Traffic accident recognition method, terminal, and storage medium

By analyzing the driving data and geographical location information before and after the vehicle stops, combining characteristic parameters, using pre-trained models to judge traffic accidents, the problem of low identification accuracy in the existing technology is solved, and high-precision traffic accidents under resource limitations are achieved Identify.

WO2024234452A1PCT designated stage expired Publication Date: 2024-11-21TIANJIN SOTEREA AUTOMOTIVE TECH LMITED CO +1

Patent Information

Application Number
PCT/CN2023/103800
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2023-05-12
Filing Date
2023-06-29
Publication Date
2024-11-21

AI Technical Summary

Technical Problem

The prior art is difficult to accurately identify traffic accidents when resources are limited, and the data loss problem leads to low recognition accuracy.

Method used

By analyzing the driving data before and after the vehicle's parking time, combining geographical location information and characteristic parameters, a pre-trained traffic accident identification model is used to determine whether a traffic accident has occurred. It only relies on the driving data of the vehicle and does not need to obtain a large amount of other traffic data.

Benefits of technology

The accuracy of traffic accident identification is improved under resource constraints, and it can accurately determine whether a vehicle has an accident. It is suitable for a variety of scenarios such as accident tracking, forward accident reminders and congestion warnings.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN2023103800_21112024_PF_FP_ABST
    Figure CN2023103800_21112024_PF_FP_ABST
Patent Text Reader

Abstract

The present application provides a traffic accident recognition method, a terminal, and a storage medium. The method comprises: according to first driving data of a first preset duration before a parking moment and second driving data of a second preset duration after the parking moment of a vehicle, determining whether the vehicle has an abnormal parking behavior, wherein the parking moment is the ending moment of the first preset duration and the starting moment of the second preset duration; for the vehicle having an abnormal parking behavior, according to the geographic location information of the vehicle at the parking moment, determining that the vehicle is in a suspected accident; for the vehicle in the suspected accident, extracting feature parameters of the vehicle according to driving data of the vehicle in at least one preset time period, wherein each time period among the at least one preset time period comprises the parking moment; and inputting the feature parameters into a pre-trained traffic accident recognition model to obtain a recognition result, wherein the recognition result comprises that the vehicle is in a traffic accident, or the vehicle is not in a traffic accident. Accurate real-time recognition of traffic accidents can be realized.
Need to check novelty before this filing date? Find Prior Art

Description

Traffic accident identification method, terminal and storage medium

[0001] This patent application claims priority to Chinese patent application No. CN 202310530375.8 filed on May 12, 2023. The disclosure of the prior application is incorporated herein by reference in its entirety. Technical Field

[0002] The present application relates to the field of safe driving technology, and in particular to a traffic accident identification method, terminal, and storage medium. Background Art

[0003] Traffic accidents are a common occurrence in complex transportation networks. Once they occur, they severely impact traffic efficiency. For example, an accident occurring during rush hour can paralyze traffic on the affected route and potentially lead to secondary accidents. Traffic accidents also indirectly contribute to complex social and transportation issues, such as increased transportation costs and increased pollution. They pose a serious threat to the safety of commuters and hinder economic development.

[0004] Rapidly identifying and providing early warnings for traffic accidents is an effective means of reducing accident losses. Traditional traffic accident identification algorithms require extensive data to establish reliable anomaly recognition models. However, acquiring large amounts of traffic data is challenging, making traffic accident identification impossible under resource-constrained conditions. Furthermore, various issues, such as missing data, exist in the acquisition of large amounts of traffic data, resulting in low accident identification accuracy using existing technologies.

[0005] Technical content

[0006] In view of this, the present application provides a traffic accident identification method, terminal and storage medium, which can solve the problem of low accident identification accuracy in the existing technology. Technical Solutions

[0007] In a first aspect, the present application provides a traffic accident identification method, comprising:

[0008] determining that the vehicle has exhibited abnormal parking behavior based on first driving data of the vehicle for a first preset time period before a parking moment and second driving data of the vehicle for a second preset time period after the parking moment, wherein the parking moment is the end time of the first preset time period and the parking moment is the start time of the second preset time period;

[0009] For a vehicle that has engaged in abnormal parking behavior, determining that the vehicle has been involved in a suspected accident based on the vehicle's geographic location information at the time of parking;

[0010] For a vehicle suspected of having an accident, extracting characteristic parameters of the vehicle based on driving data of the vehicle in at least one preset time period, each of the at least one preset time period including the parking moment;

[0011] The characteristic parameters are input into a pre-trained traffic accident recognition model to obtain a recognition result, where the recognition result includes whether the vehicle has been involved in a traffic accident or whether the vehicle has not been involved in a traffic accident.

[0012] In a second aspect, the present application provides a terminal comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the steps of the method described in the first aspect or any possible implementation of the first aspect are implemented.

[0013] In a third aspect, the present application provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the method described in the first aspect or any possible implementation of the first aspect. Beneficial effects

[0014] Compared with the prior art, the identification method provided by this application has the following beneficial effects:

[0015] This application analyzes driving data from the time period before and after a vehicle stops to identify vehicles that have engaged in abnormal parking behavior. The application then uses the vehicle's geographic location at the time of the stop to determine if the vehicle was involved in a suspected accident. Driving data from at least one time period before and after the vehicle stops is then used to extract characteristic parameters that can be used to indicate whether the vehicle was involved in a traffic accident. By inputting these characteristic parameters into a pre-trained traffic accident recognition model, a determination result is obtained regarding whether the vehicle was involved in a traffic accident. The method provided in this application determines whether a vehicle has been involved in an accident solely based on its driving data, without requiring the acquisition of other data, such as large amounts of traffic data from other vehicles. This method not only enables the determination of whether a vehicle has been involved in a traffic accident even with limited resources, but also improves the accuracy of traffic accident recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the embodiments or descriptions of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0017] FIG1 is a flow chart of a method for identifying a traffic accident according to an embodiment of the present application;

[0018] FIG2 is a flowchart of another traffic accident identification method provided in an embodiment of the present application;

[0019] FIG3 is a flowchart of another method for identifying traffic accidents according to an embodiment of the present application;

[0020] FIG4 is a schematic diagram of a terminal provided in an embodiment of the present application. Modes for Carrying Out the Invention

[0021] In the following description, specific details such as specific system structures and techniques are provided for purposes of illustration, not limitation, to facilitate a thorough understanding of the embodiments of the present application. However, it will be apparent to those skilled in the art that the present application may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid obscuring the description of the present application with unnecessary detail.

