Driving behavior safety measurement and real-time warning method driven by vehicle trajectory data

Abstract

The present invention discloses a driving behavior safety measurement and real-time warning method driven by vehicle trajectory data. First, real spatial road network data and real traffic flow data are obtained, vehicle driving trajectory data in an NGSIM data set is selected, driving behavior safety measurement indicators are selected, and a driving behavior safety measurement comprehensive model is constructed. Then, the driving behavior safety measurement indicators are subjected to correlation analysis using a Pearson correlation coefficient method, and then a driving behavior safety measurement comprehensive model is obtained using a CARTIC weighting method and weighting. The thresholds of the driving behavior safety measurement indicators are compared using an interquartile range method, and a personalized, highly operable intelligent warning scheme is formulated by performing cluster analysis on the driving behavior safety measurement comprehensive model and combining the threshold analysis results. The present invention can provide technical support for the evaluation of the driving behavior and real-time warning of drivers in a road traffic system.

Description

Technical Field

[0001] The present invention can be applied to a safe driving assistance system or a safety warning platform in a vehicle-mounted system. Background Art

[0002] With the continuous development of the automobile industry, the growth rate of motor vehicle ownership and road traffic volume in my country is accelerating, and traffic safety has received more and more attention. Traffic accidents are mainly caused by drivers' bad driving behavior. Drivers' driving behavior plays an increasingly important role in traffic safety.

[0003] Existing research shows that correct identification and timely warning of bad driving behavior can effectively remind drivers to correct their driving behavior, thereby avoiding traffic accidents. At present, most of the evaluation of driving behavior focuses on the evaluation of bad driving behavior, such as the number and magnitude of bad driving behaviors such as serpentine driving, unstable speed, and frequent overtaking, but lacks a comprehensive evaluation of the specific conditions of various driving behaviors during vehicle driving; the collection of driving behavior data focuses on vehicle-mounted detectors, in-vehicle pedal data collectors, etc., and the results obtained are difficult to popularize in practical applications. The use of vehicle driving trajectory data can intuitively represent the various driving behaviors of drivers, and the vehicle trajectory data reflects the real traffic situation, which helps to evaluate the effectiveness of various traffic operations and control strategies in terms of traffic safety.

[0004] At present, the existing technology focuses more on the study of bad driving behavior. There is a lack of comprehensive evaluation of the specific circumstances of various driving behaviors during vehicle driving using vehicle trajectory data, a lack of real-time measurement of driving behavior, and no personalized and highly operational early warning solutions for different types of bad driving behaviors. Therefore, it cannot meet the technical needs of traffic safety management and control. Summary of the invention

[0005] In view of the above-mentioned defects of the prior art, the purpose of the present invention is to provide a driving behavior safety measurement and real-time warning method driven by vehicle trajectory data. The technical solution designed by the present invention includes the following steps, specifically including:

[0006] S1: Obtain vehicle driving trajectory data in the NGSIM dataset;

[0007] S2: Select driving behavior safety metrics, which include driving stability metrics, car-following risk metrics, and lane-changing safety metrics;

[0008] S3: A comprehensive model of driving behavior safety measurement is constructed using the risk measurement method, and the driver's driving behavior is comprehensively evaluated through driving stability measurement indicators, car-following risk measurement indicators, and lane-changing safety measurement indicators;

[0009] S4: Pearson correlation coefficient method is used to conduct correlation analysis on driving stability measurement index, car-following risk measurement index and lane-changing safety measurement index;

[0010] S5: The CARTIC weighting method is used to calculate the variability and conflict of the driving stability metric index, the car-following risk metric index, and the lane-changing safety metric index, and the weight of each indicator is obtained, and then the weighting is performed to obtain the comprehensive driving behavior safety metric model;

[0011] S6: Using the interquartile range method to extract the thresholds of driving stability metrics, car-following risk metrics, and lane-changing safety metrics;

[0012] S7: By performing cluster analysis on the comprehensive model of driving behavior safety metrics and combining the thresholds of driving stability metrics, car-following risk metrics, and lane-changing safety metrics, an intelligent early warning plan is developed.

