Green wave road section traffic event perception and control system based on multi-source data fusion

Abstract

The application provides a green wave road section traffic event perception and control system based on multi-source data fusion, and belongs to the technical field of intelligent transportation systems. The system comprises: a multi-source perception and collection module, which is used for collecting multi-source traffic monitoring data from multiple traffic monitoring devices and extracting structured vehicle behavior records; an event identification module, which is used for performing multi-angle identification record operations on the vehicle behavior records to identify traffic events and form traffic event records; and a signal linkage control module, which is used for automatically adjusting signal timing parameters of green wave intersections according to the traffic event records. Through the complementation and convergence of multi-source data, the event identification can obtain more comprehensive information from the observation of multiple devices, and through the linkage of event perception and signal control, the response speed and traffic management efficiency of the green wave road section sudden event are improved.

Description

Technical Field

[0001] This application relates to the field of intelligent transportation systems technology, specifically providing a green wave road segment traffic event perception and control system based on multi-source data fusion. Background Technology

[0002] Intelligent traffic signal control systems manage and optimize traffic signals in a spatiotemporal manner, creating continuous "green waves" on main roads. This effectively increases average vehicle speed and the number of vehicles passing through per unit time, thus alleviating the road resource shortage caused by the continuous growth of motor vehicle ownership. However, the initial design of such systems mainly focuses on "efficiency optimization," emphasizing macro-level improvements in average vehicle speed and traffic capacity. They lack the ability to deeply perceive and finely manage abnormal traffic events that may occur in green wave sections, such as sudden braking, rear-end collisions, prolonged vehicle congestion, and illegal lane changes. Once a sudden event occurs in a green wave section, it often disrupts the original signal coordination rhythm, causing the entire green wave to malfunction and even triggering wider congestion and safety risks.

[0003] With the widespread application of video surveillance, vehicle-to-infrastructure (V2I) technology, and artificial intelligence algorithms in the transportation sector, many cities have already constructed high-definition video surveillance systems covering major road intersections and various types of roadside sensing devices (such as cameras, millimeter-wave radar, and geomagnetic sensors). Utilizing algorithms for video target detection, behavior analysis, and trajectory mining, some traffic incident perception systems are already able to identify typical traffic incidents such as sudden deceleration, rear-end collisions, and illegal lane changes. However, from a technical architecture perspective, traditional traffic incident perception systems are mostly still designed with a "platform-based, closed" approach. Their event confirmation methods primarily rely on the recognition results of a single device, making them prone to false alarms or missed alarms due to equipment failure, obstruction, changes in lighting, etc. They also lack multi-source data cross-validation mechanisms, resulting in insufficient reliability and robustness in event perception.

[0004] Meanwhile, the outputs of existing event perception systems are mostly scattered alarm records or alarm pop-ups, lacking structured organization of elements involved in the event such as vehicle identity, behavior trajectory, lane position, and spatiotemporal range. Furthermore, there is a lack of deep coupling between the perception results and the signal control system. Traditional green wave control systems mainly operate based on historical traffic flow and fixed timing schemes. Even if a sudden event occurs on a green wave section, the signal controller still operates according to the existing timing, unable to dynamically adjust the timing strategy based on event information.

[0005] In summary, existing technologies lack a systematic solution that can unify and integrate multi-source traffic event perception, multi-view reliable confirmation, and signal linkage control in green wave road segment scenarios. Summary of the Invention

[0006] The purpose of this application is to provide a traffic event perception and control system for green wave road sections based on multi-source data fusion.

[0007] The embodiments of this application can be implemented through the following technical solutions: A traffic event perception and control system for green wave road sections based on multi-source data fusion includes: a multi-source perception and acquisition module, used to acquire multi-source traffic monitoring data from multiple traffic monitoring devices deployed on green wave road sections and extract multiple structured vehicle behavior records from them; an event recognition module, used to perform multi-angle recognition and recording operations on at least one of the vehicle behavior records to identify the traffic event type and form a traffic event record; the traffic event record includes the identified traffic event type, the vehicle involved, and vehicle behavior record information associated with the vehicle involved; and a signal linkage control module, used to automatically adjust the signal timing parameters of the green wave intersection according to the information in the traffic event record.

[0008] Furthermore, each of the structured vehicle behavior records includes vehicle identity information and corresponding vehicle behavior information; the vehicle behavior information corresponding to any specific vehicle is the vehicle's own movement and status information or the vehicle's driving operation information observed from traffic monitoring data at a specific moment or within a specific time window.

[0009] Furthermore, the traffic incidents include Class I traffic incidents, Class II traffic incidents, and Class III traffic incidents, wherein: The first category of traffic incidents refers to traffic incidents that involve only the behavior of a single vehicle; The second category of traffic incidents involves an interaction between a single vehicle and road infrastructure that triggers traffic rules; The third category of traffic incidents involves the interaction of two or more vehicles.

[0010] Furthermore, the multi-angle recognition and recording operation includes the following steps: Step S1: Map the vehicle behavior records extracted by the multi-source perception acquisition module to a unified spatiotemporal coordinate system; Step S2: Based on the vehicle behavior record of a single vehicle, determine whether it meets the first type of condition. If it does, determine that a first type of traffic event has occurred and generate a first type of traffic event record. Step S3: Based on the vehicle behavior record of a single vehicle and the traffic facility status information within the same spatiotemporal window, determine whether it meets the second type of condition. If it does, determine that a second type of traffic event has occurred and generate a second type of traffic event record. Step S4: Based on the vehicle behavior records of multiple vehicles within the same spatiotemporal window, determine whether the third type of conditions are met among the multiple vehicles. If the conditions are met, determine that a third type of traffic event has occurred and generate a third type of traffic event record.

[0011] Furthermore, the same spatiotemporal window refers to the time and space deviation between vehicles, or between a vehicle and traffic facilities that can affect the vehicle, not exceeding a preset threshold.

[0012] Furthermore, the system also includes a green wave window prediction module, used to acquire signal timing parameters and road geometric parameters of green wave road segments, calculate one or more green wave spatiotemporal windows, and mark the green wave spatiotemporal windows as key sensing areas; the multi-source sensing acquisition module collects data within the key sensing areas at a higher sampling frequency than outside the key sensing areas.

[0013] Furthermore, the green wave spatiotemporal window is calculated in the following manner: For the intersection Given the signal period Phase difference and the recommended speed for this section of road. Then in the first Within a given period, assuming the convoy is at the upstream intersection at time... Once permitted, the estimated time interval for vehicles to reach the downstream intersection can be expressed as: ， in, For the length of the road segment, To account for the speed range after fluctuations in driving behavior, the corresponding spatial range can be simplified to: This creates a green wave spatiotemporal window. .