[0022] In order to make the objectives, technical solutions and advantages of this application clearer, the following will be described through specific implementation methods in conjunction with the accompanying drawings.

[0023] Refer to Figure 1, which shows a flow chart for implementing a traffic accident identification method provided by an embodiment of the present application. The execution subject of the traffic accident prediction method can be a vehicle-mounted terminal or a remote server, and this embodiment does not limit it. The method can be applied to commercial vehicles or passenger vehicles. In a typical case, the present application is applicable to tracking traffic accidents of vehicles based on the data of vehicles that have been driven after an accident, so as to determine whether it is necessary to contact the driver and help the driver to deal with the accident, so as to accurately understand the scene of the vehicle's danger. In another typical case, the present application is applicable to the application scenario of reminding the rear vehicle of the accident ahead after the vehicle accident to prevent the occurrence of a secondary accident. In another typical case, the present application is applicable to the scenario of assisting congestion identification and issuing congestion warnings to vehicles after the vehicle accident, etc. According to actual conditions, the method provided by the embodiment of the present application can be applied to a variety of scenarios, and this application does not limit it.

[0024] Referring to FIG1 , the traffic accident identification method includes:

[0025] In step 101, based on first driving data of the vehicle for a first preset time period before the parking moment and second driving data of the vehicle for a second preset time period after the parking moment, it is determined that the vehicle has an abnormal parking behavior.

[0026] The parking time is the end time of the first preset time, and the parking time is the start time of the second preset time.

[0027] Since vehicles are typically parked after a traffic accident, the "parking time" referred to in the embodiments of this application is the moment when the vehicle transitions from a moving state to a speed of zero. During the period before and after a traffic accident, the vehicle's state differs from a normal parking state. For example, sudden braking is often required before an accident, and the vehicle typically remains stationary for a period of time after an accident. By analyzing the vehicle's driving data before and after the parking time, it is possible to determine whether the vehicle's parking behavior was abnormal.

[0028] The specific values ​​of the first preset time length and the second preset time length can be determined according to actual needs or specific experiments. For example, the first preset time length can be 5 seconds, the second preset time length can be 15 minutes, and so on.

[0029] In a possible implementation, the first preset duration may be shorter than the second preset duration.

[0030] This embodiment does not impose any specific restrictions on the means for obtaining the first driving data of the first preset time period before the parking moment and the second driving data of the second preset time period after the parking moment, and any feasible means may be used.

[0031] For example, vehicle data can be collected at a certain frequency by relevant equipment in the vehicle and uploaded to the terminal. When the terminal needs to obtain vehicle driving data within a certain time period, it can directly extract the vehicle driving data within that time period. The data collection frequency can be set according to actual needs, for example, once every second.

[0032] In an optional embodiment, the vehicle starts to decelerate at the initial moment of the first preset time period until it reaches the stopping time. The first driving data includes the vehicle speed at the initial moment of the first preset time period, the deceleration of the vehicle in the first preset time period, and the second driving data includes the average speed of the vehicle in the second preset time period.

[0033] In an optional embodiment, whether the vehicle has undergone emergency braking is determined based on the vehicle speed at the initial moment of the first preset time period and the deceleration of the vehicle during the first preset time period; if the vehicle has undergone emergency braking and the average speed of the vehicle during the second preset time period is lower than the preset speed threshold, it is determined that the vehicle has undergone abnormal parking behavior.

[0034] For example, when a vehicle undergoes emergency braking, the vehicle speed usually changes rapidly and the vehicle deceleration is large. Combined with the initial speed when the vehicle starts to decelerate and the deceleration during the deceleration process, it can be determined whether the vehicle has undergone emergency braking. When the vehicle undergoes emergency braking and the average speed of the vehicle for a period of time after the speed reaches 0 is lower than the preset speed threshold, it can be determined that the vehicle has experienced abnormal parking behavior.

[0035] In one possible implementation, based on the deceleration of the vehicle during the first preset time period, a speed range corresponding to the deceleration is determined in a preset first judgment list, and it is determined whether the speed of the vehicle at the initial moment of the first preset time period belongs to the speed range. If so, it is determined that the vehicle has undergone emergency braking, wherein the first judgment list includes a unique mapping relationship between the vehicle deceleration and speed in an emergency braking situation; or, based on the deceleration of the vehicle during the first preset time period and the vehicle speed at the initial moment of the first preset time period, an index is performed in a preset second judgment list to determine the braking type of the vehicle, wherein the second judgment list includes a unique mapping relationship between the vehicle deceleration, speed and braking type, and the braking type includes emergency braking and non-emergency braking.

[0036] For example, the first preset duration is 5 seconds, and the second preset duration is 15 minutes. The vehicle is in a decelerating state from the start to the end of the first preset duration. The vehicle's speed at the beginning of the first preset duration is the maximum speed during the first preset duration. The preset speed threshold is set to 1 km / h, indicating that the vehicle remains substantially stationary for the second preset duration after the accident. To facilitate description of the first and second judgment lists, the vehicle's speed at the beginning of the first preset duration is represented by the maximum speed v, and the vehicle's deceleration during the first preset duration is represented by the deceleration a. For example, the first judgment list may be as shown in Figure 1 below.

[0037] Table 1

[0038] Deceleration a (m / s2) Maximum speed v (km / h) a≥a1 Any speed a1>a≥a2v≥v1a2>a≥a3v≥v2a3>a≥a4v≥v3

[0039] Among them, a1>a2>a3>a4, v1<v2<v3.

[0040] In conjunction with Table 1 above, for example, when the deceleration is greater than or equal to a1 m / s 2 When the maximum speed is any value, the vehicle stops for emergency braking; for example, when the deceleration is ax m / s 2 When a1>ax≥a2, the speed interval corresponding to the deceleration is determined to be v≥v1 according to the first judgment list, and according to the speed of the vehicle at the initial moment of the first preset time length, it is determined whether the speed is greater than or equal to v1km / h. If so, it is determined that the vehicle has undergone emergency braking. If not, it is determined that the vehicle has not undergone emergency braking.

