Spatiotemporal data feature mining method and system for smart transportation

Abstract

The embodiment of the present application provides a method and system for mining spatiotemporal data features for smart transportation. By acquiring and processing the dynamic traffic flow data of multiple traffic participants in real time, a traffic spatiotemporal knowledge network is generated. By marking and analyzing the first spatiotemporal vertex and the target leading participant in the traffic spatiotemporal knowledge network, the key participants with the greatest impact on traffic flow can be accurately identified. The subject movement state trajectory data of the key participants is deeply mined, and potential traffic accident risk points can be discovered in advance, which significantly improves the early warning capability of the smart transportation system and helps reduce the occurrence of traffic accidents. Based on the analysis results of dynamic traffic flow data and key participants, it is convenient to allocate traffic resources more scientifically and reasonably, optimize traffic signal control strategies, improve road traffic efficiency, reduce traffic congestion, and enhance the efficiency and effectiveness of urban traffic management.

Description

Technical Field

[0001] The present application relates to the field of smart transportation technology, and more specifically, to a method and system for mining spatiotemporal data features for smart transportation. Background Art

[0002] With the acceleration of urbanization and the growing demand for transportation, problems such as traffic congestion and frequent traffic accidents have become increasingly prominent, posing a severe challenge to the sustainable development of cities and the quality of life of residents. As one of the effective ways to solve these problems, the intelligent transportation system is gradually becoming an important direction for urban traffic management. By integrating advanced information technology, communication technology and control technology, the intelligent transportation system realizes comprehensive perception, intelligent analysis and decision support of traffic information, so as to optimize the allocation of traffic resources, improve traffic operation efficiency and ensure traffic safety.

[0003] In the intelligent transportation system, the collection, processing and analysis of spatiotemporal data is one of the key links. Traditional traffic data analysis methods often focus on the statistics and analysis of static data, which makes it difficult to fully capture the dynamic characteristics of traffic flow and the complex interactive relationships behind it. However, in a dynamic traffic environment, the behavior and status of traffic participants (such as vehicles, pedestrians, public transportation, etc.) change over time, and there are complex interactions between them. These dynamic characteristics are of great significance for understanding traffic flow patterns, predicting traffic conditions, and assessing traffic accident risks. Summary of the invention

[0004] In order to at least overcome the above-mentioned deficiencies in the prior art, the purpose of the present application is to provide a spatiotemporal data feature mining method and system for smart transportation.

[0005] In a first aspect, the present application provides a spatiotemporal data feature mining method for smart transportation, which is applied to a spatiotemporal data feature mining system for smart transportation. The method comprises:

[0006] Acquire dynamic traffic flow data generated by X traffic participants in a dynamic traffic participation process, configure X space-time vertices with the X traffic participants, configure space-time pointing links between the X space-time vertices with the dynamic traffic flow data, and generate traffic space-time knowledge network data;

[0007] Marking the spatiotemporal vertices in the traffic spatiotemporal knowledge network data that have interactive relationships with other traffic participants as first spatiotemporal vertices;

[0008] Extract the target leading participating subjects corresponding to each of the X space-time vertices from the traffic space-time knowledge network data, and mark the traffic participating subject corresponding to the space-time vertex with the largest participation frequency among the target leading participating subjects corresponding to each of the X space-time vertices as the key participating subject among the X traffic participating subjects; the target leading participating subject corresponding to each of the space-time vertices is used to represent the participating subjects covered in the traffic participation trajectory from the first space-time vertex to the space-time vertex;

[0009] Feature mining is performed on the subject movement state trajectory data associated with the key participating subject to determine the first traffic accident risk data in the dynamic traffic participation process.

[0010] In the second aspect, an embodiment of the present application also provides a spatiotemporal data feature mining system for smart transportation, the spatiotemporal data feature mining system for smart transportation comprising a processor and a machine-readable storage medium, the machine-readable storage medium storing a computer program, the computer program is loaded and executed in conjunction with the processor to implement the spatiotemporal data feature mining method for smart transportation of the first aspect above.

[0011] In combination with the above aspects, the embodiment of the present application generates a traffic spatiotemporal knowledge network by acquiring and processing the dynamic traffic flow data of multiple traffic participants in real time. By marking and analyzing the first spatiotemporal vertex and the target dominant participant in the traffic spatiotemporal knowledge network, it can accurately identify the key participants that have the greatest impact on traffic flow, conduct in-depth feature mining on the subject movement state trajectory data of the key participants, and discover potential traffic accident risk points in advance, significantly improve the early warning capability of the smart transportation system, and help reduce the occurrence of traffic accidents. Based on the analysis results of dynamic traffic flow data and key participants, it is convenient to allocate traffic resources more scientifically and reasonably, optimize traffic signal control strategies, improve road traffic efficiency, reduce traffic congestion, and enhance the efficiency and effectiveness of urban traffic management. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required to be used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0013] Figure 1 A flow chart of a method for mining spatiotemporal data features for smart transportation provided in an embodiment of the present application;

[0014] Figure 2A schematic block diagram of the structure of a spatiotemporal data feature mining system for smart transportation provided in an embodiment of the present application for implementing the above-mentioned spatiotemporal data feature mining method for smart transportation. DETAILED DESCRIPTION

[0015] The following description is intended to enable one of ordinary skill in the art to implement and utilize the present application, and the description is provided in the context of a specific application scenario and its requirements. It will be apparent to one of ordinary skill in the art that various changes may be made to the disclosed embodiments, and that the general principles defined in the present application may be applied to other embodiments and application scenarios without departing from the principles and scope of the present application. Therefore, the present application is not limited to the described embodiments, but should be given the broadest scope equal to the claims.

[0016] Step S110, obtaining dynamic traffic flow data generated by X traffic participants in the dynamic traffic participation process, configuring X space-time vertices with the X traffic participants, configuring space-time pointing links between the X space-time vertices with the dynamic traffic flow data, and generating traffic space-time knowledge network data.

[0017] Specifically, the dynamic traffic flow data refers to the data generated by traffic participants (such as vehicles, pedestrians, traffic lights, etc.) in a real-time traffic environment. These dynamic traffic flow data change over time and reflect the dynamic characteristics of traffic flow. These dynamic traffic flow data usually include information such as location, speed, acceleration, driving direction, signal status, etc.

[0018] The spatiotemporal vertex is used to represent the state point of a traffic participant at a specific time and space position in the traffic spatiotemporal knowledge network. Each traffic participant generates a series of spatiotemporal vertices in the dynamic traffic process, and these vertices are connected by spatiotemporal pointing links to form a traffic flow trajectory.

[0019] The space-time directed link is a directed edge connecting two space-time vertices, representing the transition process of traffic participants from one state to another. These space-time directed links contain information such as the direction, speed and time sequence of traffic flow.

[0020] The traffic spatiotemporal knowledge network data is a network structure composed of a series of spatiotemporal vertices and spatiotemporal pointing links, which is used to represent the mutual relationship, state change and spatiotemporal information of traffic participants in the dynamic traffic process.

[0021] For example, in this embodiment, for traffic data of a specific area of ​​a city (such as a large commercial area and its surrounding road network), there are multiple traffic participants in this specific area of ​​the city, such as cars, motorcycles, pedestrians, bicycles, and traffic lights. Assume X=5, and the five traffic participants are: three cars (respectively marked as car A, car B, and car C), a pedestrian (marked as pedestrian P), and a traffic light (marked as light L).

[0022] Car A is equipped with GPS positioning system, speed sensor and other devices, which send data to the server regularly (for example, every 10 seconds). During the dynamic traffic participation process, the server receives a series of data from Car A, including timestamp (for example, starting from 10:00:00 on May 1, 2023, one data point every 10 seconds), location coordinates (latitude and longitude information, such as longitude 116.39, latitude 39.90, etc., indicating the position of the car on the road), speed (such as 30km / h, 35km / h, etc. at different times), acceleration (for example, there will be corresponding acceleration values ​​when starting and braking), and driving direction (such as driving north).

[0023] Similar to car A, car B also sends data to the server. Its driving trajectory may intersect or run parallel to that of car A. For example, at 10:05:00 on May 1, 2023, car B is located at longitude 116.38, latitude 39.91, with a speed of 25 km / h and a driving direction of east.

[0024] The data of car C shows that it is moving faster in some sections and slows down when approaching an intersection. For example, at 10:10:00, its speed drops from 50 km / h to 30 km / h, and its location coordinates are close to an intersection (longitude 116.40, latitude 39.90).

[0025] The data sent by the smartphone carried by pedestrian P (assuming it has positioning and motion sensor functions and has data exchange with the server) includes timestamp (such as starting at 10:00:00), location coordinates (such as latitude and longitude information on the sidewalk), walking speed (such as 15km / h), and walking direction (such as walking towards the mall entrance).

[0026] The status information of traffic light L is also obtained by the server, including the time when each light color changes (such as the time interval when the red light is on is 10:00:00-10:00:30, the time interval when the green light is on is 10:00:30-10:01:30, etc.) and the corresponding intersection signs (such as the traffic light at intersection 1).

[0027] The server configures five space-time vertices based on the five traffic participants. For car A, the corresponding space-time vertex contains the identification of car A and a comprehensive representation of the time and space information related to it. For example, the state of car A at longitude 116.39 and latitude 39.90 at 10:00:00 is used as part of the space-time vertex information. Similarly, corresponding space-time vertices are configured for car B, car C, pedestrian P, and traffic light L.

[0028] Based on dynamic traffic flow data, the server constructs a spatiotemporal pointing link. For example, when car A travels from location 1 (longitude 116.39, latitude 39.90) to location 2 (longitude 116.395, latitude 39.902) at 10:00:0010:00:10, if it intersects with the driving trajectory of car B or is affected by traffic light L during this process, then a spatiotemporal pointing link is constructed between the spatiotemporal vertex corresponding to car A and the spatiotemporal vertex corresponding to car B and traffic light L. This link contains information such as the direction of traffic flow (for example, car A is heading in the direction of car B), the speed of traffic flow (information such as the speed of car A), and the time sequence (the time period of 10:00:0010:00:10).

[0029] Through the configuration of the above-mentioned spatiotemporal vertices and spatiotemporal pointing links, the server finally generates traffic spatiotemporal knowledge network data. This network data represents the mutual relationship, state changes, and spatiotemporal information of the five traffic participants in the dynamic traffic participation process in a structured way. For example, in this network, we can clearly see the interaction between car A and car B, traffic light L, and the potential interaction between pedestrian P and car C at certain intersections (if pedestrian P is crossing the road when car C approaches the intersection).

[0030] Step S120 , marking the spatiotemporal vertices in the traffic spatiotemporal knowledge network data that have interactive relationships with other traffic participants as first spatiotemporal vertices.

[0031] The first spatiotemporal vertex refers to a spatiotemporal vertex that has an interactive relationship with all other traffic participants in the traffic spatiotemporal knowledge network data. This first spatiotemporal vertex usually represents a core control point or key point in a traffic scene.

