] Existing technical solutions separate the road network structure characteristics from the accident impact analysis when analyzing traffic accident handling capacity, resulting in insufficient quantification of the spatial impact of accidents, inability to assess the transmission effects of key node density and distribution on traffic resilience, and inability to scientifically assess the actual degree of weakening of road traffic capacity by accidents. Summary of the Invention

[0006] In view of this, in order to solve the problems raised in the above background technology, a traffic resilience dynamic assessment and prediction system based on deep learning is proposed.

[0007] The purpose of the present invention can be achieved through the following technical solutions: a traffic resilience dynamic assessment and prediction system based on deep learning, including: a traffic data acquisition module, which obtains the driving speed, driving trajectory and vehicle type of each vehicle within the preset monitoring range of the accident scene.

[0008] The processing capacity evaluation module constructs a traffic resilience evaluation matrix including absorption capacity dimension, recovery capacity dimension and maintenance capacity dimension based on the driving speed and driving trajectory of each vehicle at the accident scene.

[0009] The road structure analysis module obtains road network information within the preset monitoring range and identifies the number and distance of key nodes.

[0010] The accident situation analysis module locates the number of accident vehicles, accident location and accident vehicle outline at the accident scene, and identifies the lateral passable distance and radial accident distance.

[0011] The traffic resilience correction module constructs the accident impact correction coefficient, combines the traffic resilience assessment matrix to calculate the corrected traffic resilience assessment index and outputs the traffic resilience level.

[0012] Compared with the existing technology, the beneficial effects of the present invention are as follows: (1) The present invention breaks through the limitations of traditional single indicators by constructing a traffic resilience assessment matrix that includes the absorption capacity dimension, recovery capacity dimension and maintenance capacity dimension, and realizes a three-dimensional characterization of traffic resilience. It combines the road network structure and accident parameter correction to improve the assessment accuracy and scenario adaptability, provides a scientific basis for traffic emergency management and road network planning, and helps to improve the comprehensive performance of the transportation system under emergencies.

[0013] (2) The present invention integrates the identification of key nodes in the road network with the calculation of accident spatial parameters to establish a multi-factor weighted accident impact correction model, breaking through the limitations of traditional single accident parameter correction, realizing the coordinated analysis of road network topology and accident scale, improving the adaptability of evaluation results to complex traffic scenarios, and providing a more reliable quantitative basis for emergency resource allocation and congestion prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0015] Figure 1 This is a schematic diagram of the connection of various modules of the system of the present invention.

[0016] Figure 2 A schematic diagram of identifying the lateral passable distance and radial accident distance corresponding to an embodiment provided by the present invention.

[0017] Figure 3 A schematic diagram of constructing a traffic resilience assessment matrix corresponding to an embodiment provided by the present invention.

[0018] Reference numerals: 1—lateral passable distance, 2—radial accident distance. DETAILED DESCRIPTION

[0019] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0020] See also Figure 1 As shown, the present invention provides a traffic resilience dynamic assessment and prediction system based on deep learning, including a traffic data acquisition module, a processing capability evaluation module, a road structure analysis module, an accident situation analysis module and a traffic resilience correction module, wherein the traffic data acquisition module is connected to the processing capability evaluation module, and the processing capability evaluation module, the road structure analysis module and the accident situation analysis module are all connected to the traffic resilience correction module.

[0021] The traffic data acquisition module is used to obtain the driving speed, driving trajectory and vehicle type of each vehicle within a preset monitoring range at the accident scene.

[0022] The processing capacity evaluation module is used to construct a traffic resilience evaluation matrix including absorption capacity dimension, recovery capacity dimension and maintenance capacity dimension based on the driving speed and driving trajectory of each vehicle at the accident scene.

[0023] See also Figure 3 As shown, the present invention provides a method for constructing a traffic resilience assessment matrix, wherein the absorptive capacity dimension, the recovery capacity dimension and the maintenance capacity dimension correspond to the absorptive capacity evaluation index, the recovery capacity evaluation index and the maintenance capacity evaluation index respectively.

[0024] In a preferred embodiment of the present invention, the specific analysis process of the absorption capacity is as follows: extract the driving speed of each vehicle, and then construct a driving speed change curve for each vehicle with time as the horizontal coordinate and driving speed as the vertical coordinate, traversing the timestamps corresponding to the driving speed change curve to obtain the driving speed of each vehicle corresponding to each time.