[0014] Furthermore, each traffic incident record stores information on the behavior of the vehicles involved that is related to the traffic incident in terms of time, space, and behavior, so as to form a directed edge between the traffic incident record and the behavior records of the vehicles involved.

[0015] Furthermore, the traffic incident record also stores the status record information of the traffic facilities involved that are related to the traffic incident in terms of time, space and behavior, so as to form a directed edge between the traffic incident record and the status record of the traffic facilities involved.

[0016] Furthermore, the traffic incident record also stores information on preceding traffic incident records that have a causal relationship with the traffic incident, so as to form a directed edge between the traffic incident record and its preceding traffic incident records.

[0017] The embodiments of this application provide a traffic event perception and control system for green wave road sections based on multi-source data fusion, which has at least the following beneficial effects: This application acquires data from multiple traffic monitoring devices through a multi-source sensing acquisition module, enabling the event recognition module to identify different types of traffic events based on the complementarity and convergence of multi-source data. This avoids the problem that the recognition results of a single device cannot independently confirm the event under conditions of obstruction or failure, and improves the reliability and robustness of event perception.

[0018] This application uses directed edges formed from the behavior records of the vehicles involved in the traffic incident stored in the traffic incident logs. This organizes the perception results into a directed graph data structure, transforming them from fragmented data into structured data with clear relationships. Starting from any identified traffic incident record, one can trace back along the directed edges to all related behavior records of the vehicles involved, achieving complete backtracking of perception elements.

[0019] This application integrates event perception with signal linkage control, automatically triggering signal timing adjustment after event confirmation, realizing a rapid closed loop from "event perception" to "control response", significantly improving the adaptability and recovery capability of green wave sections in emergency scenarios.

[0020] This application utilizes green wave timing parameters to calculate the spatiotemporal window of the green wave, and dynamically increases the data acquisition frequency of the sensing device within the window. Without significantly increasing hardware investment, it enhances the event capture capability of key spatiotemporal areas of the green wave road segment, achieving a balance between sensing accuracy and computing resources. Attached Figure Description

[0021] Figure 1 A structural block diagram of a green wave road segment traffic event perception and control system based on multi-source data fusion provided in an embodiment of this application; Figure 2 A flowchart illustrating the multi-angle recognition and recording operation of traffic incidents provided in this application embodiment; Figure 3 This is a schematic diagram of directed association relationships centered on traffic incident records, provided as an embodiment of this application. Figure 4 This is a schematic diagram of the road network, sensing devices, and event locations in the simulation experiment of this application. Figure 5 This is a schematic diagram illustrating the spatiotemporal relationship between the green wave window and the vehicle trajectory in the simulation experiment of this application. Figure 6 A bar chart comparing the event recognition performance of different sensing methods in the simulation experiment of this application; Figure 7 This is a comparison curve of the evolution of road segment speed and queue length before and after the simulation experiment of this application; Figure 8A bar chart showing the comprehensive comparison of the operating effects of the four control schemes in the simulation experiment of this application; Figure 9 This is a heat map showing the sensitivity of the linkage control parameters in the simulation experiment of this application. Detailed Implementation

[0022] The present application will now be further described based on preferred embodiments and with reference to the accompanying drawings.

[0023] Furthermore, for ease of understanding, various components on the drawings have been enlarged or reduced, but this is not intended to limit the scope of protection of this application.

[0024] Singular forms of words also include plural meanings, and vice versa.

[0025] In the description of the embodiments of this application, it should be noted that if terms such as "upper," "lower," "inner," or "outer" are used to indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship commonly used when the product of this application is in use, they are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation on this application. In addition, in the description of this application, in order to distinguish different units, the terms "first," "second," etc. are used in this specification, but these are not limited by the manufacturing order, nor should they be construed as indicating or implying relative importance. Their names may differ in the detailed description and claims of this application.

[0026] The vocabulary used in this specification is for illustrative purposes and is not intended to limit the scope of this application. It should also be noted that, unless otherwise expressly specified and limited, the terms "set," "connected," and "linked" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection, a direct connection, or an indirect connection via an intermediate medium; or they can refer to the internal communication between two components. Those skilled in the art will understand the specific meaning of these terms in this application.

[0027] like Figure 1 As shown, this application provides a traffic event perception and control system for green wave road sections based on multi-source data fusion, which mainly includes the following modules: The multi-source sensing acquisition module is used to collect multi-source sensing data from multiple traffic monitoring devices deployed on green wave road sections, and extract multiple structured vehicle behavior records from the multi-source sensing data.

[0028] The event recognition module is used to perform multi-angle recognition and recording operations on at least one of the vehicle behavior records to identify the traffic event type and form a traffic event record; the traffic event record includes the identified traffic event type, the vehicle involved, and vehicle behavior record information associated with the vehicle involved.

[0029] The signal linkage control module is used to automatically adjust the intersection signal timing parameters of the green wave road segment based on the information in the traffic event record.

[0030] Preferably, it also includes a green wave window prediction module, used to acquire signal timing parameters and road geometric parameters of the green wave road segment, calculate one or more green wave spatiotemporal windows, and mark the green wave spatiotemporal windows as key sensing areas; the multi-source sensing acquisition module collects data within the key sensing area at a higher sampling frequency than outside the key sensing area.

[0031] Multi-source sensing and acquisition module In the embodiments of this application, traffic monitoring devices refer to various monitoring equipment installed at different locations on traffic roads (e.g., on both sides of the traffic road, above the traffic road, and at intersections) to monitor vehicles traveling on the road and the resulting traffic conditions. In the embodiments of this application, the raw data directly obtained by these traffic monitoring devices through real-time monitoring of their designated monitoring areas is called traffic monitoring data. The following are exemplary lists of usable traffic monitoring devices and the traffic monitoring data obtained by these traffic monitoring devices: High-definition cameras are typically installed above roads and at traffic intersections to continuously capture images of designated areas. The monitoring data they acquire is a high-definition video data stream with a set frame rate. Generally, the acquired high-definition video data stream can be divided according to preset time intervals or data sizes to form video files that are arranged continuously in time. These video files record the dynamic information of vehicles, traffic facilities, and traffic environment (such as traffic lights, pedestrians on zebra crossings, etc.) that appear in the designated area. Speed ​​cameras with camera functions are usually deployed on road sections where vehicle speed needs to be controlled (such as accident-prone areas, near schools, etc.). They can simultaneously acquire the instantaneous speed of vehicles and the corresponding captured images or short videos to form a continuous speed record in time. Image / video capture equipment: usually set up at intersections and near schools, hospitals, etc., to capture behaviors such as running red lights and failing to yield to pedestrians. Under specific triggering conditions (such as a vehicle crossing the stop line during a red light), it captures high-resolution images or videos to record vehicle license plates, vehicle models, colors, and the status of traffic facilities and traffic environment at the time or period when the specific conditions are triggered. Cameras with night vision capabilities are suitable for nighttime or low-light environments. They acquire clear video or image data through infrared illumination or starlight-level photosensitive elements, ensuring all-weather monitoring capabilities.