[0032] For example, in this embodiment, the server first uses the space-time vertex corresponding to car A as the source node in the traffic space-time knowledge network data. The server starts to track the traffic participation trajectory of car A in this network. For example, along the space-time pointing link, check the interaction between car A and other traffic participants. For example, during 10:00:0010:05:00, car A is first controlled by traffic light L (starts after the green light comes on), and then runs parallel with car B for a period of time, and at 10:03:00, there is an intersection with the sidewalk where pedestrian P is located (although pedestrian P may not cross the road in this area at this time, there is an intersection in the traffic space-time relationship).

[0033] Next, the server uses the space-time vertex corresponding to car B as the source node for tracking. During 10:02:0010:10:00, car B is parallel and interactive with car A, and is also affected by traffic light L. It may also affect the walking path of pedestrian P near a certain intersection (because pedestrian P may need to wait for car B to pass before crossing the road).

[0034] Similarly, the space-time vertices corresponding to the car C, the pedestrian P and the traffic light L are tracked.

[0035] In the tracked traffic participation trajectory, if there is a traffic participation trajectory including these five space-time vertices, then the corresponding space-time vertex will be marked as the first space-time vertex. For example, suppose a series of traffic participation trajectories starting from traffic light L (because the state of traffic light L will affect the actions of all other traffic participants) is found, passing through car A, car B, car C and pedestrian P, then the space-time vertex corresponding to traffic light L can be marked as the first space-time vertex.

[0036] If no complete traffic participation trajectory containing five spatiotemporal vertices is found during the tracking process, for example, only a trajectory containing cars A, B, C, and traffic lights L but no pedestrian P is found, then the server will configure a simulation participant in the traffic spatiotemporal knowledge network data. This simulation participant can simulate the possible behavior of pedestrian P in this traffic scene, such as assuming that pedestrian P crosses the road at the right time according to traffic rules. Then configure the simulation spatiotemporal pointing link between this simulation participant and the traffic spatiotemporal knowledge network data, and finally mark this simulation participant as the first spatiotemporal vertex.

[0037] Among them, the server extracts the space-time vertex corresponding to car A from the traffic space-time knowledge network data (set as the a=1th space-time vertex), takes the space-time vertex corresponding to car A as the source node, and sorts the traffic space-time knowledge network data. In this process, the server will traverse the traffic participants and their interactive relationships related to car A in the order of the space-time pointing links, and generate the first basic traffic trajectory network. The root participant of this trajectory network is the space-time vertex corresponding to car A.

[0038] If in this first basic traffic trajectory network, there are space-time vertices corresponding to car B, car C, pedestrian P and traffic light L, then the space-time vertex corresponding to car A can be marked as the first space-time vertex.

[0039] If there are spatiotemporal vertices that do not belong to the first basic traffic trajectory network, for example, only car B and traffic light L are in this network, but car C and pedestrian P are not, then the server will use the spatiotemporal vertex corresponding to car B (the a+1=2nd spatiotemporal vertex) as the source node, perform walk sorting on the traffic spatiotemporal knowledge network data, and generate the second basic traffic trajectory network. Then continue to determine whether the second basic traffic trajectory network contains all 5 spatiotemporal vertices, and so on, until a basic traffic trajectory network containing all spatiotemporal vertices is found, and the corresponding source node spatiotemporal vertex is marked as the first spatiotemporal vertex.

[0040] Step S130, extracting the target leading participant corresponding to each of the X space-time vertices from the traffic space-time knowledge network data, and marking the traffic participant corresponding to the space-time vertex with the highest participation frequency among the target leading participant corresponding to each of the X space-time vertices as the key participant among the X traffic participants. The target leading participant corresponding to each space-time vertex is used to represent the participant covered in the traffic participation trajectory from the first space-time vertex to the space-time vertex.

[0041] In this embodiment, the server extracts Y traffic participation trajectories from the first spatiotemporal vertex (traffic light L) to the b=1th spatiotemporal vertex corresponding to car A from the traffic spatiotemporal knowledge network data. For example, there may be 3 different traffic participation trajectories from traffic light L to car A, and these 3 trajectories may differ due to different times or different traffic conditions.

[0042] For these three traffic participation trajectories, the server obtains the interactive participants covered by them respectively. In the first trajectory, it covers car B (because car B and car A have an interactive relationship under the control of signal light L, such as waiting for the green light at the same time or driving in the same lane, etc.), signal light L and pedestrian P (assuming that pedestrian P is on the nearby sidewalk when car A starts); in the second trajectory, it covers car C (car C may be parallel or intersecting with car A in another lane) and signal light L; in the third trajectory, it covers car B, car C, signal light L and pedestrian P.

[0043] Therefore, the server marks the interactive participants that exist in all three trajectories among the interactive participants covered by the three traffic participation trajectories as the target dominant participant of the spatiotemporal vertex corresponding to car A. In this example, traffic light L exists in all three trajectories, so traffic light L is marked as the target dominant participant of the spatiotemporal vertex corresponding to car A.

[0044] Similarly, the server will perform similar operations on the space-time vertices corresponding to car B, car C, and pedestrian P, extract the traffic participation trajectories from the first space-time vertex (traffic light L) to their respective traffic participation subjects, obtain the interactive participating subjects, and then determine the target dominant participating subject.

[0045] Alternatively, the server first marks the first spatiotemporal vertex (signal light L) as its own target dominant participant.

[0046] For car A (as the second space-time vertex, i.e., the space-time vertex other than the first space-time vertex), the server obtains the first preceding participant of car A. For example, the first preceding participant of car A may be traffic light L (because car A needs to start or stop according to the state of traffic light L).

[0047] The server obtains the secondary leading participant of the first preceding participant (traffic light L) of car A. Assuming that in this traffic scenario, it is analyzed from the traffic spatiotemporal knowledge network data that the secondary leading participant of traffic light L is the traffic management center (because the traffic management center controls the state change of traffic light L). Then the traffic management center and car A are marked as the target leading participants of car A.

[0048] Similar operations are performed on other traffic participants (car B, car C, pedestrian P) to obtain their first preceding participant, then obtain the second leading participant of the first preceding participant, and finally mark them as their respective target leading participant.

[0049] Furthermore, the server can also extract the second preceding participant (assuming it is a traffic management center) of the first preceding participant (e.g., the preceding participant signal light L of car A) from the traffic spatiotemporal knowledge network data, and then extract the first traffic participation trajectory including the signal light L and the traffic management center from the traffic spatiotemporal knowledge network data. This trajectory may include information such as the time when the traffic management center sends a control instruction to the signal light L and the state response time of the signal light L.

[0050] From this first traffic participation trajectory, the server extracts the second traffic participation trajectory. For example, this second traffic participation trajectory indicates that in a certain time period (such as 10:00:0010:01:00), the traffic management center sent a special control instruction (such as adjusting the duration of signal light L in response to sudden traffic flow), and in this time period, signal light L has no other prior participating subject's participation trajectory. Then the traffic management center corresponding to this second traffic participation trajectory is marked as the secondary dominant participating subject of signal light L.

[0051] Furthermore, the embodiment of the present application can also perform operations related to adjusting the dominant participating entity. Specifically, the server marks five space-time vertices (car A, car B, car C, pedestrian P, and traffic light L) as the basic dominant participating entity of car A (as an example of the second space-time vertex).

[0052] In the c=1th ​​tracking and extraction process of the leading participant, the Z first preceding participants of car A are extracted from the traffic spatiotemporal knowledge network data. For example, set Z=2, and through the search and screening method mentioned above, the server determines that the two first preceding participants of car A are the traffic light L and car B (because car B may enter a lane or intersection before car A, which has a certain impact on the driving of car A).

[0053] The server obtains the sub-dominant participant corresponding to each of the two first preceding participants of car A. Assume that the sub-dominant participant of signal light L is the traffic management center, and the sub-dominant participant of car B is the driver of car B (because the driver controls the driving state of car B). Then the shared participant of these two sub-dominant participants (there is no shared participant in this example) and car A are marked as the first adjustment dominant participant of car A.

[0054] If in the subsequent calculation of the cth adjustment dominant participant (for example, c=2), it is found that the second adjustment dominant participant of car A is consistent with the first adjustment dominant participant, then this adjustment dominant participant is marked as the target dominant participant of car A.

[0055] Furthermore, for the operations related to the search scope of the prior participating subject, the server stores the traffic spatiotemporal knowledge network data in a set data structure. For example, a graph structure is used to store the relevant information of spatiotemporal vertices and spatiotemporal pointing links, and they are arranged in order according to the identification of the traffic participating subject or the spatiotemporal order. For each spatiotemporal vertex, an index is established with the position of the spatiotemporal vertex in the traffic network (such as geographic coordinates or logical location identification, such as intersection number, etc.) as the index key, and the type information of the traffic participating subject (such as car, pedestrian, signal light, etc.) is indexed.

[0056] Set a search range parameter to limit the time and space range considered when searching for the first participant. This parameter is determined based on the distance of mutual influence between traffic participants in historical data (for example, according to previous data, the effective distance between cars is generally within 100 meters) and the time interval (for example, traffic events within 10 minutes are interrelated). At the same time, initialize an empty temporary storage structure, such as an array structure, to store the potential first first participant found during the search process.

[0057] Suppose you want to determine the first participant of car A. The server searches for the corresponding space-time vertex in the traffic space-time knowledge network data based on the identification of car A, and obtains the location information of the space-time vertex in the traffic network (such as longitude 116.39, latitude 39.90) and related time information (such as 10:00:00). Based on the determined search range, traverse the space-time vertices in the traffic space-time knowledge network data. For example, during the traversal process, the space-time vertex of car B is found, and the position relationship (calculate whether the distance is within 100 meters) and time relationship (whether there is a possibility of interaction within 10 minutes) between car A and car B are compared. If the conditions are met, car B is marked as a potential first first participant and added to the temporary storage structure.

[0058] For each potential first preceding participant in the temporary storage structure, obtain the type of traffic participant represented by the potential first preceding participant (car B is a car type), and determine the type of traffic participant represented by car A (also a car type). According to the pre-defined traffic participant type association rules, the judgment is made, because car A and car B are the same type of traffic participants, and have a precedence relationship in the same lane or intersection (for example, car B is driving in front of car A), so car B is retained as the potential first preceding participant.

[0059] For the remaining potential first preceding participants in the temporary storage structure, their traffic flow direction related information is analyzed, and the traffic flow direction information of car A is obtained. For example, if car A is traveling north, and car B is also traveling north and in front of car A, the flow direction and logical relationship between car B and car A are judged to meet the requirements of the first preceding participant according to the traffic rules and the logical relationship in the traffic scene, so car B is finally determined as the first preceding participant of car A.

[0060] After completing the determination of the target leading participant corresponding to each space-time vertex, for the space-time vertex corresponding to car A, there are three traffic participation trajectories from the first space-time vertex (signal light L) to the space-time vertex corresponding to car A. Among these three trajectories, it is determined through analysis that the target leading participant of the space-time vertex corresponding to car A is signal light L.

[0061] For the space-time vertex corresponding to car B, there are 4 traffic participation trajectories from traffic light L to car B. Among these 4 trajectories, it is found that each traffic participation trajectory of car B includes traffic light L and car A (for example, it is controlled by traffic light L at the same intersection and has an interactive relationship with car A, such as passing through the intersection in sequence, etc.), so the target dominant participant of the space-time vertex corresponding to car B is traffic light L and car A.