[0025] The average driving speed of each vehicle at each time is calculated to obtain the average driving speed of the vehicle at each time, and then the average driving speed change curve of the vehicles at the accident scene is obtained with time as the horizontal coordinate and the average driving speed as the vertical coordinate. The average driving speed change curve of the vehicles is evenly distributed to obtain a number of speed monitoring points.

[0026] The average vehicle speed corresponding to the time of the accident is recorded as the reference speed, and the average vehicle speed corresponding to each speed monitoring point is identified. The average vehicle speed corresponding to each speed monitoring point is compared to obtain the minimum vehicle average speed. The absolute deviation of the minimum vehicle average speed relative to the reference speed is calculated, and the inverse is taken to obtain the absorption capacity evaluation index.

[0027] It's important to note that absorptive capacity is one of the three dimensions of the traffic resilience assessment matrix. Its core objective is to measure the traffic system's ability to buffer disturbances after a traffic accident—that is, its ability to maintain operation despite the impact of an accident. The analysis process is based on the dynamic changes in vehicle speeds, using mathematical calculations to quantify the system's ability to absorb the accident. A higher absorptive capacity index indicates a smaller speed drop during an accident, a stronger ability to buffer disturbances, and a higher level of traffic resilience.

[0028] In a preferred embodiment of the present invention, the specific recovery capability analysis process is as follows: based on the average vehicle speed change curve at the accident scene, the average vehicle speed corresponding to each speed monitoring point after the minimum average vehicle speed is located, and the absolute deviation from the reference speed is calculated to obtain the average vehicle speed recovery degree corresponding to each speed monitoring point. The speed monitoring point that is first greater than or equal to the preset average vehicle speed recovery degree threshold is recorded as a recovery point.

[0029] The time corresponding to the recovery point and the time of the accident are difference calculated, and then the preset reference accident handling time and the result of the difference calculation are ratio calculated to obtain the recovery capacity evaluation index.

[0030] It's important to note that resilience is one of the three dimensions of the traffic resilience assessment matrix. Its core objective is to measure the speed and efficiency with which a transportation system recovers from a traffic incident. The analysis is based on the recovery process of average vehicle speeds. By comparing the time difference with a threshold, the resilience level of the system's "recovery speed" is quantified. A higher resilience index indicates stronger resilience.

[0031] In a preferred embodiment of the present invention, the specific analysis process of the maintenance capability is as follows: based on the average vehicle speed change curve at the accident scene, the absolute difference deviation between the average vehicle speed corresponding to each speed monitoring point and the minimum average vehicle speed is calculated, and then compared with a preset reference range, and the timestamp corresponding to each speed monitoring point is traversed.

[0032] The time between the speed monitoring point where the absolute difference deviation calculation result is first located within the preset reference range and the speed monitoring point where the absolute difference deviation calculation result is last located within the preset reference range is recorded as the accident stability interval.

[0033] The overall variance of the average vehicle speed corresponding to each speed monitoring point within the accident stability interval is calculated and the inverse is taken to obtain the maintenance capability evaluation index.

[0034] It's important to note that sustainment capability is one of the three dimensions of the traffic resilience assessment matrix. Its core objective is to measure the continuity and reliability of a traffic system's ability to maintain stable operation after a traffic accident. The analysis is based on the fluctuations in average vehicle speed within the "accident stability interval," quantifying the system's ability to maintain stable operation through variance calculations. A lower sustainment capability index indicates a stronger system sustainment capability and higher traffic resilience.

[0035] It should be noted that the invention breaks through the limitations of traditional single indicators by constructing a traffic resilience assessment matrix that includes the dimensions of absorption capacity, recovery capacity and maintenance capacity, and realizes a three-dimensional characterization of traffic resilience. It combines road network structure and accident parameter correction to improve assessment accuracy and scenario adaptability, providing a scientific basis for traffic emergency management and road network planning, and helping to improve the comprehensive performance of the transportation system under emergencies.

[0036] The road structure analysis module is used to obtain road network information within a preset monitoring range and identify the number of key nodes and the distance between key nodes.