[0013] Furthermore, in steps S1 and S2, since the vehicle driving trajectory data in the NGSIM dataset has problems such as noise and discontinuity, the present invention uses the DataFrame module in Python to analyze and process the trajectory data. In order to clarify the research object, the trajectory data is processed as follows:

[0014] S2001: Processing abnormal values ​​of vehicle driving trajectory data based on the DataFrame module, re-labeling data marked with the same vehicle ID under different vehicle trajectories, and eliminating abnormal vehicle driving trajectory data;

[0015] S2002: Convert the imperial units in the vehicle driving trajectory data to international standard units, and only keep the car data within the research scope as the research object;

[0016] S2003: recalculating the speed, acceleration and following distance of each vehicle in each frame according to the vehicle driving trajectory data coordinates;

[0017] S2004: searching for information of the front and rear adjacent vehicles of the vehicle through data matching, and only retaining vehicles with continuous frames and matching the front and rear vehicle information as the research object of the car-following risk measurement index;

[0018] S2005: Only vehicles with a single lane change behavior are retained as the research object of the lane change safety metric;

[0019] S2006: Using the Z-standardized scoring method, abnormality judgment is performed on the selected vehicle driving trajectory data and abnormal data is removed.

[0020] Furthermore, in step S2, for the driving stability metric, the lateral and longitudinal changes of the vehicle trajectory are considered, and the lateral displacement change is used to describe the driver's lateral driving behavior, and the acceleration change is used to describe the driver's longitudinal driving behavior; for the following risk metric, the inverse of the collision time is used as the metric parameter; for the lane change safety metric, the difference between the spatial distance between the front and rear vehicles and the parking distance during the lane change process is used as the judgment basis;

[0021] In order to achieve real-time safety measurement indicators for driving behavior, by comparing the trajectory data analysis results, 40 frames are used as the safety measurement range, and the comprehensive judgment of the driving behavior of each vehicle in the first 40 frames is used as the result of the driving behavior safety measurement indicator at this time.

[0022] Furthermore, in step S3, the driving stability measurement index includes a lateral displacement change index and a longitudinal speed change index; the lateral displacement change index evaluates the lateral stability of the driver during driving by describing the coefficient of variation of the lateral displacement offset of the target vehicle in each frame within 40 frames before frame t. The larger the lateral displacement change index, the greater the discrete degree of the lateral displacement, the higher the possibility of serpentine driving of the vehicle, and the worse the driving stability. The formula is as follows:

[0023] D x (t) = |X(t) - X(t-1)| (1)

[0024]

[0025] Where R1(t) is the lateral displacement change index, X(t) is the lateral coordinate of the target vehicle at frame t, and D x (t) is the lateral displacement of the target vehicle in the time period [t-1, t], std is the standard deviation, mean is the mean, and t is the preset value;

[0026] The longitudinal speed change index is calculated by dividing the absolute value of the difference between the target vehicle acceleration and the average acceleration of each vehicle in the first 40 frames by the total time period and then taking the square root to describe the change of the longitudinal speed. The larger the longitudinal speed change index, the greater the possibility of speed instability during driving and the worse the driving stability. The formula is as follows:

[0027]

[0028] Where R2(t) is the longitudinal velocity change index; a i represents the acceleration of the target vehicle at frame t; Represents the average acceleration of the target vehicle in the T time period; T is the total length of the time period before i frames; T, i are preset values;

[0029] The vehicle following risk measurement index is described by the inverse of the collision time TTC. The larger the index value is, the smaller the collision time TTC is, and the higher the risk in the following process is. The calculation formula is as follows:

[0030]

[0031] Where R3(t) is the vehicle following risk measurement index; Y F (t), V F (t) are the longitudinal coordinate and longitudinal velocity of the following vehicle respectively; Y P (t), V P (t), L P (t) are respectively the longitudinal coordinate, longitudinal speed and body length of the vehicle being followed;