[0062] For the space-time vertex corresponding to car C, there are two traffic participation trajectories from traffic light L to car C. In these two trajectories, the target dominant participants of the space-time vertex corresponding to car C are traffic light L and traffic management center (for example, traffic light L is controlled by traffic management center, and its state changes affect the driving of car C, such as the traffic management center adjusts the duration of traffic light L in a certain period of time, causing car C to slow down or stop).

[0063] For the space-time vertex corresponding to pedestrian P, there are three traffic participation trajectories from the traffic light L to pedestrian P. Among these three trajectories, the target dominant participants of the space-time vertex corresponding to pedestrian P are the traffic light L and car A (for example, pedestrian P is waiting to cross the road near car A, and the traffic light L controls the driving of car A, thereby indirectly affecting the waiting and walking timing of pedestrian P).

[0064] For the space-time vertex corresponding to the traffic light L, it is marked as the target dominant participant.

[0065] Next, the server counts the participation frequency of each target leading participant:

[0066] Traffic light L, as the target dominant participant, appeared a total of 4 times in the process of determining the target dominant participant of the space-time vertices corresponding to car A, car B, car C, and pedestrian P (once for car A, once for car B, once for car C, and once for pedestrian P).

[0067] Car A, as the target dominant participant, appeared once in the process of determining the target dominant participant of the space-time vertex corresponding to car B.

[0068] As the target-leading participant, the traffic management center appeared once in the process of determining the target-leading participant of the space-time vertex corresponding to car C.

[0069] Since the traffic light L has the highest participation frequency, the server marks the traffic participant corresponding to the traffic light L (that is, the traffic light L itself) as the key participant among the five traffic participants (car A, car B, car C, pedestrian P, and traffic light L).

[0070] Step S140, performing feature mining on the subject movement state trajectory data associated with the key participating subject to determine the first traffic accident risk data in the dynamic traffic participation process.

[0071] In this embodiment, for example, taking the subject movement state trajectory data of car A associated with the traffic light L as an example, the subject movement state trajectory data corresponding to car A can be extracted from the traffic spatiotemporal knowledge network data. These subject movement state trajectory data contain the state information of car A at different time points, for example, starting from 10:00:00 on May 1, 2023, one data point every 10 seconds, including position coordinates (such as longitude 116.39, latitude 39.90, etc.), speed (such as 30km / h, 35km / h, etc.), acceleration (such as acceleration values ​​when starting and braking), and driving direction (such as driving north) and other information.

[0072] The server obtains characteristic information of the movement status record of car A, including timestamp, location coordinates, speed, acceleration, driving direction, etc.

[0073] Specifically, the server calculates the speed fluctuation state of car A within a set time window (e.g., 10 minutes). During 10:00-10:10, the speed of car A changes from 30km / h to 35km / h and then to 32km / h, and the server analyzes the relationship between this speed fluctuation and traffic environment factors (e.g., road conditions, surrounding vehicle density, etc.). For example, it is found that the speed of car A fluctuates greatly when passing through a congested section of road, which may be because there are many vehicles around and the speed needs to be adjusted continuously.

[0074] The server detects the event frequency of acceleration events and deceleration events of car A. For example, car A has a deceleration event when approaching an intersection, and analyzes the relationship between these acceleration and deceleration events and traffic signals (such as deceleration according to the state change of the signal light L) and other traffic participants (such as deceleration when seeing pedestrian P nearby).

[0075] The server counts the relationship between the turning events of car A and the traffic scene within a set time window (such as 15 minutes) and determines the turning frequency of the turning events. For example, if car A has no turning events within these 15 minutes, it may mean that car A has been driving on a straight road and the traffic scene is relatively simple.

[0076] The server calculates the average value of the driving direction vector angles of car A at consecutive time points to obtain the driving direction stability of car A. Since car A has been driving north, the average value of the direction vector angle is close to 0, indicating that car A has high driving direction stability.

[0077] The server analyzes the smoothness of the driving trajectory of car A based on the curve fitting method. For example, the driving trajectory of car A appears as a relatively smooth curve on the map, indicating that car A is driving relatively stably without frequent sudden braking or acceleration that causes sudden changes in the trajectory.

[0078] The server determines the activity range of car A (for example, driving within a certain block) and analyzes the regularity of the traffic participation path. It is found that car A drives along a specific road network in this block with strong regularity.

[0079] The server analyzes the encounter and avoidance events between car A and other traffic participants (such as car B, pedestrian P, etc.), and counts the frequency and type of encounter and avoidance events. For example, car A and car B have an encounter event when they are driving in parallel, and there may be an avoidance event when approaching pedestrian P (even if pedestrian P has not actually crossed the road, car A may have made avoidance preparations such as slowing down).

[0080] The server analyzes the impact of these encounter and avoidance events on the changes in driving characteristics. For example, when driving parallel to car B, car A may maintain a relatively stable speed; but when avoiding pedestrian P, the speed of car A will change.

[0081] Therefore, the server marks the traffic participants whose driving characteristic change data meets the characteristic conditions of each candidate traffic accident template as traffic accident participants. For example, there is a traffic accident template with the characteristic condition of "suddenly reducing speed and pedestrians around when approaching the intersection". If the speed of car A suddenly drops from 35km / h to 30km / h when approaching the intersection (based on the position coordinates), and pedestrian P is nearby at this time (based on the position coordinates of pedestrian P and the position coordinates of car A), then car A will be marked as a traffic accident participant. The information related to these marked traffic accident participants (such as time, location, and types of traffic participants involved, etc.) constitutes the first traffic accident risk data.

[0082] Through the above feature mining of the subject movement state trajectory data of the key participant car A, the first traffic accident risk data in the dynamic traffic participation process is determined.

[0083] Based on the above steps, the embodiment of the present application generates a traffic spatiotemporal knowledge network by acquiring and processing the dynamic traffic flow data of multiple traffic participants in real time. By marking and analyzing the first spatiotemporal vertex and the target dominant participant in the traffic spatiotemporal knowledge network, it can accurately identify the key participants that have the greatest impact on traffic flow, conduct in-depth feature mining on the subject movement state trajectory data of the key participants, and discover potential traffic accident risk points in advance, significantly improve the early warning capability of the smart transportation system, and help reduce the occurrence of traffic accidents. Based on the analysis results of dynamic traffic flow data and key participants, it is convenient to allocate traffic resources more scientifically and reasonably, optimize traffic signal control strategies, improve road traffic efficiency, reduce traffic congestion, and enhance the efficiency and effectiveness of urban traffic management.

[0084] In a possible implementation, step S120 includes:

[0085] Step S121 , in the traffic spatiotemporal knowledge network data, taking the X spatiotemporal vertices as source nodes respectively, tracing out the traffic participation trajectory of each spatiotemporal vertex in the traffic spatiotemporal knowledge network data.

[0086] Step S122: If there is a traffic participation trajectory including the X space-time vertices in the tracked traffic participation trajectory, the corresponding space-time vertex is marked as the first space-time vertex.

[0087] Step S123: If there is no traffic participation trajectory including the X space-time vertices in the tracked traffic participation trajectory, a simulation participant subject is configured in the traffic space-time knowledge network data, and a simulation space-time pointing link between the simulation participant subject and the traffic space-time knowledge network data is configured, and the simulation participant subject is marked as the first space-time vertex.

[0088] In this embodiment, in the traffic spatiotemporal knowledge network data, the spatiotemporal vertex corresponding to car A is first used as the source node to start tracking. Car A is controlled by the traffic light L during 10:00:00-10:05:00 (for example, at 10:00:00, the traffic light L turns green and car A starts), and then runs parallel with car B for a period of time (such as 10:02:00-10:03:00), and at 10:03:00, there is an intersection with the sidewalk where pedestrian P is located (although pedestrian P may not cross the road in this area at this time, there is an intersection in the traffic spatiotemporal relationship). This is the traffic participation trajectory corresponding to car A.

[0089] The server then tracks the space-time vertex corresponding to car B as the source node. During 10:02:00-10:10:00, car B has parallel and interactive with car A (such as the parallel from 10:02:00 to 10:03:00 mentioned above), and is also affected by traffic light L (for example, car B waits when traffic light L is red from 10:00:00 to 10:00:30), and may affect the walking path of pedestrian P near a certain intersection (for example, at 10:08:00, pedestrian P may need to wait for car B to pass before crossing the road).

[0090] When the spatiotemporal vertex corresponding to car C is used as the source node, the data of car C shows that it has a faster speed in some sections of the road and slows down when approaching an intersection. For example, at 10:10:00, its speed drops from 50km / h to 30km / h, and its position coordinates are close to an intersection (longitude 116.40, latitude 39.90). In this process, it is also controlled by the traffic light L (such as slowing down according to the state of the traffic light L when approaching the intersection), and there may be potential interactions with cars A and B (such as interactions with cars A and B in different lanes at the intersection), and there may also be traffic spatiotemporal associations with pedestrian P (for example, pedestrian P is on the nearby sidewalk when car C approaches the intersection).

[0091] For pedestrian P, the data sent by the smartphone he carries shows that at 10:00:00, his position coordinates are on the sidewalk, his walking speed varies, such as 15km / h, and his walking direction is like walking towards the entrance of the shopping mall. While walking, pedestrian P will be affected by the traffic light L (such as waiting for the traffic light L to turn green before crossing the road), and there is also a traffic space-time interaction relationship with cars A, B, and C (such as when cars A, B, and C are driving on the road near pedestrian P, pedestrian P needs to choose the time to cross the road according to the vehicle situation).

[0092] The status information of the traffic light L is obtained by the server, including the time when each light color changes (such as the time interval of the red light from 10:00:00 to 10:00:30, the time interval of the green light from 10:00:30 to 10:01:30, etc.) and the corresponding intersection sign (such as the traffic light at intersection 1). The status change of the traffic light L directly affects the driving of cars A, B, and C (such as determining their starting, stopping, deceleration, etc.), and also indirectly affects the walking of pedestrian P (determining whether pedestrian P can cross the road).

[0093] Suppose that after tracking the traffic participation trajectories of car A, car B, car C, pedestrian P, and traffic light L, a series of traffic participation trajectories starting from traffic light L, passing through car A, car B, car C, and pedestrian P are found. For example, the traffic light L is red from 10:00:00 to 10:00:30, at which time cars A, B, and C stop, and pedestrian P waits on the roadside; from 10:00:30 to 10:01:30, the traffic light L is green, car A starts and runs parallel to car B for a distance, car C drives in another lane, and pedestrian P starts to cross the road (this sequential relationship exists in the traffic space-time relationship). Then the space-time vertex corresponding to the traffic light L can be marked as the first space-time vertex.

[0094] For example, during the tracking process, only trajectories containing cars A, B, C, and traffic lights L but no pedestrian P are found. For example, cars A, B, and C are driving on a road without a pedestrian crosswalk, interacting with traffic lights L, and no pedestrian P is involved. At this time, the server will configure a simulation participant in the traffic spatiotemporal knowledge network data. Assume that this simulation participant simulates the possible behavior of pedestrian P in this traffic scene, for example, assuming that pedestrian P crosses the road at the right time according to traffic rules. Then configure the simulation spatiotemporal pointing link between this simulation participant and the traffic spatiotemporal knowledge network data, such as setting the simulated pedestrian P to arrive at the sidewalk of a certain intersection at 10:05:00, and establish a connection with cars A, B, C, and traffic lights L in the traffic spatiotemporal relationship (for example, car A approaches this intersection at 10:05:00, and needs to pay attention to the possible appearance of simulated pedestrian P). Finally, this simulation participant is marked as the first spatiotemporal vertex.