[0032] Radar (millimeter wave / LiDAR) is used to detect the position, speed, distance to other vehicles, and trajectory of targets. Its output is generally cloud data or information such as target distance, vehicle speed, and trajectory.

[0033] Inductive loop / geomagnetic sensor is buried under the road surface. It detects changes in magnetic field or pressure to sense the time, speed and lane occupancy of vehicles, and outputs pulse signals or switching data.

[0034] A panoramic stitching camera is composed of multiple lenses and can cover a wide range of intersections or long road sections, providing a global monitoring image without blind spots. It is convenient to track the continuous driving trajectory of vehicles in multiple lanes or complex intersections, and its output is a panoramic video stream.

[0035] Understandably, the traffic monitoring data acquired by these devices all carries precise timestamps or time frame information to ensure that subsequent multi-source data can be aligned and correlated in time. For example, for continuously acquired video streams, each frame is accompanied by a corresponding timestamp or frame number; for conditionally triggered data, such as images captured by image capture devices, or pulse signals indicating lane occupancy acquired by inductive loop / geomagnetic sensors, the trigger time information is also included.

[0036] From the traffic monitoring data obtained from the various types of traffic monitoring devices mentioned above, the vehicle identity information and corresponding vehicle behavior information of each vehicle appearing within the monitoring range of each device can be identified.

[0037] In the embodiments of this application, vehicle identity information refers to information that can be used to uniquely identify each vehicle appearing in traffic monitoring data. Generally, vehicle identity information is the vehicle's license plate number. In addition, in some embodiments, vehicle identity information may also include auxiliary information such as the vehicle brand, model, and color of each vehicle appearing in traffic monitoring data. This auxiliary information can be compared with the model, color, and other information pre-registered for the license plate number (generally, this registration information can come from basic data provided by the vehicle management office or a third-party vehicle management system) to help determine the vehicle identity information and promptly detect whether the vehicle has been cloned or its registration paint has been altered.

[0038] In the embodiments of this application, the vehicle behavior information corresponding to any specific vehicle (i.e., a vehicle that can be uniquely identified by its vehicle identity information) refers to the information related only to the vehicle's own operating status or operation, which is observed from traffic monitoring data at a specific moment or within a specific time window. In some embodiments, the vehicle behavior information corresponding to any specific vehicle includes one or more of the following: The vehicle's own motion and status information includes its spatial position coordinates (latitude and longitude or relative road network coordinates), lane, speed (instantaneous speed / average speed), acceleration (acceleration / braking), direction of travel (heading angle), steering angular velocity, lights (turn signal / brake light status / hazard light status), and the vehicle's spatial position coordinates (i.e., vehicle trajectory information) in one or more consecutive time series. The vehicle's driving operation information includes perceptible driving behaviors such as not wearing a seatbelt or making / receiving a phone call during driving, as well as information that can be obtained in scenarios with vehicle-to-everything (V2X) or vehicle data access, such as gear operation information, pedal (accelerator / electric / brake / clutch) pressure information, and steering wheel operation information (rotation angle, angular velocity).

[0039] In some optional embodiments, the different types of raw traffic monitoring data acquired in real time by the various traffic monitoring devices described above can be stored in local storage devices, such as flash memory, solid-state drives, and mechanical hard drives built into high-definition cameras and image capture devices. The acquired raw traffic monitoring data can be sent to a trusted storage system (such as distributed object storage, IPFS, or a centralized evidence database) for backup via wired (such as fiber optic cable) or wireless (such as 5G) communication methods.

[0040] Obviously, for any specific vehicle, its vehicle behavior information is the state or action of the vehicle in a specific time and space. Since various traffic monitoring devices may collect or measure the state or action of the same vehicle from different positions and different perspectives at the same time, multiple traffic monitoring devices may independently identify the same vehicle behavior of the same vehicle from multiple sources of raw traffic monitoring data, thereby generating multiple vehicle behavior records corresponding to that behavior.

[0041] In the embodiments of this application, the operation of identifying the identity of each vehicle and its corresponding vehicle behavior from the raw traffic monitoring data obtained from each traffic monitoring device can be implemented by an edge computing unit set in each traffic monitoring device. Specifically, the edge computing unit can be a computer device including a processor, a readable storage medium and a data transmission module. The processor can be a microprocessor such as a single-chip microcomputer or a standard processor such as a CPU. The readable storage medium can be a storage device such as Flash or a solid-state drive, or a storage medium shared with the traffic monitoring data source. It stores an executable program for identifying vehicle identity information and vehicle behavior information. The processor calls the above-mentioned executable program to identify vehicle identity information and vehicle behavior information from the traffic monitoring data obtained from the traffic monitoring data source in real time.

[0042] In the embodiments of this application, no specific limitation is made to the algorithms for vehicle identification and vehicle behavior recognition. Those skilled in the art can construct executable programs based on the various publicly disclosed algorithms for recognizing target features and behaviors of different types of data such as images / videos / point clouds.

[0043] In the embodiments of this application, each edge computing unit performs real-time analysis on the raw traffic monitoring data (such as video streams and radar data) acquired by the traffic monitoring device it is located in, and outputs a structured vehicle behavior record arranged in chronological order. The record can be organized in the form of key-value pairs or key-value sets. Each vehicle behavior record contains at least the following fields: global ID, time information (such as the timestamp of the time when the behavior occurs, or the timestamp of the start and end times of the behavior), vehicle identity information, the vehicle's own motion and state information (position, speed, acceleration, heading, lane, driving trajectory), and driving operation related information (such as not wearing a seat belt, using a mobile phone while driving, gear position, accelerator / brake pedal opening, steering wheel angle), etc.

