Data processing method, device, equipment and storage medium

Abstract

The present application provides a data processing method, device, equipment and storage medium, firstly, using a drone to collect available video data containing images of a moving vehicle, then obtaining standard trajectory data of the moving vehicle based on the available video data, and obtaining driving characteristic data based on the standard trajectory data of the moving vehicle, and then training an initial driving scene classification model based on the driving characteristic data to obtain a target driving scene classification model. By collecting available video data through a drone, and obtaining driving characteristic data of the moving vehicle based on the available video data collected by the drone for model training, a target driving scene classification model for driving scene classification is obtained based on machine learning means and the method of using drones to collect information, which overcomes the defects and drawbacks of driving information collection in the prior art and the shortcomings of using indicator statistics to classify driving scenes, and improves classification accuracy, intelligence and classification efficiency.

Description

Technical Field

[0001] The present application relates to the field of autonomous driving technology, and in particular to a data processing method, device, equipment and storage medium. Background Art

[0002] At present, most of the research on autonomous driving is to collect information about the vehicle, the surrounding environment, the movement trajectory of surrounding vehicles and pedestrians, and the weather and lighting during driving by vehicles equipped with cameras and radars. The collected information is then imported into the autonomous driving simulation platform to restore the driving environment and simulate driving.

[0003] However, for the current form of collecting information through vehicles, only when driving and ensuring that various sensor data indicators are working normally can the collection work be carried out normally and accurately. Even so, since driving is the premise of collection work, there will inevitably be defects and disadvantages in driving collection. For example, the driving environment and driving mileage that can be collected each time are limited. But if multiple vehicles are used for collection at the same time, it will bring the problem of expensive collection costs. For example, although the information required for some driving scenes can be collected based on the car, the danger and cost of the information collection process are very high. For example, collect information related to dangerous driving scenes such as dangerous lane changes, rear-end collisions, accidents, etc. that may be encountered in daily driving. In addition, if the collected information is to support the use of autonomous driving simulation, the driving scene classification must be carried out according to this information. Moreover, the current methods of constructing driving scene classification using the information collected by driving are all achieved through indicator statistics, and the classification accuracy and intelligence are insufficient, and the efficiency is relatively low.

[0004] It can be seen that a solution is urgently needed for the above-mentioned defects and disadvantages of the current method of collecting information while driving and using the collected information to classify driving scenes. Summary of the invention

[0005] The present application provides a data processing method, device, equipment and storage medium, aiming to solve the defects and disadvantages of the prior art in collecting information while driving and classifying driving scenes using the information collected while driving.

[0006] In a first aspect, the present application provides a data processing method, comprising:

[0007] Collecting available video data by using a drone, wherein the available video data includes images of moving vehicles;

[0008] Obtaining standard trajectory data of the traveling vehicle according to the available video data, and obtaining driving characteristic data of the traveling vehicle according to the standard trajectory data of the traveling vehicle;

[0009] The initial driving scene classification model is trained according to the driving characteristic data of the driving vehicle to obtain a target driving scene classification model.

[0010] In a possible design, after obtaining the target driving scene classification model, the method further includes:

[0011] The target driving scenario classification model is used to classify the driving data of the target vehicle, and the classified driving scenarios are used for the simulation platform to simulate the autonomous driving simulation.

[0012] In a possible design, the collecting of available video data by using a drone includes:

[0013] Using the drone to obtain raw video data within a preset height range of a target road area;

[0014] The original video data is screened for video content to determine the original video data containing the image of the moving vehicle as the available video data.

[0015] In a possible design, obtaining the standard trajectory data of the moving vehicle according to the available video data includes:

[0016] Performing image jitter repair processing on the available video data to obtain standard image data;

[0017] Performing vehicle extraction and trajectory tracking processing on the standard image data to obtain trajectory data of the moving vehicle;

[0018] The trajectory data of the traveling vehicle is subjected to coordinate transformation to obtain standard trajectory data of the traveling vehicle, and the coordinate transformation is used for data conversion between the relative coordinate system of the UAV and the earth coordinate system.

[0019] In a possible design, performing image jitter repair processing on the available video data to obtain standard image data includes:

[0020] Performing the UAV standard positioning point conversion on the available video data according to the positioning information of the UAV, wherein the standard positioning point conversion is used to convert the reference acquisition point when acquiring the available video data into the standard acquisition point, so as to implement the image jitter repair processing on the available video data;

[0021] The available video data converted into the standard acquisition point is determined as the standard image data.

[0022] In a possible design, the coordinate conversion of the trajectory data of the traveling vehicle to obtain the standard trajectory data of the traveling vehicle includes:

[0023] Performing the coordinate transformation on the first coordinate matrix data according to the coordinate information of the UAV and the coordinate information of the key points in the target road area to obtain second coordinate matrix data in a geodetic coordinate system, wherein the first coordinate matrix data is used to represent the trajectory data of the moving vehicle in the relative coordinate system of the UAV;

[0024] The trajectory data of the moving vehicle represented by the second coordinate matrix is ​​determined as the standard trajectory data of the moving vehicle.

[0025] In a possible design, obtaining the driving characteristic data of the driving vehicle according to the standard trajectory data of the driving vehicle includes:

[0026] Acquiring characteristic data of each traveling vehicle at a current trajectory point according to the standard trajectory data of the traveling vehicle, and structuring the characteristic data to obtain traveling characteristic data of the traveling vehicle;

[0027] The driving characteristic data of the driving vehicles include the speed, acceleration, steering angle, vehicle number and vehicle size of each driving vehicle.

[0028] In a possible design, the training of the initial driving scene classification model according to the driving characteristic data of the driving vehicle to obtain the target driving scene classification model includes:

[0029] The initial driving scene classification model is trained according to the driving characteristic data of the driving vehicle and a plurality of classified driving scenes to obtain an intermediate driving scene classification model, wherein the plurality of classified driving scenes include a straight driving scene, a turning driving scene, a lane changing driving scene, an overtaking driving scene, a following driving scene, an avoidance driving scene, and a first dangerous driving scene;

[0030] The target driving scene classification model is obtained by training the intermediate driving scene classification model according to the second dangerous driving scene and the driving characteristic data of the accident vehicle in the second dangerous driving scene.