[0095] In a possible implementation, step S120 includes:

[0096] Step S124, extracting the ath spatiotemporal vertex from the traffic spatiotemporal knowledge network data, taking the ath spatiotemporal vertex as the source node, performing walk sorting on the traffic spatiotemporal knowledge network data, and generating the ath basic traffic trajectory network. a is not greater than X. The root participant of the ath basic traffic trajectory network is the ath spatiotemporal vertex.

[0097] Step S125: If there is a space-time vertex in the X space-time vertices that does not belong to the a-th basic traffic trajectory network, the a+1-th space-time vertex is used as the source node to perform wandering sorting on the traffic space-time knowledge network data to generate the a+1-th basic traffic trajectory network.

[0098] Step S126: If the a-th basic traffic trajectory network includes the X space-time vertices, the a-th space-time vertex is marked as the first space-time vertex.

[0099] In this embodiment, when a=1 (taking the space-time vertex corresponding to car A as an example), the server extracts the space-time vertex corresponding to car A from the traffic space-time knowledge network data (set as the a=1th space-time vertex). Then, the space-time vertex corresponding to car A is used as the source node to perform wandering sorting on the traffic space-time knowledge network data.

[0100] Car A is equipped with a GPS positioning system, a speed sensor and other equipment. Its data includes timestamp (e.g., starting from 10:00:00 on May 1, 2023, with one data point every 10 seconds), location coordinates (longitude 116.39, latitude 39.90, etc.), speed (e.g., 30km / h, 35km / h, etc.), acceleration and driving direction (driving north).

[0101] The server sorts the walks in the order of the space-time pointing links. For example, car A travels from position 1 (longitude 116.39, latitude 39.90) to position 2 (longitude 116.395, latitude 39.902) from 10:00:00 to 10:00:10. If it intersects with the driving trajectory of car B or is affected by traffic light L during this process, then a space-time pointing link is constructed between the space-time vertex corresponding to car A and the space-time vertex corresponding to car B and traffic light L.

[0102] In this wandering sorting process, the server will traverse the traffic participants and their interactions related to car A in turn to generate the first basic traffic trajectory network. The root participant of this trajectory network is the space-time vertex corresponding to car A. In this trajectory network, car A is controlled by the traffic light L at 10:00:00 (for example, the traffic light L is green and car A starts), and runs parallel to car B from 10:02:00 to 10:03:00, and at 10:03:00, there is an intersection with the sidewalk where pedestrian P is located (although pedestrian P may not cross the road in this area at this time, there is an intersection in the traffic space-time relationship).

[0103] When the first basic traffic trajectory network does not contain all space-time vertices, if there are space-time vertices that do not belong to this network in this first basic traffic trajectory network, for example, only car B and traffic light L are in this network, while car C and pedestrian P are not.

[0104] Then the server will use the space-time vertex corresponding to car B (the a+1=2nd space-time vertex) as the source node, sort the traffic space-time knowledge network data, and generate the second basic traffic trajectory network.

[0105] During the period from 10:02:00 to 10:10:00, car B runs in parallel and interacts with car A, and is also affected by traffic light L. It may also affect the walking path of pedestrian P near a certain intersection (because pedestrian P may need to wait for car B to pass before crossing the road). When the server constructs the second basic traffic trajectory network, it will follow the traffic participation trajectory of car B and include the traffic participants that have an interactive relationship with car B and their spatiotemporal relationships into this trajectory network.

[0106] If the space-time vertices corresponding to car B, car C, pedestrian P, and traffic light L are already included in the first basic traffic trajectory network, then the space-time vertex corresponding to car A can be marked as the first space-time vertex. For example, in the basic traffic trajectory network of car A, while car B is parallel to car A, car C is driving on a nearby road (having a potential interaction relationship with car A, such as at an intersection), pedestrian P has a traffic space-time intersection with car A on the sidewalk, and traffic light L controls the start of car A, etc. In this case, the space-time vertex corresponding to car A meets the condition that it has an interaction relationship with all other traffic participants, so it is marked as the first space-time vertex.

[0107] In a possible implementation, step S130 includes:

[0108] Step S131, extracting Y traffic participation trajectories e from the first spatiotemporal vertex to the bth spatiotemporal vertex from the traffic spatiotemporal knowledge network data, and obtaining the interactive participating subjects respectively covered by the Y traffic participation trajectories e. b is not greater than X.

[0109] Step S132: Among the interactive participating subjects respectively covered by the Y traffic participating trajectories e, the interactive participating subjects existing in the Y traffic participating trajectories e are marked as the target leading participating subjects corresponding to the b-th space-time vertex.

[0110] In this embodiment, the server extracts Y traffic participation trajectories from the first spatiotemporal vertex (traffic light L) to the b=1th spatiotemporal vertex corresponding to car A from the traffic spatiotemporal knowledge network data. For example, there may be 3 different traffic participation trajectories from traffic light L to car A.

[0111] In the first trajectory, car B is included (because car B and car A have an interactive relationship under the control of traffic light L, such as waiting for the green light at the same time or driving in the same lane, etc.), traffic light L and pedestrian P (assuming that pedestrian P is on the nearby sidewalk when car A starts).

[0112] In the second trajectory, car C (car C may be parallel or intersecting with car A in another lane) and traffic light L are included.

[0113] In the third trajectory, car B, car C, traffic light L and pedestrian P are covered.

[0114] Next, the server marks the interactive participants that exist in all three trajectories among the interactive participants covered by the three traffic participation trajectories as the target dominant participant of the spatiotemporal vertex corresponding to car A. In this example, traffic light L exists in all three trajectories, so traffic light L is marked as the target dominant participant of the spatiotemporal vertex corresponding to car A.

[0115] Taking car B (the bth space-time vertex) as an example, the server extracts the traffic participation trajectory from traffic light L (the first space-time vertex) to car B. Assume that there are 4 such trajectories.

[0116] In the first trajectory, it covers car A (which is controlled by traffic light L at the same intersection and has an interactive relationship with car A, such as passing through the intersection in sequence), traffic light L and pedestrian P (pedestrian P is on the sidewalk near car B and is affected by the movement of car B).

[0117] In the second trajectory, it covers car C (which has an interactive relationship with car B near the intersection), traffic light L, and car A (the driving status of car A affects the driving path selection of car B).

[0118] In the third trajectory, traffic light L, car A, car C and pedestrian P are covered (the movement of car B affects the walking path selection of pedestrian P, etc.).

[0119] In the fourth trajectory, the traffic light L, car A, and car C (the interaction of car B with other vehicles in a specific traffic scenario) are covered.

[0120] After analysis, it is found that each traffic participation trajectory of car B includes traffic light L and car A, so the target dominant participants of the space-time vertex corresponding to car B are traffic light L and car A.

[0121] Taking car C (the bth spatiotemporal vertex) as an example, the server extracts two traffic participation trajectories from traffic light L to car C.

[0122] In the first trajectory, it covers traffic light L (traffic light L controls the driving of car C), traffic management center (traffic light L is controlled by the traffic management center, and its state change affects the driving of car C) and car A (the driving of car A has an interactive relationship with car C at the intersection).

[0123] In the second trajectory, the traffic light L, the traffic management center and the car B are covered (the driving of the car B indirectly affects the driving status of the car C at the intersection).

[0124] The target leading participants of the space-time vertex corresponding to car C are the traffic light L and the traffic management center.

[0125] Taking pedestrian P (the bth space-time vertex) as an example, the server extracts three traffic participation trajectories from the traffic light L to pedestrian P.

[0126] In the first trajectory, it covers traffic light L (traffic light L controls the timing of pedestrian P crossing the road), car A (the driving state of car A affects the waiting and walking timing of pedestrian P) and car B (the driving of car B affects the waiting situation of pedestrian P at the intersection).

[0127] In the second trajectory, the traffic light L, car A and car C are covered (the movement of car C also indirectly affects the traffic conditions in the area where pedestrian P is located).

[0128] In the third trajectory, the traffic light L, cars B and C are included (the driving of cars B and C affects the walking decision of pedestrian P on the nearby sidewalk). The target dominant participants of the spatiotemporal vertex corresponding to pedestrian P are traffic light L and car A.

[0129] Taking the traffic light L (the bth space-time vertex) as an example, the space-time vertex corresponding to the traffic light L is itself marked as the target dominant participant.

[0130] In a possible implementation, step S130 may further include:

[0131] Step S133: marking the first space-time vertex as the target dominant participant of the first space-time vertex.

[0132] Step S134, obtaining the first preceding participant of the second space-time vertex, marking the second leading participant of the first preceding participant of the second space-time vertex and the second space-time vertex as the target leading participant of the second space-time vertex. The second space-time vertex is the space-time vertex other than the first space-time vertex among the X space-time vertices.

[0133] In this embodiment, for the first spatiotemporal vertex, the traffic light L, the server directly marks the traffic light L as its own target dominant participant. This is because the traffic light L has a unique position in the entire traffic scene, and its state change directly affects the behavior of other traffic participants, so it itself is its target dominant participant.

[0134] For car A (as the second space-time vertex), the server obtains the first preceding participant of car A. The first preceding participant of car A may be traffic light L (because car A needs to start or stop according to the state of traffic light L).

[0135] The server obtains the sub-dominant participant of the first preceding participant (traffic light L) of car A. Assume that in this traffic scenario, it is analyzed from the traffic spatiotemporal knowledge network data that the sub-dominant participant of traffic light L is the traffic management center (because the traffic management center controls the state change of traffic light L).

[0136] Then the traffic management center and car A are marked as the target dominant participant of car A. This is because the traffic management center indirectly affects the driving of car A by controlling the traffic light L, so when determining the target dominant participant of car A, this indirect influence relationship should be taken into account.

[0137] For the target dominant participant of car B (the second space-time vertex), the server determines its first preceding participant. For example, the first preceding participant of car B may be car A (because car A may enter a lane or intersection before car B, which has a certain impact on the driving of car B). The secondary dominant participant of car A may be the driver of car A (because the driver controls the driving state of car A). Therefore, the driver of car A and car B are marked as the target dominant participant of car B. This shows that the operation of the driver of car A indirectly affects the driving of car B through car A, so it is included in the category of the target dominant participant of car B.

[0138] For the target dominant participant of car C (the second space-time vertex), the server analyzes and concludes that its first preceding participant may be the signal light L (because the driving of car C is controlled by the signal light L). The secondary dominant participant of the signal light L is the traffic management center (as explained above). Therefore, the traffic management center and car C are marked as the target dominant participant of car C. This reflects the indirect influence of the traffic management center on the driving of car C through the signal light L.

[0139] For the target dominant participant of pedestrian P (the second space-time vertex), the first preceding participant may be the traffic light L (because pedestrian P needs to decide whether to cross the road based on the state of the traffic light L). The secondary dominant participant of the traffic light L is the traffic management center. Therefore, the traffic management center and pedestrian P are marked as the target dominant participant of pedestrian P. This means that the traffic management center indirectly affects pedestrian P's walking decision by controlling the traffic light L.