[0041] Exemplarily, the second judgment list may be as shown in the following representation 2.

[0042] Table 2

[0043] Deceleration a (m / s2) Maximum speed v (km / h) Braking type ≥a5 Emergency braking at any speed a6≤a<a5v≥v4 Emergency braking a6≤a<a5v<v4 Non-emergency braking a7≤a<a6v≥v5 Emergency braking a7≤a<a6v<v5 Non-emergency braking a8≤a<a7v≥v6 Emergency braking a8≤a<a7v<v6 Non-emergency braking a<a8 Any speed Non-emergency braking

[0044] Among them, a5>a6>a7>a8, v4<v5<v6. According to Table 2 above, for example, when the deceleration is aym / s 2 , and a7≤ay<a6, the maximum vehicle speed is vxkm / h<v5, and the braking type is non-emergency braking according to the index in Table 2.

[0045] It should be noted that the data in Table 1 and Table 2 are only examples. The data in Table 1 and / or Table 2 can be determined through historical data statistics, expert experience values, or other methods, and the implementation methods of this application do not limit this.

[0046] In step 102, for a vehicle that has abnormal parking behavior, it is determined that the vehicle has had a suspected accident based on the geographic location information of the vehicle at the time of parking.

[0047] In the embodiment of the present application, when a vehicle is involved in a traffic accident, it is usually a sudden traffic accident during driving. If, after the vehicle stops, the parking location of the vehicle is judged to be a parking lot, a community garage, a highway service area, a gas station, a roadside temporary parking space, etc. based on the vehicle's geographic location information, then the possibility of the vehicle being involved in an accident is small. If, after the vehicle stops, the parking location of the vehicle is judged to be a main road, a non-service area of ​​a highway, etc. based on the vehicle's geographic location information, then the possibility of the vehicle being involved in an accident is high.

[0048] Based on this, in an embodiment of the present application, the geographic location information of the vehicle at the time of parking is used to determine whether the vehicle with abnormal parking behavior is in a suspected accident.

[0049] In an optional embodiment, based on the geographic location information of the vehicle at the time of parking and a pre-set list of common parking points for the vehicle, it is determined whether the vehicle is located at a parking point in the list of common parking points. If not, it is determined that the vehicle has been involved in a suspected accident, wherein the list of common parking points is a collection of common parking points for the vehicle obtained based on the vehicle's historical driving data; or, based on the geographic location information of the vehicle at the time of parking, it is determined whether the vehicle is located in a road area. If so, it is determined that the vehicle has been involved in a suspected accident.

[0050] In one possible implementation, the vehicle's historical parking locations are obtained based on its historical driving data to form a list of frequent parking locations. When it is determined that the vehicle has engaged in abnormal parking behavior, the vehicle's parking location is matched against locations in the frequent parking location list. If the parking location matches a location in the frequent parking location list, the vehicle's parking is determined to be normal. It should be noted that due to errors in positioning data, a match between the parking location and a location in the frequent parking location list means that the distance between the two locations is less than or equal to a preset distance threshold.

[0051] In a possible implementation, to determine whether the vehicle is located in a road area, map matching may be performed based on the position of the vehicle at the time of parking to determine whether the vehicle is located in a road area.

[0052] In an optional embodiment, for any vehicle, the process of obtaining a list of frequent stopping points of the vehicle includes: obtaining the vehicle's historical driving data; marking the vehicle's braking and stopping conditions in the historical driving data, wherein the condition for meeting the braking and stopping conditions is that the vehicle's speed drops to 0 and the time for which the speed remains at 0 exceeds a preset time threshold; and establishing a list of frequent stopping points of the vehicle based on the vehicle's geographic location information under each braking and stopping condition.

[0053] For example, the preset time threshold is 5 minutes. The preset time threshold can be determined based on data statistics or experiments, or by other means. The embodiments of this application do not limit its specific size and acquisition method.

[0054] In an optional implementation, vehicle location information can be collected by devices such as the vehicle's AEB (Autonomous Emergency Braking) system and TBOX (Telematics Box). The vehicle's integrated GPS (Global Positioning System) or Beidou navigation system can then be used to record vehicle location information. This collected real-time vehicle location data can be transmitted back to the terminal via a wireless network via an on-device communication module, such as a SIM card (Subscriber Identity Module). For example, the transmission frequency can be once every second. The collected speed and location information for each vehicle is stored in a vehicle history database.

[0055] Based on the aforementioned vehicle driving history database, a braking stop condition is defined as a situation where the vehicle's speed drops to zero and remains at zero for a predetermined time threshold, such as five minutes. The corresponding parking location in this condition is a parking point for the vehicle. The latitude and longitude information of this parking point can be represented by Pi(Xi, Yi). Therefore, over a period of time, each vehicle has its own set of parking locations, P{}, which serves as the list of frequently stopped locations in this step.

[0056] In an optional embodiment, after establishing a list of frequent stop points for the vehicle, the method further includes: sorting the stop points in the frequent stop point list in descending order of parking time to obtain a first queue; entering an update step, the update step including sequentially selecting the first frequent stop point in the first queue as a target frequent stop point, traversing the other frequent stop points in the first queue to obtain n frequent stop points whose distance from the target frequent stop point is less than or equal to a preset distance, constructing a new frequent stop point position based on the position information of the target frequent stop point and the n frequent stop points, placing the new frequent stop point position into a second queue, recording the number of stops corresponding to the new frequent stop point position as n+1 in the second queue, deleting the target frequent stop point and the n frequent stop points from the first queue, and if there is no frequent stop point in the first queue whose distance from the target frequent stop point is less than or equal to the preset distance, deleting the target frequent stop point from the first queue, where n is a positive integer greater than or equal to 1; repeating the update step until the number of frequent stop points in the first queue is zero, and using the frequent stop points in the second queue as the optimized frequent stop point list for the vehicle.

[0057] The above process corresponds to the process of optimizing the stop point list in actual application scenarios, which reduces the matching calculation amount by merging stop points with similar distances into the same stop point.

[0058] The following is an explanation using a specific example.

[0059] The first step is to denote the initial set of frequent stops for a vehicle by P{}. The stops in the frequent stop list P{} are sorted in descending order of stop time to obtain the first queue.