[0032] Lane changing safety metric index, first calculate the difference between the distance between the lane changing target vehicle and its front and rear adjacent vehicles and the minimum safe distance, defined as DSS. When DSS>0, the distance between the front and rear vehicles is greater than the minimum safe distance, and the front and rear vehicles will not collide; when DSS<0, the distance between the front and rear vehicles is less than the minimum safe distance. The lane changing safety metric index is characterized by the ratio of the absolute value of DSS to the speed of the lane changing target vehicle. The missing reaction time of the target vehicle and its front vehicle in the same lane before the lane change, the front vehicle in the target lane, and the rear vehicle in the target lane is calculated, and the maximum value is taken to describe the lane changing safety situation. The larger the lane changing safety metric index, the longer the missing reaction time and the worse the lane changing safety. The specific calculation formula is as follows:

[0033]

[0034]

[0035] R4(t)=max(T op (t),T cp (t),T cf (t)) (7)

[0036] Where R4(t) is the lane change safety metric, u is the fraction rate, g is the acceleration of gravity, τ is the preset value of the reaction time, and v f (t) is the longitudinal speed of the following vehicle during the following process, T(t) is the missing reaction time required by the driver to avoid collision, and T op (t) represents the ratio of the absolute value of the DSS of the target vehicle and the preceding vehicle in the same lane to the longitudinal velocity of the target vehicle, T cp (t) represents the ratio of the absolute value of the DSS between the target vehicle and the vehicle in front of it in the target lane to the longitudinal velocity of the target vehicle, T cf (t) represents the ratio of the absolute DSS value of the target vehicle and the following vehicle on the target lane to the longitudinal speed of the following vehicle, and max is Top (t), T cp (t) and T cp The maximum value in (t).

[0037] Further, in step S5, the driving behavior safety evaluation of each vehicle at the i frame is calculated according to the weight, which is used to represent the driving behavior safety metric comprehensive model MOR obtained by integrating all driving behaviors of each vehicle at the i frame, where i is a preset value, and the calculation formula and distribution are as follows:

[0038] MOR=0.397R1+0.310R2+0.082R3+0.210R4 (8)

[0039] Wherein, MOR is the comprehensive model for measuring driving behavior safety; R1 is the lateral displacement change index, R2 is the longitudinal speed change index, and the two are collectively called the driving stability measurement index; R3 is the following risk measurement index; and R4 is the lane change safety measurement index.

[0040] Furthermore, in step S7, a cluster analysis is performed on the comprehensive model of driving behavior safety measurement, which is divided into four types of driving behaviors: dangerous, aggressive, safe, and conservative, to further judge the safety of vehicle driving. Combined with the threshold analysis results of the four measurement indicators, the evaluation of each measurement indicator of each type of driving behavior is studied and found. Through the threshold analysis of each indicator, a warning plan for the driver to adjust the driving behavior is formulated in a targeted manner.

[0041] Beneficial effects: The present invention is based on vehicle driving trajectory data, combined with the longitudinal and lateral driving characteristics of the vehicle, and defines four driving behavior risk measurement indicators from three perspectives, including driving stability, following and lane changing safety. A comprehensive driving behavior safety measurement model is established by weighting, and then combined with the threshold analysis of each indicator to obtain different types of bad driving behavior evaluation results, formulate a personalized and highly operational early warning plan, and provide technical support for the evaluation of drivers' driving behaviors and real-time early warnings in road traffic systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 is a schematic diagram of vehicle trajectory data processing according to a preferred embodiment of the present invention;

[0043] Figure 2 It is a schematic diagram of the interquartile range box curves of each coefficient of a preferred embodiment of the present invention;

[0044] Figure 3 It is a schematic diagram of real-time safety measurement of driving behavior according to a preferred embodiment of the present invention. DETAILED DESCRIPTION

[0045] The embodiments of the present invention are described in detail below. The following embodiments are implemented on the premise of the technical solution of the present invention, and detailed implementation methods and specific operation processes are given, but the protection scope of the present invention is not limited to the following embodiments.