[0031] In a possible design, after obtaining the target driving scene classification model, the method further includes:

[0032] Analyzing the target driving scene classification model according to the model interpretation tool to obtain a collision feature data set, wherein the collision feature data set includes a data set consisting of a plurality of distance critical values ​​between two vehicles, a data set consisting of a plurality of vehicle speed critical values, and a data set consisting of a plurality of relative speed critical values ​​between two vehicles;

[0033] Determining a collision feature data set that is positively correlated with the dangerous driving scene classified by the target driving scene classification model as a positive excitation sample;

[0034] Determine a collision feature data set that is negatively correlated with the dangerous driving scene classified by the target driving scene classification model as a negative stimulus sample;

[0035] Visually displaying the positive excitation samples and the negative excitation samples to obtain a target data range, and obtaining collision parameters using a collision feature data set within the target data range;

[0036] The collision parameters include the following vehicle distance and the collision avoidance time.

[0037] In a possible design, after obtaining the collision parameter, the method further includes:

[0038] Optimizing the target driving scene classification model according to the reacquired driving characteristic data of the driving vehicle;

[0039] The collision parameters are optimized according to the optimized target driving scenario classification model.

[0040] In a second aspect, the present application provides a data processing device, including:

[0041] An acquisition module, used to collect available video data through a drone, wherein the available video data includes images of moving vehicles;

[0042] A processing module, used to obtain standard trajectory data of the traveling vehicle according to the available video data, and obtain driving characteristic data of the traveling vehicle according to the standard trajectory data of the traveling vehicle;

[0043] The training module is used to train the initial driving scene classification model according to the driving characteristic data of the driving vehicle to obtain a target driving scene classification model.

[0044] In a possible design, the data processing device further includes: a classification module; the classification module is used to:

[0045] The target driving scenario classification model is used to classify the driving data of the target vehicle, and the classified driving scenarios are used for the simulation platform to simulate the autonomous driving simulation.

[0046] In a possible design, the acquisition module is specifically used to:

[0047] Using the drone to obtain raw video data within a preset height range of a target road area;

[0048] The original video data is screened for video content to determine the original video data containing the image of the moving vehicle as the available video data.

[0049] In a possible design, the processing module includes:

[0050] A first processing submodule, configured to perform image jitter repair processing on the available video data to obtain standard image data;

[0051] A second processing submodule is used to perform vehicle extraction and trajectory tracking processing on the standard image data to obtain trajectory data of the moving vehicle;

[0052] The third processing submodule is used to perform coordinate conversion on the trajectory data of the moving vehicle to obtain standard trajectory data of the moving vehicle. The coordinate conversion is used for data conversion between the relative coordinate system of the UAV and the earth coordinate system.

[0053] In a possible design, the first processing submodule is specifically configured to:

[0054] Performing a standard positioning point conversion of the drone on the available video data according to the positioning information of the drone, wherein the standard positioning point conversion is used to convert a reference acquisition point when acquiring the available video data into a standard acquisition point, so as to implement the image jitter repair processing on the available video data;

[0055] The available video data converted into the standard acquisition point is determined as the standard image data.

[0056] In a possible design, the third processing submodule is specifically used to:

[0057] Performing the coordinate transformation on the first coordinate matrix data according to the coordinate information of the UAV and the coordinate information of the key points in the target road area to obtain second coordinate matrix data in a geodetic coordinate system, wherein the first coordinate matrix data is used to represent the trajectory data of the moving vehicle in the relative coordinate system of the UAV;

[0058] The trajectory data of the traveling vehicle represented by the second coordinate matrix data is determined as the standard trajectory data of the traveling vehicle.

[0059] In a possible design, the processing module further includes: a fourth processing submodule; the fourth processing submodule is used to:

[0060] Acquiring characteristic data of each traveling vehicle at a current trajectory point according to the standard trajectory data of the traveling vehicle, and structuring the characteristic data to obtain traveling characteristic data of the traveling vehicle;

[0061] The driving characteristic data of the driving vehicles include the speed, acceleration, steering angle, vehicle number and vehicle size of each driving vehicle.

[0062] In a possible design, the training module includes:

[0063] A first training submodule is used to train the initial driving scene classification model according to the driving characteristic data of the driving vehicle and a plurality of classified driving scenes to obtain an intermediate driving scene classification model, wherein the plurality of classified driving scenes include a straight driving scene, a turning driving scene, a lane changing driving scene, an overtaking driving scene, a following driving scene, an avoidance driving scene, and a first dangerous driving scene;

[0064] The second training submodule is used to train the intermediate driving scene classification model according to the second dangerous driving scene and the driving characteristic data of the accident vehicle in the second dangerous driving scene to obtain the target driving scene classification model.

[0065] In a possible design, the processing module is further used for:

[0066] Analyzing the target driving scene classification model according to the model interpretation tool to obtain a collision feature data set, wherein the collision feature data set includes a data set consisting of a plurality of distance critical values ​​between two vehicles, a data set consisting of a plurality of vehicle speed critical values, and a data set consisting of a plurality of relative speed critical values ​​between two vehicles;

[0067] Determining a collision feature data set that is positively correlated with the dangerous driving scene classified by the target driving scene classification model as a positive excitation sample;

[0068] Determine a collision feature data set that is negatively correlated with the dangerous driving scene classified by the target driving scene classification model as a negative stimulus sample;

[0069] The positive excitation samples and the negative excitation samples are visualized to obtain a target data range, and collision parameters are obtained using a collision feature data set within the target data range.

[0070] In a possible design, the data processing device further includes: an optimization module; the optimization module is used to:

[0071] Optimizing the target driving scene classification model according to the reacquired driving characteristic data of the driving vehicle;

[0072] The collision parameters are optimized according to the optimized target driving scenario classification model.

[0073] In a third aspect, the present application provides an electronic device, comprising: a processor, and a memory communicatively connected to the processor;

[0074] The memory stores computer-executable instructions;

[0075] The processor executes the computer-executable instructions stored in the memory to implement any possible data processing method provided in the first aspect.

[0076] In a fourth aspect, the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are executed by a processor, they are used to implement any possible data processing method provided in the first aspect.

[0077] In a fifth aspect, the present application provides a computer program product, comprising computer execution instructions, which, when executed by a processor, are used to implement any possible data processing method provided in the first aspect.