[0044] The following is a schematic illustration of a specific vehicle behavior record in JSON format, where some data contains specific characters beginning with "". The quotation mark (”) is used to represent: json { "behavior_record_id": " beh_550e8400-e29b-41d4-a716-44665544a3f2", "timestamp": 1736926337200, "vehicle_id": " K12345", "vehicle_type": "small car", "color": "white", "motion_state": { "position": { "x": 1099.4, "y": 636.9, "coordinate_system": "Road network relative coordinates" }, "lane": 2, "speed": 30.0, "acceleration": -5.6, "brake": true, "heading": 32.0, "angular_velocity": 0.2 }, "signal_state": { "turn_signal": "off", "brake_light": true, "hazard_light": false }, "trajectory": [ { "t": 1736926336200, "x": 1090.0, "y": 631.0}, { "t": 1736926336300, "x": 1091.2, "y": 631.7}, { "t": 1736926336400, "x": 1092.3, "y": 632.4}, { "t": 1736926336500, "x": 1093.3, "y": 633.1}, { "t": 1736926336600, "x": 1094.3, "y": 633.7}, { "t": 1736926336700, "x": 1095.3, "y": 634.3}, { "t": 1736926336800, "x": 1096.2, "y": 634.9}, { "t": 1736926336900, "x": 1097.1, "y": 635.4}, { "t": 1736926337000, "x": 1097.9, "y": 635.9}, { "t": 1736926337100, "x": 1098.7, "y": 636.4} ], "driving_behavior": { "seatbelt_fastened": true, "using_phone": false }, "v2x_info": { "gear_position": "D", "accelerator_pedal": 0.2, "brake_pedal": 0.6, "steering_wheel_angle": -15.0 } } See Table 1 for explanations of the fields involved in the above vehicle behavior records.

[0045] Table 1. Description of Vehicle Behavior Record Fields Clearly, the structured vehicle behavior records generated by each edge computing unit depict the vehicle behavior it monitors from the perspective of the traffic monitoring device it is located in. The vehicle behavior records corresponding to multiple different traffic monitoring devices then depict the behavior of the same vehicle in the same time and space from different perspectives or positions, using different behavior description methods (such as speedometers mainly describing vehicle speed, acceleration, and other information, while cameras can acquire vehicle trajectory, in-vehicle operations, and other information).

[0046] <Event Recognition Module> As mentioned earlier, the structured vehicle behavior records output by each edge computing unit depict vehicle behavior from the perspective of the traffic monitoring device to which it is located. However, the same vehicle may be observed by multiple devices simultaneously, and the behavior records output by different devices differ in time reference and spatial coordinate system. Each record only reflects the observation results from a single perspective.

[0047] Therefore, this module takes the vehicle behavior records output by the multi-source perception and acquisition module as input, first establishes a unified time axis and spatial coordinate system, then identifies different traffic event types through multi-angle recognition and recording operations, and finally generates traffic event records based on the recognition results.

[0048] In the embodiments of this application, a traffic event refers to an event that occurs in a road traffic environment and involves at least one vehicle. The specific types of traffic events include: The first category of traffic incidents specifically refers to traffic incidents involving only the behavior of a single vehicle, such as sudden acceleration, sudden deceleration, or sudden turning of the vehicle itself. In addition, the aforementioned behaviors such as not wearing a seat belt or making / receiving a mobile phone also fall under the category of traffic incidents involving only a specific vehicle. The second category of traffic incidents specifically refers to incidents involving the interaction between a single vehicle and road facilities that trigger traffic rules, such as running a red light, not driving in the designated lane, changing lanes without using turn signals, occupying the overtaking lane for an extended period of time, and frequent lane changes (without affecting other vehicles). The third category of traffic incidents specifically refers to incidents involving the interaction of two or more vehicles, such as rear-end collisions, minor collisions, and illegal lane changes (such as changing lanes without using turn signals and affecting the normal driving of other vehicles).

[0049] As can be seen from the above description, in this application, vehicle behavior and traffic events are fundamentally different. Vehicle behavior refers to the movement and state or driving operation that can be directly observed by each individual vehicle; while traffic events refer to the result caused by the behavior of an individual vehicle and its interaction with the state of traffic facilities (representing the corresponding traffic rules), or by the interaction between the related behaviors of multiple vehicles.

[0050] The following combination Figure 2 The specific steps of the above multi-angle recognition and recording operation will be explained in detail.

[0051] Step S1: Map the vehicle behavior records extracted by the multi-source perception acquisition module to a unified spatiotemporal coordinate system.

[0052] Since the vehicle behavior records output by the multi-source sensing and acquisition module come from different traffic monitoring devices, and each traffic monitoring device operates independently, its data has its own time reference and spatial coordinate system. To achieve joint analysis of multi-source data, this module first maps the data collected by each traffic monitoring device to a unified time axis and a unified spatial coordinate system.

[0053] Specifically, this can be achieved through mapping functions. , will the i Traffic monitoring devices at all times t Raw data collected Converted to a set of objects in a unified spacetime reference frame : in, Indicates the number of the traffic monitoring device. =1, 2, ..., N, where N is the total number of traffic monitoring devices; Indicates the first i The device at time t The raw data collected; This represents the set of structured vehicle behavior records extracted from the traffic monitoring device at time t in a unified spatiotemporal reference frame.

[0054] Perform a union operation on the output object sets of all devices to obtain the complete object space in a unified spacetime reference frame. : , Through the above-mentioned unified spatiotemporal operation The data already includes complete observation data from all devices at the same moment. For example, when a vehicle brakes suddenly at an intersection, camera A might capture the sudden braking action from the front, camera B might capture the vehicle's motion from the side, and radar C might detect the sudden drop in speed. Before spatiotemporal unification, these three records belong to different devices, with different time references and spatial coordinate systems, making direct correlation difficult. After mapping in step S1, they are unified under the same spatiotemporal reference system. By associating the vehicle ID and timestamp, it can be determined that these three records describe the same behavior of the same vehicle, achieving complementarity and convergence of multi-source data. Subsequent steps S2 to S4 are all based on this fused multi-source data. This allows for the identification of various events by obtaining more comprehensive information from data from multiple devices.

[0055] Step S2: Based on the vehicle behavior record of a single vehicle, determine whether it meets the first type of condition. If it does, determine that a first type of traffic event has occurred and generate the corresponding first type of traffic event record.

[0056] Specifically, the first type of traffic incident refers to traffic incidents involving only the behavior of a single vehicle. This step begins from... The system extracts the vehicle behavior records of each vehicle one by one to determine whether they meet the first type of condition.

[0057] The first type of conditions includes, but is not limited to: (1) If the vehicle acceleration If the speed exceeds a set threshold (e.g., +2.5 m / s²) and the duration exceeds 1000 ms, it is determined to be a rapid acceleration event. (2) If the vehicle acceleration If the speed is less than a set threshold (e.g., -3.0 m / s²) and the duration exceeds 1000ms while the brake lights are on, it is determined to be a rapid deceleration event (emergency braking). (3) If the vehicle's angular velocity If the absolute value is greater than the threshold (e.g., 0.8 rad / s) and the duration exceeds 300 ms, it is determined to be a sharp turn event; (4) If the same vehicle records the behavior of not wearing a seat belt in several consecutive vehicle behavior records within a certain time period (i.e., the relevant fields in driving_behavior are always true), it is determined to be a seat belt not wearing event. In some specific embodiments, the determination of the first type of condition can be achieved through a first discriminant function. accomplish, = , in, For vehicles At any moment Vehicle behavior records, when When =1, a Class I traffic incident is determined to have occurred, and a corresponding Class I traffic incident record is generated. The traffic incident record includes at least an event type identifier, a specific event type, an identification identifier of the vehicle involved, a trigger timestamp, and an associated vehicle behavior record ID.