[0046] The technical solution designed by the present invention comprises the following steps, specifically including:

[0047] S1: select the vehicle trajectory data in the NGSIM dataset;

[0048] S2: Selection of driving behavior safety metrics, specifically driving stability metrics, car-following risk metrics, and lane-changing safety metrics;

[0049] S3: A comprehensive model of driving behavior safety measurement is constructed using the risk measurement method, and the driver's driving behavior is comprehensively evaluated through driving stability measurement indicators, car-following risk measurement indicators, and lane-changing safety measurement indicators;

[0050] S4: Correlation analysis was performed on driving stability metric, car-following risk metric and lane-changing safety metric using the Pearson correlation coefficient method;

[0051] S5: The CARTIC weighting method is used to calculate the variability and conflict of the driving stability metric index, the car-following risk metric index, and the lane-changing safety metric index, and the weight of each indicator is obtained, and then the weighting is performed to obtain the comprehensive driving behavior safety metric model;

[0052] S6: The thresholds of driving stability metric, car-following risk metric and lane-changing safety metric were compared using the interquartile range method;

[0053] S7: By performing cluster analysis on the comprehensive model of driving behavior safety metrics and combining it with the results of threshold analysis, a personalized and highly operational intelligent warning plan is developed.

[0054] Furthermore, combined with Figure 1 In steps S1 and S2, since the vehicle driving trajectory data in the NGSIM data set has problems such as noise and discontinuity, the present invention uses the DataFrame module in Python to analyze and process the trajectory data. In order to clarify the research object, the trajectory data is processed as follows:

[0055] S2001: Processing the abnormal values ​​of trajectory data based on the DataFrame module, re-numbering the data marked with the same vehicle ID under different vehicle trajectories, eliminating abnormal vehicles with mismatched frame numbers due to lane changing behavior; eliminating abnormal data in the data collection process such as negative vehicle spacing;

[0056] S2002: Convert the imperial units in the trajectory data to international standard units, remove the data of trucks and motorcycles, and only keep the data of cars within the research scope as the research object;

[0057] S2003: renumbering the sample data vehicle IDs, and recalculating the speed, acceleration and following distance of each vehicle in each frame according to the trajectory data coordinates;

[0058] S2004: searching for information of the front and rear adjacent vehicles of the vehicle through data matching, and only retaining vehicles with continuous frames and matching the front and rear vehicle information as the research object of the car-following risk measurement index;

[0059] S2005: Regarding lane-changing behavior, the data analysis of a vehicle changing lanes once and multiple times during the research period is quite different. The present invention only considers vehicles with a single lane-changing behavior as the research object of the lane-changing safety metric indicator;

[0060] S2006: Using the Z-standardized scoring method, abnormality judgment is performed on the screened trajectory data and abnormal data is removed.

[0061] Furthermore, in step S2, for the driving stability measurement index, the present invention takes into account the lateral and longitudinal changes of the vehicle trajectory, uses the lateral displacement change to describe the driver's lateral driving behavior, and uses the acceleration change to describe the driver's longitudinal driving behavior; for the following risk measurement index, the inverse of the collision time is used as the measurement parameter; for the lane changing safety measurement index, the difference between the spatial distance between the front and rear vehicles and the parking distance during the lane changing process is used as the basis for judgment; in order to achieve real-time safety measurement indicators for driving behavior, by comparing the trajectory data analysis results, the present invention uses 40 frames as the safety measurement range, that is, the comprehensive judgment of the driving behavior of each vehicle in the first 40 frames is used as the result of the driving behavior safety measurement index at this time.