[0037] In a preferred embodiment of the present invention, the specific analysis method for identifying the number of key nodes and the distance between key nodes is as follows: identify each intersection node based on the road network information within a preset monitoring range, and obtain the number of connecting roads, the number of lanes, and the closest distance to the location of other intersection nodes.

[0038] It should be noted that the preset monitoring range is usually centered on the accident site, with a certain radius of the area, such as 1 kilometer, or is determined based on administrative regions or road network divisions.

[0039] It's important to clarify that intersections are road junctions, serving as points of traffic flow and distribution. These can be automatically identified using electronic map data or GIS systems. The number of lanes is used to assess the vulnerability of a node. Nodes with more lanes have higher lane occupancy rates and less remaining traversable space if an accident occurs.

[0040] The intersection nodes that meet the following conditions at the same time are recorded as key nodes: the number of connected roads is greater than the preset connected road number threshold.

[0041] The number of lanes is greater than the preset lane number threshold.

[0042] The closest distance to the other intersection node positions is greater than a preset threshold value of the closest distance to the other intersection node positions.

[0043] It should be noted that key nodes are intersections within the road network within the preset monitoring range that have a significant impact on traffic resilience assessments. The core purpose of identifying key nodes is to assess the impact and transmission capacity of an accident on the surrounding traffic network by analyzing the "vulnerabilities" or "hubs" in the road network structure. For example, a greater number of key nodes or their proximity to the accident site may indicate a greater disruption to the road network, requiring significant correction in the resilience assessment.

[0044] The key nodes are counted, and the distance from each key node to the accident location is used as the key node distance corresponding to each key node, and then the mean is calculated to obtain the key node distance.

[0045] The accident situation analysis module is used to locate the number of accident vehicles at the accident scene, the accident location and the accident vehicle outline, and identify the lateral passable distance and the radial accident distance.

[0046] In a preferred embodiment of the present invention, please refer to Figure 2 As shown, the specific method of identifying the lateral passable distance and the radial accident distance is as follows: extract the contours of each accident vehicle at the accident scene, and then construct the scope of the area directly affected by the accident scene, locate the boundary points along the road direction and perpendicular to the road direction to obtain several boundary lines, record the distance between the two corresponding boundary lines along the road direction as the lateral accident width, and calculate the difference between the road width and the lateral accident width to obtain the lateral passable distance.

[0047] The distance between two corresponding boundary lines perpendicular to the road direction is recorded as the radial accident distance.

[0048] It should be noted that the lateral traversable distance refers to the remaining lateral width of the road after the accident, which is not directly occupied by the accident. It reflects the direct impact of the accident on the road's traffic capacity. A larger value indicates more remaining traversable space and higher traffic resilience. The radial accident distance refers to the vertical impact range of the accident on the road, that is, the length of road occupied by the accident. A larger value indicates a larger space occupied by the accident and a more severe traffic disruption.

[0049] It's important to note that the lateral traversable distance and radial accident distance quantify the physical occupation of roads by accidents through geometric spatial dimensions, providing intuitive spatial indicators for traffic resilience assessment. The core logic is that the greater the road space occupied by an accident, the lower the remaining capacity and the weaker the traffic resilience. Automating contour extraction and distance calculation using deep learning technology improves the real-time and accuracy of the assessment system, providing a scientific basis for emergency response and road network management.

[0050] The traffic resilience correction module is used to construct an accident impact correction coefficient, calculate the corrected traffic resilience assessment index in conjunction with the traffic resilience assessment matrix, and output the traffic resilience level.

[0051] In a preferred embodiment of the present invention, the specific method of constructing the accident impact correction coefficient is as follows: the number of key nodes and the distance between key nodes are compared with the preset reference values, and the lateral passable distance and the radial accident distance are compared with the preset reference values ​​respectively, and the above comparison results are subjected to weighted fusion analysis to obtain the accident impact correction coefficient.

[0052] It should be noted that the number of key nodes, distance reference values, and radial accident distance reference values ​​are primarily based on four criteria: first, historical data statistics: by analyzing the distribution characteristics of traffic data under different scenarios, high-frequency values ​​or central tendency values ​​are used as benchmarks; second, industry standards, ensuring that road parameters meet engineering requirements and that key node determination is compatible with road grades; third, expert experience: combining practical needs such as emergency management and congestion control, and having experts adjust parameter sensitivity; and fourth, model validation: by simulating accident scenarios with different parameter combinations and analyzing congestion indicators to optimize reference values ​​and ensure they meet actual assessment needs. This setting system balances data objectivity, standard compliance, and practical adaptability, providing a scientific benchmark for traffic resilience assessment.