[0058] The following is an illustration of a specific Type I traffic incident record in JSON format: json { "event_record_id": "evt_20250501_001", "event_type": "Not wearing a seatbelt", "timestamp_start": 1736926336200, "timestamp_end": 1736926337200, "involved_behavior_groups": [ { "vehicle_id": " A12345", "behavior_record_ids": [ "beh_001_a1b2c3d4-e5f6-7890-abcd-ef1234567890", "beh_002_f2g3h4i5-j6k7-8901-bcde-fg2345678901", "beh_003_g3h4i5j6-k7l8-9012-cdef-gh3456789012", "beh_004_h4i5j6k7-l8m9-0123-defg-hi4567890123", "beh_005_i5j6k7l8-m9n0-1234-efgh-ij5678901234", "beh_006_j6k7l8m9-n0o1-2345-fghi-jk6789012345" ] } ], "involved_infrastructure": [] } For a description of each field involved in the above traffic incident record, please refer to Table 2 (fields that are the same as those in the vehicle behavior record are not listed again).

[0059] Table 2. Description of Traffic Incident Record Fields Step S3: Based on the vehicle behavior record and traffic facility status information of a single vehicle within the same spatiotemporal window, determine whether it meets the second type of condition. If it does, determine that a second type of traffic event has occurred and generate a second type of traffic event record.

[0060] Specifically, the second type of traffic incident refers to incidents involving the interaction of a single vehicle with road infrastructure and triggering traffic rules. This step begins from... The driving status information of each vehicle is extracted one by one, and the second type of condition judgment is made in combination with the traffic facility status information within the same spatiotemporal window.

[0061] In this embodiment of the application, traffic facility status information refers to the status information of traffic facilities in the environment in which the vehicle is traveling, which can instruct and constrain the vehicle's behavior, such as the status information provided by traffic lights, lane lines, zebra crossings, and other traffic facilities at intersections.

[0062] Specifically, the status information provided by traffic facilities such as fixed lane lines and zebra crossings generally remains unchanged within each time period for identifying traffic events, while the status of traffic facilities such as traffic lights and reversible lane lines changes differently within different time periods. This traffic facility status information can be obtained through an interface connected to the traffic control system and represented in a structured manner similar to vehicle behavior records. The following is a schematic illustration of the JSON format of a traffic facility status record: json { "infrastructure_id": "light_089", "infrastructure_type": "traffic_light", "infrastructure_function": "straight", "infrastructure_state": { "color": "red", "time_window": { "start": 1736926340000, "end": 1736926340400 } }, "timestamp": 1736926340200, "location": { "x": 1050.0, "y": 600.0 } } In this embodiment of the application, "same spatiotemporal window" means that the time and space deviation between vehicles, or between a vehicle and a traffic facility that can affect the vehicle, does not exceed a preset threshold. For example, when a vehicle is in the lane corresponding to a traffic light within a certain time interval (such as 10 seconds) and is within a lane range of 50 meters before and after the traffic light, the vehicle is within the effective range of the traffic light. Accordingly, during this time interval and within this lane range, the vehicle's behavior interacts with the state of the traffic light and may result in traffic incidents such as running a red light. In this case, it can be considered that the vehicle and the traffic light are in the same spatiotemporal window during this period.

[0063] The second type of conditions includes, but is not limited to: (1) When the comparison between a vehicle behavior record and a traffic light status record under the same time and space window shows that the vehicle crossed the stop line during the red light phase (the position coordinates crossed the stop line and the timestamp overlapped with the red light interval), and at least three consecutive records correspond to the three key positions of before the stop line, crossing the stop line, and entering the intersection, it is determined to be a red light running event. (2) If the direction of a vehicle at a certain intersection (determined by the trajectory direction angle) does not match the intersection channelization marking information of its lane (e.g., left turn from a straight lane) under the same time and space window, it can be determined that a traffic incident of not driving in the designated lane has occurred.

[0064] In some specific embodiments, the determination of the second type of condition can be achieved through a second discriminant function. G Implementation, for vehicles At any moment Vehicle behavior records Extract traffic facility status records that meet the above requirements within the same spatiotemporal window. ,when G = When any one of the second-type conditions is triggered, a second-type traffic event is determined to have occurred, and a second-type traffic event record is generated.

[0065] The following is an illustration of a specific Type II traffic incident record in JSON format: json { "event_record_id": "evt_20250501_010", "event_type": "running a red light", "timestamp_start": 1736926340000, "timestamp_end": 1736926340400, "involved_behavior_groups": [ { "vehicle_id": " B98765", "behavior_record_ids": [ "beh_b1_c3e4f5g6-h7i8-9012-jklm-nopqrstuvwxy", "beh_b2_d4f5g6h7-i8j9-0123-klmn-pqrstuvwxyza", "beh_b3_e5f6g7h8-i9j0-1234-lmno-pqrstuvwxyzab" ] } ], "involved_infrastructure": [ { "infrastructure_type": "traffic_light", "infrastructure_id": "light_089", "infrastructure_function": "straight", "infrastructure_state": { "color": "red", "time_window": { "start": 1736926340000, "end": 1736926340400 } } } ] } Step S4: Based on the vehicle behavior records of multiple vehicles within the same spatiotemporal window, determine whether the third type of condition is met. If the condition is met, determine that a third type of traffic event has occurred and generate a third type of traffic event record.

[0066] The third type of traffic incident involves the interaction of two or more vehicles. This step performs joint analysis on multiple vehicle behavior records within the same spatiotemporal window to determine whether the conditions for the third type are met. The data already includes multi-vehicle behavior records from different devices. For example, camera A captures the driving trajectory of vehicle p, camera B captures the driving trajectory of vehicle q, and radar C simultaneously detects the position and speed information of both vehicles. After spatiotemporal unification, this data from different devices is aligned to the same reference frame, making joint analysis of multi-vehicle behavior records possible.

[0067] Specifically, vehicle behavior records are first grouped based on timestamps and spatial locations. Vehicle behavior records with temporal and spatial deviations not exceeding preset thresholds are grouped into potential event groups corresponding to the same spatiotemporal window. For example, multiple vehicle behavior records with a time difference of no more than 3 seconds and a spatial distance of no more than 30 meters can be regarded as potential event groups within the same spatiotemporal window.

