A traffic conflict risk assessment method fusing multi-dimensional conflict indicators and cloud model

Abstract

The application relates to a traffic conflict risk assessment method combining multi-dimensional conflict indicators and a cloud model, which comprises the following steps: obtaining original trajectory data of a vehicle and cleaning the trajectory data; then, effectively identifying traffic conflicts based on a safety distance threshold and a time threshold; continuing to construct three-dimensional traffic conflict measurement indicators according to the effective traffic conflict trajectory data; determining the cloud model digital features of each conflict measurement indicator, establishing a traffic conflict severity assessment cloud model, calculating the mean value of the membership degree, and obtaining the weight of each conflict evaluation indicator in the comprehensive risk assessment by using the principal component analysis method; finally, based on the mean value of the membership degree and the weight value, the conflict is classified according to the weighted membership degree and the confidence degree criterion, and the traffic conflict severity classification is determined. The application can more comprehensively depict the real dynamic changes of the vehicle and the conflict risk severity, provides an efficient data processing flow and a modeling framework for the non-motor vehicle red light running conflict risk assessment, and provides scientific theoretical support and technical reference for the urban intersection traffic safety optimization and the control strategy formulation.

Description

Technical Field

[0001] This invention relates to a traffic conflict risk assessment method that integrates multi-dimensional conflict indicators and cloud models, belonging to the field of urban road traffic safety technology. Background Technology

[0002] Urban intersections are frequent locations for traffic accidents. With the increase in the number of non-motorized vehicles, intersection traffic accidents show a fluctuating upward trend. While the total number of deaths across all modes of transportation has decreased, the number of deaths involving electric bicycles is increasing year by year. Running red lights by non-motorized vehicles is the main cause of intersection traffic accidents. Therefore, it is necessary to conduct conflict risk assessments of non-motorized vehicles running red lights at intersections to take appropriate measures to ensure intersection traffic safety based on the assessment results.

[0003] Conflict risk assessment mainly includes two parts: determining conflict evaluation indicators and risk assessment. Currently, traffic conflict evaluation indicators are mainly divided into three categories: first, spatiotemporal proximity indicators, including Time to Conflict (TTC) and Time to Intrusion (PET); second, avoidance behavior indicators, including deceleration and yaw rate ratio; and third, indicators of the severity of potential collision consequences, such as Delta-V and ROC (Risk of Crash). Early traffic conflict assessments used single indicators. As conflict fields have become more complex, evaluation indicators have shifted from single to two-dimensional, primarily spatiotemporal proximity indicators and avoidance behavior indicators. These two types of indicators reflect the proximity of vehicles and the driver's reaction behavior to avoid collision, focusing on the conflict process but insufficiently considering the severity of potential collision consequences, which is a key factor in measuring conflict risk.

[0004] Traffic conflict risk has the following characteristics: First, both the risk level and evaluation indicators are fuzzy concepts without precise numerical boundaries; second, traffic conflict is a transitional state from normal vehicle movement to collision, and vehicle movement is random, making the occurrence of collisions and the risk level uncertain. Traditional deterministic risk assessment methods are prone to information loss when dealing with fuzzy and uncertain issues, leading to erroneous risk assessment results. Fuzzy theory, however, can effectively handle fuzzy and uncertain issues.

[0005] Furthermore, the same risk assessment indicator expresses different risks for frontal, lateral, and rear-end collisions. Existing methods often fail to assign differentiated weights to different collision risk levels, resulting in a significant deviation between the risk assessment level and the actual risk level.

[0006] Therefore, a three-dimensional evaluation index system is formed by incorporating indicators of the severity of potential collision consequences, the expected collision time index based on spatiotemporal proximity, and the avoidance degree index based on vehicle avoidance behavior. A cloud model, which can convert between qualitative concepts and quantitative data, is chosen as the risk assessment method to achieve a fuzzy transition between risk levels and resolve uncertainty issues. Risk assessments are conducted for various collision types, including frontal, lateral, and rear-end collisions, to obtain more accurate assessment results. Summary of the Invention

[0007] This invention provides a traffic conflict risk assessment method that integrates multi-dimensional conflict indicators and cloud models. It can more comprehensively depict the real dynamic changes of vehicles and the severity of conflict risks, providing an efficient data processing flow and modeling framework for non-motorized vehicles running red lights. It also provides scientific theoretical support and technical reference for the formulation of traffic safety optimization and control strategies at urban intersections.