[0062] Furthermore, in step S3, the driving stability measurement index includes a lateral displacement change index and a longitudinal speed change index; the lateral displacement change index evaluates the lateral stability of the driver during driving by describing the coefficient of variation of the lateral displacement offset of the target vehicle in each frame within 40 frames before frame t. The larger the lateral displacement change index, the greater the discrete degree of the lateral displacement, the higher the possibility of serpentine driving of the vehicle, and the worse the driving stability. The formula is as follows:

[0063] D x (t) = |X(t) - X(t-1)| (1)

[0064]

[0065] Where R1(t) is the lateral displacement change index, X(t) is the lateral coordinate of the target vehicle at frame t, and D x(t) is the lateral displacement of the target vehicle in the time period [t-1, t], std is the standard deviation, mean is the mean, and t is the preset value;

[0066] The longitudinal speed change index is calculated by dividing the absolute value of the difference between the target vehicle acceleration and the average acceleration of each vehicle in the first 40 frames by the total time period and then taking the square root to describe the change of the longitudinal speed. The larger the longitudinal speed change index is, the greater the possibility of unstable speed during driving and the worse the driving stability. The formula is as follows:

[0067]

[0068] Where R2(t) is the longitudinal velocity change index; a i represents the acceleration of the target vehicle at frame t; Represents the average acceleration of the target vehicle in the T time period; T is the total length of the time period before i frames; T, i are preset values;

[0069] The vehicle following risk measurement index is described by the inverse of the collision time TTC. The larger the index value is, the smaller the collision time TTC is, and the higher the risk in the following process is. The calculation formula is as follows:

[0070]

[0071] Where R3(t) is the vehicle following risk measurement index; Y F (t), V F (t) are the longitudinal coordinate and longitudinal velocity of the following vehicle respectively; Y P (t), V P (t), L P (t) are respectively the longitudinal coordinate, longitudinal speed and body length of the vehicle being followed;

[0072] The lane-changing safety metric first calculates the difference between the distance between the lane-changing target vehicle and its front and rear adjacent vehicles and the minimum safe distance, which is defined as DSS. That is, when DSS>0, it means that the distance between the front and rear vehicles is greater than the minimum safe distance, and the front and rear vehicles will not collide; when DSS<0, it means that the distance between the front and rear vehicles is less than the minimum safe distance, that is, if the front vehicle stops in an emergency, the rear vehicle does not have enough distance to slow down, and there is a risk of collision between the front and rear vehicles. The safety metric is characterized by the ratio of the absolute value of DSS to the speed of the lane-changing target vehicle. The missing reaction time of the target vehicle and its front vehicle in the same lane before the lane change, the front vehicle in the target lane, and the rear vehicle in the target lane are calculated, and the maximum value is taken to describe the lane-changing safety situation. The larger the lane-changing safety metric, the longer the missing reaction time and the worse the lane-changing safety. The specific calculation formula is as follows:

[0073]

[0074]

[0075] R4(t)=max(T op (t),T cp (t),T cf (t)) (7)

[0076] Where R4(t) is the lane-changing safety metric, u is the fractional rate, g is the gravitational acceleration, τ is the preset value of the reaction time (when the rear vehicle accelerates, τ is 1.5s, when the rear vehicle decelerates or travels at a constant speed, τ is 0.7s), v f (t) is the longitudinal speed of the following vehicle during the following process, T(t) is the missing reaction time required by the driver to avoid collision, and T op (t) represents the ratio of the absolute value of the DSS of the target vehicle and the preceding vehicle in the same lane to the longitudinal velocity of the target vehicle, T cp (t) represents the ratio of the absolute value of the DSS between the target vehicle and the vehicle in front of it in the target lane to the longitudinal velocity of the target vehicle, T cf (t) represents the ratio of the absolute DSS value of the target vehicle and the following vehicle on the target lane to the longitudinal speed of the following vehicle, and max is T op (t), T cp (t) and T cp The maximum value in (t).

[0077] Furthermore, in step S4, the Pearson correlation coefficient method is used to perform correlation analysis on the above four measurement indicators, and the analysis results are shown in Table 1 below.

