Vehicle accident handling system

Abstract

The embodiment of the invention relates to a vehicle accident processing system. The system comprises a data acquisition module, a data preprocessing module, a data storage module, an accident detection early warning module, a detection model library and an accident analysis module, the data acquisition module is used for receiving uploaded data of the vehicle / roadside equipment and sending the uploaded data to the data preprocessing module; the data preprocessing module is used for preprocessing the received vehicle end / road side data and storing the processed data into the data storage module; the accident detection and early warning module is used for timely identifying whether each vehicle has an accident or not and the corresponding accident type according to the data storage module and the detection model library, and carrying out timely early warning after confirming that the vehicle has the accident; and the accident analysis module is used for analyzing the accident cause of each vehicle according to the data storage module. According to the invention, the real-time performance and the analysis efficiency of accident analysis can be improved, and the analysis precision can also be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a vehicle accident processing system. Background Art

[0002] Accident analysis of unmanned / autonomous driving vehicles is of great significance in improving the safety of unmanned / autonomous driving systems and reducing the accident rate. The current conventional accident analysis method for unmanned / autonomous driving vehicles is a non-real-time post-analysis method, that is, data is collected and saved during the driving process of the unmanned / autonomous driving vehicle, and when an accident occurs, the accident analysis is performed based on the saved vehicle-side collected data and the roadside collected data of the road / intersection where the accident occurred based on manual or machine processing. The defects of this conventional analysis method are quite obvious: poor real-time performance, low analysis efficiency, and insufficient analysis accuracy. Summary of the invention

[0003] The purpose of the present invention is to provide a vehicle accident handling system in view of the defects of the prior art, the system includes: a data acquisition module, a data preprocessing module, a data storage module, an accident detection and early warning module, a detection model library and an accident analysis module; wherein the data acquisition module is used to receive the vehicle-side / roadside data uploaded by the vehicle / roadside equipment and send it to the data preprocessing module; the data preprocessing module is used to preprocess the received vehicle-side / roadside data and store the processed data in the data storage module; the accident detection and early warning module is used to timely identify whether each vehicle has an accident and the corresponding accident type according to the data storage module and the detection model library and to timely warn when confirming that the vehicle has an accident; the accident analysis module is used to analyze the cause of the accident of each vehicle according to the data storage module. The present invention can perform real-time analysis during the driving process of the vehicle, and can use the accident detection model of the detection model library to refine the analysis accuracy; the present invention can not only improve the real-time and analysis efficiency of the accident analysis, but also improve the analysis accuracy.

[0004] To achieve the above-mentioned purpose, an embodiment of the present invention provides a vehicle accident processing system, the system comprising: a data acquisition module, a data preprocessing module, a data storage module, an accident detection and early warning module, a detection model library and an accident analysis module;

[0005] The data acquisition module is respectively connected to each first vehicle, each first roadside device and the data preprocessing module; the data storage module is respectively connected to the data preprocessing module, the accident detection and early warning module and the accident analysis module; the accident detection and early warning module is connected to the detection model library;

[0006] The data acquisition module is used to receive the first vehicle-side record uploaded by the first vehicle or the first roadside record uploaded by the first roadside device and send it to the data preprocessing module;

[0007] The data preprocessing module is used to perform data preprocessing on the first vehicle-side record received to obtain corresponding processed data and store it in the corresponding vehicle database in the data storage module; the data preprocessing module is also used to perform data preprocessing on the first roadside record received to obtain corresponding processed data and store it in the corresponding roadside database in the data storage module;

[0008] The data storage module is used to store a vehicle video database, a vehicle perception database, a vehicle status database, a vehicle log database, a roadside video database, a roadside perception database, an accident detection database and an accident analysis database;

[0009] The accident detection and warning module is used to detect whether each of the first vehicles has an accident and the corresponding accident type according to the vehicle perception database, the vehicle status database, the vehicle log database and the detection model library to obtain corresponding detection data and store it in the accident detection database; and perform corresponding accident warning processing based on the latest accident detection results of each of the first vehicles; the accident types at least include sudden acceleration abnormality, sudden deceleration abnormality, emergency start abnormality, emergency braking abnormality, emergency stop abnormality, sharp turn abnormality, sudden lane change abnormality, lane departure abnormality, manual takeover abnormality, vehicle collision abnormality, and system status abnormality;

[0010] The detection model library includes multiple types of accident detection models; the accident detection models correspond to the accident types one by one; each of the accident detection models is implemented based on a type of machine learning model or a type of deep learning model;

[0011] The accident analysis module is used to analyze the cause of the accident of each of the first vehicles according to the vehicle video database, the vehicle perception database, the vehicle status database, the vehicle log database, the roadside video database, the roadside perception database and the accident detection database to obtain corresponding analysis data and store it in the accident analysis database.

[0012] Preferably, the first vehicle-side record includes a first vehicle identification, a first timestamp, a first data type and first uploaded data;

[0013] The first data type includes video, perception, status, and log;

[0014] When the first data type is video, the corresponding first uploaded data is vehicle side video data of the current vehicle; the vehicle side video data is the front video of the current vehicle or a 360° panoramic video around the vehicle;

[0015] When the first data type is perception, the corresponding first uploaded data is the latest environmental perception data of the current vehicle, which is composed of multiple vehicle-side perception target data; the vehicle-side perception target data at least includes a target identifier, a target type, a target geometric size, a target coordinate, a target speed, a target lateral acceleration, a target longitudinal acceleration and a target orientation of the corresponding target;

[0016] When the first data type is status, the corresponding first uploaded data is the latest motion status data of the current vehicle, including at least positioning data, trajectory planning data and driving control data; the positioning data includes at least vehicle coordinates, road section identification and lane identification; the trajectory planning data is a planned trajectory, consisting of multiple trajectory points; the driving control data includes at least driving mode, driving gear, heading angle, steering wheel angle, vehicle speed, longitudinal acceleration, lateral acceleration, yaw angular velocity, roll angular velocity, accelerator pedal opening and closing degree, brake pedal opening and closing degree and headlight status;

[0017] When the first data type is log, the corresponding first uploaded data is a single or multiple latest driving logs of the current vehicle; each driving log is used to record normal, abnormal and warning status information of the driving system.

[0018] Preferably, the first roadside record includes a first device identifier, a second timestamp, a second data type, and second uploaded data;

[0019] The second data type includes video and perception;

[0020] When the second data type is video, the corresponding second uploaded data is the latest video data of the traffic intersection monitored by the current roadside equipment, which is composed of a panoramic video of the traffic intersection and / or a video of a key monitoring area;

[0021] When the second data type is perception, the corresponding second uploaded data is the latest environmental perception data of the traffic intersection monitored by the current roadside equipment, and is composed of multiple roadside perception target data; the roadside perception target data includes at least the target identification, target type, target geometric dimensions, target coordinates, target road section identification, target lane identification, target speed, target lateral acceleration, target longitudinal acceleration, target orientation and target state of the corresponding target.

[0022] Preferably, the vehicle video database includes a plurality of first vehicle video data tables; the first vehicle video data tables correspond to the first vehicles one by one; the first vehicle video data tables are composed of a plurality of first vehicle video records; the first vehicle video record at least includes a first vehicle identification field, a first time field and a first vehicle side video field;

[0023] The vehicle perception database includes a plurality of first vehicle perception data tables; the first vehicle perception data tables correspond to the first vehicles one by one; the first vehicle perception data tables are composed of a plurality of first vehicle perception records; the first vehicle perception records at least include a second vehicle identification field, a second time field and a plurality of first vehicle-side perception target fields; each of the first vehicle-side perception target fields corresponds to a vehicle-side perception target, and the field data is composed of a target identification, a target type, a target geometric size, a target coordinate, a target speed, a target lateral acceleration, a target longitudinal acceleration and a target orientation of the corresponding vehicle-side perception target;

[0024] The vehicle status database includes a plurality of first vehicle status data tables; the first vehicle status data tables correspond one to one with the first vehicles; the first vehicle status data tables are composed of a plurality of first vehicle status records; the first vehicle status record includes at least a third vehicle identification field, a third time field, a first vehicle coordinate field, a first road section identification field, a first lane identification field, a first planning trajectory field, a first driving mode field, a first driving gear field, a first heading angle field, a first steering wheel angle field, a first vehicle speed field, a first longitudinal acceleration field, a first lateral acceleration field, a first yaw angular velocity field, a first roll angular velocity field, a first accelerator pedal opening degree field, a first brake pedal opening degree field and a first headlight status field;

[0025] The vehicle log database includes a plurality of first vehicle log data tables; the first vehicle log data tables correspond to the first vehicles one by one; the first vehicle log data tables are composed of a plurality of first vehicle log records; the first vehicle log record at least includes a fourth vehicle identification field, a fourth time field and a first log sequence field; the first log sequence field is formed by sequentially sorting one or more driving logs;

[0026] The roadside video database includes a plurality of first roadside video data tables; the first roadside video data tables correspond one to one with the first roadside devices; the first roadside video data table is composed of a plurality of first roadside video records; the first roadside video record includes at least a first device identification field, a fifth time field and a first roadside video field;

[0027] The roadside perception database includes multiple first roadside perception data tables; the first roadside perception data tables correspond one-to-one to the first roadside devices; the first roadside perception data tables are composed of multiple first roadside perception records; the first roadside perception record at least includes a second device identification field, a sixth time field and multiple first roadside perception target fields; each of the first roadside perception target fields corresponds to a roadside perception target, and the field data consists of a target identification, a target type, a target geometric size, a target coordinate, a target road section identification, a target lane identification, a target speed, a target lateral acceleration, a target longitudinal acceleration, a target orientation and a target state of the corresponding roadside perception target;

[0028] The accident detection database includes a plurality of first vehicle detection data tables; the first vehicle detection data tables correspond to the first vehicles one by one; the first vehicle detection data tables are composed of a plurality of first vehicle accident detection records; the first vehicle accident detection records at least include a fifth vehicle identification field, a seventh time field and a first accident type sequence field; the first accident type sequence field is composed of one or more accident types;

[0029] The accident analysis database includes multiple first vehicle analysis data tables; the first vehicle analysis data tables correspond one-to-one to the first vehicles; the first vehicle analysis data tables are composed of multiple first vehicle accident analysis records; the first vehicle accident analysis records include at least a sixth vehicle identification field, an eighth time field and a first analysis report field.

[0030] Preferably, the data preprocessing module is specifically used for performing data preprocessing on the received first vehicle-side record to obtain corresponding processed data and storing it in the corresponding vehicle database in the data storage module.

[0031] Extracting the first vehicle identification, the first timestamp, the first data type and the first uploaded data from the first vehicle-side record received at the time as the corresponding current vehicle identification, current timestamp, current data type and current uploaded data;

[0032] and identifying the current data type;

[0033] If the current data type is video, the first vehicle video data table corresponding to the current vehicle identification in the vehicle video database of the data storage module is used as the corresponding current data table; and based on the preset vehicle-side video data quality analysis rule, the current uploaded data is subjected to video data quality analysis to obtain a corresponding first analysis result; and when the first analysis result is that the quality meets the standard, a corresponding first vehicle video record is formed by the current vehicle identification, the current timestamp and the current uploaded data as the corresponding first vehicle identification field, the first time field and the first vehicle-side video field and added to the current data table; the first analysis result includes quality meeting the standard and quality not meeting the standard;

[0034] If the current data type is perception, the first vehicle perception data table corresponding to the current vehicle identification in the vehicle perception database of the data storage module is used as the corresponding current data table; and based on the preset vehicle-side perception data quality analysis rule, the perception data quality analysis is performed on the current uploaded data to obtain a corresponding second analysis result; and when the second analysis result is that the quality meets the standard, the current uploaded data is time synchronized and data interpolated based on the time scale of the current data table to obtain one or more corresponding first vehicle perception records and add them to the current data table; the second analysis result includes quality meeting the standard and quality not meeting the standard;

[0035] If the current data type is status, the first vehicle status data table corresponding to the current vehicle identifier in the vehicle status database of the data storage module is used as the corresponding current data table; and based on the preset vehicle-side status data quality analysis rule, the status data quality analysis is performed on the current uploaded data to obtain a corresponding third analysis result; and when the third analysis result is that the quality meets the standard, the current uploaded data is time synchronized and data interpolated based on the time scale of the current data table to obtain one or more corresponding first vehicle status records and add them to the current data table; the third analysis result includes quality meeting the standard and quality not meeting the standard;

[0036] If the current data type is log, the first vehicle log data table corresponding to the current vehicle identification in the vehicle log database of the data storage module is used as the corresponding current data table; and based on the preset vehicle-side log data quality analysis rules, the current uploaded data is subjected to log data quality analysis to obtain a corresponding fourth analysis result; and when the fourth analysis result is that the quality meets the standard, a corresponding first vehicle log record consisting of the current vehicle identification, the current timestamp and the current uploaded data as the corresponding fourth vehicle identification field, the fourth time field and the first log sequence field is added to the current data table; the fourth analysis result includes quality meeting the standard and quality failing to meet the standard.

[0037] Further, the data preprocessing module is specifically used for performing time synchronization and data interpolation processing on the current uploaded data based on the time scale of the current data table to obtain the corresponding one or more first vehicle perception records and add them to the current data table,

[0038] Using the currently uploaded data as the corresponding current perception target data set;

[0039] The second time fields of the last two first vehicle perception records in the current data table are recorded as the corresponding first time and second time; the current timestamp corresponding to the current uploaded data is used as the corresponding third time; the absolute time difference between the first and second times is used as the corresponding first time scale; and the time period from the first time to the third time is used as the corresponding first time period;

[0040] And in the first time period, taking the first time as the first sampling point and performing time point sampling every first time scale to obtain corresponding multiple first sampling points; wherein the first and second first sampling points are aligned with the first and second times respectively;

[0041] And all the first vehicle-side perception target fields of the first vehicle perception record corresponding to the first and second times form a first and second sampling point perception target data set corresponding to the first sampling points;

[0042] and identifying whether the last obtained first sampling point can be aligned with the third time; if so, taking the current perceived target data set as the sampling point perceived target data set corresponding to the last first sampling point; if not, predicting the sampling point perceived target data set of the last first sampling point according to the uniform acceleration motion model based on the current perceived target data set and the time difference between the third time and the last first sampling point;

[0043] and record the other sampling points between the second and the last of the first sampling points as corresponding first intermediate sampling points; and when the number of the first intermediate sampling points is not zero, record the perception targets with the same target identifier and target type in the three sampling point perception target data sets corresponding to the first, second and last of the first sampling points as corresponding first continuous targets; and estimate the target geometric dimensions, target coordinates, target speed, target lateral acceleration, target longitudinal acceleration and target orientation of the current continuous target at each of the first intermediate sampling points based on the target geometric dimensions, target coordinates, target speed, target lateral acceleration, target longitudinal acceleration and target orientation of each of the first continuous targets at the first, second and last of the first sampling points based on the preset vehicle-side perception data interpolation algorithm; and the sampling point perception target data of the current continuous target at each of the first intermediate sampling points is composed of the target identifier, target type and the target geometric dimensions, target coordinates, target speed, target lateral acceleration, target longitudinal acceleration and target orientation of each of the first continuous targets at each of the first intermediate sampling points; and the sampling point perception target data set of the current intermediate sampling point is composed of all the sampling point perception target data at each of the first intermediate sampling points;

[0044] And traverse the sampling points one by one from the third first sampling point; and in the traversal process, take the currently traversed first sampling point as the corresponding current sampling point; and take the current vehicle identification and sampling point time corresponding to the current sampling point as the corresponding second vehicle identification field and the second time field; and take each perception target data in the sampling point perception target data set corresponding to the current sampling point as a corresponding first vehicle-side perception target field; and the second vehicle identification field, the second time field and all the first vehicle-side perception target fields corresponding to the current sampling point form a corresponding first vehicle perception record; and identify whether the current sampling point is the last first sampling point, if so, end the traversal, if not, go to the next first sampling point for traversal;

[0045] At the end of the traversal, all the first vehicle perception records obtained this time are added to the current data table one by one in chronological order.

