Vehicle accident data processing method and device and storage medium

Abstract

The invention provides a vehicle accident data processing method and device and a storage medium, and the method comprises the steps: S1, obtaining the vehicle state information of a vehicle within a preset time range before and after the accident time according to the accident time after a triggering event of an accident is received; s2, inputting the vehicle state information into a pre-trained vehicle accident analysis model, and judging whether the vehicle has an accident or not; s3, when it is judged that the vehicle has the accident, determining the latest effective parking time point; s4, according to the effective parking time point, judging whether at least one piece of alarm data of a predetermined type exists or not; if yes, determining the time corresponding to the first piece of alarm data of the pre-type as the accident occurrence time; otherwise, determining the time corresponding to the vehicle positioning record which is closest to the effective parking time point and has the vehicle speed greater than 0 as the accident occurrence time. By means of the technical scheme, the accident occurrence time can be determined more accurately.

Description

Technical Field

[0001] This invention relates to the field of intelligent transportation technology, and in particular to a method, apparatus, system, and storage medium for processing vehicle accident data. Background Technology

[0002] Vehicles are prone to traffic accidents while in motion due to factors such as the driver, vehicle condition, and road conditions. Knowing the time and location of an accident is of great significance for road rescue, traffic control, and insurance claims. In the past, to know the time and location of an accident, on-site personnel had to report it proactively. From a software perspective, this method is inconvenient for accident registration, and the authenticity of the accident is also uncontrollable. Summary of the Invention

[0003] To address the aforementioned problems, embodiments of the present invention provide a method, apparatus, system, and storage medium for processing vehicle accident data, so as to more accurately identify the specific time when a vehicle accident occurs.

[0004] To achieve the above objectives, on the one hand, a method for processing vehicle accident data is provided, including:

[0005] S1. After receiving the triggering event of the accident, according to the time of the accident event triggered by the triggering event, obtain the vehicle status information of the vehicle within a predetermined time range before and after the time of the accident event. The vehicle status information includes one or more of the following: vehicle speed, positioning location, ACC status, parking time, parking duration, acceleration, equipment alarm data and driver's driving behavior. The vehicle status information comes from the data collected by the vehicle video monitoring terminal.

[0006] S2, input the acquired vehicle status information into a pre-trained vehicle accident analysis model to determine whether the vehicle has been involved in an accident. The vehicle accident analysis model is pre-trained using historical vehicle status information and historical vehicle accident data.

[0007] S3, when the vehicle accident analysis model determines that the vehicle has been involved in an accident, determine the nearest valid parking time point before and after the time of the accident event;

[0008] S4. Based on the determined effective parking time point, determine whether there is at least one alarm data of a predetermined type within a predetermined time period before the effective parking time point; if so, determine the time corresponding to the first alarm data of the predetermined type as the accident occurrence time; otherwise, determine the time corresponding to the vehicle positioning record closest to the effective parking time point with a speed greater than 0 before the effective parking time point as the accident occurrence time.

[0009] Preferably, in the processing method, the triggering event for the occurrence of the accident includes:

[0010] Accident alarms collected from vehicle-mounted video surveillance terminals; and / or,

[0011] An accident report sent by the driver.

[0012] Preferably, in the processing method, the alarm data of the predetermined type includes: collision, rollover, and / or sudden deceleration.

[0013] Preferably, in the processing method, the vehicle accident analysis model includes a maximum likelihood estimation model, which calculates the probability of an accident based on the following maximum likelihood estimation formula:

[0014] P = (x1, x2, x3, x4) = f D (x1, x2, x3, x4|θ)

[0015] x1, x2, x3, and x4 are the four selected sample data, f D Let θ be the distribution function, and θ be the distribution parameter, where:

[0016] Sample x1 represents an independent event of sudden braking during vehicle operation;

[0017] Sample x2 represents independent events detected by the vehicle video surveillance terminal, such as collisions and rollovers.

[0018] Sample x3 is an independent event when the vehicle's acc state is 1 and the vehicle speed is 0;

[0019] Sample x4 represents an independent event where the vehicle's parking time exceeds the predetermined time.