[0008] The technical solution adopted by this invention to solve its technical problem is: A traffic conflict risk assessment method integrating multi-dimensional conflict indicators and cloud models includes the following steps: Step S1: Obtain the vehicle's original trajectory data and clean the trajectory data; Step S2: Based on the preset safe distance threshold and time threshold, perform effective conflict identification on the trajectory data obtained in step S1. First, determine whether the actual distance of the target vehicle group is less than the safe distance threshold. If it is less, determine that there is a potential conflict in the traffic conflict vehicle group. Then, based on the time threshold, determine whether there is a traffic conflict in the traffic conflict vehicle group, and determine the specific type of traffic conflict, such as a frontal conflict, a lateral conflict, or a rear-end collision. The time threshold is set to <1.5s. Step S3: Based on the effective traffic conflict trajectory data obtained in step S2, construct a traffic conflict measurement index; The traffic conflict measurement indicators include the expected collision time indicator based on spatiotemporal proximity, the degree of avoidance based on vehicle avoidance behavior, and the severity of potential collision consequences. Step S4: Based on the conflict measurement index constructed in step S3, establish a cloud model for assessing the severity of traffic conflicts, determine the digital features of the cloud model through the threshold of the conflict measurement index, and then use the FCG algorithm to obtain the membership degree. Step S5: Based on the conflict measurement index data obtained in Step S3, principal component analysis is used to obtain the weight of each conflict measurement index in the comprehensive risk assessment. Step S6: Calculate the weighted membership degree based on the membership degree obtained in step S4 and the weight value calculated in step S5, and obtain the standardized correlation coefficient through the normalization formula. Step S7: Based on the standardized correlation coefficients obtained in step S6, the conflicts are classified into levels according to the confidence criterion, and the severity of traffic conflicts is determined. Furthermore, in step S1, the process for cleaning the trajectory data is as follows: Step S11: Remove redundant trajectory data. Methods include ROI filtering, first and last point spacing filtering, and trajectory point count filtering. Step S12: Identify and remove total error. The identification method is to first use the modified KNN nearest neighbor algorithm to identify and remove outliers associated with vehicle ID errors, and then analyze the moving average and moving standard deviation of motion features such as speed to remove outliers caused by vehicle trajectory fluctuations. Step S13, remove random errors. The method for removing random errors is to use the Kalman filter method, which embeds vehicle acceleration and deceleration information, to smooth the trajectory data. Furthermore, in step S2, the steps for effective traffic conflict identification are as follows: Step S21: Based on the trajectory data cleaned in step S1, obtain the vehicle's position and speed in a continuous time series; Step S22: Calculate the Euclidean distance between each vehicle based on the current operating status of the vehicles. Calculate the safe distance by combining the driver's reaction time and braking performance. ; Step S23, if Euclidean distance <Safe distance If so, then the potential conflict is identified in the traffic conflict vehicle group; Step S24: Calculate the intrusion time for the screened potential conflict vehicle group data. If the intrusion time is later If the time threshold is exceeded, it is determined that there is a traffic conflict in the vehicle group, and the operating status and conflict characteristic information of the corresponding vehicle group are recorded. The formula for calculating the safe distance is as follows: (1) In the formula, For reaction distance, Braking distance, For vehicle speed, For reaction time, It is the acceleration due to gravity. The coefficient of friction of the ground; The formula for calculating the post-intrusion time is: (2) In the formula, For the time of subsequent intrusion, The time when the first vehicle left the conflict zone. The time when the second vehicle entered the conflict zone; Furthermore, in step S3, the calculation steps for the expected collision time index based on spatiotemporal proximity are as follows: Step S311, the motion parameters of the vehicle pair at any given time ( x, y, v x , v y , a x , a y , W, L Using this as input, a reference frame is set, and the expression for the vehicle's center point is obtained based on the input information. The formula for the center point of vehicle B is: (3) The formula for the center point of car A is: (4) Step S312, obtain the corner motion expressions. The formulas for all corners of car B are: (5) (6) (7) (8) The formula for all corner points of car A is: (9) (10) (11) (12) Step S313: Solve for t when the corner point coincides with the edge, which is the potential collision time; Step S314, select t The minimum value is the expected collision time, i.e. (13) In the formula, The potential collision time is when the point of collision is the corner of vehicle A. The potential collision time is when the point of collision is the B-corner of the vehicle. Furthermore, including time integrals, the formula for calculation is as follows: Including time integration The calculation formula is as follows: (15) In the formula, (14), Set the unsafe threshold for MACT. It lasts for 3 seconds. This is for simulating step size; Furthermore, in step S3, the formula for calculating the avoidance degree index based on vehicle avoidance behavior is as follows: (16) In the formula, a lon For the longitudinal acceleration of the vehicle, a lat This refers to the vehicle's lateral acceleration. Furthermore, in step S3, the calculation steps for the severity index of potential collision consequences are as follows: Step S331: Reconstruct the coordinate system with the orientation of the first vehicle as the X-axis. Then, according to the law of conservation of momentum, on the X-axis: (17) In the formula, , The masses of the two vehicles in the conflict vehicle group are respectively. The angle between the two vehicles before the collision. It is the angle between the directions of the two vehicles after the collision and the original direction of travel of the first vehicle. , These represent the speeds of the two vehicles in the conflict vehicle group before the collision. v The speeds of the two vehicles in the conflict vehicle group after the collision; Step S332, according to the law of conservation of momentum on the Y-axis: (18) Step S333, combining steps S331 and S332, the speeds of the two vehicles after the collision are expressed as follows: (19) Step S334, the formula for calculating the severity index of potential collision consequences is: (twenty two) In the formula, The velocity vector difference between the first vehicle in a collision group before and after the collision is expressed by the following formula: (20) The formula for calculating the velocity vector difference between the second vehicle in the collision group before and after the collision is: (twenty one) Furthermore, the specific steps of step S4 are as follows: Step S41, establish the cloud model as follows: (twenty three) set up U For the domain of quantitative discourse, i.e., concepts T The set of all possible quantitative values ​​corresponding to it. T It is the domain U The fuzzy qualitative concept above, if the quantitative value x satisfy x U and x It is a qualitative concept. T A random implementation of then x exist T The membership degree in a given number can be represented as a random number with a stable trend. μ ( x ),say( x , μ () represents a cloud droplet in a cloud model; Step S42: Calculate the digital features of the cloud model based on the 3En rule of normal clouds and the threshold values ​​of each parameter. Ex , En and He : (twenty four) In the formula, For the maximum value of the concept, The minimum value of the concept, the hyperentropy of the cloud model. He Choose a suitable constant based on the maximum range of each predictor. k ,set up He ≤0.5, k Take an empirical value of 0.01; Step S43: Generate normally distributed random numbers based on the three numerical characteristics of the normal cloud. ; Step S44: Generate normally distributed random numbers ; Step S45, based on steps S42-S44, according to the formula (25) Calculation is a qualitative concept. T membership degree ,in, i Indicates the first i The index of a cloud droplet i =1,2,..., N ; Step S46: Repeat steps S42-S45 until generated. N Cloud droplets in individual domains ( x i , μ i ); Step S47, calculate the first cloud for each cloud. The first cloud droplet The indicator belongs to the first p Membership of level And then according to X The principle of conditional cloud generator is to obtain the mean membership degree. And use this value as the final membership degree: (26) In the formula, The number of cloud droplets generated by the X-conditional cloud generator. For the first i Membership degree of each cloud droplet j For the number of indicators, j =1,2,…, m , k To determine the number of levels, k =1,2,…, p .