[0046] Further, the data preprocessing module is specifically used for performing time synchronization and data interpolation processing on the current uploaded data based on the time scale of the current data table to obtain the corresponding one or more first vehicle status records and add them to the current data table,

[0047] Using the current uploaded data as the corresponding current motion state data;

[0048] and recording the third time fields of the last two first vehicle status records in the current data table as the corresponding fourth time and fifth time; and recording the current timestamp corresponding to the current uploaded data as the corresponding sixth time; and recording the absolute time difference between the fourth and fifth times as the corresponding second time scale; and recording the time period from the fourth time to the sixth time as the corresponding second time period;

[0049] And in the second time period, taking the fourth time as the first sampling point and performing time point sampling every second time scale to obtain a corresponding plurality of second sampling points; wherein the first and second second sampling points are aligned with the fourth and fifth times respectively;

[0050] The first vehicle coordinate field, the first road section identification field, the first lane identification field, the first planning trajectory field, the first driving mode field, the first driving gear field, the first heading angle field, the first steering wheel angle field, the first vehicle speed field, the first longitudinal acceleration field, the first lateral acceleration field, the first yaw angular velocity field, the first roll angular velocity field, the first accelerator pedal opening and closing degree field, the first brake pedal opening and closing degree field and the first headlight status field of the first vehicle status record corresponding to the fourth and fifth times constitute the first and second sampling point motion state data corresponding to the second sampling points;

[0051] and identifying whether the last obtained second sampling point can be aligned with the sixth time; if so, using the current motion state data as the sampling point motion state data corresponding to the last second sampling point; if not, predicting the sampling point motion state data of the last second sampling point according to a uniform acceleration motion model based on the current motion state data and the time difference between the sixth time and the last second sampling point;

[0052] and recording other sampling points between the second and the last second sampling points as corresponding second intermediate sampling points; and when the number of the second intermediate sampling points is not zero, estimating the sampling point motion state data at each of the second intermediate sampling points based on the sampling point motion state data at the first, second and last second sampling points based on a preset vehicle-side motion state data interpolation algorithm;

[0053] And starting from the third second sampling point, traverse the sampling points one by one; and in the traversal process, take the currently traversed second sampling point as the corresponding current sampling point; and take the current vehicle identification and sampling point time corresponding to the current sampling point as the corresponding third vehicle identification field and the third time field; and take the vehicle coordinates, road section identification, lane identification, planned trajectory, driving mode, driving gear, heading angle, steering wheel angle, vehicle speed, longitudinal acceleration, lateral acceleration, yaw angular velocity, roll angular velocity, throttle pedal opening degree, brake pedal opening degree and headlight status of the sampling point motion state data corresponding to the current sampling point as a group of corresponding first vehicle coordinate fields, first road section identification fields, first lane identification fields, first planned trajectory fields, first driving mode fields, first driving gear fields, first heading angle fields, first steering wheel angle fields, first vehicle speed fields, first longitudinal acceleration fields, first lateral acceleration fields, the first yaw angular velocity field, the first roll angular velocity field, the first accelerator pedal opening and closing degree field, the first brake pedal opening and closing degree field and the first headlight status field; and the third vehicle identification field, the third time field, the first vehicle coordinate field, the first road section identification field, the first lane identification field, the first planning trajectory field, the first driving mode field, the first driving gear field, the first heading angle field, the first steering wheel angle field, the first vehicle speed field, the first longitudinal acceleration field, the first lateral acceleration field, the first yaw angular velocity field, the first roll angular velocity field, the first accelerator pedal opening and closing degree field, the first brake pedal opening and closing degree field and the first headlight status field corresponding to the current sampling point form a corresponding first vehicle status record; and identify whether the current sampling point is the last second sampling point, if so, end the traversal, if not, go to the next second sampling point for traversal;

[0054] At the end of the traversal, all the first vehicle status records obtained this time are added to the current data table one by one in chronological order.

[0055] Preferably, the data preprocessing module is specifically used for performing data preprocessing on the received first roadside record to obtain corresponding processed data and storing it in the corresponding roadside database in the data storage module.

[0056] Extracting the corresponding first device identifier, the second timestamp, the second data type and the second uploaded data from the first roadside record received at the time as the corresponding current device identifier, current timestamp, current data type and current uploaded data;

[0057] and identifying the current data type;

[0058] If the current data type is video, the first roadside video data table corresponding to the current device identifier in the roadside video database of the data storage module is used as the corresponding current data table; and based on the preset roadside video data quality analysis rule, the video data quality analysis is performed on the current uploaded data to obtain the corresponding fifth analysis result; and when the fifth analysis result is that the quality meets the standard, a corresponding first roadside video record is composed of the current device identifier, the current timestamp and the current uploaded data as the corresponding first device identifier field, the fifth time field and the first roadside video field and added to the current data table; the fifth analysis result includes whether the quality meets the standard or not;

[0059] If the current data type is perception, the first roadside perception data table corresponding to the current device identifier in the roadside perception database of the data storage module is used as the corresponding current data table; and based on the preset roadside perception data quality analysis rules, the current uploaded data is subjected to a perception data quality analysis to obtain a corresponding sixth analysis result; and when the sixth analysis result is that the quality meets the standard, the current uploaded data is subjected to time synchronization and data interpolation processing based on the time scale of the current data table to obtain one or more corresponding first roadside perception records and add them to the current data table; the sixth analysis result includes quality that meets the standard and quality that does not meet the standard.

[0060] Further, the data preprocessing module is specifically used for performing time synchronization and data interpolation processing on the current uploaded data based on the time scale of the current data table to obtain the corresponding one or more first roadside sensing records and add them to the current data table,

[0061] Using the currently uploaded data as the corresponding current perception target data set;

[0062] and recording the sixth time fields of the last two first roadside sensing records in the current data table as the corresponding seventh time and eighth time; and recording the current timestamp corresponding to the current uploaded data as the corresponding ninth time; and recording the absolute time difference between the seventh and eighth times as the corresponding third time scale; and recording the time period from the seventh time to the ninth time as the corresponding third time period;

[0063] And in the third time period, taking the seventh time as the first sampling point and performing time point sampling every the third time scale to obtain corresponding multiple third sampling points; wherein the first and second third sampling points are aligned with the seventh and eighth times respectively;

[0064] And all the first roadside perception target fields of the first roadside perception records corresponding to the seventh and eighth time points form the sampling point perception target data sets corresponding to the first and second third sampling points;

[0065] and identifying whether the last third sampling point obtained can be aligned with the ninth time; if so, taking the current perception target data set as the sampling point perception target data set corresponding to the last third sampling point; if not, predicting the sampling point perception target data set of the last third sampling point according to the uniform acceleration motion model based on the current perception target data set and the time difference between the ninth time and the last third sampling point;

[0066] and recording other sampling points between the second and the last third sampling points as corresponding third intermediate sampling points; and when the number of the third intermediate sampling points is not zero, recording the perception targets with the same target identification and target type in the three sampling point perception target data sets corresponding to the first, second and last third sampling points as corresponding second continuous targets; and based on the preset roadside perception data interpolation algorithm, according to the target geometric dimensions, target coordinates, target road section identification, target lane identification, target speed, target lateral acceleration, target longitudinal acceleration, target orientation and target state of each of the second continuous targets at the first, second and last third sampling points, the current continuous target at each of the third intermediate sampling points is interpolated. The target geometric size, target coordinates, target road section identification, target lane identification, target speed, target lateral acceleration, target longitudinal acceleration, target orientation and target state at the sampling point are estimated; and the target identification, target type and the geometric size, target coordinates, target road section identification, target lane identification, target speed, target lateral acceleration, target longitudinal acceleration, target orientation and target state of each of the second continuous targets at each of the third intermediate sampling points constitute the sampling point perception target data of the current continuous target at each of the third intermediate sampling points; and all the sampling point perception target data at each of the third intermediate sampling points constitute the sampling point perception target data set of the current intermediate sampling point;

[0067] And starting from the third sampling point, traverse the sampling points one by one; and in the traversal process, take the third sampling point currently traversed as the corresponding current sampling point; and take the current device identification and sampling point time corresponding to the current sampling point as the corresponding second device identification field and the sixth time field; and take each perception target data in the sampling point perception target data set corresponding to the current sampling point as a corresponding first roadside perception target field; and the second device identification field, the sixth time field and all the first roadside perception target fields corresponding to the current sampling point form a corresponding first roadside perception record; and identify whether the current sampling point is the last third sampling point, if so, end the traversal, if not, go to the next third sampling point for traversal;

[0068] And at the end of the traversal, all the first roadside sensing records obtained this time are added to the current data table one by one in chronological order.

[0069] Preferably, the accident detection and early warning module is specifically used for detecting whether each of the first vehicles has an accident and the corresponding accident type according to the vehicle perception database, the vehicle status database, the vehicle log database and the detection model library to obtain corresponding detection data and store them in the accident detection database.

[0070] According to the preset first time frequency, the current time is regularly used as the corresponding end time; and the time obtained by subtracting the preset first time length from the end time is used as the corresponding start time; and the current start time and the end time form the corresponding current recent time period;

[0071] and extracting the first vehicle perception records whose second time fields satisfy the current most recent time period in each of the first vehicle perception data tables in the vehicle perception database to form a corresponding first vehicle perception record sequence; and performing corresponding vehicle-side perception target feature vector extraction processing according to each of the first vehicle perception record sequences to obtain a corresponding first vehicle-side perception target vector sequence;

[0072] and extracting the first vehicle state records whose third time fields in each of the first vehicle state data tables in the vehicle state database satisfy the current most recent time period to form a corresponding first vehicle state record sequence; and performing corresponding vehicle state feature vector extraction processing according to each of the first vehicle state record sequences to obtain a corresponding first vehicle state vector;

[0073] and extracting the first vehicle log records whose fourth time fields in each of the first vehicle log data tables in the vehicle log database satisfy the current most recent time period to form a corresponding first vehicle log record sequence; and performing corresponding vehicle log feature vector extraction processing according to each of the first vehicle log record sequences to obtain a corresponding first vehicle log vector;

[0074] The accident detection models corresponding to sudden acceleration abnormality, sudden deceleration abnormality, emergency start abnormality, emergency braking abnormality, emergency stop abnormality, sharp turn abnormality, sudden lane change abnormality, lane departure abnormality, manual takeover abnormality, vehicle collision abnormality, and system status abnormality in the detection model library are recorded as corresponding sudden acceleration abnormality detection model, sudden deceleration abnormality detection model, emergency start abnormality detection model, emergency braking abnormality detection model, emergency stop abnormality detection model, sharp turn abnormality detection model, sudden lane change abnormality detection model, lane departure abnormality detection model, manual takeover abnormality detection model, vehicle collision abnormality detection model, and system status abnormality detection model;

[0075] The first vehicle state vectors corresponding to the first vehicles are respectively input into the sudden acceleration abnormality detection model, the sudden deceleration abnormality detection model, the emergency start abnormality detection model, the emergency braking abnormality detection model, the emergency stop abnormality detection model, the sharp turn abnormality detection model, the sudden lane change abnormality detection model, the lane departure abnormality detection model and the manual takeover abnormality detection model to perform corresponding sudden acceleration abnormal event recognition, sudden deceleration abnormal event recognition, emergency start abnormal event recognition, emergency braking abnormal event recognition, emergency stop abnormal event recognition, sharp turn abnormal event recognition, sudden lane change abnormal event recognition, lane departure abnormal event recognition and manual takeover abnormal event recognition. to the corresponding first, second, third, fourth, fifth, sixth, seventh, eighth and ninth event recognition results; and input the first vehicle-side perception target vector sequence and the first vehicle state vector corresponding to each of the first vehicles into the vehicle collision anomaly detection model to perform corresponding vehicle collision anomaly event recognition to obtain the corresponding tenth event recognition result; and input the first vehicle log vector corresponding to each of the first vehicles into the system state anomaly detection model to perform corresponding system state anomaly event recognition to obtain the corresponding eleventh event recognition result; the first, second, third, fourth, fifth, sixth, seventh, eighth, ninth, tenth and eleventh event recognition results all include abnormality and no abnormality;

[0076] and extracting the event recognition results that are specifically abnormal from the first, second, third, fourth, fifth, sixth, seventh, eighth, ninth, tenth and eleventh event recognition results corresponding to each of the first vehicles to form a corresponding first vehicle abnormal result set;

[0077] And perform a round of traversal on all the first vehicle abnormal result sets obtained; and in this round of traversal, use the first vehicle abnormal result set currently traversed as the corresponding current vehicle abnormal result set; and when the current vehicle abnormal result set is not empty, all the accident types corresponding to the current vehicle abnormal result set form a corresponding first accident type sequence field, and use the current end time as a corresponding seventh time field, and use the vehicle identification of the first vehicle corresponding to the current vehicle abnormal result set as a corresponding fifth vehicle identification field, and use the fifth vehicle identification field, the seventh time field and the first accident type sequence field corresponding to the current vehicle abnormal result set to form a corresponding first vehicle accident detection record, which is added to the first vehicle detection data table corresponding to the current vehicle abnormal result set in the accident detection database.