[0020] During the training phase of the accident analysis model, sample data corresponding to samples x1, x2, x3, and x4 at the time of the historical accident are obtained through historical vehicle accident data. The optimal value of the distribution parameter θ is obtained, and the trained maximum likelihood estimation model is obtained. The optimal value of the distribution parameter θ makes the estimation result of the maximum likelihood estimation module close to the real accident to a predetermined threshold.

[0021] Preferably, in the processing method, the vehicle accident analysis model further includes a support vector machine model; wherein training the accident analysis model using historical vehicle state information and historical vehicle accident data includes:

[0022] The kernel function and parameters of the support vector machine model are selected by cross-validation and network search; the kernel function includes: linear kernel function, polynomial kernel function and / or radial basis kernel function.

[0023] Preferably, in the processing method, the vehicle accident analysis model further includes a random forest model; wherein training the accident analysis model using historical vehicle state information and historical vehicle accident data includes:

[0024] The random forest model is trained using labeled accident and non-accident datasets, and learns to distinguish between vehicle accidents and non-accidents through multiple iterations and splitting of the dataset.

[0025] Preferably, the processing method further includes one or more of the following:

[0026] The video recording file of the vehicle's parking point within a predetermined time range before and after the time of the accident is obtained from the vehicle-mounted video monitoring terminal.

[0027] The vehicle status information of the vehicle is obtained from the vehicle-mounted video monitoring terminal within a predetermined time range before and after the time of the accident, and one or more of the following statistical information are determined based on the vehicle status information: number of alarms, alarms per 100 kilometers, mileage, driving time, average daily speed and driving trajectory.

[0028] Based on the required time range, generate a visual report from the video files and / or the statistical information.

[0029] Preferably, the processing method further includes:

[0030] The location of the accident was determined based on the vehicle's location records at the time of the accident.

[0031] On the other hand, a vehicle accident data processing apparatus is also provided, including a memory and a processor, wherein the memory stores at least one program, which is executed by the processor to implement the vehicle accident data processing method as described above.

[0032] In another aspect, a computer-readable storage medium is also provided, wherein at least one program is stored therein, the at least one program being executed by a processor to implement the vehicle accident data processing method as described in any of the above.

[0033] The above technical solution has the following technical effects:

[0034] The technical solution of this invention, upon receiving a trigger event indicating an accident, uses real-time vehicle status-related data such as driver, vehicle, road conditions, and alarms collected by the in-vehicle video monitoring terminal. It then utilizes a pre-trained accident analysis model to further confirm whether an accident has actually occurred and to determine the time of the accident. This allows for more accurate judgment of the accident's occurrence and determination of its time, reduces false alarms and misreporting of vehicle accidents, and improves the management efficiency of vehicles such as ride-hailing vehicles.

[0035] In a further technical solution, by actively generating visual reports, these reports can be pushed to vehicles such as ride-hailing operating companies, facilitating the operation and management of ride-hailing vehicles. In emergency situations, they can also provide necessary assistance to ride-hailing vehicles and drivers involved in accidents, such as calling the police and providing roadside assistance. Attached Figure Description

[0036] Figure 1 This is a flowchart illustrating a method for processing vehicle accident data according to an embodiment of the present invention;

[0037] Figure 2 This is a schematic diagram of the workflow of the support vector machine model used in the embodiment;

[0038] Figure 3 This is a schematic diagram of the workflow of the random forest model used in the embodiment;

[0039] Figure 4 This is a schematic diagram of a vehicle accident data processing device according to an embodiment of the present invention. Detailed Implementation

[0040] To further illustrate the various embodiments, the present invention provides accompanying drawings. These drawings are part of the disclosure of the present invention, primarily used to illustrate the embodiments and to explain the operating principles of the embodiments in conjunction with the relevant descriptions in the specification. With reference to these drawings, those skilled in the art should be able to understand other possible implementations and the advantages of the present invention. Components in the drawings are not drawn to scale, and similar component symbols are generally used to represent similar components.

[0041] The present invention will now be further described in conjunction with the accompanying drawings and specific embodiments.