[0060] Enter the update step: First, select the parking location Pi (Xi, Yi) with the longest parking time, traverse this first queue, and compare it with the remaining parking locations. Assume that there is a parking point Pj (Xj, Yj), and define the distance Dij between Pi and Pj as ||Pi(Xi, Yi)-Pj(Xj, Yj)||. If this distance Dij is within a preset distance range, such as Dij < 50 meters, it can be determined that location i and location j are the same regular parking location. Based on parking point Pi (Xi, Yi) and parking point Pj (Xj, Yj), construct a new parking location Q (Ai, Bi). Its specific latitude and longitude (Ai, Bi) are defined as: (Ai, Bi) = ((Xi+Xj) / 2, (Yi+Yj) / 2);

[0061] The number of times N that this regular stop point Q (Ai, Bi) stops at this position is recorded as 2, and it is put into the second queue. The second queue can be represented by the regular stop point set Q{}, and the Pi and Pj parking points are removed from the first queue;

[0062] If the comparison between Pi (Xi, Yi) and all points in the first queue does not meet the condition Dij < 50 meters, then the Pi parking point is considered to be a temporary parking point, is not included in the second queue, and is removed from the first queue.

[0063] The above is an example of n=1. If n=2, that is, there are two other frequent stop points in the first queue whose distance from Pi(Xi, Yi) is less than or equal to the preset distance, then the longitude value of the new frequent stop point position Q is the average of the longitude values ​​of these three frequent stop points, and the latitude value of Q is the average of the latitude values ​​of these three frequent stop points.

[0064] The update step proceeds again, selecting the first stop point Pk (Xk, Yk) in the first queue. Pk is compared with all other stop points in the first queue, and the distance between each of these stop points and Pk is calculated. Stop points whose distance from Pk is less than or equal to the preset distance are identified. New stop points are constructed using the above method and placed in the second queue. If there are no stop points in the first queue whose distance from Pk is less than or equal to the preset distance, then Pk is deleted from the first queue.

[0065] Repeat the above updating steps until the number of parking points in the first queue is 0. Then a new set of frequent parking points is obtained, namely the second queue.

[0066] In one possible implementation, based on actual business scenarios, restrictions related to actual business can be added to the second queue. For example, the historical driving data is the historical driving data within a preset period before the current moment, for example, the period is 1 month. Or, according to actual business needs, it is necessary to obtain parking locations with more than a preset number of parking times. For example, it is necessary to obtain parking locations with more than or equal to 3 parking times. Then, parking spots with more than or equal to 3 parking times are selected from the second queue to construct a new parking spot list.

[0067] In one possible implementation, for some application scenarios, such as commercial vehicle fleet scenarios, fixed pre-set regular stopping locations can be added to the second queue, such as the fleet's fixed regular vehicle operation parking lot, designated repair shop, gas station, etc.

[0068] In a possible implementation, the second queue is updated periodically.

[0069] In step 103, for a vehicle suspected of having an accident, characteristic parameters of the vehicle are extracted based on the driving data of the vehicle in at least one preset time period, where each time period in the at least one preset time period includes a parking moment.

[0070] In the embodiment of the present application, by analyzing the driving data of at least one preset time period, characteristic parameters that can be used for traffic accident identification are extracted, that is, possible characteristic parameters before and after the traffic accident occurs.

[0071] In one possible implementation, before extracting the characteristic parameters of a vehicle, some existing methods may be used to clean the driving data for at least one preset time period, and then feature extraction may be performed on the cleaned driving data to obtain the characteristic parameters of the vehicle. The embodiments of this application do not limit the means for extracting the characteristic parameters of a vehicle based on the driving data for at least one time period.

[0072] In step 104, the feature parameters are input into a pre-trained traffic accident recognition model to obtain a recognition result, which includes whether the vehicle has been involved in a traffic accident or whether the vehicle has not been involved in a traffic accident.

[0073] This embodiment does not impose any specific restrictions on the structure of the traffic accident recognition model. For example, the traffic accident recognition model can be a neural network model, a decision tree model, a random forest model, or an XGBoost (eXtreme Gradient Boosting) model, etc.

[0074] In an embodiment of the present application, the training process of the traffic accident recognition model can be: obtaining a training set, the training set includes multiple samples, for any sample, the input parameter of the sample is the characteristic parameter of the vehicle involved in the suspected accident determined by step 102, the label of the sample is an accident, and the traffic accident recognition model is trained using the training set to obtain a trained traffic accident recognition model.

[0075] The implementation methods of this application do not limit the training process of the traffic accident recognition model.

[0076] In an optional embodiment, at least one preset time period includes at least one time period before the parking moment and at least one time period after the parking moment, wherein the end moment of each time period in the at least one time period before the parking moment is the parking moment, and the start moment of each time period in the at least one time period after the parking moment is the parking moment.

[0077] For example, each time period before the parking moment is referred to as a pre-parking time period, and each time period after the parking moment is referred to as a post-parking time period.

[0078] In a possible implementation, the parameters in the feature parameter set include:

[0079] Average vehicle speed during the time period before the stop;

[0080] Maximum speed and minimum speed during the time period before stopping;

[0081] Speed ​​at the time of stopping;

[0082] Minimum deceleration during the period before parking;

[0083] Average deceleration during the period before parking;

[0084] Average speed during the period after the stop;

[0085] The maximum value of the pitch angle change rate of the vehicle in the time period before parking is obtained according to the vehicle's gyroscope;

[0086] The standard deviation of the rate of change of the vehicle's pitch angle during the period before the vehicle stops, obtained from the vehicle's gyroscope;

[0087] The maximum and minimum relative distances to the vehicle ahead during the time period before stopping;

[0088] Minimum relative speed to the vehicle ahead during the period before stopping;

[0089] The minimum value of the ratio of the relative distance to the preceding vehicle and the relative speed during the time period before stopping;

[0090] The ratio of relative distance to relative speed at the moment of stopping;

[0091] Turn signal status when parking;

[0092] The proportion of braking time in the time period before parking.