[0078] Further, the accident detection and warning module is specifically used for, when the corresponding vehicle-side perception target feature vector extraction process is performed according to each of the first vehicle perception record sequences to obtain the corresponding first vehicle-side perception target vector sequence,

[0079] The total number of records of the first vehicle perception record sequence is counted to obtain a corresponding first total number M, and the total number of target identifiers of the first vehicle perception record sequence is counted to obtain a corresponding second total number N, and a two-dimensional matrix A with a shape of M×N is constructed; and all matrix elements a of the two-dimensional matrix A are i,j The unit vector of the two-dimensional matrix A is initialized with all zero vectors; each row of the two-dimensional matrix A corresponds to a vehicle-side sensing target, and each column corresponds to a time point; the two-dimensional matrix A consists of M×N matrix units a i,j Composition, 1≤column index i≤M, 1≤row index j≤N; each of the matrix units a i,j Used to store a corresponding unit vector, the vector data of each unit vector consists of a set of corresponding target geometric dimensions, target coordinates, target speed, target lateral acceleration, target longitudinal acceleration and target orientation;

[0080] A round of traversal is performed on all the first vehicle perception records of the first vehicle perception record sequence; and during this round of traversal, the first vehicle perception record currently traversed is recorded as the corresponding current record; and the second time field of the current record is used as the corresponding current time; and each of the first vehicle-side perception target fields of the current record is polled; and during the polling process, the first vehicle-side perception target field currently polled is used as the corresponding current target, and the matrix unit a whose column index i in the two-dimensional matrix A matches the current time and whose row index j matches the target is recorded. i,j Recorded as the corresponding current matrix unit, and the unit vector of the current matrix unit is reset by forming a corresponding current data vector with the target geometric size, target coordinates, target speed, target lateral acceleration, target longitudinal acceleration and target orientation of the current target;

[0081] When all the first vehicle perception records of the first vehicle perception record sequence are traversed, the M matrix elements a in each row of the two-dimensional matrix A are replaced by i,j The extracted ones are sorted in chronological order to form a corresponding first vehicle side perception target vector; and the obtained N first vehicle side perception target vectors are sorted in sequence to form a corresponding first vehicle side perception target vector sequence.

[0082] Furthermore, the accident detection and warning module is specifically used to extract the first vehicle coordinate field, the first road section identification field, the first lane identification field, the first planning trajectory field, the first driving mode field, the first driving gear field, the first heading angle field, the first steering wheel angle field, the first vehicle speed field, the first longitudinal acceleration field, the first lateral acceleration field, the first yaw angular velocity field, the first roll angular velocity field, the first accelerator pedal opening degree field, the first brake pedal opening degree field and the first headlight status field of each of the first vehicle state records in the first vehicle state record sequence to form a corresponding first single-point state vector when the corresponding vehicle state feature vector is extracted and processed according to each of the first vehicle state record sequences to obtain the corresponding first vehicle state vector; and all the obtained first single-point state vectors are sorted in chronological order to form a corresponding first vehicle state vector.

[0083] Furthermore, the accident detection and warning module is specifically used to extract the first log sequence field of each of the first vehicle log records in the first vehicle log record sequence as a corresponding first single-point log vector when the corresponding vehicle log feature vector extraction and processing is performed according to each of the first vehicle log record sequences to obtain the corresponding first vehicle log vector; and all the obtained first single-point log vectors are sorted in chronological order to form a corresponding first vehicle log vector.

[0084] Furthermore, the accident detection and warning module is specifically used to monitor the record addition events of each of the first vehicle detection data tables in the accident detection database when performing corresponding accident warning processing based on the latest accident detection results of each of the first vehicles; and when each time a first vehicle accident detection record is added to each of the first vehicle detection data tables, the first vehicle accident detection record most recently added in the current first vehicle detection data table is recorded as the corresponding current newly added record; and the current newly added record is brought into a preset warning message template for warning message conversion to obtain the corresponding current warning message; and the current warning message is sent to a preset accident warning interface.

[0085] Preferably, the accident analysis module is specifically used for analyzing the cause of the accident of each of the first vehicles according to the vehicle video database, the vehicle perception database, the vehicle status database, the vehicle log database, the roadside video database, the roadside perception database and the accident detection database to obtain corresponding analysis data and store them in the accident analysis database,

[0086] Monitor the record addition events of each of the first vehicle detection data tables of the accident detection database; and when each of the first vehicle accident detection records is added to each of the first vehicle detection data tables, record the first vehicle accident detection record most recently added in the current first vehicle detection data table as the corresponding current newly added record; and use the first vehicle corresponding to the current newly added record as the corresponding current vehicle;

[0087] The seventh time field of the newly added record is used as the corresponding first accident end time, and the time obtained by subtracting the preset second time length from the first accident end time is used as the corresponding first accident start time; and the first accident start time and the first accident end time constitute the corresponding first accident period;

[0088] and extracting the first vehicle video records in the first vehicle video data table corresponding to the current vehicle in the vehicle video database, in which the first time field meets the first accident period, and sorting them in chronological order to form a corresponding first record sequence;

[0089] and extracting the first vehicle perception records whose second time field satisfies the first accident period from the first vehicle perception data table corresponding to the current vehicle in the vehicle perception database and sorting them in chronological order to form a corresponding second record sequence;

[0090] and extracting the first vehicle status records whose third time field in the first vehicle status data table corresponding to the current vehicle in the vehicle status database satisfies the first accident period and sorting them in chronological order to form a corresponding third record sequence;

[0091] and extracting the first vehicle log records whose fourth time field in the first vehicle log data table corresponding to the current vehicle in the vehicle log database satisfies the first accident period and sorting them in chronological order to form a corresponding fourth record sequence;

[0092] and extracting the first vehicle coordinate field, the first road section identification field and the first lane identification field of the last first vehicle status record of the third record sequence to form corresponding current positioning data; and taking the first roadside device closest to the current positioning data as the corresponding current roadside device;

[0093] and extracting the first roadside video records whose fifth time field in the first roadside video data table corresponding to the current roadside device in the roadside video database satisfies the first accident period, and sorting them in chronological order to form a corresponding fifth record sequence;

[0094] and extracting the first roadside sensing records whose sixth time field in the first roadside sensing data table corresponding to the current roadside device in the roadside sensing database satisfies the first accident period, and sorting them in chronological order to form a corresponding sixth record sequence;

[0095] The first, second, third, fourth, fifth and sixth record sequences are used to form a corresponding first accident analysis material set; the first accident analysis material set is sent to a preset accident cause / responsibility analysis interface; and the accident cause / responsibility analysis text sent back by the accident cause / responsibility analysis interface is received;

[0096] The fifth vehicle identification field and the seventh time field of the currently newly added record are used as the corresponding sixth vehicle identification field and the eighth time field; the first accident type sequence field and the first accident time period, the first accident analysis material set and the accident cause / responsibility analysis text of the currently newly added record form a corresponding first analysis report field; and the sixth vehicle identification field, the eighth time field and the first analysis report field corresponding to the currently newly added record form a corresponding first vehicle accident analysis record, which is added to the first vehicle analysis data table corresponding to the current vehicle in the accident analysis database.

[0097] The embodiment of the present invention provides a vehicle accident handling system. As can be seen from the above content, the system includes: a data acquisition module, a data preprocessing module, a data storage module, an accident detection and early warning module, a detection model library and an accident analysis module; wherein the data acquisition module is used to receive the vehicle-side / roadside data uploaded by the vehicle / roadside equipment and send it to the data preprocessing module; the data preprocessing module is used to preprocess the received vehicle-side / roadside data and store the processed data in the data storage module; the accident detection and early warning module is used to timely identify whether each vehicle has an accident and the corresponding accident type according to the data storage module and the detection model library, and to timely warn when confirming that the vehicle has an accident; the accident analysis module is used to analyze the cause of the accident of each vehicle according to the data storage module. The embodiment of the present invention can perform real-time analysis during the driving process of the vehicle, and can use the accident detection model of the detection model library to refine the analysis accuracy; the embodiment of the present invention not only improves the real-time and analysis efficiency of the accident analysis, but also improves the accuracy of the accident analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0098] Figure 1 A module structure diagram of a vehicle accident handling system provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0099] In order to make the purpose, technical solution and advantages of the present invention clearer, the present invention will be further described in detail below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0100] The embodiment of the present invention provides a vehicle accident handling system 1 such as Figure 1As shown in the module structure diagram of a vehicle accident handling system provided in an embodiment of the present invention, the system mainly includes: a data acquisition module 11, a data preprocessing module 12, a data storage module 13, an accident detection and warning module 14, a detection model library 15 and an accident analysis module 16.

[0101] The connection relationship of each component of the embodiment of the present invention is as follows: the data acquisition module 11 is respectively connected to each first vehicle 2, each first roadside device 3 and the data preprocessing module 12; the data storage module 13 is respectively connected to the data preprocessing module 12, the accident detection and early warning module 14 and the accident analysis module 16; the accident detection and early warning module 14 is connected to the detection model library 15. The components of the embodiment of the present invention can be as follows.

[0102] (I) Data collection module 11:

[0103] The data acquisition module 11 in the embodiment of the present invention is used to receive a first vehicle-side record uploaded by a first vehicle 2 or a first roadside record uploaded by a first roadside device 3 and send it to a data preprocessing module 12 .

[0104] Here, the first vehicle 2 of the embodiment of the present invention is an unmanned or automatic driving vehicle; the first roadside device 3 is an intelligent roadside device (Road Side Un it, RSU) used in the Internet of Vehicles.

[0105] The first vehicle-side record uploaded by the first vehicle 2 includes a first vehicle identifier, a first timestamp, a first data type and first uploaded data; wherein:

[0106] 1) The first vehicle identification is the unique identification of the current vehicle;

[0107] 2) The first timestamp is a time information;

[0108] 3) The first data type includes video, perception, status, and log;

[0109] 4) The data content of the first uploaded data is related to the first data type, specifically:

[0110] a. When the first data type is video, the corresponding first uploaded data is the vehicle side video data of the current vehicle; the vehicle side video data is the front video of the current vehicle or the 360° panoramic video around the vehicle;

[0111] b. When the first data type is perception, the corresponding first uploaded data is the latest environmental perception data of the current vehicle, which is composed of multiple vehicle-side perception target data; each vehicle-side perception target data at least includes a target identifier, a target type, a target geometric size, a target coordinate, a target speed, a target lateral acceleration, a target longitudinal acceleration and a target orientation of the corresponding target;

[0112] c. When the first data type is status, the corresponding first uploaded data is the latest motion status data of the current vehicle, including at least positioning data, trajectory planning data and driving control data; the positioning data includes at least vehicle coordinates, road section identification and lane identification; the trajectory planning data is a planned trajectory, consisting of multiple trajectory points; the driving control data includes at least driving mode, driving gear, heading angle, steering wheel angle, vehicle speed, longitudinal acceleration, lateral acceleration, yaw angular velocity, roll angular velocity, accelerator pedal opening and closing degree, brake pedal opening and closing degree and headlight status;

[0113] d. When the first data type is log, the corresponding first uploaded data is a single or multiple latest driving logs of the current vehicle; each driving log is used to record normal, abnormal and warning status information of the driving system.

[0114] The first roadside record uploaded by the first roadside device 3 includes a first device identifier, a second timestamp, a second data type, and second uploaded data; wherein:

[0115] 1) The first device identifier is the unique identifier of the current roadside device;

[0116] 2) The second timestamp is a time information;

[0117] 3) The second data type includes video and perception;

[0118] 4) The data content of the second uploaded data is related to the second data type, specifically:

[0119] a. When the second data type is video, the corresponding second uploaded data is the latest video data of the traffic intersection monitored by the current roadside equipment, which is composed of a panoramic video of the traffic intersection and / or a video of a key monitoring area;

[0120] b. When the second data type is perception, the corresponding second uploaded data is the latest environmental perception data of the traffic intersection monitored by the current roadside equipment, which is composed of multiple roadside perception target data; the roadside perception target data at least includes the target identification, target type, target geometric dimensions, target coordinates, target road section identification, target lane identification, target speed, target lateral acceleration, target longitudinal acceleration, target orientation and target state of the corresponding target; the target state here corresponds to the target type, for example, if the target state is a traffic light, the target state is used to mark the on and off of the red / yellow / green light on the traffic light.

[0121] (II) Data preprocessing module 12:

[0122] The data preprocessing module 12 of the embodiment of the present invention is used to perform data preprocessing on the received first vehicle-side record to obtain corresponding processed data and store it in the corresponding vehicle database in the data storage module 13 .

[0123] The data preprocessing module 12 of the embodiment of the present invention is further used to perform data preprocessing on the received first roadside record to obtain corresponding processed data and store it into the corresponding roadside database in the data storage module 13 .

[0124] In a specific implementation of the embodiment of the present invention, the data preprocessing module 12 is specifically used to perform data preprocessing on the received first vehicle-side record to obtain corresponding processed data and store it in the corresponding vehicle database in the data storage module 13:

[0125] Step A1, extracting the corresponding first vehicle identification, first timestamp, first data type and first uploaded data from the first vehicle-side record received at the time as the corresponding current vehicle identification, current timestamp, current data type and current uploaded data;

[0126] Step A2, and identify the current data type;

[0127] Step A3, if the current data type is video, the first vehicle video data table corresponding to the current vehicle identification in the vehicle video database 131 of the data storage module 13 is used as the corresponding current data table; and based on the preset vehicle-side video data quality analysis rule, the video data quality analysis is performed on the current uploaded data to obtain the corresponding first analysis result; and when the first analysis result is that the quality meets the standard, the current vehicle identification, the current timestamp and the current uploaded data are used as the corresponding first vehicle identification field, the first time field and the first vehicle-side video field to form a corresponding first vehicle video record and add it to the current data table;

[0128] The first analysis result includes quality reaching the standard and quality failing to reach the standard;

[0129] Here, the vehicle-side video data quality analysis rule is a pre-set video data quality judgment standard, which can be set based on actual application requirements;

[0130] Step A4: if the current data type is perception, the first vehicle perception data table corresponding to the current vehicle identification in the vehicle perception database 132 of the data storage module 13 is used as the corresponding current data table; and based on the preset vehicle-side perception data quality analysis rule, the perception data quality analysis is performed on the current uploaded data to obtain the corresponding second analysis result; and when the second analysis result is that the quality meets the standard, the current uploaded data is time synchronized and data interpolated based on the time scale of the current data table to obtain one or more corresponding first vehicle perception records and add them to the current data table;

[0131] The second analysis results include quality compliance and quality non-compliance;

[0132] Here, the vehicle-side perception data quality analysis rule is a pre-set perception data quality judgment standard, which can be set based on actual application requirements;

[0133] Step A5, if the current data type is status, the first vehicle status data table corresponding to the current vehicle identifier in the vehicle status database 133 of the data storage module 13 is used as the corresponding current data table; and based on the preset vehicle-side status data quality analysis rule, the status data quality analysis is performed on the current uploaded data to obtain the corresponding third analysis result; and when the third analysis result is that the quality meets the standard, the current uploaded data is time synchronized and data interpolated based on the time scale of the current data table to obtain one or more corresponding first vehicle status records and add them to the current data table;

[0134] Among them, the third analysis results include quality reaching the standard and quality not reaching the standard;

[0135] Here, the vehicle side state data quality analysis rule is a preset vehicle motion state data quality judgment standard, which can be set based on actual application requirements;

[0136] Step A6, if the current data type is log, the first vehicle log data table corresponding to the current vehicle identification in the vehicle log database 134 of the data storage module 13 is used as the corresponding current data table; and based on the preset vehicle-side log data quality analysis rule, the current uploaded data is subjected to log data quality analysis to obtain a corresponding fourth analysis result; and when the fourth analysis result is up to standard, a corresponding first vehicle log record is formed by the current vehicle identification, the current timestamp and the current uploaded data as the corresponding fourth vehicle identification field, the fourth time field and the first log sequence field and added to the current data table;

[0137] Among them, the fourth analysis result includes quality reaching the standard and quality failing to reach the standard;

[0138] Here, the vehicle-side log data quality analysis rule is a preset vehicle log data quality judgment standard, which can be set based on actual application requirements.