[0042] Example 1:

[0043] Figure 1 This is a flowchart illustrating a method for processing vehicle accident data according to an embodiment of the present invention. Figure 1 The vehicle accident data processing method in this embodiment includes the following steps:

[0044] S1, after receiving the triggering event of the accident, obtains the vehicle status information within a predetermined time range before and after the accident event time according to the time of the accident event triggered by the triggering event. The vehicle status information includes one or more of the following: vehicle speed, positioning location, ACC status, parking time, parking duration, acceleration, equipment alarm data and driver driving behavior. The vehicle status information comes from the data collected by the vehicle video monitoring terminal.

[0045] In one specific implementation, when the vehicle speed is less than a preset value or the ACC is turned off, the reporting time of the positioning data is marked as the parking time; the alarm data of the above-mentioned devices includes: alarm data of ADAS and DSM;

[0046] S2, input the acquired vehicle status information into the pre-trained vehicle accident analysis model to determine whether the vehicle has been involved in an accident. The vehicle accident analysis model is pre-trained using historical vehicle status information and historical vehicle accident data.

[0047] S3, when the vehicle accident analysis model determines that a vehicle has been involved in an accident, determine the nearest valid parking time point before and after the time of the accident event;

[0048] In one specific implementation, for the vehicle status information (i.e., the vehicle status dataset) used in response to the accident-triggered event, it preferably includes parking time, ADAS, and DSM alarm data, where the ADAS and DSM alarm data include alarm times. The parking time data is used in the alarm time dataset of the ADAS and DSM alarm data to employ a binary search algorithm to find valid parking time points that meet predetermined rules. These predetermined search rules include: when the error between a parking time point and the alarm time is within a preset time range, such as a preset number of seconds, the parking time point is considered a valid parking time point. If the binary search algorithm does not find a parking time point that meets the error between the parking time point and the alarm time, the parking time points in the parking time point dataset are sorted in order, and the earliest parking time is taken as the valid parking time point. For example, the preset number of seconds is 2 seconds; parking time points with a time difference of less than 2 seconds from the alarm time are determined as valid parking time points. Those skilled in the art will understand that, depending on the actual scenario and requirements, the preset number of seconds can be other time ranges.

[0049] S4. Based on the determined valid parking time point, determine whether there is at least one alarm data of a predetermined type within the predetermined time period before the valid parking time point; if so, determine the time corresponding to the first alarm data of the predetermined type as the accident time; otherwise, determine the time corresponding to the vehicle positioning record of the vehicle closest to the valid parking time point with a speed greater than 0 before the valid parking time point as the accident time.

[0050] Furthermore, the method also includes determining the location of the accident based on the determined time of the accident and the vehicle's location records.

[0051] In one implementation, the vehicle accident analysis model includes a maximum likelihood estimation model, which calculates the probability of an accident based on the following maximum likelihood estimation formula:

[0052] P = (x1, x2, x3, x4) = f D (x1, x2, x3, x4|θ)

[0053] x1, x2, x3, and x4 are the four selected sample data, f D Let θ be the distribution function, and θ be the distribution parameter, where:

[0054] Sample x1 represents an independent event of sudden braking during vehicle operation;

[0055] Sample x2 represents independent events detected by the vehicle video surveillance terminal, such as collisions and rollovers.

[0056] Sample x3 is an independent event when the vehicle's acc state is 1 and the vehicle speed is 0;

[0057] Sample x4 represents an independent event where a vehicle is parked for more than 5 minutes;

[0058] During the training phase of the accident analysis model, sample data corresponding to samples x1, x2, x3, and x4 at the time of the historical accident are obtained through historical vehicle accident data. By adjusting the distribution parameter, the optimal value of the distribution parameter θ is obtained, and the trained maximum likelihood estimation model is obtained. The optimal value of the distribution parameter θ makes the estimation result of the maximum likelihood estimation module close to the real accident to a predetermined threshold.

[0059] Example 2:

[0060] The following is an exemplary description of a specific implementation of an embodiment of the present invention.