[0093] In an optional embodiment, the at least one preset time period includes a first preset duration, a second preset duration, and a third preset duration; the parking time is the end time of the third preset duration; the third preset duration is greater than the first preset duration; the characteristic parameter of the vehicle includes at least one item in a characteristic parameter set; and the parameters in the characteristic parameter set include:

[0094] The average speed of the vehicle during the first preset time period and the average speed of the vehicle during the third preset time period;

[0095] the maximum speed and the minimum speed of the vehicle during the first preset time period, and the maximum speed and the minimum speed of the vehicle during the third preset time period;

[0096] The speed of the vehicle at the end of the first preset time period and the third preset time period;

[0097] a minimum deceleration of the vehicle during a first preset time period and a minimum deceleration of the vehicle during a third preset time period;

[0098] an average deceleration of the vehicle during the first preset time period and an average deceleration of the vehicle during the third preset time period;

[0099] The average speed of the vehicle during the second preset time period;

[0100] a maximum value of a pitch angle change rate of the vehicle within a first preset time period and a maximum value of a pitch angle change rate within a third preset time period, obtained by a gyroscope of the vehicle;

[0101] A standard deviation of a rate of change of the vehicle's pitch angle within a first preset time period and a standard deviation of a rate of change of the vehicle's pitch angle within a third preset time period, obtained from a gyroscope of the vehicle;

[0102] the maximum relative distance and the minimum relative distance between the vehicle and the vehicle in front during the first preset time period, and the maximum relative distance and the minimum relative distance between the vehicle and the vehicle in front during the third preset time period;

[0103] the minimum relative speed of the vehicle with respect to the preceding vehicle within the first preset time period and the minimum relative speed of the vehicle with respect to the preceding vehicle within the third preset time period;

[0104] the minimum value of the ratio of the relative distance to the preceding vehicle to the relative speed within the first preset time period, and the minimum value of the ratio of the relative distance to the preceding vehicle to the relative speed within the third preset time period;

[0105] The ratio of the relative distance to the relative speed of the vehicle at the end of the first preset time period and the third preset time period;

[0106] The vehicle's turn signal status at the end of the first preset duration and the third preset duration;

[0107] The braking time ratio of the vehicle in the first preset time period and the braking time ratio in the third preset time period are used to represent the ratio of the time period in the braking state to the total time period.

[0108] For example, the first preset time period is 5 seconds before the parking time, the second preset time period is 15 minutes after the parking time, and the third preset time period is 25 seconds before the parking time. The moment when the vehicle speed drops to 0, that is, the parking time, is the end time of the first and third preset time periods.

[0109] Taking this as an example, the parameters in the above characteristic parameter set are explained.

[0110] The first parameter: the average speed of the vehicle during the first preset time period and the average speed during the third preset time period, i.e., the average speed of the vehicle during the 5 seconds before the stop and the average speed during the 25 seconds before the stop;

[0111] The second parameter: the maximum speed and minimum speed of the vehicle during the first preset time period, and the maximum speed and minimum speed of the vehicle during the third preset time period, i.e., the maximum speed and minimum speed of the vehicle 5 seconds before the stop, and the maximum speed and minimum speed of the vehicle 25 seconds before the stop;

[0112] The third parameter: the vehicle speed at the end of the first preset time period and the third preset time period, that is, the vehicle speed at the time of parking, which is usually 0 or a value very close to 0;

[0113] Fourth parameter: the minimum deceleration of the vehicle during the first preset time period and the minimum deceleration during the third preset time period, i.e., the minimum deceleration of the vehicle 5 seconds before the stop and the minimum deceleration 25 seconds before the stop;

[0114] The fifth parameter: the average deceleration of the vehicle during the first preset time period and the average deceleration during the third preset time period, i.e., the average deceleration of the vehicle 5 seconds before the stop and the average deceleration 25 seconds before the stop;

[0115] The sixth parameter: the average speed of the vehicle during the second preset time period, i.e., the average speed of the vehicle within 15 minutes after the stop;

[0116] The seventh parameter is the maximum rate of change of the vehicle's pitch angle within a first preset time period and the maximum rate of change of the vehicle's pitch angle within a third preset time period, as measured by the vehicle's gyroscope. The gyroscope is installed inside the vehicle and collects the vehicle's pitch angle at equal intervals, for example, every one second. In this parameter, the rate of change of the pitch angle is the difference between two consecutive pitch angles divided by the time interval.

[0117] Eighth parameter: the standard deviation of the vehicle's pitch angle change rate within a first preset time period and the standard deviation of the vehicle's pitch angle change rate within a third preset time period, as obtained by the vehicle's gyroscope, i.e., the standard deviation of the vehicle's pitch angle change rate within 5 seconds and within 25 seconds before the vehicle stops. The definition of the pitch angle change rate can be found in the explanation of the seventh parameter and is not repeated here.

[0118] Ninth parameter: the maximum relative distance and minimum relative distance between the vehicle and the vehicle ahead within the first preset time period, and the maximum relative distance and minimum relative distance between the vehicle and the vehicle ahead within the third preset time period. That is, the maximum relative distance and minimum relative distance between the vehicle and the vehicle ahead within 5 seconds before stopping, and the maximum relative distance and minimum relative distance between the vehicle and the vehicle ahead within 25 seconds before stopping;

[0119] The tenth parameter: the minimum relative speed of the vehicle with respect to the vehicle in front within the first preset time period and the minimum relative speed of the vehicle with respect to the vehicle in front within the third preset time period, i.e., the minimum relative speed of the vehicle with respect to the vehicle in front within 5 seconds before stopping and the minimum relative speed of the vehicle with respect to the vehicle in front within 25 seconds before stopping;

[0120] Parameter 11: The minimum ratio of the vehicle's relative distance to the preceding vehicle to its relative speed within the first preset time period, and the minimum ratio of the vehicle's relative distance to its relative speed within the third preset time period, i.e., the minimum TTA value 5 seconds before the vehicle stops and the minimum TTA value 25 seconds before the vehicle stops, where TTA = relative distance ÷ relative speed.

[0121] The twelfth parameter is the ratio of the relative distance to the relative speed of the vehicle at the end of the first preset time period and the third preset time period. This is the TTA value of the vehicle at the time of parking. It is usually a relatively large value. The concept of TTA can be found in the explanation of the eleventh parameter and will not be repeated here.