[0139] In another specific implementation of the embodiment of the present invention, the data preprocessing module 12 is specifically used to perform time synchronization and data interpolation processing on the current uploaded data based on the time scale of the current data table to obtain the corresponding one or more first vehicle perception records and add them to the current data table:

[0140] Step B1, taking the currently uploaded data as the corresponding current perception target data set;

[0141] Step B2, record the second time fields of the last two first vehicle perception records in the current data table as the corresponding first time and second time; and record the current timestamp corresponding to the current uploaded data as the corresponding third time; and record the absolute time difference between the first and second times as the corresponding first time scale; and record the time period from the first time to the third time as the corresponding first time period;

[0142] Step B3, in the first time period, taking the first time as the first sampling point and performing time point sampling every first time scale to obtain corresponding multiple first sampling points;

[0143] Wherein, the first and second first sampling points are aligned with the first and second times respectively;

[0144] Step B4, forming a sampling point perception target data set corresponding to the first and second first sampling points from all first vehicle-side perception target fields of the first vehicle perception records corresponding to the first and second time periods;

[0145] Step B5, identifying whether the obtained last first sampling point can be aligned with the third time; if so, taking the current perception target data set as the sampling point perception target data set corresponding to the last first sampling point; if not, predicting the sampling point perception target data set of the last first sampling point according to the uniform acceleration motion model based on the current perception target data set and the time difference between the third time and the last first sampling point;

[0146] Here, the embodiment of the present invention predicts information such as target geometric size, target coordinates, target speed, target lateral acceleration, target longitudinal acceleration and target orientation of the last first sampling point according to the uniform acceleration motion model based on the current sensed target data set and the time difference between the third time and the last first sampling point;

[0147] Step B7, and record the other sampling points between the second and the last first sampling points as corresponding first intermediate sampling points; and when the number of the first intermediate sampling points is not zero, record the perception targets with the same target identification and target type in the three sampling point perception target data sets corresponding to the first, second and last first sampling points as corresponding first continuous targets; and based on the preset vehicle-side perception data interpolation algorithm, estimate the target geometric dimensions, target coordinates, target speed, target lateral acceleration, target longitudinal acceleration and target orientation of each first continuous target at the first, second and last first sampling points; and the sampling point perception target data of the current continuous target at each first intermediate sampling point is composed of the target identification, target type of each first continuous target and the target geometric dimensions, target coordinates, target speed, target lateral acceleration, target longitudinal acceleration and target orientation at each first intermediate sampling point; and the sampling point perception target data set of the current intermediate sampling point is composed of all the sampling point perception target data at each first intermediate sampling point;

[0148] Here, the vehicle-side sensing data interpolation algorithm is a pre-set interpolation algorithm, which can be set based on actual application requirements; commonly used algorithms include linear interpolation algorithm, polynomial interpolation algorithm, quadratic interpolation algorithm, etc.;

[0149] Step B8, and traverse the sampling points one by one from the third first sampling point; and in the traversal process, the first sampling point currently traversed is used as the corresponding current sampling point; and the current vehicle identification and sampling point time corresponding to the current sampling point are used as the corresponding second vehicle identification field and second time field; and each perception target data in the sampling point perception target data set corresponding to the current sampling point is used as a corresponding first vehicle-side perception target field; and the second vehicle identification field, the second time field and all the first vehicle-side perception target fields corresponding to the current sampling point form a corresponding first vehicle perception record; and identify whether the current sampling point is the last first sampling point, if so, end the traversal, if not, go to the next first sampling point for traversal;

[0150] Step B9, at the end of the traversal, all first vehicle perception records obtained this time are added to the current data table one by one in chronological order.

[0151] In another specific implementation of the embodiment of the present invention, the data preprocessing module 12 is specifically used to perform time synchronization and data interpolation processing on the current uploaded data based on the time scale of the current data table to obtain one or more corresponding first vehicle status records and add them to the current data table:

[0152] Step C1, using the current uploaded data as the corresponding current motion state data;

[0153] Step C2, and record the third time fields of the last two first vehicle status records in the current data table as the corresponding fourth time and fifth time; and record the current timestamp corresponding to the current uploaded data as the corresponding sixth time; and record the absolute time difference between the fourth and fifth times as the corresponding second time scale; and record the time period from the fourth time to the sixth time as the corresponding second time period;

[0154] Step C3: In the second time period, the fourth time is used as the first sampling point and time point sampling is performed every second time scale to obtain a plurality of corresponding second sampling points; wherein the first and second second sampling points are aligned with the fourth and fifth times respectively;

[0155] Step C4, the first vehicle coordinate field, the first road section identification field, the first lane identification field, the first planning trajectory field, the first driving mode field, the first driving gear field, the first heading angle field, the first steering wheel angle field, the first vehicle speed field, the first longitudinal acceleration field, the first lateral acceleration field, the first yaw angular velocity field, the first roll angular velocity field, the first accelerator pedal opening degree field, the first brake pedal opening degree field and the first headlight state field of the first vehicle state record corresponding to the fourth and fifth time constitute the sampling point motion state data corresponding to the first and second second sampling points;

[0156] Step C5, identifying whether the obtained last second sampling point can be aligned with the sixth time; if so, using the current motion state data as the sampling point motion state data corresponding to the last second sampling point; if not, predicting the sampling point motion state data of the last second sampling point according to the uniform acceleration motion model based on the current motion state data and the time difference between the sixth time and the last second sampling point;

[0157] Here, when the embodiment of the present invention predicts the sampling point motion state data of the last second sampling point according to the uniform acceleration motion model based on the current motion state data and the time difference between the sixth time and the last second sampling point: first, according to the uniform acceleration motion model based on the current motion state data and the time difference between the sixth time and the last second sampling point, the coordinates, planned trajectory, driving gear, heading angle, steering wheel angle, vehicle speed, longitudinal acceleration, lateral acceleration, yaw angular velocity, roll angular velocity, accelerator pedal opening and closing degree, brake pedal opening and closing degree and other information of the last second sampling point are predicted; then, based on the predicted coordinates, a preset high-precision road map is queried to obtain the predicted road section identification and lane identification; then, the driving mode and headlight status in the current motion state data are used as the driving mode and headlight status at the last second sampling point;

[0158] Step C6, and record the other sampling points between the second and the last second sampling points as corresponding second intermediate sampling points; and when the number of the second intermediate sampling points is not zero, estimate the sampling point motion state data at each second intermediate sampling point based on the sampling point motion state data at the first, second and last second sampling points based on a preset vehicle-side motion state data interpolation algorithm;

[0159] Here, when the embodiment of the present invention estimates the motion state data of the sampling points at each second intermediate sampling point based on the vehicle-side motion state data interpolation algorithm: first, based on a conventional interpolation algorithm (such as a linear interpolation algorithm, a polynomial interpolation algorithm, a quadratic interpolation algorithm, etc.), the coordinates, planned trajectory, driving gear, heading angle, steering wheel angle, vehicle speed, longitudinal acceleration, lateral acceleration, yaw angular velocity, roll angular velocity, accelerator pedal opening and closing, brake pedal opening and closing, and other information at each second intermediate sampling point are interpolated and estimated; then, based on the estimated coordinates, a preset high-precision road map is queried to obtain the road map at each second intermediate sampling point. segment identification and lane identification; then identify whether the driving modes at the first, second and last second sampling points are the same, if they are the same, uniformly set the driving modes at each second intermediate sampling point based on the driving mode at the first second sampling point, if they are not the same, set the driving modes at each second intermediate sampling point to empty; then identify whether the vehicle light states at the first, second and last second sampling points are the same, if they are the same, uniformly set the vehicle light states at each second intermediate sampling point based on the vehicle light state at the first second sampling point, if they are not the same, set the vehicle light states at each second intermediate sampling point to empty;

[0160] Step C7, and traverse the sampling points one by one starting from the third second sampling point; and in the traversal process, the second sampling point currently traversed is used as the corresponding current sampling point; and the current vehicle identification and sampling point time corresponding to the current sampling point are used as the corresponding third vehicle identification field and third time field; and the vehicle coordinates, road section identification, lane identification, planned trajectory, driving mode, driving gear, heading angle, steering wheel angle, vehicle speed, longitudinal acceleration, lateral acceleration, yaw angular velocity, roll angular velocity, throttle pedal opening degree, brake pedal opening degree and headlight state of the sampling point motion state data corresponding to the current sampling point are used as a group of corresponding first vehicle coordinate fields, first road section identification fields, first lane identification fields, first planned trajectory fields, first driving mode fields, first driving gear fields, first heading angle fields, first steering wheel angle fields, first vehicle speed fields, first longitudinal acceleration fields, and third time fields. a lateral acceleration field, a first yaw angular velocity field, a first roll angular velocity field, a first accelerator pedal opening and closing degree field, a first brake pedal opening and closing degree field and a first headlight status field; and a corresponding first vehicle status record is formed by the third vehicle identification field, the third time field, the first vehicle coordinate field, the first road section identification field, the first lane identification field, the first planning trajectory field, the first driving mode field, the first driving gear field, the first heading angle field, the first steering wheel angle field, the first vehicle speed field, the first longitudinal acceleration field, the first lateral acceleration field, the first yaw angular velocity field, the first roll angular velocity field, the first accelerator pedal opening and closing degree field, the first brake pedal opening and closing degree field and the first headlight status field corresponding to the current sampling point; and identify whether the current sampling point is the last second sampling point, if so, end the traversal, if not, go to the next second sampling point for traversal;

[0161] Step C8: At the end of the traversal, all first vehicle status records obtained this time are added to the current data table one by one in chronological order.

[0162] In another specific implementation of the embodiment of the present invention, the data preprocessing module 12 is specifically used to perform data preprocessing on the received first roadside record to obtain corresponding processed data and store it in the corresponding roadside database in the data storage module 13:

[0163] Step D1, extracting the corresponding first device identifier, second timestamp, second data type and second uploaded data from the first roadside record received at the time as the corresponding current device identifier, current timestamp, current data type and current uploaded data;

[0164] Step D2, and identifying the current data type;

[0165] Step D3, if the current data type is video, the first roadside video data table corresponding to the current device identifier in the roadside video database 135 of the data storage module 13 is used as the corresponding current data table; and based on the preset roadside video data quality analysis rule, the current uploaded data is subjected to video data quality analysis to obtain a corresponding fifth analysis result; and when the fifth analysis result is that the quality meets the standard, a corresponding first roadside video record is formed by the current device identifier, the current timestamp and the current uploaded data as the corresponding first device identifier field, the fifth time field and the first roadside video field and added to the current data table;

[0166] Among them, the fifth analysis result includes quality reaching the standard and quality not reaching the standard;

[0167] Here, the roadside video data quality analysis rule is a preset roadside video data quality judgment standard, which can be set based on actual application requirements;

[0168] Step D4, if the current data type is perception, the first roadside perception data table corresponding to the current device identifier in the roadside perception database 136 of the data storage module 13 is used as the corresponding current data table; and based on the preset roadside perception data quality analysis rule, the perception data quality analysis is performed on the current uploaded data to obtain the corresponding sixth analysis result; and when the sixth analysis result is that the quality meets the standard, the current uploaded data is time synchronized and data interpolated based on the time scale of the current data table to obtain one or more corresponding first roadside perception records and add them to the current data table;

[0169] Among them, the sixth analysis result includes quality meeting the standard and quality not meeting the standard.

[0170] Here, the roadside perception data quality analysis rule is a pre-set roadside perception data quality judgment standard, which can be set based on actual application requirements.