[0061] In embodiments of the present invention, upon receiving a trigger event indicating an accident involving a vehicle, such as a ride-hailing vehicle, the system can collect data from an onboard video surveillance terminal installed in the vehicle. Based on alarm videos and images before and after the accident trigger event, and by fusing features such as vehicle status characteristics, device alarm characteristics, and surrounding environmental features like road conditions, the system uses algorithms and decision trees to determine the actual time of the accident. Then, according to business needs, the system sends instructions to the terminal to upload data such as surrounding environment videos, vehicle status, and driver status information recorded by ADAS and DSM within a certain time range to the system, generating a visualized accident report for subsequent accident analysis.

[0062] Vehicles such as ride-hailing vehicles are equipped with Advanced Driving Assistance Systems (ADAS) and Driver State Monitoring Systems (DSM) onboard video surveillance terminals to monitor the vehicle's surroundings. The DSM can be used to detect the driver's state. The onboard video surveillance terminal primarily collects real-time data on the vehicle and driver, including speed, location, direction, ACC status, device alarm data, and driver behavior. The ACC status indicates the vehicle's power status: ACC = 0 indicates the off state (no power), and ACC = 1 indicates the on state (power is on).

[0063] A data acquisition system will be built to collect data from vehicle-mounted video surveillance terminals in real time and extract features of the collected vehicle data, including vehicle status (such as vehicle speed, ACC status, parking time, and parking duration), device alarm features, and surrounding environmental features (such as road conditions). The data will then be stored, for example, in Elasticsearch.

[0064] After obtaining the data, preferably, it is preprocessed, including data cleaning, standardization, and normalization, to remove invalid data, duplicate data, and outliers, thereby improving data quality and consistency. Simultaneously, the obtained data is labeled to clarify which data pertains to vehicle accidents and which to non-accidents, thus obtaining a historical accident database for training the accident analysis model.

[0065] Supported by a large amount of pre-collected historical accident data, and based on the maximum likelihood estimation method, combined with support vector machines and random forest models, an accident analysis model is finally obtained. The accident analysis model includes a maximum likelihood estimation model, a support vector machine model, and / or a random forest model. The maximum likelihood estimation model, support vector machine model, and / or random forest model can be used individually to determine whether an accident has occurred based on input real-time vehicle status data or historical vehicle status data and historical accident data; they can also be used in combination to determine whether an accident has occurred based on input real-time vehicle status data or historical vehicle status data and historical accident data. The following sections explain each of these three models.

[0066] 1) Maximum likelihood estimation model; wherein, the maximum likelihood estimation formula used is shown in the following equation:

[0067] P=(x1,x2,x3,x4,…,xn)=f D (x1, x2, x3, x4..., xn|θ)

[0068] Where x1, x2, x3, x4, ..., xn are the n selected sample data, f D Let θ be the distribution function, and θ be the distribution parameter.

[0069] In one specific implementation, the following maximum likelihood formula is used:

[0070] P = (x1, x2, x3, x4) = f D (x1, x2, x3, x4|θ)

[0071] In this implementation, the probability of an accident occurring is calculated using four types of sample data and the distribution parameter θ:

[0072] in:

[0073] Sample x1 represents an independent event of sudden braking during vehicle operation;

[0074] Sample x2 represents independent events detected by the vehicle video surveillance terminal, such as collisions and rollovers.

[0075] Sample x3 is an independent event when the vehicle's acc state is 1 and the vehicle speed is 0; when acc is 1, it means that the vehicle starts normally and the vehicle power supply is on.

[0076] Sample x4 is an independent event where the vehicle is parked for a period of time exceeding a predetermined duration; the predetermined duration can be, for example, 5 minutes or other shorter or longer durations.

[0077] In other implementations of this invention, more independent events can be set that can be used for the maximum likelihood estimation model.

[0078] By using historical vehicle accident data, sample data at the time of the accident is obtained. Then, by modifying the value of the distribution parameter θ, the maximum likelihood estimation result is made to best approximate the actual accident. Finally, the optimal value of the distribution parameter θ is determined. In specific implementation, the optimal value of the distribution parameter θ is the value that makes the estimation result of the maximum likelihood estimation module approach the actual accident at a predetermined threshold.