[0009] Furthermore, the specific steps of step S5 are as follows: Step S51: Based on the data obtained in step S3 for the three indicators, convert them into the original data matrix, calculate their mean and standard deviation, and obtain the standardized matrix. Z and its transpose matrix The calculation formula is: (27) In the formula, It is the first l The first collision type j The standardized value of an indicator object, It is the first l The first collision type j The actual value of the object of each indicator ， It is the first l The mean of the index across all collision types. It is the first l Standard deviation of the index in each collision type; Step S52, based on the standardized matrix obtained in step S51 Z and its transpose matrix The covariance matrix is ​​calculated to reflect the correlation between indicators. M The calculation formula is: (28) In the formula, M It is the covariance matrix; Step S53: Solve the characteristic equation to obtain the eigenvalues ​​of the covariance matrix. λand eigenvectors v : (29) In the formula, λ It is an eigenvalue. u It is the corresponding feature vector; Step S54: Select the components with a cumulative contribution rate of 70% or higher as principal components. Based on the eigenvectors of the calculated principal components, calculate the weights of each indicator in the comprehensive risk assessment. (30) (31) In the formula, Indicates the first r The variance contribution rate of each principal component r =1,2,…, m , Indicates the first r The eigenvalues ​​corresponding to each principal component are obtained by formula (29). t Auxiliary subscripts indicating summation. For the front q The cumulative variance contribution rate of each principal component q Indicates the number of principal components selected. q m ; (32) In the formula, To extract the principal component coefficient matrix that satisfies a cumulative contribution rate of over 70%, It can be obtained from formula (29); Step S55: Calculate the weight of each indicator in the comprehensive risk assessment. (33) In the formula, For the first j The relative importance of each indicator in the principal components For the first The first indicator in the The degree of contribution of each principal component For the t-th indicator at the th... Loading coefficients in each principal component; Furthermore, the step in step S6 to determine the severity classification is as follows: Step S61: Based on the membership degree of each conflict measurement index obtained in step S4 and the weight of each conflict index in the comprehensive risk assessment obtained in step S5, calculate the object... With severity level The normalized correlation coefficient between them The calculation process is as follows: (34) In the formula, ω j As an indicator j The weights; The step in step S7 to classify the severity of the conflict is as follows: Step S71: Assess the severity of the conflict using the confidence criterion, setting the confidence level to 1. * The current level of conflict severity for (35) In the formula, Take the empirical value of 0.6.