[0171] In another specific implementation of the embodiment of the present invention, the data preprocessing module 12 is specifically used to perform time synchronization and data interpolation processing on the current uploaded data based on the time scale of the current data table to obtain one or more corresponding first roadside perception records and add them to the current data table:

[0172] Step E1, taking the currently uploaded data as the corresponding current perception target data set;

[0173] Step E2, and record the sixth time fields of the last two first roadside sensing records in the current data table as the corresponding seventh time and eighth time; and record the current timestamp corresponding to the current uploaded data as the corresponding ninth time; and record the absolute time difference between the seventh and eighth times as the corresponding third time scale; and record the time period from the seventh time to the ninth time as the corresponding third time period;

[0174] Step E3: In the third time period, the seventh time is used as the first sampling point and time point sampling is performed every third time scale to obtain corresponding multiple third sampling points; wherein the first and second third sampling points are aligned with the seventh and eighth time respectively;

[0175] Step E4, forming a sampling point perception target data set corresponding to the first and second third sampling points from all first roadside perception target fields of the first roadside perception records corresponding to the seventh and eighth time points;

[0176] Step E5, and identify whether the obtained last third sampling point can be aligned with the ninth time; if so, use the current perception target data set as the sampling point perception target data set corresponding to the last third sampling point; if not, predict the sampling point perception target data set of the last third sampling point according to the uniform acceleration motion model based on the current perception target data set and the time difference between the ninth time and the last third sampling point;

[0177] Here, when the embodiment of the present invention predicts the sampling point perception target data set of the last third sampling point according to the uniform acceleration motion model based on the current perception target data set and the time difference between the ninth time and the last third sampling point: firstly, the target geometric size, target coordinates, target speed, target lateral acceleration, target longitudinal acceleration, target orientation and other information of the last third sampling point are predicted according to the uniform acceleration motion model based on the current perception target data set and the time difference between the ninth time and the last third sampling point; then, based on the predicted target coordinates, the preset road high-precision map is queried to obtain the predicted target road section mark and target lane mark; then, the target state in the current perception target data set is used as the target state at the last third sampling point;

[0178] Step E6, and record the other sampling points between the second and the last third sampling points as corresponding third intermediate sampling points; and when the number of the third intermediate sampling points is not zero, record the perception targets with the same target identification and target type in the three sampling point perception target data sets corresponding to the first, second and last third sampling points as corresponding second continuous targets; and based on the preset roadside perception data interpolation algorithm, according to the target geometric dimensions, target coordinates, target road section identification, target lane identification, target speed, target lateral acceleration, target longitudinal acceleration, target orientation and target state of each second continuous target at the first, second and last third sampling points, calculate the current continuous target at each third intermediate sampling point. The target geometric size, target coordinates, target road section identification, target lane identification, target speed, target lateral acceleration, target longitudinal acceleration, target orientation and target state at the intermediate sampling point are estimated; and the target identification, target type and the geometric size, target coordinates, target road section identification, target lane identification, target speed, target lateral acceleration, target longitudinal acceleration, target orientation and target state of each second continuous target at each third intermediate sampling point constitute the sampling point perception target data of the current continuous target at each third intermediate sampling point; and the sampling point perception target data set of the current intermediate sampling point is constituted by all the sampling point perception target data at each third intermediate sampling point;

[0179] Here, the embodiment of the present invention estimates the target geometric size, target coordinates, target road segment identification, target lane identification, target speed, target lateral acceleration, target longitudinal acceleration, target orientation and target state of the current continuous target at each third intermediate sampling point based on the roadside perception data interpolation algorithm: firstly, based on a conventional interpolation algorithm (such as a linear interpolation algorithm, a polynomial interpolation algorithm, The target geometric size, target coordinates, target speed, target lateral acceleration, target longitudinal acceleration, target orientation and other information at each third intermediate sampling point are interpolated and estimated using a quadratic interpolation algorithm, etc.; based on the estimated target coordinates, a preset high-precision road map is queried to obtain the target road section identification and target lane identification at each third intermediate sampling point; whether the target states at the first, second and last third sampling points are the same is identified; if they are the same, the target states at each third intermediate sampling point are uniformly set based on the target state at the first third sampling point; if they are not the same, the target states at each third intermediate sampling point are set to null;

[0180] Step E7, and traverse the sampling points one by one from the third sampling point; and in the traversal process, the third sampling point currently traversed is used as the corresponding current sampling point; and the current device identification and sampling point time corresponding to the current sampling point are used as the corresponding second device identification field and sixth time field; and each perception target data in the sampling point perception target data set corresponding to the current sampling point is used as a corresponding first roadside perception target field; and the second device identification field, the sixth time field and all the first roadside perception target fields corresponding to the current sampling point form a corresponding first roadside perception record; and identify whether the current sampling point is the last third sampling point, if so, end the traversal, if not, go to the next third sampling point for traversal;

[0181] Step E8: At the end of the traversal, all first roadside sensing records obtained this time are added to the current data table one by one in chronological order.

[0182] (III) Data storage module 13:

[0183] The data storage module 13 of the embodiment of the present invention is used to store a vehicle video database 131, a vehicle perception database 132, a vehicle status database 133, a vehicle log database 134, a roadside video database 135, a roadside perception database 136, an accident detection database 137 and an accident analysis database 138.

[0184] Here, the vehicle video database 131 of the embodiment of the present invention includes multiple first vehicle video data tables; the first vehicle video data tables correspond one-to-one to the first vehicles 2; the first vehicle video data tables are composed of multiple first vehicle video records; the first vehicle video record includes at least a first vehicle identification field, a first time field and a first vehicle side video field.

[0185] The vehicle perception database 132 of the embodiment of the present invention includes multiple first vehicle perception data tables; the first vehicle perception data tables correspond one-to-one to the first vehicles 2; the first vehicle perception data tables are composed of multiple first vehicle perception records; the first vehicle perception records at least include a second vehicle identification field, a second time field and multiple first vehicle-side perception target fields; each first vehicle-side perception target field corresponds to a vehicle-side perception target, and the field data is composed of the target identification, target type, target geometric size, target coordinates, target speed, target lateral acceleration, target longitudinal acceleration and target orientation of the corresponding vehicle-side perception target.

[0186] The vehicle status database 133 of the embodiment of the present invention includes multiple first vehicle status data tables; the first vehicle status data tables correspond one-to-one to the first vehicle 2; the first vehicle status data tables are composed of multiple first vehicle status records; the first vehicle status record at least includes a third vehicle identification field, a third time field, a first vehicle coordinate field, a first road section identification field, a first lane identification field, a first planned trajectory field, a first driving mode field, a first driving gear field, a first heading angle field, a first steering wheel angle field, a first vehicle speed field, a first longitudinal acceleration field, a first lateral acceleration field, a first yaw angular velocity field, a first roll angular velocity field, a first accelerator pedal opening degree field, a first brake pedal opening degree field and a first headlight status field.

[0187] The vehicle log database 134 of the embodiment of the present invention includes multiple first vehicle log data tables; the first vehicle log data tables correspond one-to-one to the first vehicles 2; the first vehicle log data tables are composed of multiple first vehicle log records; the first vehicle log record includes at least a fourth vehicle identification field, a fourth time field and a first log sequence field; the first log sequence field is formed by sequentially sorting one or more driving logs.

[0188] The roadside video database 135 of the embodiment of the present invention includes multiple first roadside video data tables; the first roadside video data tables correspond one-to-one to the first roadside devices 3; the first roadside video data tables are composed of multiple first roadside video records; the first roadside video record includes at least a first device identification field, a fifth time field and a first roadside video field.

[0189] The roadside perception database 136 of the embodiment of the present invention includes multiple first roadside perception data tables; the first roadside perception data tables correspond one-to-one to the first roadside devices 3; the first roadside perception data tables are composed of multiple first roadside perception records; the first roadside perception record at least includes a second device identification field, a sixth time field and multiple first roadside perception target fields; each first roadside perception target field corresponds to a roadside perception target, and the field data is composed of the target identification, target type, target geometric size, target coordinates, target road section identification, target lane identification, target speed, target lateral acceleration, target longitudinal acceleration, target orientation and target state of the corresponding roadside perception target.

[0190] The accident detection database 137 of the embodiment of the present invention includes multiple first vehicle detection data tables; the first vehicle detection data tables correspond one-to-one to the first vehicles 2; the first vehicle detection data tables are composed of multiple first vehicle accident detection records; the first vehicle accident detection records include at least the fifth vehicle identification field, the seventh time field and the first accident type sequence field; the first accident type sequence field is composed of one or more accident types.

[0191] The accident analysis database 138 of the embodiment of the present invention includes multiple first vehicle analysis data tables; the first vehicle analysis data tables correspond one-to-one to the first vehicles 2; the first vehicle analysis data tables are composed of multiple first vehicle accident analysis records; the first vehicle accident analysis records include at least the sixth vehicle identification field, the eighth time field and the first analysis report field.

[0192] (IV) Accident detection and early warning module 14:

[0193] The accident detection and warning module 14 of the embodiment of the present invention is used to detect whether an accident has occurred in each first vehicle 2 and the corresponding accident type based on the vehicle perception database 132, the vehicle status database 133, the vehicle log database 134 and the detection model library 15 to obtain the corresponding detection data and store it in the accident detection database 137; and perform corresponding accident warning processing based on the latest accident detection results of each first vehicle 2.

[0194] Here, the accident types in the embodiment of the present invention include at least abnormal sudden acceleration, abnormal sudden deceleration, abnormal emergency start, abnormal emergency braking, abnormal emergency stop, abnormal sharp turn, abnormal sudden lane change, abnormal lane departure, abnormal manual takeover, abnormal vehicle collision, and abnormal system status.

[0195] In another specific implementation of the embodiment of the present invention, the accident detection and warning module 14 is specifically used to detect whether each first vehicle 2 has an accident and the corresponding accident type according to the vehicle perception database 132, the vehicle status database 133, the vehicle log database 134 and the detection model library 15 to obtain corresponding detection data and store them in the accident detection database 137:

[0196] Step F1, regularly taking the current time as the corresponding end time according to the preset first time frequency; and taking the time obtained by subtracting the preset first duration from the current end time as the corresponding start time; and the current start time and the current end time form the corresponding current recent time period;

[0197] Here, the first time frequency is a preset frequency parameter; the first duration is a preset time length;

[0198] Step F2, extract the first vehicle perception records whose second time fields satisfy the current most recent time period in each first vehicle perception data table of the vehicle perception database 132 to form a corresponding first vehicle perception record sequence; and perform corresponding vehicle side perception target feature vector extraction processing according to each first vehicle perception record sequence to obtain a corresponding first vehicle side perception target vector sequence;

[0199] Step F3, extract the first vehicle state records in the third time field of each first vehicle state data table of the vehicle state database 133 that meet the current most recent time period to form a corresponding first vehicle state record sequence; and perform corresponding vehicle state feature vector extraction processing according to each first vehicle state record sequence to obtain a corresponding first vehicle state vector;

[0200] Step F4, extract the first vehicle log records whose fourth time fields satisfy the current most recent time period in each first vehicle log data table of the vehicle log database 134 to form a corresponding first vehicle log record sequence; and perform corresponding vehicle log feature vector extraction processing according to each first vehicle log record sequence to obtain a corresponding first vehicle log vector;

[0201] Step F5, and record the accident detection models corresponding to the sudden acceleration abnormality, sudden deceleration abnormality, emergency start abnormality, emergency braking abnormality, emergency stop abnormality, sharp turn abnormality, sudden lane change abnormality, lane departure abnormality, manual takeover abnormality, vehicle collision abnormality, and system status abnormality in the detection model library 15 as the corresponding sudden acceleration abnormality detection model, sudden deceleration abnormality detection model, emergency start abnormality detection model, emergency braking abnormality detection model, emergency stop abnormality detection model, sharp turn abnormality detection model, sudden lane change abnormality detection model, lane departure abnormality detection model, manual takeover abnormality detection model, vehicle collision abnormality detection model, and system status abnormality detection model;

[0202] Step F6, and input the first vehicle state vector corresponding to each first vehicle 2 into the sudden acceleration abnormality detection model, the sudden deceleration abnormality detection model, the emergency start abnormality detection model, the emergency braking abnormality detection model, the emergency stop abnormality detection model, the sharp turn abnormality detection model, the sudden lane change abnormality detection model, the lane departure abnormality detection model and the manual takeover abnormality detection model respectively to perform corresponding sudden acceleration abnormal event recognition, sudden deceleration abnormal event recognition, emergency start abnormal event recognition, emergency braking abnormal event recognition, emergency stop abnormal event recognition, sharp turn abnormal event recognition, sudden lane change abnormal event recognition, lane departure abnormal event recognition and manual takeover abnormal event recognition to obtain the corresponding first, second, third, fourth, fifth, sixth, seventh, eighth and ninth event recognition results; and input the first vehicle side perception target vector sequence and the first vehicle state vector corresponding to each first vehicle 2 into the vehicle collision abnormality detection model to perform corresponding vehicle collision abnormal event recognition to obtain the corresponding tenth event recognition result; and input the first vehicle log vector corresponding to each first vehicle 2 into the system state abnormality detection model to perform corresponding system state abnormal event recognition to obtain the corresponding eleventh event recognition result;

[0203] Among them, the first, second, third, fourth, fifth, sixth, seventh, eighth, ninth, tenth and eleventh event recognition results all include abnormality and no abnormality;

[0204] Step F7, extracting the event recognition results that are specifically abnormal from the first, second, third, fourth, fifth, sixth, seventh, eighth, ninth, tenth and eleventh event recognition results corresponding to each first vehicle 2 to form a corresponding first vehicle abnormal result set;

[0205] Step F8, and perform a round of traversal on all the first vehicle abnormal result sets obtained; and in this round of traversal, use the currently traversed first vehicle abnormal result set as the corresponding current vehicle abnormal result set; and when the current vehicle abnormal result set is not empty, all accident types corresponding to the current vehicle abnormal result set are used to form a corresponding first accident type sequence field, and the end time of this time is used as a corresponding seventh time field, and the vehicle identification of the first vehicle 2 corresponding to the current vehicle abnormal result set is used as a corresponding fifth vehicle identification field, and the fifth vehicle identification field, the seventh time field and the first accident type sequence field corresponding to the current vehicle abnormal result set are used to form a corresponding first vehicle accident detection record, which is added to the first vehicle detection data table corresponding to the current vehicle abnormal result set in the accident detection database 137.

[0206] In another specific implementation of the embodiment of the present invention, the accident detection and warning module 14 is specifically configured to extract corresponding vehicle-side perception target feature vectors according to each first vehicle perception record sequence to obtain a corresponding first vehicle-side perception target vector sequence:

[0207] Step G1, the total number of records of the first vehicle perception record sequence is counted to obtain a corresponding first total number M, and the total number of target identifiers of the first vehicle perception record sequence is counted to obtain a corresponding second total number N, and a two-dimensional matrix A with a shape of M×N is constructed; and all matrix elements a of the two-dimensional matrix A are i,j The unit vector of is initialized with all zero vectors;

[0208] Each row of the two-dimensional matrix A corresponds to a vehicle-side sensing target, and each column corresponds to a time point; the two-dimensional matrix A consists of M×N matrix units a i,j Composition, 1≤column index i≤M, 1≤row index j≤N; each matrix unit a i,j Used to store a corresponding unit vector, the vector data of each unit vector consists of a set of corresponding target geometric dimensions, target coordinates, target speed, target lateral acceleration, target longitudinal acceleration and target orientation;

[0209] Step G2, perform a round of traversal of all first vehicle perception records of the first vehicle perception record sequence; and in this round of traversal, record the first vehicle perception record currently traversed as the corresponding current record; and use the second time field of the current record as the corresponding current time; and poll each first vehicle-side perception target field of the current record; and in the polling process, use the first vehicle-side perception target field currently polled as the corresponding current target, and match the column index i in the two-dimensional matrix A with the current time and the row index j with the target with the matrix unit a i,j The corresponding current matrix unit is recorded as the target geometric size, target coordinates, target speed, target lateral acceleration, target longitudinal acceleration and target orientation of the current target to form a corresponding current data vector to reset the unit vector of the current matrix unit;

[0210] Step G3: At the end of traversing all the first vehicle perception records of the first vehicle perception record sequence, the M matrix elements a of each row of the two-dimensional matrix A are replaced by i,j The extracted ones are sorted in chronological order to form a corresponding first vehicle side perception target vector; and the obtained N first vehicle side perception target vectors are sorted in sequence to form a corresponding first vehicle side perception target vector sequence.