[0079] 2) Support Vector Machine (SVM) model: This model uses the SVM algorithm to extract vehicle state information strongly correlated with vehicle accidents from preprocessed data, including vehicle speed, acceleration / deceleration (ACC) status, stopping time, stopping duration, acceleration, direction, and location. In this process, it's necessary to select appropriate kernel functions and parameters to optimize the accident analysis model's performance. Kernel functions include linear kernel functions, polynomial kernel functions, and radial basis function kernel functions. Selecting appropriate kernel functions and parameters can be achieved through methods such as cross-validation and grid search. Real-time vehicle data is then matched with the trained model to determine whether an accident has occurred. Figure 2 This is a schematic diagram illustrating the workflow related to the Support Vector Machine (SVM) model. Figure 2 The workflow for support vector machine models includes:

[0080] Choose an appropriate kernel function;

[0081] Find a separating hyperplane in the feature space that maximizes the distance between positive and negative samples;

[0082] Input a large amount of pre-collected historical data, determine its category based on its position in the feature space, and verify the accuracy of the learning model.

[0083] The support vector machine algorithm used in the support vector machine model is an existing technology algorithm, and the specific use of the algorithm will not be elaborated here.

[0084] 3) Random Forest Model Using the Random Forest Algorithm. The random forest model is trained using labeled accident and non-accident datasets from pre-collected historical data. Through multiple iterations and dataset splits, the model learns to distinguish between vehicle accidents and non-accidents. Specifically, a portion of the collected historical data is extracted as the training set, and the remaining samples are used as the test set. Then, for each training sample, it selects some random data for training, thereby constructing a decision tree. Finally, all decision trees vote to determine the final prediction result, judging whether an accident has occurred. The specific workflow of the random forest model is as follows: Figure 3 As shown. Figure 3 It uses n training samples, each training sample corresponding to a decision tree.

[0085] The random forest model uses the existing random forest algorithm, which will not be described in detail here.

[0086] In this embodiment of the invention, the triggering events for an accident include: the vehicle-mounted video monitoring terminal collecting an alarm for the accident; and / or, the driver actively reporting the accident.

[0087] Based on the accident event time triggered by the above two scenarios, vehicle status data within a predetermined time range before and after the accident event time is extracted, including: driving speed, ACC status, parking time, parking duration, acceleration, direction, and / or location. A pre-trained accident analysis model determines whether an accident has occurred based on this data. The nearest valid parking time point before and after the accident event time is found, for example, through a divide-and-conquer algorithm using recursion. Based on the obtained valid parking time points, the actual accident time is obtained by comprehensively analyzing the following scenarios: All alarms or alerts of a predetermined type are retrieved from the parking time point prior to the predetermined time; where the predetermined time can be 5 minutes, or other longer or shorter durations; the predetermined type of alarm data includes: alarm data of predetermined hazard types, such as: collision, rollover, sudden deceleration, etc. In one implementation, the above alarm data is obtained from an in-vehicle video monitoring terminal. When alarm data of the aforementioned predetermined type exists, the time of the first alarm data is determined as the time of the accident. When alarm data of the aforementioned predetermined type does not exist, the time corresponding to the nearest location data with a speed greater than 0 before the aforementioned valid parking time point is taken as the time of the accident. Based on the time of the accident, the location of the accident can then be determined based on the location data or location records, because the location data or location records usually include the corresponding vehicle speed, location, and location time.

[0088] In one implementation, after determining the time of the incident, the method further includes: automatically collecting relevant video and statistical data; specifically including:

[0089] 1) Automated collection of video files: Based on the time of the accident, obtain video files and other resource data of the parking point where the accident occurred on the vehicle video monitoring terminal within a selected time range;

[0090] 2) Automated data processing: Based on the time of the accident, obtain vehicle status information from the data collection system of vehicles such as ride-hailing vehicles, including the vehicle's own status information, equipment alarms, and driver status, and determine one or more of the following statistical information: alarm count, alarms per 100 kilometers, mileage, driving time, average daily speed, and driving trajectory.

[0091] In one implementation, the system also includes generating a visual report from the obtained vehicle status information, video files, and / or statistical information based on the required time frame, such as the time frame required by business needs, and presenting it in an intuitive way.