[0078] The present application provides a data processing method, device, equipment and storage medium. First, available video data is collected by a drone, and the collected available video data includes images of moving vehicles. Then, standard trajectory data of the moving vehicle is obtained based on the available video data, and driving characteristic data of the moving vehicle is obtained based on the standard trajectory data of the moving vehicle. Then, the initial driving scene classification model is trained based on the driving characteristic data of the moving vehicle to obtain a target driving scene classification model. Available video data is collected by a drone, and the driving characteristic data of the moving vehicle is obtained based on the available video data collected by the drone for model training, and a target driving scene classification model for driving scene classification is obtained, thereby obtaining a target driving scene classification model based on machine learning means and using drones to collect information, overcoming the defects and disadvantages of driving for information collection in the prior art and the shortcomings of using indicator statistics for driving scene classification, and improving classification accuracy and intelligence as well as classification efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0080] Figure 1 A schematic diagram of an application scenario provided for an embodiment of the present application;

[0081] Figure 2A flowchart of a data processing method provided in an embodiment of the present application;

[0082] Figure 3 A flowchart of another data processing method provided in an embodiment of the present application;

[0083] Figure 4 A flowchart of another data processing method provided in an embodiment of the present application;

[0084] Figure 5 A flowchart of another data processing method provided in an embodiment of the present application;

[0085] Figure 6 A flowchart of another data processing method provided in an embodiment of the present application;

[0086] Figure 7 A flowchart of another data processing method provided in an embodiment of the present application;

[0087] Figure 8 A schematic diagram of the structure of a data processing device provided in an embodiment of the present application;

[0088] Fig. 9 A schematic diagram of the structure of another data processing device provided in an embodiment of the present application;

[0089] Fig.10 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0090] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of methods and devices consistent with some aspects of the present application as detailed in the appended claims.

[0091] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein, for example. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0092] At present, most of the research on autonomous driving is done by collecting information while driving, which has many defects and disadvantages. Moreover, if the collected information is to support the use of autonomous driving simulation, the driving scenes need to be classified. In addition, in the prior art, the classification of driving scenes using the information collected by driving is achieved through the method of indicator statistics, which makes the classification accuracy and intelligence insufficient, and the classification efficiency is relatively low.

[0093] In view of the above problems existing in the prior art, the present application provides a data processing method, device, equipment and storage medium. The inventive concept of the data processing method provided in the present application is: using a drone to collect images containing moving vehicles to obtain available video data, and then first obtaining the standard trajectory data of the moving vehicle based on the available video data, and then obtaining the driving characteristic data based on the standard trajectory data, and finally using the driving characteristic data of the moving vehicle to train the initial driving scene classification model to obtain a target driving scene classification model that can be used for driving scene classification. The target driving scene classification model for driving scene classification is obtained based on machine learning means and the use of drones for information collection, which overcomes the defects and disadvantages of driving for information collection in the prior art and the shortcomings of using indicator statistics for driving scene classification, and improves the classification accuracy and intelligence as well as the classification efficiency.

[0094] The following is an introduction to exemplary application scenarios of the embodiments of the present application.

[0095] Figure 1 A schematic diagram of an application scenario provided in an embodiment of the present application, such as Figure 1As shown, the drone 10 can be equipped with a high-definition camera to collect video data within a preset height range of the target road area. The target road area can be, for example, a road where vehicles 20 with different characteristics travel, such as a highway, an inter-city expressway, a service area, an urban area, and other areas that allow the drone 10 to collect images. The preset height range is, for example, different heights within a range of 100 meters to 200 meters in the sky. The drone 10 can collect images of moving vehicles in the video data. The electronic device 30 is configured to execute the data processing method provided in the embodiment of the present application, first obtaining the available video data collected by the drone 10, then obtaining the standard trajectory data of the moving vehicle based on the available video data, and obtaining the driving feature data based on the standard trajectory data, and then using the driving trajectory feature data of the moving vehicle to train the initial driving scene classification model to obtain the target driving scene classification model. The target driving scene classification model can classify the driving data of the target vehicle, and the classified driving scenes can be used for the autonomous driving simulation platform 40 to perform simulation operations of autonomous driving simulation.

[0096] It should be noted that the model of the drone 10 can be set according to actual working conditions, and the embodiment of the present application does not limit this. The electronic device 30 can be a computer, a server or a server cluster, and the embodiment of the present application does not limit this. Figure 1 The electronic device 30 in the figure is shown by taking a computer as an example.

[0097] It should be noted that the above application scenarios are merely illustrative, and the data processing methods, devices, equipment, and storage media provided in the embodiments of the present application include but are not limited to the above application scenarios.

[0098] Figure 2 A flow chart of a data processing method provided in an embodiment of the present application. Figure 2 As shown, the data processing method provided in the embodiment of the present application includes:

[0099] S101: Collect usable video data via drone.

[0100] The available video data includes images of moving vehicles.

[0101] The image is captured by using a drone equipped with a high-definition camera, and usable video data is obtained based on the video data of the image of the moving vehicle captured by the drone.

[0102] In a possible design, step S101 may be implemented as follows: Figure 3 shown. Figure 3 A flow chart of another data processing method provided in an embodiment of the present application. Figure 3 As shown, the embodiment of the present application includes:

[0103] S1011: Using a drone to obtain raw video data within a preset height range of a target road area.

[0104] A drone equipped with a high-definition camera shoots within a preset height range in an area where vehicles are traveling and drone shooting is allowed, and the captured video data is defined as raw video data.

[0105] The target road area is the area where vehicles are traveling and drone photography is allowed, such as roads where vehicles of different characteristics are traveling, such as expressways, inter-city expressways, service areas, urban areas, etc. The preset altitude range can include different altitudes within the range of 100 meters to 200 meters above the ground.

[0106] S1012: Screening the original video data for video content, so as to determine the original video data containing images of a moving vehicle as usable video data.

[0107] The video content of the original video data captured by the drone is screened to filter out the original video data containing images of moving vehicles, and the filtered original video data is determined as usable video data, and the video content of the usable video data contains images of moving vehicles.

[0108] It is understandable that if images containing moving vehicles cannot be screened out from the original video data, the original video data is recaptured using a drone.

[0109] The data processing method provided in the embodiment of the present application uses a drone to shoot at a preset height range in the target road area, and screens the video content of the captured raw video data to screen out the raw video data containing images of moving vehicles, and determines the screened raw video data as available video data, so as to realize the use of drones to collect information and avoid the defects and deficiencies of collecting information by driving. In addition, the use of drones for information collection can collect a large amount of driving behavior data in a short time, which is conducive to obtaining the data required for model training based on the available video data.

[0110] S102: Obtain standard trajectory data of the moving vehicle according to the available video data.

[0111] After obtaining available video data, standard trajectory data such as vehicle GPS data that can characterize the driving trajectory of the moving vehicle is obtained based on the available video data, so as to obtain data such as the speed and acceleration of the moving vehicle.

[0112] For example, obtaining standard trajectory data of a moving vehicle based on available video data can be achieved by performing image jitter repair processing, vehicle extraction and trajectory tracking processing, and coordinate conversion preprocessing on the available video data.