[0053] In a preferred embodiment of the present invention, the calculation formula of the accident impact correction coefficient is: ,in represents the accident impact correction coefficient, They represent the preset reference values ​​of the number of key nodes, the distance between key nodes, the width of the road at the accident site, and the radial accident distance. Clearly indicate the number of key nodes, key node distance, lateral passable distance and radial accident distance. Represent the preset weight factors respectively.

[0054] It should be noted that the basis for setting the preset weight factors mainly includes: the actual impact of each resilience dimension (absorption, recovery, and maintenance capacity) on system resilience in historical traffic data; traffic management business needs, such as focusing on the weight of recovery capacity during peak hours; expert experience in judging different types of accidents; and optimizing weight distribution through deep learning model training to adapt to differences in road types, weather, and other scenarios, to achieve a precise match between the evaluation results and the actual resilience level. For example, .

[0055] It should be noted that the accident impact correction factor integrates multi-dimensional parameters, transforming road network structural characteristics and accident spatial characteristics into a comprehensive factor affecting traffic resilience. Its core logic is that the denser the density of key nodes, the closer they are to the accident site, and the larger the space occupied by the accident, the weaker the traffic system's resilience, requiring a negative correction in the assessment. Combining dynamic weight optimization with reference value adaptation through deep learning significantly improves the assessment system's adaptability to complex traffic scenarios, providing key support for accurately predicting traffic resilience levels.

[0056] In a preferred embodiment of the present invention, the specific calculation method of the modified traffic resilience assessment index is as follows: extract the absorption capacity evaluation index, the recovery capacity evaluation index and the maintenance capacity evaluation index, and then combine the accident impact correction coefficient to perform traffic resilience assessment index correction analysis to calculate the modified traffic resilience assessment index.

[0057] In a preferred embodiment of the present invention, the calculation formula of the modified traffic resilience assessment index is: ,in represents the accident impact correction coefficient, Represents the absorption capacity evaluation index, recovery capacity evaluation index and maintenance capacity evaluation index respectively, They respectively represent the weight factors corresponding to the preset absorption capacity evaluation index, recovery capacity evaluation index and maintenance capacity evaluation index.

[0058] For example, .

[0059] It should be noted that the definition of the revised traffic resilience assessment index is the core output indicator of the traffic resilience dynamic assessment and prediction system, which is used to comprehensively reflect the resilience level of the traffic system after an accident.

[0060] It should be noted that the construction idea of ​​the calculation formula corresponding to the above-mentioned revised traffic resilience assessment index is as follows: 1. Traffic resilience is determined by the absorption, recovery and maintenance capabilities. Quantify these three capabilities separately by assigning different weight factors , reflecting the different importance of each capability in traffic resilience, and the weighted summation is used to obtain the basic traffic resilience assessment value, which comprehensively reflects the performance of the system in responding to different stages of accidents.

[0061] 2. Introducing the accident impact correction factor It comprehensively analyzes the impact of factors such as key nodes in the road network and the roads at the accident site on the accident. The basic assessment value is adjusted to reflect the additional impact of the actual scenario after the accident on traffic resilience, so that the assessment results are more in line with the actual traffic conditions under the accident.

[0062] In a preferred embodiment of the present invention, the specific analysis method of the traffic resilience level is as follows: comparing the modified traffic resilience assessment index with the preset traffic resilience assessment index thresholds.

[0063] If the revised traffic resilience assessment index is greater than or equal to the first traffic resilience assessment index, the traffic resilience level is identified as level one. If the revised traffic resilience assessment index is greater than or equal to the second traffic resilience assessment index and less than the first traffic resilience assessment index, the traffic resilience level is identified as level two. If the revised traffic resilience assessment index is less than the second traffic resilience assessment index, the traffic resilience level is identified as level three.

[0064] It's important to note that the Traffic Resilience Rating (TRR) is the system's final, intuitive assessment result. By comparing the modified TRI with preset thresholds, the system's resilience level is categorized into different levels. This provides a clear basis for decision-making by traffic management departments and emergency response agencies, such as whether to initiate traffic control measures or allocate resources.