[0068] Within each potential event group, records belonging to different vehicles within the group are compared pairwise. A comprehensive analysis is performed on the correlation between the two vehicles across multiple dimensions, including location coordinates, lane number, speed, acceleration, direction of travel, and trajectory shape. If any of the following third-category conditions are met, a third-category traffic event is determined to have occurred, and a third-category traffic event record is generated: (1) In a potential event group corresponding to a spatiotemporal window, if the longitudinal distance between the two vehicles (determined by their position coordinates and lane number) decreases sharply to less than 2 meters within 200 milliseconds, and the rear vehicle decelerates rapidly (acceleration < -3.0 m / s²) with its brake lights on, while the front vehicle's speed drops sharply (deceleration < -2.0 m / s²), then it is determined to be a rear-end collision event. (2) If the lateral distance between the two vehicles (perpendicular to the direction of travel) is less than 1.5 times the sum of their widths and their relative speed is not zero, and the trajectories of the two vehicles change abruptly (such as a position jump or abnormal speed), then it is determined to be a collision event. (3) When a lane change trajectory of a vehicle is detected (the lane number changes in several consecutive behavior records), and at the same time it is detected that the distance of the following vehicle is less than 5 meters during the lane change process, and the following vehicle decelerates suddenly (acceleration < -3.0m / s²) or brakes suddenly, the lane change is determined to be an illegal lane change and affects the normal driving of other vehicles. In some specific embodiments, a consistency scoring function can be used to quantify the correlation between two vehicle behavior records. For any two vehicle behavior records within a group... and (corresponding to vehicle p and vehicle q respectively), define the consistency scoring function. This scoring function quantifies the temporal proximity, spatial proximity, and trajectory shape similarity of two vehicle behavior records within the same spatiotemporal window. A higher value indicates a greater likelihood that the two records point to the same interaction event. ( If the value is 0.75, it is determined to be a third-class traffic incident, and then the specific third-class sub-event (i.e., rear-end collision, scrape, etc.) is determined according to the relative motion pattern.

[0069] The involved_behavior_groups in the third category of traffic incident records will contain grouped vehicle behavior records of all involved vehicles. Each group corresponds to an involved vehicle and its related vehicle behavior record information (such as a list of related vehicle behavior record IDs).

[0070] Understandably, in the process of identifying the third type of traffic incident, the vehicle behavior record of the vehicle involved can be compared with the status information of traffic facilities within the same spatiotemporal window to determine whether there is also interaction with traffic facilities. If there is, the interaction information is stored in involved_infrastructure; otherwise, it means that the third type of traffic incident does not involve interaction with traffic facilities, and involved_infrastructure can be set to an empty array.

[0071] The following is an illustration of a specific Type 3 traffic incident record in JSON format: json { "event_record_id": "evt_20250501_021", "event_type": "rear-end collision", "timestamp_start": 1736926340000, "timestamp_end": 1736926342000, "involved_behavior_groups": [ { "vehicle_id": " C12345", "behavior_record_ids": [ "beh_f1_a1b2c3d4-e5f6-7890-abcd-ef1234567890", "beh_f2_b2c3d4e5-f6g7-8901-bcde-fg2345678901", "beh_f3_c3d4e5f6-g7h8-9012-cdef-gh3456789012" ] }, { "vehicle_id": " D67890", "behavior_record_ids": [ "beh_r1_d4e5f6g7-h8i9-0123-defg-hi4567890123", "beh_r2_e5f6g7h8-i9j0-1234-efgh-ij5678901234", "beh_r3_f6g7h8i9-j0k1-2345-fghi-jk6789012345" ] } ], "involved_infrastructure": [] } <Signal Linkage Control Module> In the embodiments of this application, after the event recognition module identifies a traffic event and generates a traffic event record, it transmits the traffic event record to the signal linkage control module. As mentioned above, the traffic event record includes the event type, the time of the event, the event location (including the intersection, road segment, and specific lane), the identity information of the vehicle involved, and the corresponding behavioral information (such as sudden braking, prolonged stillness, abnormal lane change, etc.). The signal linkage control module can dynamically adjust the green wave timing parameters based on the above information.

[0072] Specifically, the execution process of the signal linkage control module is as follows: First, the module receives traffic incident records from the incident recognition module and parses out key information such as incident type, incident location, the behavior of the vehicles involved, and the severity of the incident.

[0073] Secondly, based on the event type and severity, and considering the event location and the current green wave operation plan, the adjusted signal timing parameters are calculated using a preset control strategy. For example, when the event type is a rear-end collision occurring at an intersection within the green wave band, the green wave phase at that intersection can be temporarily canceled, the red light time at the upstream intersection in the direction of the accident can be increased to cut off subsequent traffic flow, and the green light time at the downstream intersection in the direction of the accident can be increased to allow the accident vehicles to leave quickly. When the event type is sudden braking or prolonged stationary states, the recommended speed for that section of road can be reduced to prompt following vehicles to slow down in advance. When the event type is running a red light or illegal lane changing, the red light time of the corresponding phase can be extended to curb the spread of illegal behavior. The control strategy can be adjusted accordingly for different types and severity of events.

[0074] In some specific embodiments, when the spatial location of a traffic incident falls within a certain green wave spatiotemporal window and the incident type reaches a preset severity level, the signal linkage control module can, according to a preset control function, [follow a specific control function]. Calculate the adjusted signal period, phase difference, and recommended vehicle speed: , in, For the current signal period, Current phase difference, For the current phase sequence, For recording traffic incidents, These are the adjusted signal period, phase difference, and phase sequence, respectively.

[0075] Finally, the calculated adjusted timing parameters are sent to the traffic signal controllers at the corresponding intersections for execution, thereby achieving dynamic optimization of the green wave operation strategy.

[0076] <Green Wave Window Prediction Module> Existing traffic incident perception systems typically employ a uniform perception strategy for the entire road segment, where all traffic monitoring devices operate continuously at a fixed sampling frequency, and edge computing units perform real-time analysis of all video streams. However, maintaining such high-frequency perception across the entire road segment and all time periods requires enormous computing power. Actual traffic flow is highly unevenly distributed in time and space. During green wave traffic, vehicles pass through intersections in convoys, making the probability of incidents during this period and segment far higher than in other time and space areas. Furthermore, during periods with little or no traffic, continuous high-frequency perception actually wastes a significant amount of computing and storage resources.

[0077] Based on this, this application introduces the parameters used for signal timing in green wave control into the sensing and scheduling stage. In some preferred embodiments, the system also includes a green wave spatiotemporal window prediction module, which can pre-calculate the spatiotemporal areas where vehicles may concentrate their passage based on the signal timing scheme and road geometry parameters of the green wave road segment, and mark these areas as key sensing areas, so that the multi-source sensing acquisition module can collect data within the key sensing areas at a higher sampling frequency than outside the key sensing areas.