[0122] Thirteenth parameter: the vehicle's turn signal status at the end of the first preset time period and the third preset time period; that is, the vehicle's turn signal status when the vehicle is parked;

[0123] The fourteenth parameter is the ratio of the vehicle's braking time during the first preset duration and the third preset duration. The braking time ratio represents the ratio of the duration of the braking state to the total duration of the time period; that is, the ratio of the vehicle's braking time in the 5 seconds before the stop and the ratio of the braking time in the 25 seconds before the stop. The time the vehicle is in the braking state can be calculated by the time the brake pedal is depressed. For example, if the brake pedal is depressed for 4 seconds out of the 5 seconds before the stop, the braking time ratio in the 5 seconds before the stop is 0.8.

[0124] The embodiments of the present application analyze driving data from a time period before and after a vehicle stops to identify a vehicle that has engaged in abnormal parking behavior. The vehicle's geographic location at the time of parking is used to determine that the vehicle is in a suspected accident. Driving data from at least one time period before and after the vehicle stops is used to extract characteristic parameters that can be used to indicate that the vehicle has been in a traffic accident. By inputting these characteristic parameters into a pre-trained traffic accident recognition model, a determination result is obtained as to whether the vehicle has been in a traffic accident. The method provided by the embodiments of the present application determines whether a vehicle has been in an accident solely based on the vehicle's driving data, without requiring the acquisition of other data, such as large amounts of traffic data from other vehicles. This method not only enables the determination of whether a vehicle has been in a traffic accident under resource-constrained conditions, but also improves the accuracy of traffic accident recognition.

[0125] FIG2 shows a flowchart of another traffic accident identification method provided by an embodiment of the present application, which is described in detail as follows:

[0126] In step 201, based on first driving data of the vehicle for a first preset time period before the parking moment and second driving data of the vehicle for a second preset time period after the parking moment, it is determined that the vehicle has an abnormal parking behavior.

[0127] In step 202, for a vehicle that has abnormal parking behavior, it is determined that the vehicle has had a suspected accident based on the geographic location information of the vehicle at the time of parking.

[0128] In step 203, for a vehicle suspected of having an accident, characteristic parameters of the vehicle are extracted based on the driving data of the vehicle in at least one preset time period, where each time period in the at least one preset time period includes a parking moment.

[0129] In step 204, the feature parameters are input into a pre-trained traffic accident recognition model to obtain a recognition result, which includes whether the vehicle has been involved in a traffic accident or whether the vehicle has not been involved in a traffic accident.

[0130] Specifically, the specific implementation of steps 201 to 204 can refer to the specific implementation of steps 101 to 104 in the implementation method corresponding to Figure 1, and the implementation method of this application will not be repeated here.

[0131] In step 205, if the recognition result is that the vehicle has not been involved in a traffic accident, it is determined that the recognition result is wrong according to the accident list pushed by the insurance company.

[0132] If the identification result of step 204 indicates that the vehicle has not been involved in a traffic accident, but within a period of time after the identification result, an accident list is received from the insurance company, and if the accident list shows that the vehicle has been involved in an accident, it means that the vehicle has actually been involved in an accident. By obtaining the vehicle accident video and related message data, it is verified that the accident corresponding to the vehicle's current accident is the same event as the event identified in step 204. This indicates that the identification result of step 204 is incorrect.

[0133] In this step, the insurance company refers to the insurance company. In an optional implementation, other methods can be used to obtain whether the vehicle has encountered an accident, such as through the driver's active reporting. The implementation method of this application does not limit the method of obtaining the accident list.

[0134] In step 206, the first training set corresponding to the traffic accident recognition model is updated. The first training set includes a sample of a vehicle corresponding to the recognition result. The input parameters of the sample are characteristic parameters of the vehicle, and the label of the sample is that a traffic accident occurred.

[0135] By adding the misjudged cases as new samples to the first training set corresponding to the accident recognition model and retraining the accident recognition model, the recognition accuracy of the model can be effectively improved.

[0136] In step 207 , the traffic accident recognition model is retrained using the updated first training set.

[0137] In this step, there is no restriction on the training method of the model, and any training method that can be implemented based on existing technology can be used.

[0138] In the implementation mode of the present application, by obtaining the accident list of the vehicle, the misjudgment cases of the accident recognition model are obtained, the misjudgment cases of the model are added to the training set corresponding to the model, and the model is retrained to improve the recognition accuracy of the model.

[0139] FIG3 shows a flowchart of another traffic accident identification method provided in an embodiment of the present application, which is described in detail as follows:

[0140] In step 301, based on first driving data of the vehicle for a first preset time period before the parking moment and second driving data of the vehicle for a second preset time period after the parking moment, it is determined that the vehicle has an abnormal parking behavior.

[0141] In step 302, for a vehicle that has abnormal parking behavior, it is determined that the vehicle has had a suspected accident based on the geographic location information of the vehicle at the time of parking.

[0142] In step 303, for a vehicle suspected of having an accident, characteristic parameters of the vehicle are extracted based on the driving data of the vehicle in at least one preset time period, where each time period in the at least one preset time period includes a parking moment.

[0143] Specifically, the specific implementation of steps 301 to 303 can refer to the specific implementation of steps 101 to 103 in the implementation method corresponding to Figure 1, and the implementation method of this application will not be repeated here.

[0144] In step 304, all vehicles involved in accidents are obtained based on the accident list pushed by the insurance company; for any vehicle, the accident severity level of the vehicle is determined based on the amount of the vehicle's accident.

[0145] Among them, accidents are pre-divided into multiple different severity levels, and each insurance amount corresponds uniquely to a severity level.

[0146] For example, accidents with an insured amount or claim amount greater than or equal to 20,000 yuan are marked as severe accidents, while accidents with an insured amount or claim amount less than 20,000 yuan are marked as minor accidents. In other words, the severity of accidents is divided into two levels: the first level is severe accidents, and the second level is minor accidents.

[0147] It should be noted that the above is only an example, and the classification of accident severity levels can be set according to actual conditions, and the implementation methods of this application do not limit this.

[0148] In step 305, a second training set is established. In the second training set, each sample corresponds to an accident, the input parameters of the sample are the characteristic parameters of the vehicle involved in the accident, and the label of the sample is the severity level corresponding to the accident.

[0149] According to the result of step 304, each sample in the second training set is obtained.

[0150] In step 306, the accident severity discrimination model is trained using the second training set to obtain a trained accident severity discrimination model.