[0078] Table 1 Index correlation analysis table

[0079]

[0080] In Table 1, ** indicates that the correlation is significant at the 0.01 level (two-tailed), and * indicates that the correlation is significant at the 0.05 level (two-tailed).

[0081] It can be seen from Table 1 that the correlation coefficient between the lateral displacement change index and the longitudinal speed change index is 0.160, which is a very weak positive correlation. For other indicators, such as the correlation coefficient between the following risk index and the lane changing safety index, it is only 0.072, and the correlation coefficients are all less than 0.1. It is concluded that the four driving behavior evaluation indicators are extremely weakly correlated or unrelated.

[0082] Furthermore, in step S5, each indicator is normalized, and then the variability and conflict of each indicator are calculated using the CARTIC weighting method, and the weights of each indicator are obtained as shown in Table 2:

[0083] Table 2 Calculation results of weights of various indicators

[0084]

[0085]

[0086] The driving behavior safety evaluation of each vehicle at the i-frame is calculated based on the weight, which is used to represent the driving behavior safety measurement comprehensive model MOR obtained by integrating all driving behaviors of each vehicle at the i-frame, where i is a preset value. The calculation formula and distribution are as follows:

[0087] MOR=0.397R1+0.310R2+0.082R3+0.210R4 (8)

[0088] Wherein, MOR is the comprehensive model for measuring driving behavior safety; R1 is the lateral displacement change index, R2 is the longitudinal speed change index, and the two are collectively called the driving stability measurement index; R3 is the following risk measurement index; and R4 is the lane change safety measurement index.

[0089] Furthermore, in step S6, by comparing the driving behavior safety metric index threshold results, the two threshold methods are analyzed to what extent the index sample data is divided, as shown in the following table:

[0090] Table 3 Comparison of driving behavior safety measurement coefficient thresholds

[0091]

[0092] As can be seen from Table 3, the interquartile difference method is more detailed and accurate than the 96th percentile method. The 96th percentile threshold uniformly regards the last 4% of the value as abnormal data, while the interquartile difference method can extract different thresholds according to different data, so as to better judge the distribution of abnormal data. This study adopts the interquartile difference method. The box plots of the interquartile difference method of each coefficient are as follows: Figure 2 shown.

[0093] Furthermore, in step S7, the vehicle trajectory data of each vehicle can be used to calculate the driving behavior safety metric evaluation of the vehicle at each frame, and the driving behavior safety metric comprehensive model is clustered and divided into dangerous type, aggressive type, safe type and conservative type, so as to judge the safety of vehicle driving, such as Figure 3As shown. During the entire driving process, the driver showed four types of driving behaviors: dangerous, aggressive, safe, and conservative. Among them, the safe type accounted for the most, and the dangerous type of driving behavior only appeared in about 220 frames. Based on the clustering results of driving behaviors and combined with the threshold analysis results of four metrics, the study found the evaluation of each metric for each type of driving behavior, and formulated a targeted warning plan for the driver to adjust the driving behavior. If a driving behavior is clustered as dangerous, through the threshold analysis of each indicator, it is found that the values ​​of its MOR1 and MOR3 exceed the safety threshold. The warning plan provided to the driver is: stabilize the steering wheel and try to keep the longitudinal trajectory of the vehicle parallel to the lane line; reduce the speed of the vehicle and keep the vehicle and the vehicle in front at a safe following distance.

[0094] The preferred specific embodiments of the present invention are described in detail above. It should be understood that ordinary technicians in the field can make many modifications and changes based on the concept of the present invention without creative work. Therefore, all technical solutions that can be obtained by technicians in the technical field based on the concept of the present invention through logical analysis, reasoning or limited experiments on the basis of the prior art should be within the scope of protection determined by the claims.