[0078] In some specific embodiments, the spatiotemporal window of the green wave is calculated in the following way: For the intersection Given the signal period Phase difference and the recommended speed for this section of road. Then in the first Within a given period, assuming the convoy is at the upstream intersection at time... Once permitted, the estimated time interval for vehicles to reach the downstream intersection can be expressed as: ， in, For the length of the road segment, To account for the speed range after fluctuations in driving behavior, the corresponding spatial range can be simplified to: This creates a green wave spatiotemporal window. .

[0079] It is understood that the identification of green wave spatiotemporal windows is not limited to the calculation methods based on signal period, phase difference, and recommended vehicle speed mentioned above. In some other optional embodiments, data-driven or predictive green wave identification methods can also be used. For example, based on historical traffic flow trajectories and signal control logs, high-probability zones for vehicle traffic within a certain time period can be obtained through cluster analysis or time series prediction, and these "high-probability zones" can be used as key sensing areas equivalent to green wave windows. In road sections without strict traditional green wave timing, "temporary green wave segments" can also be adaptively identified based on vehicle passage time distribution and dynamic timing strategies, and the sensing frequency and algorithm processing priority can be increased in these areas. The core of this application is "using the traffic rhythm characteristics of road sections to drive the scheduling of sensing resources." Any method that can characterize the traffic rhythm and delineate key areas accordingly can be used as an alternative implementation of this application.

[0080] Furthermore, in some preferred embodiments, such as Figure 3 As shown, the generated traffic event records can also be organized using a directed graph data structure to establish the association between the traffic event and the behavior records of the vehicles involved, the status records of related facilities, and the preceding events, thus forming a traffic event record containing directed associations.

[0081] Specifically, each traffic incident record stores information about the behavior of the vehicles involved in the incident (such as the ID of the vehicle behavior record) through the `involved_behavior_groups` field, which is related to the traffic incident in terms of time, space, and behavior. This forms directed edges between the traffic incident record and the behavior records of each involved vehicle. The same traffic incident record can be associated with multiple vehicle behavior records through multiple directed edges; conversely, the same vehicle behavior record can also be associated with multiple traffic incident records through multiple directed edges. Through this complex many-to-many association, a directed graph topology is formed with the traffic incident record as the center and the vehicle behavior records as the periphery. This allows one to trace back to all related vehicle behavior records along directed edges from any identified traffic incident record.

[0082] In some preferred embodiments, the traffic event record also includes an involved_infrastructure field to store the status record information of the traffic facilities involved that are related to the traffic event in terms of time-space-behavior (such as the ID of the status record of the traffic facility involved), thereby establishing the association between the vehicle behavior in the traffic event and the status of the traffic facilities in the same time-space window, forming a directed edge between the traffic event record and the traffic facility status record.

[0083] It is understandable that multiple traffic events that are close in time or space are often not isolated, but rather a set of traffic events with causal relationships formed by a series of vehicle behaviors or traffic events coupled together, such as: illegal lane change → minor collision, sudden deceleration → rear-end collision. Therefore, in some optional embodiments, a related_events field can be added to each traffic event record to store information about preceding traffic event records that have a causal relationship with the current traffic event (such as the ID of the preceding traffic event record), thereby forming a directed edge between the current traffic event record and its preceding traffic event records. The causal relationship between two traffic events can be determined based on preset rules. For example, if the same vehicle is involved in both a first-type event (illegal lane change) and a third-type event (minor collision) within a short time window, and the time sequence is obvious, then the two can be considered to have a causal relationship.

[0084] Through the aforementioned directed graph data structure, the system can not only respond to events and control signals in real time, but also achieve complete backtracking of event perception elements when needed, providing structured data support for scenarios such as tracing the cause of signal adjustments, determining responsibility for multi-vehicle accidents, and reviewing and analyzing traffic conditions.

[0085] To verify the perception and linkage control effect of this application in the event scenario of a sudden incident on a green wave road section, the following method was adopted: Figure 4 The simulation of a three-intersection micro-traffic flow scenario was used for verification. The simulated road network consists of intersections A, B, and C on the main road, with intersection spacing of 400m and 350m respectively, and three approach lanes in the main direction. A coordinated control scheme with a cycle of 90s is adopted for the three intersections, with an effective green light time of 38s in the main direction, and phase differences of 0s, 24s, and 45s respectively. The recommended speed is 50km / h. Roadside cameras, millimeter-wave radar, and geomagnetic sensors are deployed in the simulation scenario, with sampling frequencies of 10Hz, 20Hz, and 1Hz respectively. The simulation lasts for 3600s, with the first 600s as a warm-up phase. Each scheme is tested repeatedly with 10 different random seeds. Specific simulation parameters are shown in Table 3.

[0086] Table 3. Main Simulation Parameters The green wave spatiotemporal window is calculated based on the phase difference and recommended vehicle speed at each intersection, and the spatiotemporal area where vehicles arrive in concentrated areas is marked as the key perception area. At 1200 seconds, a disabled vehicle is placed downstream of section A-B, occupying one lane for 240 seconds to simulate a sudden lane obstruction event. After the event, the camera is briefly blocked by the vehicle in front, the radar continuously outputs the target speed and position, and the geomagnetic sensor outputs the lane occupancy status. Data from all sources are synchronized in time, mapped in coordinates, and associated with vehicle identities to form a unified vehicle behavior record. Figure 5 The spatiotemporal relationship between the green wave window and the vehicle trajectory is shown.

[0087] To verify the multi-source fusion sensing and linkage control effects of this application, four comparison schemes were set up: Scheme S0 is a fixed green wave scheme and does not involve event linkage; Scheme S1 is an event-linked control based on single-source video recognition; Scheme S2 employs multi-source fusion identification while maintaining the original signal timing; Scheme S3 adopts the multi-source fusion sensing and signal linkage control scheme of this application. After the event is confirmed, the red light in the main direction at the upstream intersection is extended by 8 seconds and the green light in the main direction at the downstream intersection is extended by 10 seconds, and the recommended speed is temporarily adjusted to 35 km / h. The above adjustment lasts for two cycles, and then the original green wave scheme is smoothly restored according to the event status and the degree of queue dissipation.

[0088] Statistical analysis of event identification results under 10 different random seeds. For example... Figure 6 As shown, the F1 scores (harmonic mean of precision and recall) for the single-source video, radar, and geomagnetic schemes are approximately 0.82, 0.83, and 0.72, respectively; while the multi-source fusion scheme achieves precision, recall, and F1 scores of 0.95, 0.93, and 0.94, respectively. Multi-source fusion utilizes continuous motion observations from radar during the obstruction phase and geomagnetic occupancy information to cross-validate video results, thus maintaining high recognition stability even under short-term obstruction conditions.