[0211] In another specific implementation of the embodiment of the present invention, the accident detection and warning module 14 is specifically used to extract the first vehicle coordinate field, the first road section identification field, the first lane identification field, the first planning trajectory field, the first driving mode field, the first driving gear field, the first heading angle field, the first steering wheel angle field, the first vehicle speed field, the first longitudinal acceleration field, the first lateral acceleration field, the first yaw angular velocity field, the first roll angular velocity field, the first accelerator pedal opening degree field, the first brake pedal opening degree field and the first headlight status field of each first vehicle state record in the first vehicle state record sequence to form a corresponding first single-point state vector; and all the obtained first single-point state vectors are sorted in chronological order to form a corresponding first vehicle state vector.

[0212] In another specific implementation of the embodiment of the present invention, the accident detection and warning module 14 is specifically used to extract the first log sequence field of each first vehicle log record in the first vehicle log record sequence as a corresponding first single-point log vector when performing corresponding vehicle log feature vector extraction processing according to each first vehicle log record sequence to obtain a corresponding first vehicle log vector; and all the obtained first single-point log vectors are sorted in chronological order to form a corresponding first vehicle log vector.

[0213] In another specific implementation method of the embodiment of the present invention, the accident detection and warning module 14 is specifically used to monitor the record addition events of each first vehicle detection data table of the accident detection database 137 when performing corresponding accident warning processing based on the latest accident detection results of each first vehicle 2; and when each first vehicle accident detection record is added to each first vehicle detection data table, the latest added first vehicle accident detection record in the current first vehicle detection data table is recorded as the corresponding current newly added record; and the current newly added record is brought into the preset warning message template for warning message conversion to obtain the corresponding current warning message; and the current warning message is sent to the preset accident warning interface.

[0214] Here, the warning message template is a pre-set message text setting template. By bringing the various fields of the current newly added record into the corresponding template items in the warning message template, a corresponding current warning message can be obtained; the accident warning interface is a pre-set warning processing interface, which can be set according to specific application requirements, such as email notification interface, SMS notification interface, voice notification interface, WeChat notification interface, office system notification interface, etc.

[0215] (V) Detection model library 15:

[0216] The detection model library 15 of the embodiment of the present invention includes multiple types of accident detection models; the accident detection models correspond one-to-one to the accident types; each accident detection model is implemented based on a type of machine learning model or a type of deep learning model.

[0217] (VI) Accident Analysis Module 16:

[0218] The accident analysis module 16 of the embodiment of the present invention is used to analyze the cause of the accident of each first vehicle 2 according to the vehicle video database 131, the vehicle perception database 132, the vehicle status database 133, the vehicle log database 134, the roadside video database 135, the roadside perception database 136 and the accident detection database 137 to obtain the corresponding analysis data and store it in the accident analysis database 138.

[0219] In another specific implementation of the embodiment of the present invention, the accident analysis module 16 is specifically used to analyze the cause of the accident of each first vehicle 2 according to the vehicle video database 131, the vehicle perception database 132, the vehicle status database 133, the vehicle log database 134, the roadside video database 135, the roadside perception database 136 and the accident detection database 137 to obtain corresponding analysis data and store them in the accident analysis database 138:

[0220] Step H1, monitoring the record adding events of each first vehicle detection data table of the accident detection database 137; and when each first vehicle accident detection record is added to each first vehicle detection data table, recording the first vehicle accident detection record newly added in the current first vehicle detection data table as the corresponding current newly added record; and recording the first vehicle 2 corresponding to the current newly added record as the corresponding current vehicle;

[0221] Step H2, the seventh time field of the newly added record is used as the corresponding first accident end time, and the time obtained by subtracting the preset second time length from the first accident end time is used as the corresponding first accident start time; and the first accident start time and the first accident end time form the corresponding first accident period;

[0222] Here, the second duration is a preset time length;

[0223] Step H3, extract the first vehicle video record in the first vehicle video data table corresponding to the current vehicle in the vehicle video database 131, in which the first time field meets the first accident period, and sort them in chronological order to form a corresponding first record sequence;

[0224] Step H4, extract the first vehicle perception records in the first vehicle perception data table corresponding to the current vehicle in the vehicle perception database 132, whose second time field meets the first accident period, and sort them in chronological order to form a corresponding second record sequence;

[0225] Step H5, extract the first vehicle status records in the first vehicle status data table corresponding to the current vehicle in the vehicle status database 133, whose third time field satisfies the first accident period, and sort them in chronological order to form a corresponding third record sequence;

[0226] Step H6, extract the first vehicle log records in the first vehicle log data table corresponding to the current vehicle in the vehicle log database 134, whose fourth time field meets the first accident period, and sort them in chronological order to form a corresponding fourth record sequence;

[0227] Step H7, extract the first vehicle coordinate field, the first road section identification field and the first lane identification field of the last first vehicle status record of the third record sequence to form the corresponding current positioning data; and use the first roadside device 3 closest to the current positioning data as the corresponding current roadside device;

[0228] Step H8, extract the first roadside video record in the first roadside video data table corresponding to the current roadside device in the roadside video database 135, in which the fifth time field meets the first accident period, and sort them in chronological order to form a corresponding fifth record sequence;

[0229] Step H9, extract the first roadside sensing records whose sixth time field meets the first accident period in the first roadside sensing data table corresponding to the current roadside device in the roadside sensing database 136, and sort them in chronological order to form a corresponding sixth record sequence;

[0230] Step H10, the first, second, third, fourth, fifth and sixth record sequences are used to form a corresponding first accident analysis material set; the first accident analysis material set is sent to a preset accident cause / responsibility analysis interface; and the accident cause / responsibility analysis text sent back by the accident cause / responsibility analysis interface is received;

[0231] Here, the accident cause / responsibility analysis interface is a pre-set analysis processing interface; the analysis processing interface is used to classify and identify all possible causes of the accident according to the input first accident analysis material set, and score the correlation between each identified cause type and the accident, and estimate the accident responsibility ratio corresponding to each cause type based on the score, and finally form a corresponding first cause analysis text from each cause type and its corresponding correlation score and accident responsibility ratio, and return the corresponding accident cause / responsibility analysis text from all the obtained first cause analysis texts; the accident cause / responsibility analysis interface of the embodiment of the present invention can be a processing interface of an analysis function module implemented based on a deep learning model, and can also be a processing interface of an analysis function module implemented based on a large language model, and can also be a machine analysis interface implemented based on a fixed analysis process, and can also be a manual analysis interface;

[0232] Step H11, and use the fifth vehicle identification field and the seventh time field of the current newly added record as the corresponding sixth vehicle identification field and the eighth time field; and the first accident type sequence field of the current newly added record and the first accident time period, the first accident analysis material set and the accident cause / responsibility analysis text form a corresponding first analysis report field; and the sixth vehicle identification field, the eighth time field and the first analysis report field corresponding to the current newly added record form a corresponding first vehicle accident analysis record and add it to the first vehicle analysis data table corresponding to the current vehicle in the accident analysis database 138.

[0233] It should be noted that the division of the various system components / modules of the above system can be regarded as a division of logical functions. In actual implementation, they can be fully or partially integrated into one physical entity, or they can be physically separated. And these system components / modules can all be implemented in the form of software calling through processing elements; they can also be all implemented in the form of hardware; some modules can also be implemented in the form of software calling through processing elements, and some modules can be implemented in the form of hardware. For example, the data acquisition module can be a separately established processing element, or it can be integrated in a chip of the above system for implementation. In addition, it can also be stored in the memory of the above system in the form of program code, and called and executed by a processing element of the above system. The implementation of other system components / modules is similar. In addition, these modules can be fully or partially integrated together, or they can be implemented independently. The processing element described here can be an integrated circuit with signal processing capabilities. In the implementation process, the processing steps of each module of the aforementioned system can be completed by the hardware integrated logic circuit in the processor element or the instructions in the form of software.

[0234] For example, the system components / modules of the above system may be one or more integrated circuits configured to implement the functions of the corresponding components / modules, such as: one or more specific integrated circuits (ASIC), or one or more digital signal processors (DSP), or one or more field programmable gate arrays (FPGA), etc. For another example, when a system component / module of the above system is implemented in the form of a processing element scheduling program code, the processing element may be a general-purpose processor, such as a central processing unit (CPU) or other processor that can call program code. For another example, these modules may be integrated together and implemented in the form of a system-on-a-chip (SOC).

[0235] In the above embodiments, all or part of the embodiments may be implemented by software, hardware, firmware or any combination thereof. When implemented by software, all or part of the embodiments may be implemented in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the above system are generated. The above computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The above-mentioned computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the above-mentioned computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center through wired (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (such as infrared, wireless, Bluetooth, microwave, etc.) methods. The above-mentioned computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more available media integrated. The above-mentioned available medium can be a magnetic medium (such as a floppy disk, a hard disk, a tape), an optical medium (such as a DVD), or a semiconductor medium (such as a solid state drive (SSD)), etc.

[0236] In summary, the embodiment of the present invention provides a vehicle accident handling system; it can be seen from the above content that the system includes: a data acquisition module, a data preprocessing module, a data storage module, an accident detection and warning module, a detection model library and an accident analysis module; wherein the data acquisition module is used to receive the vehicle-side / roadside data uploaded by the vehicle / roadside equipment and send it to the data preprocessing module; the data preprocessing module is used to preprocess the received vehicle-side / roadside data and store the processed data in the data storage module; the accident detection and warning module is used to timely identify whether each vehicle has an accident and the corresponding accident type according to the data storage module and the detection model library and to timely warn when confirming that the vehicle has an accident; the accident analysis module is used to analyze the cause of the accident of each vehicle according to the data storage module. The embodiment of the present invention can perform real-time analysis during the driving process of the vehicle, and can use the accident detection model of the detection model library to refine the analysis accuracy; through the embodiment of the present invention, not only the real-time and analysis efficiency of the accident analysis are improved, but also the accuracy of the accident analysis is improved.

[0237] The professionals should further realize that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to the function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0238] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be implemented using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0239] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A vehicle accident handling system, characterized in that: The system includes: a data acquisition module, a data preprocessing module, a data storage module, an accident detection and early warning module, a detection model library and an accident analysis module; The data acquisition module is respectively connected to each first vehicle, each first roadside device and the data preprocessing module; the data storage module is respectively connected to the data preprocessing module, the accident detection and early warning module and the accident analysis module; the accident detection and early warning module is connected to the detection model library; The data acquisition module is used to receive the first vehicle-side record uploaded by the first vehicle or the first roadside record uploaded by the first roadside device and send it to the data preprocessing module; The data preprocessing module is used to perform data preprocessing on the first vehicle-side record received to obtain corresponding processed data and store it in the corresponding vehicle database in the data storage module; the data preprocessing module is also used to perform data preprocessing on the first roadside record received to obtain corresponding processed data and store it in the corresponding roadside database in the data storage module; The data storage module is used to store a vehicle video database, a vehicle perception database, a vehicle status database, a vehicle log database, a roadside video database, a roadside perception database, an accident detection database and an accident analysis database; The accident detection and warning module is used to detect whether each of the first vehicles has an accident and the corresponding accident type according to the vehicle perception database, the vehicle status database, the vehicle log database and the detection model library to obtain corresponding detection data and store it in the accident detection database; and perform corresponding accident warning processing based on the latest accident detection results of each of the first vehicles; the accident types at least include sudden acceleration abnormality, sudden deceleration abnormality, emergency start abnormality, emergency braking abnormality, emergency stop abnormality, sharp turn abnormality, sudden lane change abnormality, lane departure abnormality, manual takeover abnormality, vehicle collision abnormality, and system status abnormality; The detection model library includes multiple types of accident detection models; the accident detection models correspond to the accident types one by one; each of the accident detection models is implemented based on a type of machine learning model or a type of deep learning model; The accident analysis module is used to analyze the cause of the accident of each of the first vehicles according to the vehicle video database, the vehicle perception database, the vehicle status database, the vehicle log database, the roadside video database, the roadside perception database and the accident detection database to obtain corresponding analysis data and store it in the accident analysis database.

2. The vehicle accident handling system according to claim 1, characterized in that: The first vehicle-side record includes a first vehicle identifier, a first timestamp, a first data type, and first uploaded data; The first data type includes video, perception, status, and log; When the first data type is video, the corresponding first uploaded data is vehicle side video data of the current vehicle; the vehicle side video data is the front video of the current vehicle or a 360° panoramic video around the vehicle; When the first data type is perception, the corresponding first uploaded data is the latest environmental perception data of the current vehicle, which is composed of multiple vehicle-side perception target data; the vehicle-side perception target data at least includes a target identifier, a target type, a target geometric size, a target coordinate, a target speed, a target lateral acceleration, a target longitudinal acceleration and a target orientation of the corresponding target; When the first data type is status, the corresponding first uploaded data is the latest motion status data of the current vehicle, including at least positioning data, trajectory planning data and driving control data; the positioning data includes at least vehicle coordinates, road section identification and lane identification; the trajectory planning data is a planned trajectory, consisting of multiple trajectory points; the driving control data includes at least driving mode, driving gear, heading angle, steering wheel angle, vehicle speed, longitudinal acceleration, lateral acceleration, yaw angular velocity, roll angular velocity, accelerator pedal opening and closing degree, brake pedal opening and closing degree and headlight status; When the first data type is log, the corresponding first uploaded data is a single or multiple latest driving logs of the current vehicle; each driving log is used to record normal, abnormal and warning status information of the driving system.

3. The vehicle accident handling system according to claim 2, characterized in that: The first roadside record includes a first device identifier, a second timestamp, a second data type, and second uploaded data; The second data type includes video and perception; When the second data type is video, the corresponding second uploaded data is the latest video data of the traffic intersection monitored by the current roadside equipment, which is composed of a panoramic video of the traffic intersection and / or a video of a key monitoring area; When the second data type is perception, the corresponding second uploaded data is the latest environmental perception data of the traffic intersection monitored by the current roadside equipment, and is composed of multiple roadside perception target data; the roadside perception target data includes at least the target identification, target type, target geometric dimensions, target coordinates, target road section identification, target lane identification, target speed, target lateral acceleration, target longitudinal acceleration, target orientation and target state of the corresponding target.