Claims

1. A driving behavior safety measurement and real-time warning method driven by vehicle trajectory data, characterized in that: include: S1: Obtain vehicle driving trajectory data in the NGSIM dataset; S2: Select driving behavior safety metrics, which include driving stability metrics, car-following risk metrics, and lane-changing safety metrics; S3: A comprehensive model of driving behavior safety measurement is constructed using the risk measurement method, and the driver's driving behavior is comprehensively evaluated through driving stability measurement indicators, car-following risk measurement indicators, and lane-changing safety measurement indicators; S4: Pearson correlation coefficient method is used to conduct correlation analysis on driving stability measurement index, car-following risk measurement index and lane-changing safety measurement index; S5: The CARTIC weighting method is used to calculate the variability and conflict of the driving stability metric index, the car-following risk metric index, and the lane-changing safety metric index, and the weight of each indicator is obtained, and then the weighting is performed to obtain the comprehensive driving behavior safety metric model; S6: Using the interquartile range method to extract the thresholds of driving stability metrics, car-following risk metrics, and lane-changing safety metrics; S7: By clustering the comprehensive model of driving behavior safety metrics, combining the thresholds of driving stability metrics, car-following risk metrics, and lane-changing safety metrics, an intelligent early warning solution is developed; The S3 specifically includes: Driving stability metrics, including lateral displacement change and longitudinal speed change; The lateral displacement variation index evaluates the lateral stability of the driver during driving by describing the coefficient of variation of the lateral displacement offset of the target vehicle in each frame within the 40 frames before the t frame. The larger the lateral displacement variation index, the greater the discrete degree of the lateral displacement, the higher the possibility of serpentine driving of the vehicle, and the worse the driving stability. The formula is as follows: D x (t)=|X(t)-X(t-1)| (1) Where R1(t) is the lateral displacement change index, X(t) is the lateral coordinate of the target vehicle at frame t, and D x (t) is the lateral displacement of the target vehicle in the time period [t-1, t], std is the standard deviation, mean is the mean, and t is the preset value; The longitudinal speed change index is calculated by dividing the absolute value of the difference between the target vehicle acceleration and the average acceleration of each vehicle in the first 40 frames by the total time period and then taking the square root to describe the change of the longitudinal speed. The larger the longitudinal speed change index, the greater the possibility of speed instability during driving and the worse the driving stability. The formula is as follows: Where R2(t) is the longitudinal velocity change index; a i represents the acceleration of the target vehicle at frame t; Represents the average acceleration of the target vehicle in the T time period; T is the total length of the time period before i frames; T, i are preset values; The vehicle following risk measurement index is described by the inverse of the collision time TTC. The larger the index value is, the smaller the collision time TTC is, and the higher the risk in the following process is. The calculation formula is as follows: Where R3(t) is the vehicle following risk measurement index; Y F (t), V F (t) are the longitudinal coordinate and longitudinal velocity of the following vehicle respectively; Y P (t), V P (t), L P (t) are respectively represented by the longitudinal coordinate, longitudinal speed and body length of the vehicle being followed; Lane changing safety metric index, first calculate the difference between the distance between the lane changing target vehicle and its front and rear adjacent vehicles and the minimum safe distance, defined as DSS. When DSS>0, the distance between the front and rear vehicles is greater than the minimum safe distance, and the front and rear vehicles will not collide; when DSS<0, the distance between the front and rear vehicles is less than the minimum safe distance. The lane changing safety metric index is characterized by the ratio of the absolute value of DSS to the speed of the lane changing target vehicle. The missing reaction time of the target vehicle and its front vehicle in the same lane before the lane change, the front vehicle in the target lane, and the rear vehicle in the target lane is calculated, and the maximum value is taken to describe the lane changing safety situation. The larger the lane changing safety metric index, the longer the missing reaction time and the worse the lane changing safety. The specific calculation formula is as follows: R4(t)=max(T op (t),T cp (t),T cf (t)) (7) Where R4(t) is the lane-changing safety metric, u is the fraction rate, g is the gravitational acceleration, τ is the preset value of the reaction time, and v f (t) is the longitudinal speed of the following vehicle during the following process, T(t) is the missing reaction time required by the driver to avoid collision, and T op (t) represents the ratio of the absolute value of the DSS of the target vehicle and the preceding vehicle in the same lane to the longitudinal velocity of the target vehicle, T cp (t) represents the ratio of the absolute value of the DSS between the target vehicle and the vehicle in front of it in the target lane to the longitudinal velocity of the target vehicle, T cf (t) represents the ratio of the absolute DSS value of the target vehicle and the following vehicle on the target lane to the longitudinal speed of the following vehicle, and max is T op (t), T cp (t) and T cp The maximum value in (t).