[0113] Image jitter repair processing is to locate and repair the available video data to ensure that the image in the video content is not distorted, deformed, inconsistent, etc. Vehicle extraction and trajectory tracking processing is to extract the moving vehicle and track its trajectory to ensure that the moving vehicle in consecutive video frames is the same moving vehicle. Coordinate transformation is to convert the relative position of the moving vehicle based on the drone's perspective into the real position in physical space.

[0114] In a possible design, a possible implementation of step S102 is as follows: Figure 4 shown. Figure 4 A flowchart of another data processing method provided in an embodiment of the present application is shown below. Figure 4 As shown, the embodiment of the present application includes:

[0115] S1021: Perform image jitter repair processing on available video data to obtain standard image data.

[0116] S1022: Perform vehicle extraction and trajectory tracking processing on the standard image data to obtain trajectory data of the moving vehicle.

[0117] S1023: Perform coordinate conversion on the trajectory data of the moving vehicle to obtain standard trajectory data of the moving vehicle.

[0118] Among them, coordinate transformation is used for data conversion between the relative coordinate system of the drone and the earth coordinate system.

[0119] For example, firstly, image jitter repair processing is performed on the available video data, and the processed result is determined as standard image data. In other words, the image in the standard image data does not have distortion, deformation, inconsistency, etc. The standard image data is further subjected to vehicle extraction and trajectory tracking processing, and the processed result is determined as the trajectory data of the moving vehicle.

[0120] Among them, the trajectory data of the moving vehicle can be, for example, the position and movement trajectory of the moving vehicle relative to the perspective of the drone, and these positions and movement trajectories are not the real positions and movement trajectories of the moving vehicle. Therefore, the trajectory data of the moving vehicle needs to be converted into coordinates to obtain the real position and movement trajectory of the moving vehicle, that is, the standard trajectory data of the moving vehicle.

[0121] The coordinate conversion processing method is used to convert data between the relative coordinate system of the drone and the earth coordinate system. Through the coordinate conversion processing, the position and movement trajectory relative to the drone's perspective can be converted into the real position and movement trajectory relative to the real physical space, that is, the standard trajectory data of the moving vehicle. The relative coordinate system of the drone is used to represent the drone's perspective, and the earth coordinate system is used to represent the perspective of the real physical space.

[0122] The data processing method provided in the embodiment of the present application performs image jitter repair processing, vehicle extraction and trajectory tracking processing, and coordinate conversion on the available video data to obtain standard trajectory data of the moving vehicle. The driving characteristic data obtained by using the standard trajectory data of the moving vehicle can then be used to perform model training based on the driving characteristic data.

[0123] S103: Obtaining driving characteristic data of the driving vehicle according to the standard trajectory data of the driving vehicle.

[0124] The standard trajectory data of the moving vehicles may be, for example, GPS data. The characteristic data of each moving vehicle at the current trajectory point may be obtained based on the standard trajectory data of the moving vehicles. The characteristic data may be, for example, the speed, acceleration, steering angle, vehicle number, and vehicle size of each moving vehicle.

[0125] In order to make the feature data have a highly unified data structure for the convenience of model training, the feature data is structured to obtain the driving feature data of the driving vehicle. In other words, the feature data and the driving feature data include the same data, but the driving feature data is structured data, which is represented by a unified data structure compared to the feature data, such as represented by an Excel format.

[0126] S104: Training the initial driving scene classification model according to the driving characteristic data to obtain a target driving scene classification model.

[0127] The driving characteristic data is used as a training sample to train the initial driving scene classification model to obtain a target driving scene classification model. The initial driving scene classification model may be a classification model with preliminary driving scene classification capabilities. For example, the driving scene corresponding to the driving characteristic data may be manually annotated with reference to the image of the driving vehicle in the video data corresponding to the driving characteristic data, and the original classification model may be trained using the manually annotated driving scene and the driving characteristic data to obtain an initial driving scene classification model with preliminary driving scene classification capabilities.

[0128] The initial driving scene classification model is trained to obtain a target driving scene classification model, and the obtained target driving scene classification model is used to classify the driving data of the target vehicle, thereby realizing the classification of driving scenes based on machine learning. Different from the existing technology of driving scene classification through indicator statistics, it can effectively improve the classification accuracy, intelligence and classification efficiency.

[0129] Among them, the original classification model used to obtain the initial driving scene classification model can be, for example, LR (Logistic Regression), LightGBM (Light Gradient Boosting Machine, gradient boosting machine algorithm), xgboost (eXtreme Gradient Boosting, gradient boosting decision tree) and other models, which are not limited in the embodiments of the present application.

[0130] The data processing method provided in the embodiment of the present application first collects available video data through a drone, and the collected available video data contains images of moving vehicles, then obtains standard trajectory data of the moving vehicle based on the available video data, and obtains driving characteristic data of the moving vehicle based on the standard trajectory data, and trains the initial driving scene classification model based on the driving characteristic data of the moving vehicle to obtain a target driving scene classification model. Available video data is collected through a drone, and the driving characteristic data of the moving vehicle is obtained based on the available video data collected by the drone for model training, and a target driving scene classification model for driving scene classification is obtained. The classification model for driving scene classification is obtained based on machine learning means and the method of using drones to collect information, which overcomes the defects and disadvantages of driving for information collection in the prior art and the shortcomings of using indicator statistics for driving scene classification, and improves classification accuracy and intelligence as well as classification efficiency.

[0131] Figure 5 A flowchart of another data processing method provided in an embodiment of the present application is shown below. Figure 5 As shown, the data processing method provided in the embodiment of the present application includes:

[0132] S201: Collect available video data through drones.

[0133] The available video data includes images of moving vehicles.

[0134] The possible implementation method, principle and technical effect of step S201 are similar to the possible implementation method, principle and technical effect of step S101. For details, please refer to the above description and will not be repeated here.

[0135] S202a: Convert the available video data to the standard positioning point of the drone according to the positioning information of the drone. S202b: Determine the available video data after conversion to the standard acquisition point as the standard image data.

[0136] The standard positioning point conversion is used to convert the reference acquisition points when acquiring available video data into standard acquisition points, so as to implement image jitter repair processing on the available video data.

[0137] After obtaining available video data, image jitter repair processing is performed on the available video data to obtain standard image data.

[0138] Since the drone may be affected by environmental factors such as temperature, wind force, and wind speed during the shooting process, which may cause the collection position points to shift during the shooting process, it is necessary to convert the standard positioning points of the drone for the available video data based on the drone's positioning information, and convert the reference collection points when collecting the available video data into standard collection points.

[0139] The conversion from the reference acquisition point to the standard acquisition point can be implemented by performing jitter correction through the drone's own hardware module, or by manual jitter correction. For example, through manual inspection, the video clips with jitter in the available video data are selected, and then the video data of the acquisition points before and after the jitter occurs are replaced to convert the inconsistent, distorted, and deformed images that appear in the acquisition process into standard images.