[0092] In one specific implementation, the information displayed in the visual report includes one or more of the following:

[0093] (1) Alarm analysis: During the vehicle's operation, data collected by the vehicle video monitoring terminal, including alarm data such as fatigue driving, frequent lane changes, and forward collision alarms, are used to calculate the number of alarms and alarm distribution of the vehicle.

[0094] (2) 100km alarm: First calculate the number of vehicle alarms, then multiply by 100, and then divide by the total mileage of the vehicle within the statistical period.

[0095] (3) Mileage: The number of kilometers the vehicle traveled within the statistical period;

[0096] (4) Driving time: The duration of time a driver drives a vehicle within the statistical period;

[0097] (5) Average daily speed: During the statistical period, the total mileage of the vehicle is divided by the total driving time.

[0098] (6) Alarm video files: Based on the accurate time of the identified accident, the alarm video files are automatically obtained 15 minutes before the accident time to facilitate the tracing of the accident.

[0099] (7) Video files: Based on the accurate time of the identified accident, automatically obtain video files 10 minutes before the accident time to facilitate the tracing of the accident.

[0100] (8) Driving trajectory: The driving trajectory is drawn in map form using the location data.

[0101] Visualized reports can be presented in the form of charts, graphs, videos, and / or maps. They allow users to quickly, intuitively, vividly, and easily understand drivers' recent driving behavior and vehicle operation status, helping them better understand the data, grasp the operational situation, and analyze whether drivers have any dangerous driving behaviors or weak safety awareness. Ride-hailing companies can then conduct targeted safety training.

[0102] In other implementations of the embodiments of the present invention, one or more of the following may also be included:

[0103] (1) Add error correction and deviation correction functions for accident time and location to improve the accuracy of the system in extracting the time and location of the accident;

[0104] (2) Dynamically configure the parameters of the visualization report. Through dynamic configuration, different types of visualization report data can be exported to adapt to the business needs of different regions.

[0105] (3) For vehicles that have recently been involved in traffic accidents, such as ride-hailing vehicles, they are included in the list of key vehicles. The operation status of the accident vehicles for two consecutive days is pushed out at 9:00 am every day. The data mainly includes: average daily speed, average daily driving time, average daily mileage, number of alarms per 100 kilometers, number of alarms, etc., to facilitate further observation of the accident vehicles.

[0106] Example 3:

[0107] The present invention also provides a vehicle accident data processing device, such as... Figure 4 As shown, the device includes a processor 401, a memory 402, a bus 403, and a computer program stored in the memory 402 and executable on the processor 401. The processor 401 includes one or more processing cores. The memory 402 is connected to the processor 401 via the bus 403. The memory 402 is used to store program instructions. When the processor executes the computer program, it implements the steps in the above-described method embodiment of Embodiment 1 of the present invention.

[0108] Furthermore, as an executable solution, the vehicle accident data processing device can be a computer unit, which can be a desktop computer, laptop, handheld computer, or cloud server, etc. The computer unit may include, but is not limited to, a processor and memory. Those skilled in the art will understand that the above-described computer unit structure is merely an example and does not constitute a limitation on the computer unit. It may include more or fewer components, or combine certain components, or use different components. For example, the computer unit may also include input / output devices, network access devices, buses, etc., and this embodiment of the invention does not limit this.

[0109] Furthermore, as an executable solution, the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc. The processor is the control center of the computer unit, connecting various parts of the entire computer unit via various interfaces and lines.

[0110] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the computer unit by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system and at least one application program required for a function; the data storage area may store data created based on the use of the mobile phone, etc. In addition, the memory may include high-speed random access memory and may also include non-volatile memory, such as hard disk, RAM, plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0111] Example 4:

[0112] The present invention also 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 embodiments of the present invention.

[0113] If the modules / units integrated in the computer unit are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), and software distribution media, etc. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction.

[0114] Although the invention has been specifically shown and described in conjunction with preferred embodiments, those skilled in the art should understand that various changes in form and detail may be made to the invention without departing from the spirit and scope of the invention as defined in the appended claims, all of which shall be within the scope of protection of the invention.