[0089] Figure 7 The changes in average speed and maximum queue length before and after the incident are shown. Under the fixed green wave scheme S0, the average speed continued to decrease after the incident, and the maximum queue length reached approximately 312m. Under the scheme S3 of this application, through upstream flow restriction, downstream evacuation, and recommended speed adjustment, the speed decrease and queue growth were suppressed, the maximum queue length was approximately 156m, and the road could return to a stable state more quickly after the incident was resolved.

[0090] The simulation results (mean values ​​of 10 experimental groups) for the four schemes are shown in Table 4 and Figure 8 As shown.

[0091] Table 4 Simulation results of different schemes (mean values ​​of 10 experimental groups) From Table 4 and Figure 8 As can be seen, compared with the fixed green wave scheme S0, the proposed scheme S3 shows significant improvement in all evaluation indicators. Specifically, the average travel time decreased from 181.4s to 132.9s, a reduction of approximately 26.7%; the average number of stops decreased from 2.41 to 1.18, a reduction of approximately 51.0%; the maximum queue length decreased from 312m to 156m, a reduction of 50.0%; the throughput per simulated time period increased from 989veh to 1168veh, an increase of approximately 18.1%; and the traffic recovery time was shortened from 684s to 286s, a reduction of approximately 58.2%. In contrast, although scheme S2 can accurately detect events, its improvement is significantly smaller than that of scheme S3 because it does not adjust signal timing in conjunction with the event. This indicates that forming a closed loop between event perception and signal control can further improve the road segment's recovery capability.

[0092] A comprehensive benefit index was constructed using the normalized weighted results of average travel time, maximum queue length, and recovery time. The upstream red light extension and downstream green light extension were then adjusted accordingly. The results are as follows: Figure 9 As shown in the simulation, under the traffic demand scenario, the comprehensive benefit index reaches its optimal value when the upstream red light extension is approximately 8 seconds and the downstream green light extension is approximately 10 seconds. Too small an extension makes it difficult to suppress subsequent inflows from the blocked section, while too large an extension causes additional queuing upstream. In real-world scenarios, the signal linkage control module can adaptively calculate the control parameters based on the severity of the event, the event location, the real-time queue length, and the remaining capacity, rather than being limited to the fixed values ​​used in this simulation.

[0093] The simulation results show that this application can quickly form a reliable traffic event record under the condition that the multi-source observations are not completely consistent or partially obstructed, and dynamically adjust the timing parameters of adjacent intersections according to the location and severity of the event, so that the event perception and signal control form a closed loop, effectively suppressing the speed reduction, queue spread and recovery delay caused by sudden events in the green wave section.

[0094] The specific embodiments of this application have been described in detail above. For those skilled in the art, several improvements and modifications can be made to this application without departing from the principle of this application, and these improvements and modifications also fall within the protection scope of the claims of this application.

Claims

1. A traffic event perception and control system for green wave road sections based on multi-source data fusion, characterized in that, include: The multi-source sensing and acquisition module is used to collect multi-source traffic monitoring data from multiple traffic monitoring devices deployed on green wave road sections, and extract multiple structured vehicle behavior records from them. The event recognition module is used to perform multi-angle recognition and recording operations on at least one of the vehicle behavior records to identify the traffic event type and form a traffic event record; the traffic event record includes the identified traffic event type, the vehicle involved, and vehicle behavior record information associated with the vehicle involved; The signal linkage control module is used to automatically adjust the intersection signal timing parameters of the green wave road segment based on the information in the traffic event record.

2. The system according to claim 1, characterized in that, Each structured vehicle behavior record includes vehicle identity information and corresponding vehicle behavior information; the vehicle behavior information corresponding to any specific vehicle is the vehicle's own movement and status information or the vehicle's driving operation information observed from traffic monitoring data at a specific moment or within a specific time window.

3. The system according to claim 1, characterized in that, The traffic incidents mentioned include Category I traffic incidents, Category II traffic incidents, and Category III traffic incidents, among which: The first category of traffic incidents refers to traffic incidents that involve only the behavior of a single vehicle; The second category of traffic incidents involves an interaction between a single vehicle and road infrastructure that triggers traffic rules; The third category of traffic incidents involves the interaction of two or more vehicles.

4. The system according to claim 1, characterized in that, The multi-angle recognition and recording operation includes the following steps: Step S1: Map the vehicle behavior records extracted by the multi-source perception acquisition module to a unified spatiotemporal coordinate system; Step S2: Based on the vehicle behavior record of a single vehicle, determine whether it meets the first type of condition. If it does, determine that a first type of traffic event has occurred and generate a first type of traffic event record. Step S3: Based on the vehicle behavior record of a single vehicle and the traffic facility status information within the same spatiotemporal window, determine whether it meets the second type of condition. If it does, determine that a second type of traffic event has occurred and generate a second type of traffic event record. Step S4: Based on the vehicle behavior records of multiple vehicles within the same spatiotemporal window, determine whether the third type of conditions are met among the multiple vehicles. If the conditions are met, determine that a third type of traffic event has occurred and generate a third type of traffic event record.

5. The system according to claim 4, characterized in that, The same spatiotemporal window refers to the time and space deviation between vehicles, or between a vehicle and traffic facilities that can affect the vehicle, not exceeding a preset threshold.

6. The system according to claim 1, characterized in that, It also includes a green wave window prediction module, which is used to obtain the signal timing parameters and road geometry parameters of the green wave road segment, calculate one or more green wave spatiotemporal windows, and mark the green wave spatiotemporal windows as key sensing areas; The multi-source sensing acquisition module acquires data within the key sensing area at a higher sampling frequency than outside the key sensing area.

7. The system according to claim 6, characterized in that, The green wave spatiotemporal window is calculated in the following way: For the intersection Given the signal period Phase difference and the recommended speed for this section of road. Then in the first Within a given period, assuming the convoy is at the upstream intersection at time... Once permitted, the estimated time interval for vehicles to reach the downstream intersection can be expressed as: ， in, For the length of the road segment, To account for the speed range after fluctuations in driving behavior, the corresponding spatial range can be simplified to: This creates a green wave spatiotemporal window. .

8. The system according to claim 1, characterized in that, Each traffic incident record stores information on the behavior of the vehicles involved in the incident that is related to the traffic incident in terms of time, space, and behavior, thus forming a directed edge between the traffic incident record and the behavior records of the vehicles involved.

9. The system according to claim 8, characterized in that, The traffic incident record also stores the status record information of the traffic facilities involved that are related to the traffic incident in terms of time, space and behavior, so as to form a directed edge between the traffic incident record and the status record of the traffic facilities involved.

10. The system according to claim 9, characterized in that, The traffic incident record also stores information on preceding traffic incident records that have a causal relationship with the traffic incident, so as to form a directed edge between the traffic incident record and its preceding traffic incident records.

Images