[0140] It can be understood that the reference acquisition point and the standard acquisition point are essentially the same as the selected acquisition point position in the picture to be shot by the drone. Due to the jitter in shooting, there is an offset between the reference acquisition point and the standard acquisition point in the available video data, so it is necessary to convert the reference acquisition point to the standard acquisition point, that is, to convert the standard positioning point of the drone to achieve image jitter repair processing for the available video data. Then, the available video data after the reference acquisition point is converted to the standard acquisition point is determined as the corresponding result of the image jitter repair processing, that is, the standard image data.

[0141] S203a: extracting the vehicle image from the standard image data using an object recognition and positioning algorithm to obtain a moving vehicle.

[0142] S203b: Track the position of the moving vehicle based on the target tracking algorithm to obtain the moving trajectory of the moving vehicle.

[0143] S203c: Determine the location and movement trajectory of the traveling vehicle as trajectory data of the traveling vehicle.

[0144] After completing the image jitter repair process on the available video data, the obtained standard image data is subjected to vehicle extraction and trajectory tracking processes to obtain the trajectory data of the moving vehicle.

[0145] For example, firstly, an object recognition and positioning algorithm is used to extract the vehicle image in the standard image data to extract the moving vehicle. In other words, the object recognition and positioning algorithm is used to extract the target object in the standard image data, and the target object is the moving vehicle in the available video data. The object recognition and positioning algorithm can use, for example, the YOLO (You Only Look Once) algorithm or other target detection algorithms, and the embodiments of the present application do not limit the specific content of the object recognition and positioning algorithm.

[0146] After extracting the moving vehicle, the position of the extracted moving vehicle is tracked based on the target tracking algorithm, and the moving trajectory of the current moving vehicle is obtained by ensuring that the moving vehicle in the consecutive video frames is the same moving vehicle, thereby obtaining the moving trajectory of the moving vehicle. The target tracking algorithm can be, for example, the DeepSort (Detection Based Tracking) algorithm.

[0147] The location and movement trajectory of the moving vehicle are then determined as trajectory data of the moving vehicle.

[0148] S204a: Perform coordinate transformation on the first coordinate matrix data according to the positioning information of the UAV and the coordinate information of the key points in the target road area to obtain second coordinate matrix data in the geodetic coordinate system.

[0149] The first coordinate matrix data is used to represent the trajectory data of the moving vehicle in the relative coordinate system of the UAV.

[0150] S204b: Determine the trajectory data of the moving vehicle represented by the second coordinate matrix data as the standard trajectory data of the moving vehicle.

[0151] Since the trajectory data of the moving vehicle obtained from the available video data is based on the perspective of the drone, the position and movement trajectory of the moving vehicle are not the real position and movement trajectory in the real physical space. Therefore, it is necessary to perform coordinate conversion on the trajectory data of the moving vehicle. The converted trajectory data of the moving vehicle is the standard trajectory data of the moving vehicle, so that the trajectory data of the moving vehicle can be matched with the map data in the real physical space, so as to know the position and trajectory points of the moving vehicle in the real physical space, such as the geographic coordinates of the moving vehicle and the lane it is in.

[0152] For example, the trajectory data of a moving vehicle can be transformed according to the positioning information of the UAV and the coordinate information of key points in the target road area. That is, the relative coordinate system of the UAV is transformed into the earth coordinate system, and the converted coordinate matrix data is the second coordinate matrix data. The trajectory data of the moving vehicle represented by the second coordinate matrix data is determined as the standard trajectory data of the moving vehicle.

[0153] Among them, the trajectory data of the moving vehicle in the drone coordinate system is represented by the first coordinate matrix data. In the relative coordinate system of the drone, the drone is the origin, the right side of the drone is the horizontal axis, and the top is the vertical axis. The second coordinate matrix data represents the trajectory data of the moving vehicle in the geodetic coordinate system, and the data represented by the second coordinate matrix data is defined as the standard trajectory data of the moving vehicle. The position on the ground in the geodetic coordinate system is represented by geodetic longitude, geodetic latitude and geodetic height. In addition, the positioning information of the drone is used to represent the location of the drone.

[0154] Optionally, the coordinate information of key points in the target road area may be the coordinate information of a reference object selected during the drone shooting process. The embodiment of the present application does not limit the specific content of the selected reference object, for example, it may be the geometric center of a moving vehicle, the left headlight or the right headlight of a moving vehicle, etc.

[0155] S205: Obtaining driving characteristic data of the driving vehicle according to the standard trajectory data of the driving vehicle.

[0156] The possible implementation method, principle and technical effect of step S205 are similar to the possible implementation method, principle and technical effect of step S103. For details, please refer to the above description and will not be repeated here.

[0157] S206a: Training an initial driving scene classification model according to the driving characteristic data of the driving vehicle and a plurality of classified driving scenes to obtain an intermediate driving scene classification model.

[0158] Among them, multiple classified driving scenarios include straight driving scenarios, turning driving scenarios, lane changing driving scenarios, overtaking driving scenarios, following driving scenarios, avoidance driving scenarios and the first dangerous driving scenarios.

[0159] The initial driving scene classification model has the preliminary driving scene classification capability. The driving feature data of the moving vehicle is used to predict the driving scene, and the prediction results are multiple classified driving scenes, which may include but are not limited to straight driving scenes, turning driving scenes, lane changing driving scenes, overtaking driving scenes, following driving scenes, avoiding driving scenes, and the first dangerous driving scene. Then, the driving feature data and multiple classified driving scenes are used as training samples to train the initial driving scene classification model to obtain an intermediate driving scene classification model.

[0160] S206b: Training the intermediate driving scene classification model according to the second dangerous driving scene and the driving characteristic data of the accident vehicle in the second dangerous driving scene to obtain a target driving scene classification model.

[0161] In order to enhance the classification accuracy of the intermediate driving scene classification model for dangerous driving scenes, the intermediate driving scene classification model is further trained using dangerous driving data, wherein the dangerous driving data includes the second dangerous driving scene and the driving characteristic data of the vehicle in which the dangerous event occurs in the second dangerous driving scene, i.e., the accident vehicle.

[0162] Optionally, the second dangerous driving scenario refers to a driving scenario in which a dangerous event occurs, such as a car crash, etc. The second dangerous driving scenario may come from a third data source, and the embodiment of the present application does not limit the source of the second dangerous driving scenario. The driving characteristic data of the accident vehicle in the second dangerous driving scenario may include data such as the speed, acceleration, steering angle, vehicle number, and vehicle size of the accident vehicle. It should be noted that the driving characteristic data of the accident vehicle is also structured data.