[0065] It should be noted that the threshold values ​​of each traffic resilience assessment index are set based on: mainly statistical analysis of historical traffic data, and the grade dividing points are determined by studying the distribution of corrected assessment indices under historical accidents or normal conditions; combined with traffic management business needs, connected with emergency response processes and referring to industry standards; considering the collaborative logic of system modules, matching the range of correction coefficient values ​​and the original index range, and taking into account the weights of each resilience dimension; potentially through deep learning dynamic optimization, adapting to differences in road types, time periods, weather and other scenarios, to achieve a balance between the scientificity, rationality and dynamic adaptability of the thresholds, and ensure that the grade division meets actual needs.

[0066] It should be noted that the present invention establishes a multi-factor weighted accident impact correction model by integrating the identification of key nodes in the road network and the calculation of accident spatial parameters, breaking through the limitations of traditional single accident parameter correction, realizing the coordinated analysis of road network topology and accident scale, and improving the adaptability of evaluation results to complex traffic scenarios, providing a more reliable quantitative basis for emergency resource allocation and congestion prediction.

[0067] The above contents are merely examples and explanations of the concept of the present invention. Those skilled in the art may make various modifications or additions to the described specific embodiments or replace them in a similar manner. As long as they do not deviate from the concept of the invention or exceed the scope defined by the present invention, they should all fall within the scope of protection of the present invention.

Claims

1. A traffic resilience dynamic assessment and prediction system based on deep learning, characterized by: include: Traffic data collection module, which obtains the driving speed, driving trajectory and vehicle type of each vehicle within the preset monitoring range of the accident scene; A handling capacity evaluation module constructs a traffic resilience evaluation matrix including absorption capacity, recovery capacity, and maintenance capacity dimensions based on the driving speed and driving trajectory of each vehicle at the accident scene; The road structure analysis module obtains road network information within the preset monitoring range and identifies the number and distance of key nodes; The accident situation analysis module locates the number of accident vehicles at the accident scene, the accident location, and the accident vehicle outline, and identifies the lateral passable distance and radial accident distance; The traffic resilience correction module constructs the accident impact correction coefficient, combines the traffic resilience assessment matrix to calculate the corrected traffic resilience assessment index and outputs the traffic resilience level.

2. The deep learning-based traffic resilience dynamic assessment and prediction system according to claim 1, characterized in that: The specific analysis process of the absorption capacity is as follows: Extract the driving speed of each vehicle, and then construct a driving speed change curve for each vehicle with time as the horizontal coordinate and driving speed as the vertical coordinate. Traverse the timestamps corresponding to the driving speed change curve to obtain the driving speed of each vehicle at each time; The average speed of each vehicle at each time is calculated to obtain the average speed of the vehicle at each time, and then the average speed change curve of the vehicles at the accident scene is obtained with time as the horizontal axis and the average speed as the vertical axis. The average speed change curve of the vehicles is evenly distributed to obtain a number of speed monitoring points; The average vehicle speed corresponding to the time of the accident is recorded as the reference speed, and the average vehicle speed corresponding to each speed monitoring point is identified. The average vehicle speed corresponding to each speed monitoring point is compared to obtain the minimum vehicle average speed. The absolute deviation of the minimum vehicle average speed relative to the reference speed is calculated, and the inverse is taken to obtain the absorption capacity evaluation index.

3. The deep learning-based traffic resilience dynamic assessment and prediction system according to claim 2, characterized in that: The specific analysis process of the recovery capability is as follows: Based on the average vehicle speed change curve at the accident scene, locate the average vehicle speed corresponding to each speed monitoring point after the minimum average vehicle speed, and calculate the absolute deviation from the reference speed to obtain the average vehicle speed recovery degree corresponding to each speed monitoring point. The speed monitoring point whose speed is greater than or equal to the preset average vehicle speed recovery degree threshold for the first time is recorded as the recovery point; The time corresponding to the recovery point and the time of the accident are difference calculated, and then the preset reference accident handling time and the result of the difference calculation are ratio calculated to obtain the recovery capacity evaluation index.