4. The vehicle accident handling system according to claim 3, characterized in that: The vehicle video database includes a plurality of first vehicle video data tables; the first vehicle video data tables correspond to the first vehicles one by one; the first vehicle video data tables are composed of a plurality of first vehicle video records; the first vehicle video records at least include a first vehicle identification field, a first time field and a first vehicle side video field; The vehicle perception database includes a plurality of first vehicle perception data tables; the first vehicle perception data tables correspond to the first vehicles one by one; the first vehicle perception data tables are composed of a plurality of first vehicle perception records; the first vehicle perception records at least include a second vehicle identification field, a second time field and a plurality of first vehicle-side perception target fields; each of the first vehicle-side perception target fields corresponds to a vehicle-side perception target, and the field data is composed of a target identification, a target type, a target geometric size, a target coordinate, a target speed, a target lateral acceleration, a target longitudinal acceleration and a target orientation of the corresponding vehicle-side perception target; The vehicle status database includes a plurality of first vehicle status data tables; the first vehicle status data tables correspond one to one with the first vehicles; the first vehicle status data tables are composed of a plurality of first vehicle status records; the first vehicle status record includes at least a third vehicle identification field, a third time field, a first vehicle coordinate field, a first road section identification field, a first lane identification field, a first planning trajectory field, a first driving mode field, a first driving gear field, a first heading angle field, a first steering wheel angle field, a first vehicle speed field, a first longitudinal acceleration field, a first lateral acceleration field, a first yaw angular velocity field, a first roll angular velocity field, a first accelerator pedal opening degree field, a first brake pedal opening degree field and a first headlight status field; The vehicle log database includes a plurality of first vehicle log data tables; the first vehicle log data tables correspond to the first vehicles one by one; the first vehicle log data tables are composed of a plurality of first vehicle log records; the first vehicle log record at least includes a fourth vehicle identification field, a fourth time field and a first log sequence field; the first log sequence field is formed by sequentially sorting one or more driving logs; The roadside video database includes a plurality of first roadside video data tables; the first roadside video data tables correspond one to one with the first roadside devices; the first roadside video data table is composed of a plurality of first roadside video records; the first roadside video record includes at least a first device identification field, a fifth time field and a first roadside video field; The roadside perception database includes multiple first roadside perception data tables; the first roadside perception data tables correspond one-to-one to the first roadside devices; the first roadside perception data tables are composed of multiple first roadside perception records; the first roadside perception record at least includes a second device identification field, a sixth time field and multiple first roadside perception target fields; each of the first roadside perception target fields corresponds to a roadside perception target, and the field data consists of a target identification, a target type, a target geometric size, a target coordinate, a target road section identification, a target lane identification, a target speed, a target lateral acceleration, a target longitudinal acceleration, a target orientation and a target state of the corresponding roadside perception target; The accident detection database includes a plurality of first vehicle detection data tables; the first vehicle detection data tables correspond to the first vehicles one by one; the first vehicle detection data tables are composed of a plurality of first vehicle accident detection records; the first vehicle accident detection records at least include a fifth vehicle identification field, a seventh time field and a first accident type sequence field; the first accident type sequence field is composed of one or more accident types; The accident analysis database includes multiple first vehicle analysis data tables; the first vehicle analysis data tables correspond one-to-one to the first vehicles; the first vehicle analysis data tables are composed of multiple first vehicle accident analysis records; the first vehicle accident analysis records include at least a sixth vehicle identification field, an eighth time field and a first analysis report field.

5. The vehicle accident handling system according to claim 4, characterized in that: The data preprocessing module is specifically used for performing data preprocessing on the received first vehicle-side record to obtain corresponding processed data and storing it in the corresponding vehicle database in the data storage module. Extracting the first vehicle identification, the first timestamp, the first data type and the first uploaded data from the first vehicle-side record received at the time as the corresponding current vehicle identification, current timestamp, current data type and current uploaded data; and identifying the current data type; If the current data type is video, the first vehicle video data table corresponding to the current vehicle identifier in the vehicle video database of the data storage module is used as the corresponding current data table; and based on the preset vehicle-side video data quality analysis rule, the video data quality analysis is performed on the current uploaded data to obtain a corresponding first analysis result; When the first analysis result is that the quality meets the standard, the current vehicle identification, the current timestamp and the current uploaded data are used as the corresponding first vehicle identification field, the first time field and the first vehicle side video field to form a corresponding first vehicle video record and add it to the current data table; the first analysis result includes the quality meeting the standard and the quality not meeting the standard; If the current data type is perception, the first vehicle perception data table corresponding to the current vehicle identifier in the vehicle perception database of the data storage module is used as the corresponding current data table; and based on the preset vehicle-side perception data quality analysis rule, the perception data quality analysis is performed on the current uploaded data to obtain a corresponding second analysis result; When the second analysis result is that the quality meets the standard, the current uploaded data is time synchronized and data interpolated based on the time scale of the current data table to obtain one or more corresponding first vehicle perception records and add them to the current data table; the second analysis result includes quality meeting the standard and quality not meeting the standard; If the current data type is status, the first vehicle status data table corresponding to the current vehicle identifier in the vehicle status database of the data storage module is used as the corresponding current data table; and based on the preset vehicle-side status data quality analysis rule, the status data quality analysis is performed on the current uploaded data to obtain a corresponding third analysis result; When the third analysis result is that the quality meets the standard, the current uploaded data is time synchronized and data interpolated based on the time scale of the current data table to obtain one or more corresponding first vehicle status records and add them to the current data table; the third analysis result includes quality meeting the standard and quality not meeting the standard; If the current data type is a log, the first vehicle log data table corresponding to the current vehicle identifier in the vehicle log database of the data storage module is used as the corresponding current data table; and based on a preset vehicle-side log data quality analysis rule, a log data quality analysis is performed on the current uploaded data to obtain a corresponding fourth analysis result; When the fourth analysis result is that the quality meets the standard, a corresponding first vehicle log record consisting of the current vehicle identification, the current timestamp and the current uploaded data as the corresponding fourth vehicle identification field, the fourth time field and the first log sequence field is added to the current data table; the fourth analysis result includes quality meeting the standard and quality not meeting the standard.

6. The vehicle accident handling system according to claim 5, characterized in that: The data preprocessing module is specifically used for performing time synchronization and data interpolation processing on the current uploaded data based on the time scale of the current data table to obtain the corresponding one or more first vehicle perception records and add them to the current data table. Using the currently uploaded data as the corresponding current perception target data set; and recording the second time fields of the last two first vehicle perception records in the current data table as the corresponding first time and second time; and recording the current timestamp corresponding to the current uploaded data as the corresponding third time; and recording the absolute time difference between the first and second times as the corresponding first time scale; and taking the time period from the first time to the third time as the corresponding first time period; And in the first time period, taking the first time as the first sampling point and performing time point sampling every first time scale to obtain corresponding multiple first sampling points; wherein the first and second first sampling points are aligned with the first and second times respectively; And all the first vehicle-side perception target fields of the first vehicle perception record corresponding to the first and second times form a first and second sampling point perception target data set corresponding to the first sampling points; and identifying whether the last obtained first sampling point can be aligned with the third time; if so, taking the current perceived target data set as the sampling point perceived target data set corresponding to the last first sampling point; if not, predicting the sampling point perceived target data set of the last first sampling point according to the uniform acceleration motion model based on the current perceived target data set and the time difference between the third time and the last first sampling point; and record the other sampling points between the second and the last of the first sampling points as corresponding first intermediate sampling points; and when the number of the first intermediate sampling points is not zero, record the perception targets with the same target identifier and target type in the three sampling point perception target data sets corresponding to the first, second and last of the first sampling points as corresponding first continuous targets; and estimate the target geometric dimensions, target coordinates, target speed, target lateral acceleration, target longitudinal acceleration and target orientation of the current continuous target at each of the first intermediate sampling points based on the target geometric dimensions, target coordinates, target speed, target lateral acceleration, target longitudinal acceleration and target orientation of each of the first continuous targets at the first, second and last of the first sampling points based on the preset vehicle-side perception data interpolation algorithm; and the sampling point perception target data of the current continuous target at each of the first intermediate sampling points is composed of the target identifier, target type and the target geometric dimensions, target coordinates, target speed, target lateral acceleration, target longitudinal acceleration and target orientation of each of the first continuous targets at each of the first intermediate sampling points; and the sampling point perception target data set of the current intermediate sampling point is composed of all the sampling point perception target data at each of the first intermediate sampling points; And traverse the sampling points one by one from the third first sampling point; and in the traversal process, take the currently traversed first sampling point as the corresponding current sampling point; and take the current vehicle identification and sampling point time corresponding to the current sampling point as the corresponding second vehicle identification field and the second time field; and take each perception target data in the sampling point perception target data set corresponding to the current sampling point as a corresponding first vehicle-side perception target field; and the second vehicle identification field, the second time field and all the first vehicle-side perception target fields corresponding to the current sampling point form a corresponding first vehicle perception record; and identify whether the current sampling point is the last first sampling point, if so, end the traversal, if not, go to the next first sampling point for traversal; At the end of the traversal, all the first vehicle perception records obtained this time are added to the current data table one by one in chronological order.

7. The vehicle accident handling system according to claim 5, characterized in that: The data preprocessing module is specifically used for performing time synchronization and data interpolation processing on the current uploaded data based on the time scale of the current data table to obtain the corresponding one or more first vehicle status records and add them to the current data table. Using the current uploaded data as the corresponding current motion state data; and recording the third time fields of the last two first vehicle status records in the current data table as the corresponding fourth time and fifth time; and recording the current timestamp corresponding to the current uploaded data as the corresponding sixth time; and recording the absolute time difference between the fourth and fifth times as the corresponding second time scale; and recording the time period from the fourth time to the sixth time as the corresponding second time period; And in the second time period, taking the fourth time as the first sampling point and performing time point sampling every second time scale to obtain a corresponding plurality of second sampling points; wherein the first and second second sampling points are aligned with the fourth and fifth times respectively; The first vehicle coordinate field, the first road section identification field, the first lane identification field, the first planning trajectory field, the first driving mode field, the first driving gear field, the first heading angle field, the first steering wheel angle field, the first vehicle speed field, the first longitudinal acceleration field, the first lateral acceleration field, the first yaw angular velocity field, the first roll angular velocity field, the first accelerator pedal opening and closing degree field, the first brake pedal opening and closing degree field and the first headlight status field of the first vehicle status record corresponding to the fourth and fifth times constitute the first and second sampling point motion state data corresponding to the second sampling points; and identifying whether the last obtained second sampling point can be aligned with the sixth time; if so, using the current motion state data as the sampling point motion state data corresponding to the last second sampling point; if not, predicting the sampling point motion state data of the last second sampling point according to a uniform acceleration motion model based on the current motion state data and the time difference between the sixth time and the last second sampling point; and recording other sampling points between the second and the last second sampling points as corresponding second intermediate sampling points; and when the number of the second intermediate sampling points is not zero, estimating the sampling point motion state data at each of the second intermediate sampling points based on the sampling point motion state data at the first, second and last second sampling points based on a preset vehicle-side motion state data interpolation algorithm; And starting from the third second sampling point, traverse the sampling points one by one; and in the traversal process, take the currently traversed second sampling point as the corresponding current sampling point; and take the current vehicle identification and sampling point time corresponding to the current sampling point as the corresponding third vehicle identification field and the third time field; and take the vehicle coordinates, road section identification, lane identification, planned trajectory, driving mode, driving gear, heading angle, steering wheel angle, vehicle speed, longitudinal acceleration, lateral acceleration, yaw angular velocity, roll angular velocity, throttle pedal opening degree, brake pedal opening degree and headlight status of the sampling point motion state data corresponding to the current sampling point as a group of corresponding first vehicle coordinate fields, first road section identification fields, first lane identification fields, first planned trajectory fields, first driving mode fields, first driving gear fields, first heading angle fields, first steering wheel angle fields, first vehicle speed fields, first longitudinal acceleration fields, first lateral acceleration fields, the first yaw angular velocity field, the first roll angular velocity field, the first accelerator pedal opening and closing degree field, the first brake pedal opening and closing degree field and the first headlight status field; and the third vehicle identification field, the third time field, the first vehicle coordinate field, the first road section identification field, the first lane identification field, the first planning trajectory field, the first driving mode field, the first driving gear field, the first heading angle field, the first steering wheel angle field, the first vehicle speed field, the first longitudinal acceleration field, the first lateral acceleration field, the first yaw angular velocity field, the first roll angular velocity field, the first accelerator pedal opening and closing degree field, the first brake pedal opening and closing degree field and the first headlight status field corresponding to the current sampling point form a corresponding first vehicle status record; and identify whether the current sampling point is the last second sampling point, if so, end the traversal, if not, go to the next second sampling point for traversal; At the end of the traversal, all the first vehicle status records obtained this time are added to the current data table one by one in chronological order.

8. The vehicle accident handling system according to claim 4, characterized in that: The data preprocessing module is specifically used for performing data preprocessing on the received first roadside record to obtain corresponding processed data and storing it in the corresponding roadside database in the data storage module. Extracting the corresponding first device identifier, the second timestamp, the second data type and the second uploaded data from the first roadside record received at the time as the corresponding current device identifier, current timestamp, current data type and current uploaded data; and identifying the current data type; If the current data type is video, the first roadside video data table corresponding to the current device identifier in the roadside video database of the data storage module is used as the corresponding current data table; and based on the preset roadside video data quality analysis rule, the video data quality analysis is performed on the current uploaded data to obtain a corresponding fifth analysis result; When the fifth analysis result is that the quality meets the standard, the current device identifier, the current timestamp and the current uploaded data are used as the corresponding first device identifier field, the fifth time field and the first roadside video field to form a corresponding first roadside video record and add it to the current data table; the fifth analysis result includes the quality meeting the standard and the quality not meeting the standard; If the current data type is perception, the first roadside perception data table corresponding to the current device identifier in the roadside perception database of the data storage module is used as the corresponding current data table; and based on the preset roadside perception data quality analysis rule, the perception data quality analysis is performed on the current uploaded data to obtain a corresponding sixth analysis result; When the sixth analysis result is that the quality meets the standard, the current uploaded data is time synchronized and data interpolated based on the time scale of the current data table to obtain the corresponding one or more first roadside perception records and add them to the current data table; the sixth analysis result includes quality meeting the standard and quality not meeting the standard.