Claims

1. A method for processing vehicle accident data, characterized in that, include: S1. After receiving the triggering event of the accident, according to the time of the accident event triggered by the triggering event, obtain the vehicle status information of the vehicle within a predetermined time range before and after the time of the accident event. The vehicle status information includes one or more of the following: vehicle speed, positioning location, ACC status, parking time, parking duration, acceleration, equipment alarm data and driver's driving behavior. The vehicle status information comes from the data collected by the vehicle video monitoring terminal. S2, input the acquired vehicle status information into a pre-trained vehicle accident analysis model to determine whether the vehicle has been involved in an accident. The vehicle accident analysis model is pre-trained using historical vehicle status information and historical vehicle accident data. S3, when the vehicle accident analysis model determines that the vehicle has been involved in an accident, determine the nearest valid parking time point before and after the time of the accident event; S4. Based on the determined effective parking time point, determine whether there is at least one alarm data of a predetermined type within a predetermined time period before the effective parking time point; if so, determine the time corresponding to the first alarm data of the predetermined type as the accident occurrence time; otherwise, determine the time corresponding to the vehicle positioning record closest to the effective parking time point with a speed greater than 0 before the effective parking time point as the accident occurrence time.

2. The processing method according to claim 1, characterized in that, The events that triggered the accident include: Accident alarms collected from vehicle-mounted video surveillance terminals; and / or, An accident report sent by the driver.

3. The processing method according to claim 1, characterized in that, The predetermined types of alarm data include: collision, rollover, and / or sudden deceleration.

4. The processing method according to claim 1, characterized in that, The vehicle accident analysis model includes a maximum likelihood estimation model, which calculates the probability of an accident based on the following maximum likelihood estimation formula: P=(x1,x2,x3,x4)=f D (x1,x2,x3,x4|θ) x1, x2, x3, and x4 are the four selected sample data, f D Let θ be the distribution function, and θ be the distribution parameter, where: Sample x1 represents an independent event of sudden braking during vehicle operation; Sample x2 represents independent events detected by the vehicle video surveillance terminal, such as collisions and rollovers. Sample x3 is an independent event when the vehicle's acc state is 1 and the vehicle speed is 0; Sample x4 represents an independent event where the vehicle's parking time exceeds the predetermined time. During the training phase of the accident analysis model, sample data corresponding to samples x1, x2, x3, and x4 at the time of the historical accident are obtained through historical vehicle accident data. The optimal value of the distribution parameter θ is obtained, and the trained maximum likelihood estimation model is obtained. The optimal value of the distribution parameter θ makes the estimation result of the maximum likelihood estimation module close to the real accident to a predetermined threshold.

5. The processing method according to claim 4, characterized in that, The vehicle accident analysis model further includes a support vector machine model; wherein training the accident analysis model using historical vehicle state information and historical vehicle accident data includes: The kernel function and parameters of the support vector machine model are selected by cross-validation and network search; the kernel function includes: linear kernel function, polynomial kernel function and / or radial basis kernel function.

6. The processing method according to claim 4, characterized in that, The vehicle accident analysis model also includes a random forest model; wherein training the accident analysis model using historical vehicle state information and historical vehicle accident data includes: The random forest model is trained using labeled accident and non-accident datasets, and learns to distinguish between vehicle accidents and non-accidents through multiple iterations and splitting of the dataset.

7. The processing method according to claim 1, characterized in that, It also includes one or more of the following: The video recording file of the vehicle's parking point within a predetermined time range before and after the time of the accident is obtained from the vehicle-mounted video monitoring terminal. The vehicle status information of the vehicle is obtained from the vehicle-mounted video monitoring terminal within a predetermined time range before and after the time of the accident, and one or more of the following statistical information are determined based on the vehicle status information: number of alarms, alarms per 100 kilometers, mileage, driving time, average daily speed and driving trajectory. Based on the required time range, the video files and / or the statistical information are used to generate a visual accident report.

8. The processing method according to claim 1, characterized in that, Also includes: The location of the accident was determined based on the vehicle's location records at the time of the accident.

9. A device for processing vehicle accident data, characterized in that, The system includes a memory and a processor, the memory storing at least one program, which is executed by the processor to implement the vehicle accident data processing method as described in any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that, The storage medium stores at least one program, which is executed by a processor to implement the vehicle accident data processing method as described in any one of claims 1 to 8.

Images