[0010] By employing the above technical solutions, the present invention has the following beneficial effects compared to the prior art: 1. The traffic conflict risk assessment method provided by this invention integrates multi-dimensional conflict indicators and cloud models, constructs a three-dimensional indicator system, makes up for the shortcomings of traditional single indicators that cannot cover the entire conflict process, and truly assesses the traffic risks caused by non-motorized vehicles running red lights. 2. The traffic conflict risk assessment method that integrates multi-dimensional conflict indicators and cloud models provided by this invention takes into account the dynamic changes in vehicle speed and orientation, making risk identification more accurate and more adaptable to the dynamic scenarios of complex intersections. 3. The traffic conflict risk assessment method provided by this invention integrates multi-dimensional conflict indicators and cloud models. It constructs a cloud model to handle uncertainty, uses expectation, entropy, and hyperentropy to describe fuzzy boundaries, and achieves a natural conversion between quantitative and qualitative analysis, resulting in higher robustness. 4. The traffic conflict risk assessment method provided by this invention integrates multi-dimensional conflict indicators and cloud models, and assigns weights to each conflict evaluation indicator in the three types of collisions: frontal, lateral, and rear-end collisions, which are in line with the differences in driver behavior under different conflicts, and the assessment results are closer to reality. Attached Figure Description

[0011] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0012] Figure 1 This is a schematic diagram of the traffic conflict risk assessment method that integrates multi-dimensional conflict indicators and cloud models provided by the present invention. Figure 2 This invention provides a first scenario in analyzing vehicle collision situations, which is the potential collision point when the corner of vehicle B touches the side of vehicle A. Figure 3This invention provides a second scenario in analyzing vehicle collision situations, which is the potential collision point when the corner of vehicle A touches the side of vehicle B. Figure 4 yes MACT and ACT A comparative diagram of the conflict trajectory (4a), conflict value (4b), and conflict vehicle speed (4c); Figure 5 This is a schematic diagram of an inelastic collision when calculating the severity index of potential collision consequences according to the present invention; Figure 6 These are example data of the results calculated from the expected collision time, avoidance degree index, and potential collision consequence severity index in the embodiments provided by the present invention; Figure 7 This is a schematic diagram of the distribution of conflicts in the three-dimensional index space provided by the present invention. Detailed Implementation

[0013] The invention will now be described in further detail with reference to the accompanying drawings.