[0140] In a possible implementation, the method further includes:

[0141] Step A110: extracting a second preceding participant of the first preceding participant from the traffic spatiotemporal knowledge network data, and extracting a first traffic participation trajectory including the first preceding participant and the second preceding participant from the traffic spatiotemporal knowledge network data.

[0142] Step A120: extract a second traffic participation trajectory from the first traffic participation trajectory, and mark the second preceding participant corresponding to the second traffic participation trajectory as the sub-dominant participant of the first preceding participant. The second traffic participation trajectory is used to indicate that the first preceding participant and the second preceding participant have a participation interaction relationship, and the first preceding participant does not have a participation trajectory of a preceding participant other than the second preceding participant.

[0143] In this embodiment, taking car A and its prior participants as an example, its first prior participant is the traffic light L. The server extracts the second prior participant of the traffic light L (the first prior participant) from the traffic spatiotemporal knowledge network data. Assuming that in this traffic scenario, it is analyzed from the traffic spatiotemporal knowledge network data that the second prior participant of the traffic light L is the traffic management center (because the traffic management center controls the state change of the traffic light L).

[0144] The server extracts the first traffic participation trajectory including the traffic light L and the traffic management center from the traffic spatiotemporal knowledge network data. This trajectory may include information such as the time when the traffic management center sends the control instruction to the traffic light L and the state response time of the traffic light L. For example, the traffic management center sends a control instruction at 10:00:00, and the traffic light L begins to respond to this instruction at 10:00:05, changing from red to green. The relevant information of this process constitutes part of the first traffic participation trajectory.

[0145] From this first traffic participation trajectory, the server extracts the second traffic participation trajectory. For example, this second traffic participation trajectory indicates that in a certain time period (such as 10:00:00-10:01:00), the traffic management center sent a special control instruction (such as adjusting the duration of signal light L in response to sudden traffic flow), and in this time period, signal light L has no participation trajectory of other prior participants. Then the traffic management center corresponding to this second traffic participation trajectory is marked as the sub-dominant participant of signal light L. This means that in this specific time period and traffic scenario, the influence of the traffic management center on signal light L is unique and dominant, so it is marked as a sub-dominant participant.

[0146] Taking car C and its prior participants as an example, its first prior participant is traffic light L. The server extracts the second prior participant of traffic light L as the traffic management center from the traffic spatiotemporal knowledge network data. The server extracts the first traffic participation trajectory including traffic light L and traffic management center. For example, the traffic management center sends an instruction to adjust the duration of traffic light L at 10:10:00, and traffic light L responds at 10:10:05. The relevant information of this process constitutes the first traffic participation trajectory.

[0147] The second traffic participation trajectory is extracted from this first traffic participation trajectory. For example, during the time period of 10:10:00-10:10:30, the traffic management center sends a command to extend the red light time of signal light L to control traffic flow, and during this time period, there is no participation trajectory of other prior participants of signal light L. Therefore, the traffic management center corresponding to this second traffic participation trajectory is marked as the secondary dominant participant of signal light L (the first prior participant of car C). This shows that in this specific traffic scenario, the control of signal light L by the traffic management center has an indirect but important impact on the driving of car C.

[0148] In a possible implementation, the method further includes:

[0149] Step B110: marking the X space-time vertices as the basic dominant participating entities of the second space-time vertex.

[0150] Step S134 includes:

[0151] Step S1341, in the c-th tracking and extraction process of the leading participating subject, extracting Z first preceding participating subjects of the second spatiotemporal vertex from the traffic spatiotemporal knowledge network data, where c is a positive integer and Z is a positive integer.

[0152] Step S1342, obtain the sub-dominant participating entities corresponding to the Z first preceding participating entities of the second space-time vertex, and mark the shared participating entity of the sub-dominant participating entity corresponding to the Z first preceding participating entities and the second space-time vertex as the cth adjusted dominant participating entity of the second space-time vertex.

[0153] Step S1343: If the cth adjustment leading participant of the second space-time vertex is consistent with the c-1th adjustment leading participant of the second space-time vertex, the cth adjustment leading participant of the second space-time vertex is marked as the target leading participant of the second space-time vertex. When c is the default value, the c-1th adjustment leading participant of the second space-time vertex is the basic leading participant of the second space-time vertex.

[0154] In this embodiment, taking car A (as an example of the second space-time vertex) as an example, the server marks five space-time vertices (car A, car B, car C, pedestrian P, and signal light L) as the basic dominant participant of car A (as the second space-time vertex). This means that in the initial stage, these traffic participants are all regarded as subjects that may have an impact on the dominance of car A.

[0155] In the c=1th ​​tracking and extraction process of the leading participant, the Z first preceding participants of car A are extracted from the traffic spatiotemporal knowledge network data. For example, set Z=2, and through the search and screening method mentioned above, the server determines that the two first preceding participants of car A are the traffic light L and car B (because car B may enter a lane or intersection before car A, which has a certain impact on the driving of car A).

[0156] The server obtains the sub-dominant participating entities corresponding to the two first prior participating entities of car A. It is assumed that the sub-dominant participating entity of traffic light L is the traffic management center, and the sub-dominant participating entity of car B is the driver of car B (because the driver controls the driving status of car B).

[0157] Then the shared participants of these two secondary dominant participants (in this example, there is no shared participant) and car A are marked as the first adjustment dominant participant of car A. At this time, the first adjustment dominant participant of car A is only car A itself (because there is no shared participant).

[0158] In the subsequent c=2nd adjustment leading participant calculation, it is assumed that the server again determines that the first leading participant of car A is still the signal light L and car B, and their secondary leading participants are still the traffic management center and the driver of car B. At this time, the second leading participant of car A is still car A itself (because there is no shared participant).

[0159] Since the second adjustment dominant participant of Car A is the same as the first adjustment dominant participant (both are Car A itself), this adjustment dominant participant is marked as the target dominant participant of Car A.

[0160] Taking car C (as an example of the second space-time vertex), the server labels five space-time vertices (car A, car B, car C, pedestrian P, traffic light L) as the basic dominant participating entities of car C (as the second space-time vertex).

[0161] In the c=1th ​​tracking and extraction process of the dominant participant, Z=2 is set, and the server extracts the two first preceding participants of car C from the traffic spatiotemporal knowledge network data as the traffic light L and car A (because the driving status of car A may affect the driving of car C at the intersection).

[0162] The server obtains the sub-dominant participating entities corresponding to the two first preceding participating entities. The sub-dominant participating entity of the traffic light L is the traffic management center, and the sub-dominant participating entity of the car A is the driver of the car A.

[0163] The shared participants of these two secondary dominant participants (in this example, there is no shared participant) and car C are marked as the first adjustment dominant participant of car C. At this time, the first adjustment dominant participant of car C is only car C itself.

[0164] In the calculation of the c=2nd adjustment leading participant, assuming that the first leading participant of car C is still the signal light L and car A, and their second leading participants remain unchanged. Then the second leading participant of car C is still car C itself.

[0165] Since the second adjustment dominant participant of Car C is consistent with the first adjustment dominant participant, this adjustment dominant participant is marked as the target dominant participant of Car C.

[0166] In a possible implementation, step S1341 includes:

[0167] Step S1341-1, the traffic spatiotemporal knowledge network data is stored in a set data structure, wherein the spatiotemporal vertices and the related information of the spatiotemporal pointing links are arranged in order, an index is established for each spatiotemporal vertex with the position of the spatiotemporal vertex in the traffic network as the index key, and an index is established for the traffic participant type information. The position includes geographic coordinates or a logical location identifier.

[0168] Step S1341-2, set a search range parameter for limiting the time and space range considered when searching for a prior participant, the search range parameter is determined based on the distance and time interval factors of the mutual influence between traffic participants in historical data, and initialize an empty temporary storage structure at the same time. The empty temporary storage structure is an array or a linked list, which is used to store potential first prior participants found during the search process.

[0169] Step S1341-3, according to the identifier of the second space-time vertex, search for the corresponding space-time vertex in the traffic space-time knowledge network data, obtain the position information and related time information of the space-time vertex in the traffic network, and traverse the space-time vertices in the traffic space-time knowledge network data based on the determined search range. During the traversal process, by comparing the position relationship and time relationship between the space-time vertices, mark the space-time vertices in the search range as potential first prior participating entities, and add the potential first prior participating entity to the temporary storage structure.

[0170] Step S1341-4, for each potential first prior participant in the temporary storage structure, obtain the type of traffic participant represented by the potential first prior participant, and determine the type of traffic participant represented by the second space-time vertex, and make a judgment based on the predefined traffic participant type association rule. When the second space-time vertex is a vehicle, if the potential first prior participant is the type of traffic participant of the preceding vehicle in the same lane or the traffic light at the intersection that has a leading role in guiding the vehicle, then retain the potential first prior participant. When the second space-time vertex is a pedestrian, if the potential first prior participant is a traffic coordinator who guides pedestrians to cross the road or a vehicle in front of the pedestrian path that affects the pedestrian's walking, then retain the potential first prior participant, and remove other potential first prior participants that do not meet the type association rule from the temporary storage structure.

[0171] Step S1341-5, for the remaining potential first prior participating entities in the temporary storage structure, analyze the traffic flow direction related information of the first prior participating entity, and obtain the traffic flow direction information of the second space-time vertex, and judge whether the flow direction and logical relationship between the potential first prior participating entity and the second space-time vertex meet the requirements of the prior participating entity based on the traffic rules and the logical relationship in the traffic scene, and remove the potential first prior participating entity that does not meet the requirements of the prior participating entity from the temporary storage structure.

[0172] Step S1341-6, count the target number of potential first prior participants remaining in the temporary storage structure, if the target number is less than or equal to the set number, directly use all elements in the temporary storage structure as the final first prior participants. If the target number is greater than the set number, sort the elements in the temporary storage structure according to the set sorting rule, and select the first Z potential first prior participants as the final result according to the sorting result, the set sorting rule is based on the spatiotemporal proximity between the spatiotemporal vertex and the second spatiotemporal vertex, or based on the potential influence of the traffic participant on the second spatiotemporal vertex.

[0173] In this embodiment, taking car A (as the second space-time vertex) as an example, the server stores the traffic space-time knowledge network data in a set data structure. For example, a graph structure is used to store the relevant information of the space-time vertices and the space-time pointing links, and they are arranged in order according to the identification of the traffic participants or the space-time sequence. For each space-time vertex, an index is established with the position of the space-time vertex in the traffic network as the index key, and an index is established for the type information of the traffic participants. For car A, its location information may be geographic coordinates such as longitude and latitude (such as longitude 116.39, latitude 39.90), which serves as an index key to facilitate subsequent searches.

[0174] Set a search range parameter to limit the time and space range considered when searching for the first participant. Assuming that according to historical data, the effective distance between cars that affect each other is generally within 100 meters, and traffic events with a time interval of less than 10 minutes are interrelated, then set the search range parameter based on this standard. At the same time, initialize an empty temporary storage structure, where an array is used to store the potential first first participant found during the search process.