[0163] The above-mentioned model training process can be any process that can realize model training. Specifically, the corresponding training process can be set according to the characteristics of the selected original classification model. The embodiment of the present application does not limit the specific training process.

[0164] In a possible design, the second dangerous driving scenario and the driving characteristic data of the accident vehicle are obtained as follows: Figure 6 shown. Figure 6 A flowchart of another data processing method provided in an embodiment of the present application is shown below. Figure 6 As shown, the embodiment of the present application includes:

[0165] S301: Acquire accident video data corresponding to a driving scene where a dangerous event occurs.

[0166] The video data corresponding to the driving scene where the dangerous event occurs is obtained, and the video data is the accident video data, and the accident video data may come from a third data source.

[0167] S302: Acquire standard trajectory data of the accident vehicle according to the accident video data, and obtain driving characteristic data of the accident vehicle according to the standard trajectory data of the accident vehicle.

[0168] The accident video data may be collected by a drone, and the standard trajectory data of the accident vehicle may be acquired based on the accident video data by adopting steps such as steps S202a to S204b, and the standard trajectory data of the accident vehicle may be acquired by adopting steps such as step S103 to obtain the driving characteristic data of the accident vehicle based on the standard trajectory data of the accident vehicle.

[0169] It should be noted that when selecting accident video data, the selected video clips may include video clips of a preset length before the dangerous event occurs. For example, video clips 30 seconds before the collision are also part of the accident video data, which is beneficial to improving the sensitivity of the trained model in classifying driving scenes and also helps to improve the classification accuracy.

[0170] S207: Using the target driving scenario classification model to classify the driving data of the target vehicle, the classified driving scenarios are used for simulating the autonomous driving simulation on the simulation platform.

[0171] The target driving scene classification model is used to classify the driving data of the target vehicle to classify the driving scene of the target vehicle. The classified driving scene can be used for the autonomous driving simulation platform to simulate the autonomous driving simulation to provide data support for the autonomous driving simulation.

[0172] The target vehicle is the vehicle to be classified into driving scenarios, and the driving data is the video data representing the driving behavior of the target vehicle. The simulation platform runs the autonomous driving simulation software for simulation tests such as driving prototypes and driving simulations.

[0173] The data processing method provided in the embodiment of the present application collects available video data through a drone, and obtains standard trajectory data of a moving vehicle based on the available video data collected by the drone and then obtains driving characteristic data, trains an initial driving scene classification model according to the driving characteristic data to obtain a target driving scene classification model, and uses the target driving scene classification model to classify the driving data of the target vehicle, and the classification results are used for the simulation of automated driving simulation, which overcomes the deficiencies in data collection and driving scene classification in the prior art, improves classification accuracy, intelligence and classification efficiency, and provides strong support for automated driving technology.

[0174] After obtaining the target driving scene classification model, the target driving scene classification model can also be used to determine the collision parameters in the dangerous driving scenes classified by it and optimize the collision parameters, so that the determined collision parameters have better accuracy and controllability than the collision parameters obtained through empirical values ​​in the prior art.

[0175] In a possible design, after obtaining the target driving scene classification model, the following can also be included: Figure 7 Steps shown. Figure 7 A flowchart of another data processing method provided in an embodiment of the present application is shown below. Figure 7 As shown, the embodiment of the present application includes:

[0176] S401: Analyze the target driving scene classification model according to the model interpretation tool to obtain a collision feature data set.

[0177] The collision feature data set includes a data set consisting of a plurality of critical values ​​of the distance between two vehicles, a data set consisting of a plurality of critical values ​​of the vehicle speeds, and a data set consisting of a plurality of critical values ​​of the relative speeds of the two vehicles.

[0178] For example, when the target driving scene classification model classifies the driving scene each time, the model interpretation tool Shap is used to analyze the target driving scene classification model, and a shap value can be obtained each time. The shap value obtained each time can include a critical value of the distance between the two vehicles, a critical value of the vehicle speed, and a critical value of the relative speed of the two vehicles. During the multiple operations of the target driving scene classification model, multiple critical values ​​of the distance between the two vehicles, multiple critical values ​​of the vehicle speed, and multiple critical values ​​of the relative speed of the two vehicles can be obtained. The data set consisting of multiple critical values ​​of the distance between the two vehicles, the data set consisting of multiple critical values ​​of the vehicle speed, and the data set consisting of multiple critical values ​​of the relative speed of the two vehicles are collectively referred to as a collision feature data set.

[0179] S402: Determine a collision feature data set that is positively correlated with a dangerous driving scene classified by the target driving scene classification model as a positive excitation sample.

[0180] S403: Determine a collision feature data set that is negatively correlated with the dangerous driving scene classified by the target driving scene classification model as a negative excitation sample.

[0181] S404: Visually display the positive excitation samples and the negative excitation samples to obtain a target data range, and obtain collision parameters using the collision feature data set within the target data range.

[0182] Among them, the collision parameters include the following distance and the collision avoidance time.

[0183] After obtaining the collision feature data set, the critical values ​​of the distance between the two vehicles, the critical values ​​of the vehicle speed, and the critical values ​​of the relative speed of the two vehicles in the collision feature data set are screened according to the dangerous driving scenarios classified by the target driving scenario classification model. For example, positive excitation samples and negative excitation samples are obtained respectively and visualized to facilitate screening out the range of the critical values ​​of the distance between the two vehicles, the critical values ​​of the vehicle speed, and the critical values ​​of the relative speed of the two vehicles that meet the requirements, i.e., the target data range, and then the critical values ​​of the distance between the two vehicles, the critical values ​​of the vehicle speed, and the relative speed of the two vehicles in the selected range are used to calculate the collision parameters, which include the time headway (THW) and the time to collision (TTC).

[0184] The data processing method provided in the embodiment of the present application utilizes model training to obtain a target driving scene classification model, and then determines the collision parameters based on the target driving scene classification model and the data distribution of the data used to calculate the collision parameters. This is different from the prior art in which the collision parameters are set based on empirical values, and improves the accuracy and controllability of the classification of dangerous driving scenes.

[0185] Optionally, after the collision parameters are obtained, they can be further optimized. For example, the target driving scene classification model is first optimized according to the re-acquired driving characteristic data of the driving vehicle, and then the collision parameters are optimized according to the optimized target driving scene classification model to further improve accuracy and controllability.