[0014] As described in the background section, non-motorized vehicles running red lights has become a major cause of traffic conflicts, directly threatening traffic safety at intersections. However, existing methods for traffic conflict assessment have several significant shortcomings. For example, risk assessments do not consider the severity of potential collision consequences, resulting in an inability to comprehensively characterize conflict risks; conflict risks are inherently fuzzy and random, and existing deterministic risk assessment methods cannot effectively handle these fuzzy and uncertainties; the same risk assessment index expresses different risk levels for different collision types, and existing risk assessments use the same risk threshold for different collision types, leading to significant discrepancies between assessment results and actual risks.

[0015] Therefore, to address the aforementioned issues, this application provides a traffic conflict risk assessment method that integrates multi-dimensional conflict indicators and a cloud model. This method aims to more scientifically determine the severity of traffic conflicts caused by non-motorized vehicles running red lights, thereby improving the safety of urban intersections. First, this method combines computer vision trajectory extraction technology to obtain effective trajectory information, constructs a cloud theory assessment model integrating multiple indicators, then uses the cumulative frequency method to determine the cloud droplet characteristic parameters of each evaluation indicator, and applies principal component analysis to differentiate the weight allocation of three conflict indicators for three collision types (frontal, lateral, and rear-end collisions). Finally, based on weighted membership and confidence criteria, the conflicts are divided into several risk levels, achieving a natural transformation between quantitative measurement and qualitative assessment of conflict risk. Figure 1 The diagram shows the overall process of a traffic conflict risk assessment method that integrates multi-dimensional conflict indicators and cloud models, as provided in this application. The method includes the following steps: Step S1: Obtain the vehicle's raw trajectory data and clean the trajectory data. Due to factors such as algorithm misjudgment and tree occlusion, even after extracting the vehicle trajectory using an optimized high-quality model, the trajectory data still contains a large number of measurement errors. Measurement errors can lead to inaccurate calculations of vehicle speed and acceleration. Therefore, before further analysis, outlier removal and trajectory smoothing processing are required for the extracted trajectory data to reduce the impact of detection errors on subsequent analysis results.

[0016] Specifically, the process for cleaning trajectory data is as follows: Step S11: Remove redundant trajectory data. Methods include ROI filtering, first and last point spacing filtering, and trajectory point count filtering. Step S12: Identify and remove total error. The identification method is to first use the modified KNN nearest neighbor algorithm to identify and remove outliers associated with vehicle ID errors, and then analyze the moving average and moving standard deviation of motion features such as speed to remove outliers caused by vehicle trajectory fluctuations. Step S13: Remove random errors. The method for removing random errors is to use a Kalman filter method that embeds vehicle acceleration and deceleration information to smooth the trajectory data.

[0017] Step S2: Based on the preset safe distance threshold and time threshold, perform effective conflict identification on the trajectory data obtained in step S1. First, determine whether the actual distance of the target vehicle group is less than the safe distance threshold. If it is less, determine that there is a potential conflict in the traffic conflict vehicle group. Then, combined with the time threshold of less than 1.5 seconds, finally determine whether there is a traffic conflict in the traffic conflict vehicle group and determine which of the three collision types, namely frontal conflict, lateral conflict and rear-end collision, the conflict belongs to. To improve the efficiency of trajectory data processing, this application employs a traffic conflict identification method based on the mass point assumption in the traffic conflict discrimination stage. This method simplifies vehicles as point masses for calculation and comprehensively considers both safe distance and time thresholds for traffic conflict identification. The entire process is as follows: Step S21: Based on the trajectory data cleaned in step S1, obtain the vehicle's position and speed in a continuous time series; Step S22: Calculate the Euclidean distance between each vehicle based on the current operating status of the vehicles. Calculate the safe distance by combining the driver's reaction time and braking performance. ; Step S23, if Euclidean distance <Safe distance If so, then the potential conflict is identified in the traffic conflict vehicle group; Step S24: Calculate the intrusion time for the screened potential conflict vehicle group data. If the intrusion time is later If the time threshold is exceeded, it is determined that there is a traffic conflict in the vehicle group, and the operating status and conflict characteristic information of the corresponding vehicle group are recorded. The formula for calculating the safe distance is as follows: (1) In the formula, For reaction distance, Braking distance, For vehicle speed, For reaction time, It is the acceleration due to gravity. The coefficient of friction of the ground; After automatically filtering out traffic conflict vehicle groups with a Euclidean distance less than the safe distance using Python code, the PET (Peak Intrusion Time) of all filtered traffic conflict vehicle groups is calculated. Traffic conflict vehicle groups with a time threshold below the threshold are marked as conflicting vehicles, and the distance, PET, and operating status of the two vehicles are recorded. The formula for calculating the post-intrusion time is: (2) In the formula, For the time of subsequent intrusion, The time when the first vehicle left the conflict zone. The time when the second vehicle entered the conflict zone is recorded. After identifying potential traffic conflict events, various conflict indicators are further calculated. Subsequently, the trajectories of the conflicting vehicles are identified and matched according to the proposed conflict detection method, providing data support for the next step of conflict analysis.