[0151] This embodiment does not impose any specific restrictions on the structure of the accident severity discrimination model. For example, the accident severity discrimination model can be a neural network model, a decision tree model, a random forest model, or an XGBoost (eXtreme Gradient Boosting) model, etc.

[0152] In one embodiment, the traffic accident identification model is an XGBoost model and the accident severity discrimination model is an extreme gradient boosting XGBoost model.

[0153] The XGBoost model is an ensemble learning method based on the CART (Classification and Regression Tree) classifier. Due to its excellent computational efficiency and predictive accuracy, it has been widely used in various intelligent scenarios. Unlike random forests, which assign equal voting weight to each decision tree, the generation of the next decision tree in XGBoost is dependent on the training and prediction of the previous decision tree (model accuracy is improved by assigning higher learning weights to samples with lower training accuracy in the previous round of decision trees). Compared to other ensemble learning algorithms, XGBoost improves model robustness by introducing regularization terms and column sampling, and also significantly increases model speed by employing a parallel strategy when selecting split points for each tree.

[0154] In step 307, the feature parameters are input into a pre-trained traffic accident recognition model to obtain a recognition result, which includes whether the vehicle has been involved in a traffic accident or whether the vehicle has not been involved in a traffic accident.

[0155] The specific implementation of this step can be found in step 104 of the implementation method corresponding to Figure 1, and the implementation method of this application will not be repeated here.

[0156] In step 308, if the recognition result is that the vehicle is involved in a traffic accident, the characteristic parameters of the vehicle are input into the trained accident severity discrimination model to obtain the severity level of the accident.

[0157] Typically, if the accident severity judgment model determines that the accident is a serious accident, it will be pushed to the relevant handling personnel to give priority to handling the accident so that the driver and passengers in the car can receive timely help and treatment.

[0158] The implementation method of the present application labels the severity level of the accident according to the vehicle accident situation, establishes a training set for the accident severity discrimination model and trains the accident severity discrimination model. When the traffic accident recognition model recognizes that a traffic accident has occurred, the characteristic parameters of the vehicle obtained based on the driving data of the vehicle before and after the accident are input into the accident severity discrimination model to obtain the severity level of the accident, so as to achieve priority handling of major and serious accidents, so that drivers and passengers of vehicles involved in serious traffic accidents can receive timely help and treatment.

[0159] FIG4 is a schematic diagram of a terminal provided in an embodiment of the present application. As shown in FIG4 , the terminal 6 of this embodiment includes: a processor 60, a memory 61, and a computer program 62 stored in the memory 61 and executable on the processor 60. When the processor 60 executes the computer program 62, the steps in the above-mentioned various traffic accident identification method implementations are implemented, such as steps 101 to 104 shown in FIG1 . The present application also provides a computer-readable storage medium, which can store a computer program, and when the computer program is executed by the processor, the steps in the above-mentioned various traffic accident identification method implementations are implemented, such as steps 101 to 104 shown in FIG1 . Exemplarily, the computer-readable storage medium can be a non-volatile readable storage medium.

Claims

1. A traffic accident identification method, characterized in that: include: Determining that the vehicle has abnormal parking behavior according to first driving data of the vehicle for a first preset time period before the parking time and second driving data of the vehicle for a second preset time period after the parking time, wherein the parking time is the end time of the first preset time period and the parking time is the start time of the second preset time period; For a vehicle that has abnormal parking behavior, determining that the vehicle has a suspected accident based on the geographical location information of the vehicle at the time of parking; For a vehicle suspected of having an accident, extracting characteristic parameters of the vehicle according to driving data of the vehicle in at least one preset time period, each time period in the at least one preset time period including the parking time; The characteristic parameters are input into a pre-trained traffic accident recognition model to obtain a recognition result, wherein the recognition result includes whether the vehicle has had a traffic accident or whether the vehicle has not had a traffic accident.

2. The traffic accident identification method according to claim 1, characterized in that: The vehicle starts to decelerate at the initial moment of the first preset time until it reaches the stopping time. The first driving data includes the vehicle speed at the initial moment of the first preset time and the deceleration of the vehicle during the first preset time. The second driving data includes the average speed of the vehicle during the second preset time.

3. The traffic accident identification method according to claim 2, characterized in that: Determining that the abnormal parking behavior of the vehicle occurs according to first driving data of the vehicle for a first preset time period before the parking moment and second driving data of the vehicle for a second preset time period after the parking moment includes: Determining whether emergency braking occurs on the vehicle according to the vehicle speed at the initial moment of the first preset time and the deceleration of the vehicle during the first preset time; If the vehicle undergoes emergency braking and the average speed of the vehicle during the second preset time period is lower than a preset speed threshold, it is determined that the vehicle has undergone abnormal parking behavior.

4. The traffic accident identification method according to claim 3, characterized in that: The determining whether the vehicle has emergency braking according to the vehicle speed at the initial moment of the first preset time and the deceleration of the vehicle during the first preset time comprises: According to the deceleration of the vehicle during the first preset time period, a speed interval corresponding to the deceleration is determined in a preset first judgment list, and whether the speed of the vehicle at the initial moment of the first preset time period belongs to the speed interval, and if so, it is determined that the vehicle has emergency braking, wherein the first judgment list includes a unique mapping relationship between the deceleration and the speed of the vehicle in an emergency braking situation; Alternatively, based on the deceleration of the vehicle during the first preset time period and the vehicle speed at the initial moment of the first preset time period, an index is performed in a preset second judgment list to determine the braking type of the vehicle, wherein the second judgment list includes a unique mapping relationship between vehicle deceleration, speed and braking type, and the braking type includes emergency braking and non-emergency braking.

5. The method according to claim 1, characterized in that For the vehicle that has abnormal parking behavior, determining that the vehicle has a suspected accident according to the geographical location information of the vehicle at the time of parking includes: According to the geographical location information of the vehicle at the parking time and a preset list of common parking points of the vehicle, it is determined whether the vehicle is located at a parking point in the list of common parking points, and if not, it is determined that a suspected accident has occurred in the vehicle, wherein the list of common parking points is a set of common parking point locations of the vehicle obtained according to historical driving data of the vehicle; Alternatively, based on the geographical location information of the vehicle at the time of parking, it is determined whether the vehicle is located in a road area, and if so, it is determined that a suspected accident has occurred with the vehicle.