9. The vehicle accident handling system according to claim 8, characterized in that: The data preprocessing module is specifically used for performing time synchronization and data interpolation processing on the current uploaded data based on the time scale of the current data table to obtain the corresponding one or more first roadside sensing records and add them to the current data table. Using the currently uploaded data as the corresponding current perception target data set; and recording the sixth time fields of the last two first roadside sensing records in the current data table as the corresponding seventh time and eighth time; and recording the current timestamp corresponding to the current uploaded data as the corresponding ninth time; and recording the absolute time difference between the seventh and eighth times as the corresponding third time scale; and taking the time period from the seventh time to the ninth time as the corresponding third time period; And in the third time period, taking the seventh time as the first sampling point and performing time point sampling every the third time scale to obtain corresponding multiple third sampling points; wherein the first and second third sampling points are aligned with the seventh and eighth times respectively; And all the first roadside perception target fields of the first roadside perception records corresponding to the seventh and eighth time points form the sampling point perception target data sets corresponding to the first and second third sampling points; and identifying whether the last third sampling point obtained can be aligned with the ninth time; if so, taking the current perception target data set as the sampling point perception target data set corresponding to the last third sampling point; if not, predicting the sampling point perception target data set of the last third sampling point according to the uniform acceleration motion model based on the current perception target data set and the time difference between the ninth time and the last third sampling point; and recording other sampling points between the second and the last third sampling points as corresponding third intermediate sampling points; and when the number of the third intermediate sampling points is not zero, recording the perception targets with the same target identification and target type in the three sampling point perception target data sets corresponding to the first, second and last third sampling points as corresponding second continuous targets; and based on the preset roadside perception data interpolation algorithm, according to the target geometric dimensions, target coordinates, target road section identification, target lane identification, target speed, target lateral acceleration, target longitudinal acceleration, target orientation and target state of each of the second continuous targets at the first, second and last third sampling points, the current continuous target at each of the third intermediate sampling points is interpolated. The target geometric size, target coordinates, target road section identification, target lane identification, target speed, target lateral acceleration, target longitudinal acceleration, target orientation and target state at the sampling point are estimated; and the target identification, target type and the geometric size, target coordinates, target road section identification, target lane identification, target speed, target lateral acceleration, target longitudinal acceleration, target orientation and target state of each of the second continuous targets at each of the third intermediate sampling points constitute the sampling point perception target data of the current continuous target at each of the third intermediate sampling points; and all the sampling point perception target data at each of the third intermediate sampling points constitute the sampling point perception target data set of the current intermediate sampling point; And starting from the third sampling point, traverse the sampling points one by one; and in the traversal process, take the third sampling point currently traversed as the corresponding current sampling point; and take the current device identification and sampling point time corresponding to the current sampling point as the corresponding second device identification field and the sixth time field; and take each perception target data in the sampling point perception target data set corresponding to the current sampling point as a corresponding first roadside perception target field; and the second device identification field, the sixth time field and all the first roadside perception target fields corresponding to the current sampling point form a corresponding first roadside perception record; and identify whether the current sampling point is the last third sampling point, if so, end the traversal, if not, go to the next third sampling point for traversal; And at the end of the traversal, all the first roadside sensing records obtained this time are added to the current data table one by one in chronological order.

10. The vehicle accident handling system according to claim 4, characterized in that: The accident detection and early warning module is specifically used to detect whether each of the first vehicles has an accident and the corresponding accident type according to the vehicle perception database, the vehicle status database, the vehicle log database and the detection model library to obtain corresponding detection data and store them in the accident detection database. The current time is regularly used as the corresponding end time according to the preset first time frequency; and the time obtained by subtracting the preset first duration from the current end time is used as the corresponding start time; The current start time and the current end time form the corresponding current most recent time period; and extracting the first vehicle perception records whose second time fields satisfy the current most recent time period in each of the first vehicle perception data tables in the vehicle perception database to form a corresponding first vehicle perception record sequence; and performing corresponding vehicle-side perception target feature vector extraction processing according to each of the first vehicle perception record sequences to obtain a corresponding first vehicle-side perception target vector sequence; and extracting the first vehicle state records whose third time fields in each of the first vehicle state data tables in the vehicle state database satisfy the current most recent time period to form a corresponding first vehicle state record sequence; and performing corresponding vehicle state feature vector extraction processing according to each of the first vehicle state record sequences to obtain a corresponding first vehicle state vector; and extracting the first vehicle log records whose fourth time fields in each of the first vehicle log data tables in the vehicle log database satisfy the current most recent time period to form a corresponding first vehicle log record sequence; and performing corresponding vehicle log feature vector extraction processing according to each of the first vehicle log record sequences to obtain a corresponding first vehicle log vector; The accident detection models corresponding to sudden acceleration abnormality, sudden deceleration abnormality, emergency start abnormality, emergency braking abnormality, emergency stop abnormality, sharp turn abnormality, sudden lane change abnormality, lane departure abnormality, manual takeover abnormality, vehicle collision abnormality, and system status abnormality in the detection model library are recorded as corresponding sudden acceleration abnormality detection model, sudden deceleration abnormality detection model, emergency start abnormality detection model, emergency braking abnormality detection model, emergency stop abnormality detection model, sharp turn abnormality detection model, sudden lane change abnormality detection model, lane departure abnormality detection model, manual takeover abnormality detection model, vehicle collision abnormality detection model, and system status abnormality detection model; and inputting the first vehicle state vector corresponding to each of the first vehicles into the sudden acceleration abnormality detection model, the sudden deceleration abnormality detection model, the emergency start abnormality detection model, the emergency braking abnormality detection model, the emergency stop abnormality detection model, the sharp turn abnormality detection model, the sudden lane change abnormality detection model, the lane departure abnormality detection model and the manual takeover abnormality detection model respectively to perform corresponding sudden acceleration abnormal event recognition, sudden deceleration abnormal event recognition, emergency start abnormal event recognition, emergency braking abnormal event recognition, emergency stop abnormal event recognition, sharp turn abnormal event recognition, sudden lane change abnormal event recognition, lane departure abnormal event recognition and manual takeover abnormal event recognition to obtain corresponding first, second, third, fourth, fifth, sixth, seventh, eighth and ninth event recognition results; and inputting the first vehicle-side perception target vector sequence and the first vehicle state vector corresponding to each of the first vehicles into the vehicle collision abnormality detection model to perform corresponding vehicle collision abnormal event recognition to obtain the corresponding tenth event recognition result; and inputting the first vehicle log vector corresponding to each of the first vehicles into the system state abnormality detection model to perform corresponding system state abnormality event recognition to obtain a corresponding eleventh event recognition result; the first, second, third, fourth, fifth, sixth, seventh, eighth, ninth, tenth and eleventh event recognition results all include abnormality occurrence and no abnormality occurrence; and extracting the event recognition results that are specifically abnormal from the first, second, third, fourth, fifth, sixth, seventh, eighth, ninth, tenth and eleventh event recognition results corresponding to each of the first vehicles to form a corresponding first vehicle abnormal result set; and performing a round of traversal on all the obtained first vehicle abnormal result sets; and in this round of traversal, taking the currently traversed first vehicle abnormal result set as the corresponding current vehicle abnormal result set; And when the current vehicle abnormal result set is not empty, all the accident types corresponding to the current vehicle abnormal result set are composed of a corresponding first accident type sequence field, and the current end time is used as a corresponding seventh time field, and the vehicle identification of the first vehicle corresponding to the current vehicle abnormal result set is used as a corresponding fifth vehicle identification field, and the fifth vehicle identification field, the seventh time field and the first accident type sequence field corresponding to the current vehicle abnormal result set are used to form a corresponding first vehicle accident detection record, which is added to the first vehicle detection data table corresponding to the current vehicle abnormal result set in the accident detection database.

11. The vehicle accident handling system according to claim 10, characterized in that: The accident detection and warning module is specifically used for extracting corresponding vehicle-side perception target feature vectors according to each of the first vehicle perception record sequences to obtain corresponding first vehicle-side perception target vector sequences. The total number of records of the first vehicle perception record sequence is counted to obtain a corresponding first total number M, and the total number of target identifiers of the first vehicle perception record sequence is counted to obtain a corresponding second total number N, and a two-dimensional matrix A with a shape of M×N is constructed; and all matrix elements a of the two-dimensional matrix A are i,j The unit vector of the two-dimensional matrix A is initialized with all zero vectors; each row of the two-dimensional matrix A corresponds to a vehicle-side sensing target, and each column corresponds to a time point; the two-dimensional matrix A consists of M×N matrix units a i,j Composition, 1≤column index i≤M, 1≤row index j≤N; each of the matrix units a i,j Used to store a corresponding unit vector, the vector data of each unit vector consists of a set of corresponding target geometric dimensions, target coordinates, target speed, target lateral acceleration, target longitudinal acceleration and target orientation; Performing a round of traversal of all the first vehicle perception records of the first vehicle perception record sequence; During this round of traversal, the first vehicle perception record currently traversed is recorded as the corresponding current record; the second time field of the current record is used as the corresponding current time; and each of the first vehicle-side perception target fields of the current record is polled; In the polling process, the first vehicle-side perception target field currently being polled is used as the corresponding current target, and the matrix unit a whose column index i in the two-dimensional matrix A matches the current time and whose row index j matches the target is used as the matrix unit a i,j Recorded as the corresponding current matrix unit, and the unit vector of the current matrix unit is reset by forming a corresponding current data vector with the target geometric size, target coordinates, target speed, target lateral acceleration, target longitudinal acceleration and target orientation of the current target; When all the first vehicle perception records of the first vehicle perception record sequence are traversed, the M matrix elements a in each row of the two-dimensional matrix A are replaced by i,j The extracted ones are sorted in chronological order to form a corresponding first vehicle side perception target vector; and the obtained N first vehicle side perception target vectors are sorted in sequence to form a corresponding first vehicle side perception target vector sequence.

12. The vehicle accident handling system according to claim 10, characterized in that: The accident detection and warning module is specifically used to extract the first vehicle coordinate field, the first road section identification field, the first lane identification field, the first planning trajectory field, the first driving mode field, the first driving gear field, the first heading angle field, the first steering wheel angle field, the first vehicle speed field, the first longitudinal acceleration field, the first lateral acceleration field, the first yaw angular velocity field, the first roll angular velocity field, the first accelerator pedal opening degree field, the first brake pedal opening degree field and the first headlight status field of each of the first vehicle status records in the first vehicle status record sequence to form a corresponding first single-point state vector when the corresponding vehicle status feature vector extraction processing is performed according to each of the first vehicle status record sequences to obtain the corresponding first vehicle status vector; and all the obtained first single-point state vectors are sorted in chronological order to form a corresponding first vehicle status vector.

13. The vehicle accident handling system according to claim 10, characterized in that: The accident detection and warning module is specifically used to extract the first log sequence field of each of the first vehicle log records in the first vehicle log record sequence as a corresponding first single-point log vector when the corresponding vehicle log feature vector extraction processing is performed according to each of the first vehicle log record sequences to obtain the corresponding first vehicle log vector; and all the obtained first single-point log vectors are sorted in chronological order to form a corresponding first vehicle log vector.

14. The vehicle accident handling system according to claim 4, characterized in that: The accident detection and warning module is specifically used to monitor the record addition events of each of the first vehicle detection data tables in the accident detection database when performing corresponding accident warning processing based on the latest accident detection results of each of the first vehicles; and when each of the first vehicle accident detection records is added to each of the first vehicle detection data tables, the first vehicle accident detection record most recently added in the current first vehicle detection data table is recorded as the corresponding current newly added record; and the current newly added record is brought into a preset warning message template for warning message conversion to obtain the corresponding current warning message; And send the current warning message to the preset accident warning interface.

15. The vehicle accident handling system according to claim 4, characterized in that: The accident analysis module is specifically used for analyzing the cause of the accident of each of the first vehicles according to the vehicle video database, the vehicle perception database, the vehicle status database, the vehicle log database, the roadside video database, the roadside perception database and the accident detection database to obtain corresponding analysis data and store them in the accident analysis database. Monitor the record addition events of each of the first vehicle detection data tables of the accident detection database; and when each of the first vehicle accident detection records is added to each of the first vehicle detection data tables, record the first vehicle accident detection record most recently added in the current first vehicle detection data table as the corresponding current newly added record; and use the first vehicle corresponding to the current newly added record as the corresponding current vehicle; The seventh time field of the newly added record is used as the corresponding first accident end time, and the time obtained by subtracting the preset second time length from the first accident end time is used as the corresponding first accident start time; and the first accident start time and the first accident end time constitute the corresponding first accident period; and extracting the first vehicle video records in the first vehicle video data table corresponding to the current vehicle in the vehicle video database, in which the first time field meets the first accident period, and sorting them in chronological order to form a corresponding first record sequence; and extracting the first vehicle perception records whose second time field satisfies the first accident period from the first vehicle perception data table corresponding to the current vehicle in the vehicle perception database and sorting them in chronological order to form a corresponding second record sequence; and extracting the first vehicle status records whose third time field in the first vehicle status data table corresponding to the current vehicle in the vehicle status database satisfies the first accident period and sorting them in chronological order to form a corresponding third record sequence; and extracting the first vehicle log records whose fourth time field in the first vehicle log data table corresponding to the current vehicle in the vehicle log database satisfies the first accident period and sorting them in chronological order to form a corresponding fourth record sequence; and extracting the first vehicle coordinate field, the first road section identification field and the first lane identification field of the last first vehicle state record of the third record sequence to form corresponding current positioning data; and taking the first roadside device which is closest to the current positioning data as the corresponding current roadside device; and extracting the first roadside video records whose fifth time field in the first roadside video data table corresponding to the current roadside device in the roadside video database satisfies the first accident period, and sorting them in chronological order to form a corresponding fifth record sequence; and extracting the first roadside sensing records whose sixth time field in the first roadside sensing data table corresponding to the current roadside device in the roadside sensing database satisfies the first accident period, and sorting them in chronological order to form a corresponding sixth record sequence; The first, second, third, fourth, fifth and sixth record sequences are used to form a corresponding first accident analysis material set; the first accident analysis material set is sent to a preset accident cause / responsibility analysis interface; and the accident cause / responsibility analysis text sent back by the accident cause / responsibility analysis interface is received; The fifth vehicle identification field and the seventh time field of the currently newly added record are used as the corresponding sixth vehicle identification field and the eighth time field; the first accident type sequence field and the first accident time period, the first accident analysis material set and the accident cause / responsibility analysis text of the currently newly added record form a corresponding first analysis report field; and the sixth vehicle identification field, the eighth time field and the first analysis report field corresponding to the currently newly added record form a corresponding first vehicle accident analysis record, which is added to the first vehicle analysis data table corresponding to the current vehicle in the accident analysis database.

Images