[0175] According to the identification of car A, the corresponding space-time vertex is searched in the traffic space-time knowledge network data, and the location information of the space-time vertex in the traffic network (such as longitude 116.39, latitude 39.90) and related time information (such as 10:00:00) are obtained. Based on the determined search range, the space-time vertices in the traffic space-time knowledge network data are traversed.

[0176] During the traversal process, for example, the spatiotemporal vertex of car B is found, and the positional relationship (calculate whether the distance is within 100 meters) and time relationship (whether there is a possibility of interaction within 10 minutes) between car A and car B are compared. Assuming that car B is located at longitude 116.38 and latitude 39.91 at 10:02:00, it is close to car A and within 10 minutes, then car B is marked as a potential first prior participant and added to the temporary storage structure. Similarly, the spatiotemporal vertex of traffic light L is found. If its state change is within 10 minutes and within the influence range of car A (for example, car A is within the intersection controlled by traffic light L), traffic light L is also marked as a potential first prior participant and added to the temporary storage structure.

[0177] For each potential first preceding participant in the temporary storage structure, obtain the type of traffic participant represented by the potential first preceding participant, and determine the type of traffic participant represented by car A (both are car types). According to the pre-defined traffic participant type association rules, the judgment is made. Because car A and car B are the same type of traffic participants and have a precedence relationship in the same lane or intersection (for example, car B is driving in front of car A), car B is retained as the potential first preceding participant. For signal light L, it is a traffic signal light type, which has a leading role in guiding the driving of car A at the intersection, so signal light L is also retained as the potential first preceding participant.

[0178] For the remaining potential first prior participants (car B and traffic light L) in the temporary storage structure, their traffic flow direction related information is analyzed, and the traffic flow direction information of car A is obtained. For example, if car A is traveling north, car B is also traveling north and is in front of car A. According to the traffic rules and the logical relationship in the traffic scene, the flow direction and logical relationship between car B and car A meet the requirements of the prior participant, so car B is retained. Although traffic light L has no driving direction, its driving control of car A meets the requirements of the prior participant, so it is also retained.

[0179] The target number of potential first precedent participants (car B and traffic light L) remaining in the temporary storage structure is counted, where the target number is 2. Assuming the set number is 2, since the target number is equal to the set number, all elements in the temporary storage structure (car B and traffic light L) are directly used as the final first precedent participants.

[0180] Taking pedestrian P (as the second spatiotemporal vertex) as an example, the server also stores the traffic spatiotemporal knowledge network data in a suitable data structure, establishes indexes, and other operations. For pedestrian P, its location information (such as longitude and latitude on the sidewalk) is used as part of the index key. Set search range parameters, such as the effective distance and time range of pedestrians affected by surrounding vehicles or traffic coordinators. Initialize an array as a temporary storage structure.

[0181] Find the corresponding space-time vertex based on the identification of pedestrian P, and obtain its location (such as the longitude and latitude of the sidewalk near a shopping mall) and time information (such as 10:00:00). Traverse the space-time vertices in the traffic space-time knowledge network data. If the space-time vertex of car A is found within the search range, and car A is in front of the path of pedestrian P and within the influence range (for example, car A is approaching when pedestrian P is about to cross the road), car A is marked as a potential first priority participant and added to the temporary storage structure. If the traffic light L is found to be within the range that affects pedestrian P crossing the road, it is also added to the temporary storage structure.

[0182] For the car A and traffic light L in the temporary storage structure, obtain the types of traffic participants they represent. Pedestrian P is a pedestrian type. For car A, it is a vehicle type that is in front of the pedestrian path and affects the pedestrian's walking. It meets the type association rule and is retained. Traffic light L is a traffic facility type that guides pedestrians to cross the road. It also meets the rule and is retained.

[0183] For car A, its traffic flow direction is analyzed (assuming that car A is traveling northward and close to where pedestrian P is crossing the road). Pedestrian P is walking towards the mall entrance (perpendicular to the direction of car A). According to traffic rules and scene logic, car A has a prior influence on pedestrian P, so car A is retained. The state of traffic light L affects the walking of pedestrian P, which also meets the requirements and is retained.

[0184] Assume that there are only two potential first prior participants in the temporary storage structure, namely car A and traffic light L, and the number is set to 2. Since the target number is equal to the set number, car A and traffic light L are directly taken as the final first prior participants of pedestrian P.

[0185] In a possible implementation manner, the second space-time vertices include Q. The method further includes:

[0186] Step C110 , taking the first space-time vertex as a source node, performing reverse tracking and polling on the traffic space-time knowledge network data to generate a set of participating entities corresponding to the X space-time vertices.

[0187] Step S134 includes:

[0188] Step S1344, extracting the first preceding participant of the d-th second space-time vertex from the participant set, marking the second leading participant of the first preceding participant of the d-th second space-time vertex and the d-th second space-time vertex as the target leading participant of the d-th second space-time vertex. d is not greater than Q.

[0189] Step S1345: if the dth second space-time vertex is not the last space-time vertex in the participating subject set, extract the d+1th second space-time vertex from the participating subject set to obtain the target dominant participating subject of the d+1th second space-time vertex.

[0190] Step S1346, if the d-th second space-time vertex is the last space-time vertex in the set of participating entities, then enter the step of marking the traffic participating entity corresponding to the space-time vertex with the highest participation frequency among the target dominant participating entities corresponding to each of the X space-time vertices as a key participating entity among the X traffic participating entities.

[0191] In this embodiment, the traffic light L (the first spatiotemporal vertex) is used as the source node for reverse tracking polling. The server uses the traffic light L as the source node to perform reverse tracking polling on the traffic spatiotemporal knowledge network data. For car A, the server reversely searches for information related to car A along the spatiotemporal pointing link. For example, car A starts at 10:00:00 because the traffic light L turns green, and this associated information is recorded. At the same time, car A has a parallel relationship with car B during driving, and this relationship is also considered in reverse tracking. For pedestrian P, the server finds that pedestrian P crossed the road at 10:05:00 because the previous state change of the traffic light L allowed pedestrians to pass, and this relationship is also included.

[0192] Through such reverse tracking polling, the server generates a set of participating entities corresponding to X spatiotemporal vertices (car A, car B, car C, pedestrian P, and traffic light L). This set contains the association information between various traffic participating entities in the traffic spatiotemporal knowledge network, such as the association between car A and traffic light L, car B, and the association between pedestrian P and traffic light L.

[0193] Taking car A (the d=1 second space-time vertex, assuming Q>=1) as an example, the first preceding participant of car A is extracted from the participant set. In this set, the server finds that the first preceding participant of car A is the traffic light L (because the previous reverse tracking polling shows that the start of car A is controlled by the traffic light L).

[0194] Then, the sub-dominant participant of the traffic light L is obtained. It is known that the sub-dominant participant of the traffic light L is the traffic management center (because the traffic management center controls the state change of the traffic light L).

[0195] Therefore, the traffic management center and car A are marked as the target leading participants of car A.

[0196] Taking car B (the d=2nd second space-time vertex, assuming Q>=2) as an example, extract the first preceding participant of car B from the participant set. It is found in the set that the first preceding participant of car B is car A (because car A may enter a lane or intersection before car B, affecting the driving of car B).

[0197] The secondary dominant participant in obtaining car A is the driver of car A (because the driver controls the driving state of car A).

[0198] The driver of car A and car B are marked as the target dominant participants of car B.

[0199] Taking car C (the d=3th second space-time vertex, assuming Q>=3) as an example, the first preceding participant of car C is extracted from the set of participating entities, and it is found to be traffic light L (because the driving of car C is controlled by traffic light L).

[0200] The secondary dominant participant in obtaining the traffic light L is the traffic management center.

[0201] The traffic management center and car C are marked as the target leading participants of car C.

[0202] Taking pedestrian P (the d=4th second space-time vertex, assuming Q>=4) as an example, the first preceding participant of pedestrian P extracted from the set of participants is the traffic light L (because pedestrian P crosses the road based on the state of the traffic light L).

[0203] The secondary dominant participant in obtaining the traffic light L is the traffic management center.

[0204] The traffic management center and pedestrian P are marked as the target leading participants of pedestrian P.

[0205] Assuming that pedestrian P is the last space-time vertex (the dth second space-time vertex is the last space-time vertex in the set of participating entities), the labeling of the target dominant participating entities of all second space-time vertices (car A, car B, car C, pedestrian P) has been completed.

[0206] Next, we enter the step of marking the traffic participant corresponding to the space-time vertex with the highest participation frequency among the target dominant participant corresponding to each of the X space-time vertices as the key participant among the X traffic participants. For example, according to statistics, the traffic light L as the target dominant participant appeared a total of 4 times (1 time for car A, 1 time for car B, 1 time for car C, and 1 time for pedestrian P) in the process of determining the target dominant participant of the space-time vertices corresponding to car A, car B, car C, and pedestrian P, with the highest frequency of occurrence, so the traffic participant corresponding to the traffic light L (that is, the traffic light L itself) is marked as the key participant among the five traffic participants (car A, car B, car C, pedestrian P, traffic light L).

[0207] In a possible implementation, the first traffic accident risk data includes the participants of the traffic accident. Step S140 includes:

[0208] Step S141: extracting the subject movement state trajectory data associated with the key participating subject. The subject movement state trajectory data includes movement state records corresponding to each of the X traffic participating subjects.

[0209] Step S142: using the movement status records corresponding to the X traffic participants, respectively, to extract the driving characteristic change data corresponding to the X traffic participants respectively.

[0210] Step S143, marking the traffic participants whose driving characteristic change data meet the characteristic conditions of each candidate traffic accident template as traffic accident participants.

[0211] For example, in a possible implementation, the step of using the movement status records corresponding to the X traffic participants to extract the driving characteristic change data corresponding to the X traffic participants includes:

[0212] Acquire characteristic information of the movement status records of the X traffic participants, wherein the characteristic information includes a timestamp, a position coordinate, a speed, an acceleration, a driving direction, and a turn signal state of a vehicle, and a timestamp, a position coordinate, a walking speed, and a walking direction of a pedestrian.

[0213] According to the characteristic information of the movement status records of the X traffic participants, driving characteristic change data of the X traffic participants are determined, and the driving characteristic change data include: speed-related characteristic changes, driving direction-related characteristic changes, position-related characteristic changes, and characteristic changes related to interactions between traffic participants.

[0214] Wherein, the step of determining the change of the speed-related characteristic comprises:

[0215] Calculate the speed fluctuation state of the traffic participant within a set time window, analyze the relationship between the speed fluctuation state and traffic environment factors, detect the event frequency of acceleration events and deceleration events of the traffic participant, and analyze the relationship between the acceleration events and deceleration events and traffic signals and other traffic participants.

[0216] The step of determining the change of the driving direction related characteristic comprises:

[0217] The relationship between the turning events of the traffic participant within a set time window and the traffic scene is counted and the turning frequency of the turning events is determined, and the average value of the driving direction vector angles of the traffic participant at consecutive time points is calculated to obtain the driving direction stability of the traffic participant.

[0218] The step of determining the change of the position-related feature comprises:

[0219] Based on the curve fitting method, the smoothness of the driving trajectory of the traffic participant is analyzed, and the activity range of the traffic participant is determined and the regularity of the traffic participation path is obtained by analysis.