[0186] Figure 8 A schematic diagram of the structure of a data processing device provided in an embodiment of the present application. Figure 8 As shown, the data processing device 400 provided in the embodiment of the present application includes:

[0187] An acquisition module 401 is used to collect available video data through a drone, where the available video data includes images of moving vehicles;

[0188] The processing module 402 is used to obtain standard trajectory data of the moving vehicle according to the available video data, and obtain driving characteristic data of the moving vehicle according to the standard trajectory data of the moving vehicle;

[0189] The training module 403 is used to train the initial driving scene classification model according to the driving characteristic data of the driving vehicle to obtain a target driving scene classification model.

[0190] exist Figure 8 On the basis of Fig. 9 This is a schematic diagram of the structure of another data processing device provided in an embodiment of the present application. Fig. 9 As shown, the data processing device 400 provided in the embodiment of the present application further includes: a classification module 404. The classification module 404 is used to:

[0191] The target driving scenario classification model is used to classify the driving data of the target vehicle, and the classified driving scenarios are used for the simulation platform to simulate the autonomous driving simulation.

[0192] In a possible design, the acquisition module 401 is specifically used to:

[0193] Use drones to obtain raw video data within a preset height range of the target road area;

[0194] The original video data is screened for video content to determine the original video data containing images of a moving vehicle as usable video data.

[0195] In one possible design, the processing module 402 includes:

[0196] The first processing submodule is used to perform image jitter repair processing on the available video data to obtain standard image data;

[0197] The second processing submodule is used to perform vehicle extraction and trajectory tracking processing on the standard image data to obtain trajectory data of the moving vehicle;

[0198] The third processing submodule is used to perform coordinate conversion on the trajectory data of the moving vehicle to obtain standard trajectory data of the moving vehicle. The coordinate conversion is used for data conversion between the relative coordinate system of the UAV and the earth coordinate system.

[0199] In a possible design, the first processing submodule is specifically used for:

[0200] The standard positioning point conversion of the drone is performed on the available video data according to the positioning information of the drone. The standard positioning point conversion is used to convert the reference acquisition point when acquiring the available video data into the standard acquisition point, so as to implement image jitter repair processing on the available video data;

[0201] The available video data after conversion to the standard acquisition point is determined as the standard image data.

[0202] In a possible design, the third processing submodule is specifically used for:

[0203] The first coordinate matrix data is converted according to the coordinate information of the UAV and the coordinate information of the key points in the target road area to obtain the second coordinate matrix data in the earth coordinate system, and the first coordinate matrix data is used to represent the trajectory data of the moving vehicle in the relative coordinate system of the UAV;

[0204] The trajectory data of the moving vehicle represented by the second coordinate matrix data is determined as the standard trajectory data of the moving vehicle.

[0205] In a possible design, the processing module 402 further includes: a fourth processing submodule. The fourth processing submodule is used to:

[0206] Acquire characteristic data of each traveling vehicle at a current trajectory point according to the standard trajectory data of the traveling vehicle, and structure the characteristic data to obtain traveling characteristic data of the traveling vehicle;

[0207] The driving characteristic data of the driving vehicles include the speed, acceleration, steering angle, vehicle number and vehicle size of each driving vehicle.

[0208] In one possible design, the training module 403 includes:

[0209] A first training submodule is used to train an initial driving scene classification model according to driving characteristic data of a moving vehicle and a plurality of classified driving scenes to obtain an intermediate driving scene classification model, wherein the plurality of classified driving scenes include a straight driving scene, a turning driving scene, a lane changing driving scene, an overtaking driving scene, a following driving scene, an avoidance driving scene, and a first dangerous driving scene;

[0210] The second training submodule is used to train the intermediate driving scene classification model according to the second dangerous driving scene and the driving characteristic data of the accident vehicle in the second dangerous driving scene to obtain a target driving scene classification model.

[0211] In one possible design, the processing module 402 is further configured to:

[0212] The target driving scene classification model is analyzed according to the model interpretation tool to obtain a collision feature data set, which includes a data set consisting of a plurality of distance critical values ​​between two vehicles, a data set consisting of a plurality of vehicle speed critical values, and a data set consisting of a plurality of relative speed critical values ​​between two vehicles;

[0213] A collision feature dataset that is negatively correlated with the dangerous driving scene classified by the target driving scene classification model is determined as a positive stimulus sample;

[0214] A collision feature dataset that is negatively correlated with the dangerous driving scene classified by the target driving scene classification model is determined as a negative stimulus sample;

[0215] The positive excitation samples and the negative excitation samples are visualized to obtain the target data range, and the collision parameters are obtained using the collision feature data set within the target data range.

[0216] In a possible design, the data processing device 400 further includes an optimization module. The optimization module is used to:

[0217] Optimizing the target driving scene classification model according to the reacquired driving characteristic data of the moving vehicle;

[0218] The collision parameters are optimized according to the optimized target driving scenario classification model.

[0219] The data processing device provided in the embodiment of the present application can execute the corresponding steps of the data processing method in the above method embodiment. Its implementation principle and technical effects are similar and will not be repeated here.

[0220] Fig.10 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. Fig.10 As shown, the electronic device 500 may include: a processor 501 , and a memory 502 communicatively connected to the processor 501 .

[0221] The memory 502 is used to store programs. Specifically, the programs may include program codes, and the program codes include computer-executable instructions.

[0222] The memory 502 may include a high-speed RAM memory, and may also include a non-volatile memory (MoM-volatile memory), such as at least one disk memory.

[0223] The processor 501 is used to execute the computer-executable instructions stored in the memory 502 to implement the data processing method.

[0224] The processor 501 may be a central processing unit (CPU), or an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application.

[0225] Optionally, the memory 502 may be independent or integrated with the processor 501. When the memory 502 is a device independent of the processor 501, the electronic device 500 may further include:

[0226] The bus 503 is used to connect the processor 501 and the memory 502. The bus can be an industry standard architecture (ISA) bus, a peripheral component interconnect (PCI) bus, or an extended industry standard architecture (EISA) bus. The bus can be divided into an address bus, a data bus, a control bus, etc., but it does not mean that there is only one bus or one type of bus.

[0227] Optionally, in a specific implementation, if the memory 502 and the processor 501 are integrated on a chip, the memory 502 and the processor 501 can communicate through an internal interface.

[0228] The present application also provides a computer-readable storage medium, which may include: a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a disk or an optical disk, and other media that can store program codes. Specifically, the computer-readable storage medium stores computer execution instructions, and the computer execution instructions are used for the data processing method in the above-mentioned embodiment.

[0229] The present application also provides a computer program product, including computer execution instructions, which implement the data processing method in the above embodiment when executed by a processor.