6. The traffic accident identification method according to claim 5, characterized in that: The process of obtaining the list of frequent stopping points of the vehicle includes: Acquiring historical driving data of the vehicle; Marking the braking and stopping conditions of the vehicle in the historical driving data, wherein the braking and stopping conditions are that the speed of the vehicle drops to 0, and the time for which the speed is kept at 0 exceeds a preset time threshold; A list of frequent stopping points of the vehicle is established based on the geographical location information of the vehicle under each braking and stopping condition.

7. The traffic accident identification method according to claim 6, characterized in that: After establishing the list of frequent stopping points of the vehicle, the method further comprises: The parking spots in the frequent parking spot list are sorted in descending order of parking time to obtain a first queue; Entering an updating step, the updating step includes selecting the first frequent stop point in the first queue as the target frequent stop point in order, traversing other frequent stop points in the first queue, obtaining n frequent stop points whose distance from the target frequent stop point is less than or equal to a preset distance, constructing a new frequent stop point position according to the position information of the target frequent stop point and the n frequent stop points, and putting the new frequent stop point position into the second queue, and recording the number of stops corresponding to the new frequent stop point position as n+1 in the second queue, deleting the target frequent stop point and the n frequent stop points in the first queue, and if there is no frequent stop point whose distance from the target frequent stop point is less than or equal to the preset distance in the first queue, deleting the target frequent stop point in the first queue, wherein n is a positive integer greater than or equal to 1; Repeat the updating step until the number of frequent stop points in the first queue is zero, and use the frequent stop points in the second queue as the optimized frequent stop point list of the vehicle.

8. The traffic accident identification method according to any one of claims 1 to 7, characterized in that: The at least one preset time period includes the first preset time, the second preset time and the third preset time, the parking time is the end time of the third preset time, the third preset time is greater than the first preset time, and the characteristic parameter of the vehicle includes at least one item in a characteristic parameter set, and the parameters in the characteristic parameter set include: The average speed of the vehicle during the first preset time period and the average speed of the vehicle during the third preset time period; the maximum speed and the minimum speed of the vehicle during the first preset time period, and the maximum speed and the minimum speed of the vehicle during the third preset time period; The speed of the vehicle at the end of the first preset time period and the third preset time period; The minimum deceleration of the vehicle during the first preset time period and the minimum deceleration of the vehicle during the third preset time period; The average deceleration of the vehicle during the first preset time period and the average deceleration of the vehicle during the third preset time period; The average speed of the vehicle during the second preset time period; a maximum value of a pitch angle change rate of the vehicle within the first preset time period and a maximum value of a pitch angle change rate within the third preset time period obtained by the gyroscope of the vehicle; A standard deviation of a pitch angle change rate of the vehicle within the first preset time period and a standard deviation of a pitch angle change rate within the third preset time period obtained by a gyroscope of the vehicle; the maximum relative distance and the minimum relative distance between the vehicle and the vehicle ahead during the first preset time period, and the maximum relative distance and the minimum relative distance between the vehicle and the vehicle ahead during the third preset time period; The minimum relative speed of the vehicle with respect to the vehicle in front during the first preset time period and the minimum relative speed of the vehicle with respect to the vehicle in front during the third preset time period; the minimum value of the ratio of the relative distance to the preceding vehicle to the relative speed of the vehicle within the first preset time period, and the minimum value of the ratio of the relative distance to the preceding vehicle to the relative speed within the third preset time period; a ratio of the relative distance to the relative speed of the vehicle at the end of the first preset time period and the third preset time period; The turn signal status of the vehicle at the end of the first preset time period and the third preset time period; The braking time ratio of the vehicle in the first preset time period and the braking time ratio in the third preset time period, the braking time ratio is used to indicate the ratio of the duration in the braking state to the total duration of the time period.

9. The method according to any one of claims 1 to 7, characterized in that: If the recognition result is that the vehicle has not been involved in a traffic accident, the method further includes: According to the accident list pushed by the insurance company, it is determined that the identification result is wrong; Updating a first training set corresponding to the traffic accident recognition model, wherein the first training set includes a sample of a vehicle corresponding to the recognition result, the input parameter of the sample is a characteristic parameter of the vehicle, and the label of the sample is that a traffic accident occurred; The traffic accident recognition model is retrained using the updated first training set.

10. The method according to any one of claims 1 to 7, characterized in that: Before inputting the characteristic parameters into the pre-trained traffic accident recognition model to obtain the recognition result, the method further includes: According to the accident list pushed by the insurance company, obtain all vehicles involved in accidents; For any vehicle, the accident severity level of the vehicle is determined according to the amount of the vehicle insured, wherein the accident is pre-classified into a plurality of different severity levels, and each amount of the insured uniquely corresponds to a severity level; Establishing a second training set, in which each sample corresponds to an accident, the input parameter of the sample is a characteristic parameter of the vehicle involved in the accident, and the label of the sample is a severity level corresponding to the accident; The accident severity discrimination model is trained by using the second training set to obtain a trained accident severity discrimination model; If the recognition result is that the vehicle has a traffic accident, the method further includes: The characteristic parameters of the vehicle are input into the trained accident severity discrimination model to obtain the severity level of the accident.

11. A terminal comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 10 are implemented.

12. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 10 are implemented.

Citation Information

Patent Citations

  • Online hailing car abnormal stay behavior identification method and device, equipment and storage medium

    CN111860561A

  • System for judging vehicle risk scene based on GPS risk data

    CN114742635A

  • Data processing method and device, electronic equipment and vehicle

    CN114970897A

  • Accident vehicle detection method and device, electronic equipment and computer readable storage medium

    CN115272939A

  • Traffic accident prediction method and device, storage medium and electronic equipment

    CN115797403A

Cited By

  • Traffic accident auxiliary responsibility judgment method based on large language model

    CN119961611A

  • Vehicle safety state prediction and emergency call system and method

    CN120018100A

  • System and method for vehicle safety state prediction and emergency call

    CN120018100B

  • Abnormal slow vehicle detection method and abnormal slow vehicle detection system

    CN120260293A

  • Vehicle abnormal event real-time regulation and control method for urban complex road conditions

    CN121617248A