[0220] The step of determining the characteristic changes related to the interaction between the traffic participants includes:

[0221] Analyze the encounter and avoidance events between all types of traffic participants, count the frequency and type of the encounter and avoidance events, and analyze the impact of the encounter and avoidance events on the change of driving characteristics.

[0222] In this embodiment, the server extracts the subject movement state trajectory data of car A associated with the traffic light L. Car A is equipped with GPS positioning system, speed sensor and other equipment, and its data includes timestamp (such as one data point every 10 seconds starting from 10:00:00 on May 1, 2023), location coordinates (longitude 116.39, latitude 39.90, etc.), speed (such as 30km / h, 35km / h, etc.), acceleration and driving direction (driving north). These data constitute the movement state record of car A and are also part of the subject movement state trajectory data.

[0223] For car B, the data it sends also contains similar information, such as its location at longitude 116.38, latitude 39.91 at 10:02:00, speed 25km / h, direction of travel east, etc. The data for car C shows that at 10:10:00, its speed dropped from 50km / h to 30km / h, and its location coordinates were close to an intersection (longitude 116.40, latitude 39.90), etc. The data sent by the smartphone carried by pedestrian P includes timestamps (such as starting at 10:00:00), location coordinates (such as longitude and latitude information on the sidewalk), walking speed (such as varying from 15km / h), and walking direction (such as walking towards the mall entrance). The status information of the traffic light L itself (such as the time interval when the red light is on 10:00:00-10:00:30, the time interval when the green light is on 10:00:30-10:01:30, etc.) and the corresponding intersection signs (such as the traffic light at intersection 1) are also included as part of the subject movement status trajectory data.

[0224] The server calculates the speed fluctuation of car A within a set time window (e.g., 10 minutes). During the period from 10:00 to 10:10, the speed of car A changes from 30km / h to 35km / h to 32km / h, etc. The server analyzes the relationship between this speed fluctuation and traffic environment factors (such as road conditions, surrounding vehicle density, etc.). For example, it is found that the speed of car A fluctuates greatly when passing through a congested section of road. This may be because there are many vehicles around and the speed needs to be adjusted continuously.

[0225] The server detects the event frequency of acceleration events and deceleration events of car A. For example, car A has a deceleration event when approaching an intersection, and analyzes the relationship between these acceleration and deceleration events and traffic signals (such as deceleration according to the state change of the signal light L) and other traffic participants (such as deceleration when seeing pedestrian P nearby).

[0226] The server counts the relationship between the turning events of car A and the traffic scene within a set time window (such as 15 minutes) and determines the turning frequency of the turning events. For example, if car A has no turning events within these 15 minutes, it may mean that car A has been driving on a straight road and the traffic scene is relatively simple.

[0227] The server calculates the average value of the driving direction vector angles of car A at consecutive time points to obtain the driving direction stability of car A. Since car A has been driving north, the average value of the direction vector angle is close to 0, indicating that car A has high driving direction stability.

[0228] The server analyzes the smoothness of the driving trajectory of car A based on the curve fitting method. For example, the driving trajectory of car A appears as a relatively smooth curve on the map, indicating that car A is driving relatively stably without frequent sudden braking or acceleration that causes sudden changes in the trajectory.

[0229] The server determines the activity range of car A (for example, driving within a certain block) and analyzes the regularity of the traffic participation path. It is found that car A drives along a specific road network in this block with strong regularity.

[0230] The server analyzes encounters and avoidance events between all types of traffic participants, and counts the frequency and types of encounters and avoidance events. For example, car A and car B have an encounter event when they are driving in parallel, and there may be an avoidance event when approaching pedestrian P (even if pedestrian P has not actually crossed the road, car A may have made avoidance preparations such as slowing down).

[0231] The server analyzes the impact of these encounter and avoidance events on the changes in driving characteristics. For example, when driving parallel to car B, car A may maintain a relatively stable speed; but when avoiding pedestrian P, the speed of car A will change.

[0232] There is a characteristic condition of a traffic accident template: "When approaching an intersection, the speed suddenly decreases and there are pedestrians nearby." If the speed of car A suddenly decreases from 35km / h to 30km / h when approaching an intersection (based on the position coordinates), and pedestrian P is nearby at this time (based on the position coordinates of pedestrian P and the position coordinates of car A), then car A will be marked as a participant in the traffic accident.

[0233] Similarly, for traffic participants such as car B and car C, if their driving characteristic change data meets the characteristic conditions of other candidate traffic accident templates, they will also be marked as traffic accident participants. The information related to these marked traffic accident participants (such as time, location, and types of traffic participants involved, etc.) constitutes the first traffic accident risk data.

[0234] For example, in a possible implementation, the method further includes:

[0235] After the traffic accident participants are warned, the dynamic traffic participation process continues.

[0236] Alternatively, obtain the traffic cluster strategy on which the dynamic traffic participation process is based, obtain the first traffic cluster where the traffic accident participant is located under the traffic cluster strategy, obtain the traffic linkage request corresponding to the first traffic cluster in the dynamic traffic participation process, perform traffic accident prediction on the first traffic cluster in the process of responding to the traffic linkage request, and determine the second traffic accident risk data of the dynamic traffic participation process.

[0237] In this embodiment, taking car A being marked as a participant in a traffic accident as an example, when car A is marked as a participant in a traffic accident (for example, as mentioned above, car A suddenly slows down when approaching an intersection and there are pedestrians P around), the server will issue an early warning to car A.

[0238] The server may send warning information through a communication device with car A (such as the vehicle-mounted system of car A). The warning information may include content such as "There is a risk of traffic accidents when your vehicle is approaching an intersection. Please drive carefully."

[0239] At the same time, for other relevant traffic participants, such as car B and pedestrian P near car A, the server can also send early warning information through corresponding devices (such as car B's on-board system or pedestrian P's smartphone application). For example, it can send a message to car B saying "There is a risk of traffic accident with car A near you, please keep a safe distance" and send a message to pedestrian P saying "The driving status of car A in your area is risky, please pass with caution."

[0240] After sending the warning information, the server continues to monitor the entire dynamic traffic participation process, continuously obtains the dynamic traffic flow data of each traffic participant, updates the traffic spatiotemporal knowledge network data, and other operations to ensure the normal operation of traffic.

[0241] Next, it is assumed that there is a traffic clustering strategy in this traffic scene, for example, clustering is performed according to road areas or traffic flow directions, etc. The server obtains the traffic clustering strategy based on which this dynamic traffic participation process is based.

[0242] Assume that car A, car B, and car C are classified into one traffic cluster (the first traffic cluster) because they travel on the same main road and have similar traffic flow directions.

[0243] The server obtains the first traffic cluster (including car A, car B and car C) where the subject involved in the traffic accident (such as car A) is located under this traffic cluster strategy.

[0244] Then, the traffic linkage request corresponding to the first traffic cluster in the dynamic traffic participation process is obtained. For example, the traffic management center may issue a traffic linkage request, requiring the cluster to adjust the speed or change lanes when approaching a congested area to relieve traffic pressure.

[0245] The server predicts traffic accidents for the first traffic cluster in the process of responding to the traffic linkage request. For example, when car A responds to the traffic linkage request (such as slowing down and changing lanes), the server analyzes the interaction between car A and car B and car C.

[0246] If car A slows down and changes lanes, but car B does not receive the information in time or does not make corresponding adjustments (such as not maintaining a safe distance or not slowing down appropriately), this may increase the risk of a traffic accident.

[0247] The server analyzes changes in speed-related features of cars A, B and C in the process of responding to traffic linkage requests (such as whether the speed of car B is appropriate when car A slows down), changes in driving direction-related features (such as whether the driving direction of car C needs to be adjusted when car A changes lanes), changes in position-related features (such as whether the distance between cars A, B and C remains within a safe range), and changes in features related to interactions between traffic participants (such as whether the avoidance and coordination actions between cars A, B and C are reasonable).

[0248] Through such analysis, the second traffic accident risk data of this dynamic traffic participation process is determined. For example, if the analysis shows that there is a high risk of collision between car A and car B when responding to the traffic linkage request, then the information related to this risk (such as the vehicles involved, location, time, risk type, etc.) will be included in the second traffic accident risk data.

[0249] Figure 2 In an embodiment of the present application, a spatiotemporal data feature mining system 100 for smart transportation is provided, including a processor 1001 and a memory 1003 and an application code stored in the memory 1003. The processor 1001 executes the above application code to implement the steps of a sensor data processing method applied to smart photovoltaics.

[0250] Figure 2The spatiotemporal data feature mining system 100 for smart transportation shown includes: a processor 1001 and a memory 1003. The processor 1001 and the memory 1003 are connected, such as through a bus 1002. Optionally, the spatiotemporal data feature mining system 100 for smart transportation may also include a transceiver 1004, which may be used for data interaction between the server and other servers, such as data transmission and / or data reception. It should be noted that the transceiver 1004 is not limited to one in actual scheduling, and the structure of the spatiotemporal data feature mining system 100 for smart transportation does not constitute a limitation on the embodiments of the present application.

[0251] Processor 1001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array) or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. It may implement or execute various exemplary logic blocks, modules and circuits described in conjunction with the disclosure of this application. Processor 1001 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0252] The bus 1002 may include a path to transmit information between the above components. The bus 1002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. The bus 1002 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 2 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0253] The memory 1003 may be a ROM (Read Only Memory) or other types of static storage devices that can store static information and instructions, a RAM (Random Access Memory) or other types of dynamic storage devices that can store information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory) or other optical disk storage, optical disk storage (including compressed optical disk, laser disk, optical disk, digital versatile disk, Blu-ray disk, etc.), magnetic disk storage medium, other magnetic storage devices, or any other medium that can be used to carry or store application code and can be read by a computer, without limitation herein.

[0254] The memory 1003 is used to store application code for executing the embodiment of the present application, and the execution is controlled by the processor 1001. The processor 1001 is used to execute the application code stored in the memory 1003 to implement the steps shown in the above method embodiment.

[0255] The server includes but is not limited to: mobile terminals such as mobile phones, notebook computers, PADs, etc. and fixed terminals such as digital TVs, desktop computers, etc.

[0256] An embodiment of the present application provides a computer-readable storage medium on which application code is stored. When the application code is executed by a processor, the steps and corresponding contents of the aforementioned method embodiment can be implemented.

[0257] It should be understood that, although each operation step is indicated by arrows in the flowchart of the embodiment of the present application, the implementation order of these steps is not limited to the order indicated by the arrows. Unless there is clear explanation in this article, in some implementation scenarios of the embodiment of the present application, the implementation steps in each flowchart can be performed in other orders according to demand. In addition, some or all of the steps in each flowchart may include multiple sub-steps or multiple stages according to actual implementation scenarios, and some or all of these sub-steps or stages may be executed at the same time, and each sub-step or stage in these sub-steps or stages may also be executed at different times respectively. In different scenarios at the execution time, the execution order of these sub-steps or stages can be flexibly configured based on demand, and the embodiment of the present application does not limit this.

[0258] The above is only an optional implementation method for some implementation scenarios of the present application. It should be pointed out that for ordinary technicians in this technical field, without departing from the technical concept of the scheme of the present application, other similar implementation methods based on the technical ideas of the present application are also within the protection scope of the embodiments of the present application.