4. The deep learning-based traffic resilience dynamic assessment and prediction system according to claim 3, characterized in that: The specific analysis process of the maintenance capability is as follows: Based on the average vehicle speed change curve at the accident scene, the absolute difference between the average vehicle speed corresponding to each speed monitoring point and the minimum average vehicle speed is calculated, and then compared with the preset reference range, and the timestamps corresponding to each speed monitoring point are traversed; The time between the speed monitoring point where the absolute difference deviation calculation result is first located within the preset reference range and the speed monitoring point where the absolute difference deviation calculation result is last located within the preset reference range is recorded as the accident stability interval; The overall variance of the average vehicle speed corresponding to each speed monitoring point within the accident stability interval is calculated and the inverse is taken to obtain the maintenance capability evaluation index.

5. The deep learning-based traffic resilience dynamic assessment and prediction system according to claim 1, characterized in that: The specific analysis method for identifying the number of key nodes and the distance between key nodes is as follows: Identify each intersection node based on the road network information within the preset monitoring range, and obtain the number of connected roads and lanes of each intersection node and the closest distance to other intersection nodes; The intersection nodes that meet the following conditions at the same time are recorded as key nodes: The number of connected roads is greater than a preset threshold value of the number of connected roads; The number of lanes is greater than the preset lane number threshold; The closest distance of the other intersection node positions is greater than the preset closest distance threshold of the other intersection node positions; The key nodes are counted, and the distance from each key node to the accident location is used as the key node distance corresponding to each key node, and then the mean is calculated to obtain the key node distance.

6. The deep learning-based traffic resilience dynamic assessment and prediction system according to claim 1, characterized in that: The specific method of identifying the transverse passable distance and the radial accident distance is as follows: The outlines of each accident vehicle at the accident scene are extracted to construct the scope of the directly affected area of ​​the accident scene. Boundary points are located along the road direction and perpendicular to the road direction to obtain several boundary lines. The distance between two corresponding boundary lines along the road direction is recorded as the lateral accident width. The difference between the road width and the lateral accident width is calculated to obtain the lateral drivable distance. The distance between two corresponding boundary lines perpendicular to the road direction is recorded as the radial accident distance.

7. The deep learning-based traffic resilience dynamic assessment and prediction system according to claim 4, characterized in that: The specific method of constructing the accident impact correction coefficient is as follows: The number of key nodes and the distance between key nodes are compared with the preset reference values. At the same time, the lateral passable distance and radial accident distance are compared with the preset reference values. The comparison results are subjected to weighted fusion analysis to obtain the accident impact correction coefficient.

8. The deep learning-based traffic resilience dynamic assessment and prediction system according to claim 7, characterized in that: The specific calculation method of the modified traffic resilience assessment index is as follows: The absorption capacity evaluation index, the recovery capacity evaluation index and the maintenance capacity evaluation index are extracted, and then the traffic resilience evaluation index correction analysis and calculation are performed in combination with the accident impact correction coefficient to correct the traffic resilience evaluation index.

9. The deep learning-based traffic resilience dynamic assessment and prediction system according to claim 1, characterized in that: The specific analysis method of the traffic resilience level is as follows: Comparing the modified traffic resilience assessment index with preset traffic resilience assessment index thresholds; If the revised traffic resilience assessment index is greater than or equal to the first traffic resilience assessment index, the traffic resilience level is identified as level one. If the revised traffic resilience assessment index is greater than or equal to the second traffic resilience assessment index and less than the first traffic resilience assessment index, the traffic resilience level is identified as level two. If the revised traffic resilience assessment index is less than the second traffic resilience assessment index, the traffic resilience level is identified as level three.

10. The deep learning-based traffic resilience dynamic assessment and prediction system according to claim 8, characterized in that: The calculation formula of the accident impact correction coefficient is: ； Where, represents the accident impact correction coefficient, They represent the preset reference values ​​of the number of key nodes, the distance between key nodes, the width of the road at the accident site, and the radial accident distance. Clearly indicate the number of key nodes, key node distance, lateral passable distance and radial accident distance. Respectively represent the preset weight factors; The calculation formula of the modified traffic resilience assessment index is: ； Where, represents the accident impact correction coefficient, Represents the absorption capacity evaluation index, recovery capacity evaluation index and maintenance capacity evaluation index respectively, They respectively represent the weight factors corresponding to the preset absorption capacity evaluation index, recovery capacity evaluation index and maintenance capacity evaluation index.

Images