2. The method for driving behavior safety measurement and real-time warning driven by vehicle trajectory data according to claim 1 is characterized in that: The S1 and S2 specifically include: The vehicle driving trajectory data is processed as follows: S2001: Processing abnormal values ​​of vehicle driving trajectory data based on the DataFrame module, re-labeling data marked with the same vehicle ID under different vehicle trajectories, and eliminating abnormal vehicle driving trajectory data; S2002: Convert the imperial units in the vehicle driving trajectory data to international standard units, and only keep the car data within the research scope as the research object; S2003: recalculating the speed, acceleration and following distance of each vehicle in each frame according to the vehicle driving trajectory data coordinates; S2004: searching for information of the front and rear adjacent vehicles of the vehicle through data matching, and only retaining vehicles with continuous frames and matching the front and rear vehicle information as the research object of the car-following risk measurement index; S2005: Only vehicles with a single lane change behavior are retained as the research object of the lane change safety metric; S2006: Using the Z-standardized scoring method, abnormality judgment is performed on the selected vehicle driving trajectory data and abnormal data is removed.

3. The method for driving behavior safety measurement and real-time warning driven by vehicle trajectory data according to claim 1 is characterized in that: The S2 specifically includes: For the driving stability metric, the lateral and longitudinal changes of the vehicle trajectory are taken into account. The lateral displacement change is used to describe the driver's lateral driving behavior, and the acceleration change is used to describe the driver's longitudinal driving behavior. For the following risk metric, the inverse of the collision time is used as the measurement parameter. For the lane change safety metric, the difference between the spatial distance between the front and rear vehicles and the parking distance during the lane change process is used as the judgment basis. In order to achieve real-time safety measurement indicators for driving behavior, by comparing the trajectory data analysis results, 40 frames are used as the safety measurement range, and the comprehensive judgment of the driving behavior of each vehicle in the first 40 frames is used as the result of the driving behavior safety measurement indicator at this time.

4. The method for driving behavior safety measurement and real-time warning driven by vehicle trajectory data according to claim 1, characterized in that: The S5 specifically includes: The driving behavior safety evaluation of each vehicle at the i-frame is calculated based on the weight, which is used to represent the driving behavior safety measurement comprehensive model MOR obtained by integrating all driving behaviors of each vehicle at the i-frame, where i is a preset value. The calculation formula and distribution are as follows: MOR=0.397R1+0.310R2+0.082R3+0.210R4 (8) Wherein, MOR is the comprehensive model for measuring driving behavior safety; R1 is the lateral displacement change index, R2 is the longitudinal speed change index, and the two are collectively called the driving stability measurement index; R3 is the following risk measurement index; and R4 is the lane change safety measurement index.

5. The method for driving behavior safety measurement and real-time warning driven by vehicle trajectory data according to claim 1 is characterized in that: The S7 specifically includes: A cluster analysis is performed on the comprehensive model of driving behavior safety measurement, which is divided into four types of driving behaviors: dangerous, aggressive, safe, and conservative. The safety of vehicle driving is then judged. Combined with the threshold analysis results of the four measurement indicators, the study finds the evaluation of each measurement indicator for each type of driving behavior. Through the threshold analysis of each indicator, a targeted early warning plan is formulated for the driver to adjust his driving behavior.

Images