Claims

1. A spatiotemporal data feature mining method for smart transportation, characterized in that: The method includes: Acquire dynamic traffic flow data generated by X traffic participants in the dynamic traffic participation process, configure X space-time vertices with the X traffic participants, configure space-time pointing links between the X space-time vertices with the dynamic traffic flow data, and generate traffic space-time knowledge network data, wherein the space-time vertex refers to a state point used to represent a traffic participant at a specific time and space position in the traffic space-time knowledge network; Marking the spatiotemporal vertices in the traffic spatiotemporal knowledge network data that have interactive relationships with other traffic participants as the first spatiotemporal vertices; Extract the target leading participating subjects corresponding to each of the X space-time vertices from the traffic space-time knowledge network data, and mark the traffic participating subject corresponding to the space-time vertex with the largest participation frequency among the target leading participating subjects corresponding to each of the X space-time vertices as the key participating subject among the X traffic participating subjects; the target leading participating subject corresponding to each of the space-time vertices is used to represent the participating subjects covered in the traffic participation trajectory from the first space-time vertex to the space-time vertex; Performing feature mining on the subject movement state trajectory data associated with the key participating subject to determine the first traffic accident risk data in the dynamic traffic participation process; The step of extracting the target leading participating entities corresponding to each of the X spatiotemporal vertices from the traffic spatiotemporal knowledge network data includes: marking the first space-time vertex as the target leading participant of the first space-time vertex; Obtaining a first preceding participant of a second space-time vertex, marking a second leading participant of the first preceding participant of the second space-time vertex and the second space-time vertex as target leading participants of the second space-time vertex; the second space-time vertex is a space-time vertex other than the first space-time vertex among the X space-time vertices; Wherein, the method further comprises: Extracting a second preceding participant of the first preceding participant from the traffic spatiotemporal knowledge network data, and extracting a first traffic participation trajectory including the first preceding participant and the second preceding participant from the traffic spatiotemporal knowledge network data; A second traffic participation trajectory is extracted from the first traffic participation trajectory, and the second preceding participant corresponding to the second traffic participation trajectory is marked as the sub-dominant participant of the first preceding participant; the second traffic participation trajectory is used to indicate that the first preceding participant and the second preceding participant have a participation interaction relationship, and the first preceding participant does not have a participation trajectory of any preceding participant other than the second preceding participant.

2. The method for mining spatiotemporal data features for smart transportation according to claim 1 is characterized in that: The spatiotemporal vertices in the traffic spatiotemporal knowledge network data that have interactive relationships with other traffic participants are marked as first spatiotemporal vertices, including: In the traffic spatiotemporal knowledge network data, the X spatiotemporal vertices are respectively used as source nodes to track the traffic participation trajectory of each spatiotemporal vertex in the traffic spatiotemporal knowledge network data; If there is a traffic participation trajectory including the X space-time vertices in the tracked traffic participation trajectory, the corresponding space-time vertex is marked as the first space-time vertex; If there is no traffic participation trajectory including the X space-time vertices in the tracked traffic participation trajectory, a simulation participating subject is configured in the traffic space-time knowledge network data, and a simulation space-time pointing link between the simulation participating subject and the traffic space-time knowledge network data is configured, and the simulation participating subject is marked as the first space-time vertex.

3. The spatiotemporal data feature mining method for smart transportation according to claim 1 is characterized in that: The spatiotemporal vertices in the traffic spatiotemporal knowledge network data that have interactive relationships with other traffic participants are marked as first spatiotemporal vertices, including: Extracting the ath space-time vertex from the traffic space-time knowledge network data, taking the ath space-time vertex as the source node, performing walk sorting on the traffic space-time knowledge network data to generate the ath basic traffic trajectory network; a is not greater than X; the root participant of the ath basic traffic trajectory network is the ath space-time vertex; If there is a space-time vertex among the X space-time vertices that does not belong to the a-th basic traffic trajectory network, then the a+1-th space-time vertex is used as the source node, and the traffic space-time knowledge network data is sorted to generate the a+1-th basic traffic trajectory network; If the a-th basic traffic trajectory network includes the X space-time vertices, the a-th space-time vertex is marked as the first space-time vertex.

4. The spatiotemporal data feature mining method for smart transportation according to claim 1 is characterized in that: The step of extracting the target leading participating entities corresponding to each of the X spatiotemporal vertices from the traffic spatiotemporal knowledge network data includes: Extracting Y traffic participation trajectories e from the first space-time vertex to the bth space-time vertex from the traffic space-time knowledge network data, and obtaining the interactive participating subjects respectively covered by the Y traffic participation trajectories e; b is not greater than X; Among the interactive participating subjects respectively covered by the Y traffic participating trajectories e, the interactive participating subjects existing in the Y traffic participating trajectories e are marked as the target leading participating subjects corresponding to the b-th space-time vertex.

5. The method for mining spatiotemporal data features for smart transportation according to claim 1, characterized in that: The method further comprises: Marking the X space-time vertices as the basic leading participating entities of the second space-time vertex; The obtaining of the first preceding participant of the second space-time vertex and marking the second leading participant of the first preceding participant of the second space-time vertex and the second space-time vertex as target leading participants of the second space-time vertex include: In the c-th tracking and extraction process of the leading participating subject, Z first preceding participating subjects of the second spatiotemporal vertex are extracted from the traffic spatiotemporal knowledge network data; c is a positive integer; Z is a positive integer; Obtain the sub-dominant participating entities corresponding to the Z first preceding participating entities of the second space-time vertex, and mark the shared participating entity of the sub-dominant participating entity corresponding to the Z first preceding participating entities and the second space-time vertex as the cth adjusted dominant participating entity of the second space-time vertex; If the cth adjustment leading participant of the second space-time vertex is consistent with the c-1th adjustment leading participant of the second space-time vertex, the cth adjustment leading participant of the second space-time vertex is marked as the target leading participant of the second space-time vertex; when c is the default value, the c-1th adjustment leading participant of the second space-time vertex is the basic leading participant of the second space-time vertex; The step of extracting the Z first preceding participating entities of the second spatiotemporal vertex from the traffic spatiotemporal knowledge network data during the tracking and extraction process of the cth leading participating entity comprises: The traffic spatiotemporal knowledge network data is stored in a set data structure, wherein the spatiotemporal vertices and the related information of the spatiotemporal pointing links are arranged in order, an index is established for each spatiotemporal vertex with the position of the spatiotemporal vertex in the traffic network as the index key, and an index is established for the traffic participant type information; wherein the position includes geographic coordinates or a logical position identifier; A search range parameter is set for limiting the time and space range considered when searching for a first participant, the search range parameter is determined based on the distance and time interval factors of the mutual influence between traffic participants in historical data, and an empty temporary storage structure is initialized at the same time, the empty temporary storage structure is an array or a linked list, and is used to store potential first first participants found during the search process; Searching for a corresponding space-time vertex in the traffic space-time knowledge network data according to the identifier of the second space-time vertex, obtaining the position information and related time information of the space-time vertex in the traffic network, traversing the space-time vertices in the traffic space-time knowledge network data based on the determined search range, marking the space-time vertices in the search range as potential first prior participating entities by comparing the position relationship and time relationship between the space-time vertices during the traversal process, and adding the potential first prior participating entity to a temporary storage structure; For each potential first prior participant in the temporary storage structure, the type of traffic participant represented by the potential first prior participant is obtained, and the type of traffic participant represented by the second space-time vertex is determined, and the judgment is made according to the predefined traffic participant type association rule. When the second space-time vertex is a vehicle, if the potential first prior participant is the type of the traffic participant of the preceding vehicle in the same lane or the traffic light at the intersection that has a leading role in guiding the driving of vehicles, then the potential first prior participant is retained; when the second space-time vertex is a pedestrian, if the potential first prior participant is the traffic coordinator who guides pedestrians to cross the road or the type of the vehicle in front of the pedestrian path and affecting the walking of pedestrians, then the potential first prior participant is retained, and other potential first prior participants that do not meet the type association rule are removed from the temporary storage structure; For the remaining potential first prior participants in the temporary storage structure, analyze the traffic flow direction related information of the first prior participants, obtain the traffic flow direction information of the second space-time vertex, determine whether the flow direction and logical relationship between the potential first prior participants and the second space-time vertex meet the requirements of the prior participants according to the traffic rules and the logical relationship in the traffic scene, and remove the potential first prior participants that do not meet the requirements of the prior participants from the temporary storage structure; Count the target number of potential first prior participants remaining in the temporary storage structure; if the target number is less than or equal to the set number, directly use all elements in the temporary storage structure as the final first prior participants; if the target number is greater than the set number, sort the elements in the temporary storage structure according to a set sorting rule, and select the first Z potential first prior participants as the final result according to the sorting result, wherein the set sorting rule is based on the spatiotemporal proximity between a space-time vertex and a second space-time vertex, or based on the potential influence of the traffic participant on the second space-time vertex.

6. The method for mining spatiotemporal data features for smart transportation according to claim 1, characterized in that: The second space-time vertices include Q vertices; the method further includes: Taking the first space-time vertex as a source node, performing reverse tracking and polling on the traffic space-time knowledge network data to generate a set of participating entities corresponding to the X space-time vertices; The obtaining of the first preceding participant of the second space-time vertex and marking the second leading participant of the first preceding participant of the second space-time vertex and the second space-time vertex as target leading participants of the second space-time vertex include: Extract the first preceding participating subject of the d-th second space-time vertex from the participating subject set, and mark the second leading participating subject of the first preceding participating subject of the d-th second space-time vertex and the d-th second space-time vertex as the target leading participating subject of the d-th second space-time vertex; d is not greater than Q; If the d-th second space-time vertex is not the last space-time vertex in the set of participating entities, extracting the d+1-th second space-time vertex from the set of participating entities, and obtaining the target leading participating entity of the d+1-th second space-time vertex; If the d-th second space-time vertex is the last space-time vertex in the set of participating entities, then the step of marking the traffic participating entity corresponding to the space-time vertex with the highest participation frequency among the target dominant participating entities corresponding to each of the X space-time vertices as a key participating entity among the X traffic participating entities is entered.

7. The method for mining spatiotemporal data features for smart transportation according to claim 1, characterized in that: The first traffic accident risk data includes traffic accident participants; the feature mining of the subject movement state trajectory data associated with the key participants to determine the first traffic accident risk data in the dynamic traffic participation process includes: Extracting the subject movement state trajectory data associated with the key participating subject; the subject movement state trajectory data includes the movement state records corresponding to each of the X traffic participating subjects; Using the movement state records corresponding to the X traffic participants, respectively, extracting the driving characteristic change data corresponding to the X traffic participants; Traffic participants whose driving characteristic change data meet the characteristic conditions of each candidate traffic accident template are marked as traffic accident participants.

8. A spatiotemporal data feature mining system for smart transportation, characterized in that: The spatiotemporal data feature mining system for smart transportation includes a processor and a machine-readable storage medium, in which machine-executable instructions are stored. The machine-executable instructions are loaded and executed by the processor to implement the spatiotemporal data feature mining method for smart transportation according to any one of claims 1 to 7.

Images