[0230] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the invention disclosed herein. The present application is intended to cover any modification, use or adaptation of the present application, which follows the general principles of the present application and includes common knowledge or customary techniques in the art that are not disclosed in the present application. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present application are indicated by the claims.

[0231] It should be understood that the present application is not limited to the precise structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present application is limited only by the appended claims.

Claims

1. A data processing method, It is characterized in that include: Collecting available video data by using a drone, wherein the available video data includes images of moving vehicles; Obtaining standard trajectory data of the traveling vehicle according to the available video data, and obtaining driving characteristic data of the traveling vehicle according to the standard trajectory data of the traveling vehicle; Training an initial driving scene classification model according to the driving characteristic data of the driving vehicle to obtain a target driving scene classification model; After obtaining the target driving scene classification model, the method further includes: Analyzing the target driving scene classification model according to the model interpretation tool to obtain a collision feature data set, wherein the collision feature data set includes a data set consisting of a plurality of distance critical values ​​between two vehicles, a data set consisting of a plurality of vehicle speed critical values, and a data set consisting of a plurality of relative speed critical values ​​between two vehicles; Determining a collision feature data set that is positively correlated with the dangerous driving scene classified by the target driving scene classification model as a positive excitation sample; Determine a collision feature data set that is negatively correlated with the dangerous driving scene classified by the target driving scene classification model as a negative stimulus sample; The positive excitation samples and the negative excitation samples are visualized to obtain a target data range, and collision parameters are obtained using a collision feature data set within the target data range; wherein the collision parameters include a following vehicle distance and a collision avoidance time.

2. The data processing method according to claim 1, It is characterized in that After obtaining the target driving scene classification model, the method further includes: The target driving scenario classification model is used to classify the driving data of the target vehicle, and the classified driving scenarios are used for the simulation platform to simulate the autonomous driving simulation.

3. The data processing method according to claim 2, It is characterized in that The method of collecting available video data through a drone includes: Using the drone to obtain raw video data within a preset height range of a target road area; The original video data is screened for video content to determine the original video data containing the image of the moving vehicle as the available video data.

4. The data processing method according to claim 3, It is characterized in that The step of obtaining the standard trajectory data of the moving vehicle according to the available video data includes: Performing image jitter repair processing on the available video data to obtain standard image data; The standard image data is subjected to vehicle extraction and trajectory tracking processing to obtain the trajectory data of the moving vehicle, The trajectory data of the traveling vehicle is subjected to coordinate transformation to obtain standard trajectory data of the traveling vehicle, and the coordinate transformation is used for data conversion between the relative coordinate system of the UAV and the earth coordinate system.

5. The data processing method according to claim 4, It is characterized in that The performing image jitter repair processing on the available video data to obtain standard image data includes: Performing a standard positioning point conversion of the drone on the available video data according to the positioning information of the drone, wherein the standard positioning point conversion is used to convert a reference acquisition point when acquiring the available video data into a standard acquisition point, so as to implement the image jitter repair processing on the available video data; The available video data converted into the standard acquisition point is determined as the standard image data.

6. The data processing method according to claim 5, It is characterized in that The coordinate conversion of the moving vehicle trajectory data to obtain the standard trajectory data of the moving vehicle includes: Performing the coordinate transformation on the first coordinate matrix data according to the positioning information of the UAV and the coordinate information of the key points in the target road area to obtain second coordinate matrix data in the earth coordinate system, wherein the first coordinate matrix data is used to represent the trajectory data of the moving vehicle in the relative coordinate system of the UAV; The trajectory data of the traveling vehicle represented by the second coordinate matrix data is determined as the standard trajectory data of the traveling vehicle.

7. The data processing method according to claim 6, It is characterized in that The obtaining the driving characteristic data of the driving vehicle according to the standard trajectory data of the driving vehicle comprises: Acquiring characteristic data of each traveling vehicle at a current trajectory point according to the standard trajectory data of the traveling vehicle, and structuring the characteristic data to obtain traveling characteristic data of the traveling vehicle; The driving characteristic data of the driving vehicles include the speed, acceleration, steering angle, vehicle number and vehicle size of each driving vehicle.

8. The data processing method according to any one of claims 1 to 7, It is characterized in that The step of training the initial driving scene classification model according to the driving characteristic data of the driving vehicle to obtain a target driving scene classification model includes: The initial driving scene classification model is trained according to the driving characteristic data of the driving vehicle and a plurality of classified driving scenes to obtain an intermediate driving scene classification model, wherein the plurality of classified driving scenes include a straight driving scene, a turning driving scene, a lane changing driving scene, an overtaking driving scene, a following driving scene, an avoidance driving scene, and a first dangerous driving scene; The target driving scene classification model is obtained by training the intermediate driving scene classification model according to the second dangerous driving scene and the driving characteristic data of the accident vehicle in the second dangerous driving scene.

9. The data processing method according to claim 1, It is characterized in that After obtaining the collision parameters, the method further includes: Optimizing the target driving scene classification model according to the reacquired driving characteristic data of the driving vehicle; The collision parameters are optimized according to the optimized target driving scenario classification model.

10. A data processing device, It is characterized in that include: An acquisition module, used to collect available video data through a drone, wherein the available video data includes images of moving vehicles; A first processing module, configured to obtain standard trajectory data of the traveling vehicle according to the available video data, and obtain driving characteristic data of the traveling vehicle according to the standard trajectory data of the traveling vehicle; A second processing module is used to train the initial driving scene classification model according to the driving characteristic data of the driving vehicle to obtain a target driving scene classification model; The processing module is further used for: Analyzing the target driving scene classification model according to the model interpretation tool to obtain a collision feature data set, wherein the collision feature data set includes a data set consisting of a plurality of distance critical values ​​between two vehicles, a data set consisting of a plurality of vehicle speed critical values, and a data set consisting of a plurality of relative speed critical values ​​between two vehicles; Determining a collision feature data set that is positively correlated with the dangerous driving scene classified by the target driving scene classification model as a positive excitation sample; Determine a collision feature data set that is negatively correlated with the dangerous driving scene classified by the target driving scene classification model as a negative stimulus sample; The positive excitation samples and the negative excitation samples are visualized to obtain a target data range, and collision parameters are obtained using a collision feature data set within the target data range; wherein the collision parameters include a following vehicle distance and a collision avoidance time.

11. An electronic device, It is characterized in that include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory to implement the data processing method according to any one of claims 1 to 9.

12. A computer-readable storage medium, It is characterized in that The computer-readable storage medium stores computer-executable instructions, which are used to implement the data processing method according to any one of claims 1 to 9 when executed by a processor.

13. A computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the computer program implements the data processing method according to any one of claims 1 to 9.

Images