[0018] Step S3: Based on the effective traffic conflict trajectory data obtained in step S2, construct traffic conflict measurement indicators; in order to comprehensively characterize the traffic conflict risk features, the traffic conflict measurement indicators described in this application include expected collision time indicators based on spatiotemporal proximity, avoidance degree indicators based on vehicle avoidance behavior, and potential collision consequence severity indicators.

[0019] The first expected collision time index based on spatiotemporal proximity is mainly used to describe the temporal and spatial proximity between traffic participants. Most studies use acceleration and angular direction as evidence and corrections for the prediction results, failing to directly input acceleration into the calculation. Furthermore, they all treat the vehicle as a point mass, ignoring its external dimensions, leading to significant deviations between the vehicle's future trajectory and its actual trajectory, thus affecting the collision time calculation. Therefore, this application is subject to... MTTC Inspired by ACT Based on this, a conflict measurement index that considers the dynamic changes in vehicle speed and direction is proposed. MACT A collision time calculation method that can cover all collision types in a two-dimensional plane is proposed, taking into account the vehicle's shape and its avoidance behavior.

[0020] The calculation steps are as follows: Step S311, the motion parameters of the vehicle pair at any given time ( x, y, v x , v y , a x , a y , W, L Using this as input, a reference frame is set, and the expression for the vehicle's center point is obtained based on the input information; Step S312: Obtain the corner motion expression; Step S313, solve for the case where the corner point coincides with the edge. This is the potential collision time; Step S314, for The minimum value is selected as the expected collision time. (13) In the formula, The potential collision time is when the point of collision is the corner of vehicle A. The potential collision time is when the collision point is the corner of vehicle B.

[0021] To clarify, the applicant provides the specific calculation process here, assuming the vehicles are A and B, with vehicle A as the reference vehicle and vehicle B as the moving vehicle, and setting the initial position of vehicle A as ( x A , y A ), the initial velocity is ( v Ax , v Ay ), acceleration is ( a Ax , a Ay The length and width of the rectangle are ( ), L A , W A Let the initial position of car B be ( ). x B , y B ), the initial velocity is ( v Bx , v By ), acceleration is ( a Bx , a By The length and width of the rectangle are ( ), L B , WB ); Then relative position ( x BA , y BA )for: , relative velocity ( v BAx , v BAy )for: , Relative acceleration ( a BAx , a BAy )for: , relative angle θ BA for: ; Car B at any time relative velocity ( v BAx ( t ), v BAy ( t )) is represented as: , Car B at any time θ BA ( t The relative angle of ) θ BA ( t ) is represented as: , The center point of car B at any time relative position ( x BA ( t ), y BA ( t )) is represented as: (3); Let the coordinates of the four corner points of car B be ( x FL ( t ), y FL ( t )), ( x FR ( t ), y FR ( t )), ( x RL (t ), y RL ( t )), ( x RR ( t ), y RR ( t The coordinates of the corner points are related to the trajectory and direction of movement of the center point, and can be obtained through rotation and translation. The rotation matrix is: ; Taking the four corners of car B as an example, (5) (6) (7) (8) The formula for all corner points of car A is: (9) (10) (11) (12) The first type, such as Figure 2 As shown, the corner of car B touches the edge of car A. Based on the motion equations of the corner points mentioned above, the trajectory of each corner point will intersect the extensions of the four sides of the reference vehicle A at four points. Input the initial state parameters of the two vehicles ( x, y, v x , v y , a x , a y , W, L ), and each corner point X or Y Substitute the coordinates into formula (5) and solve the system of equations to obtain the coordinates of the intersection point and the time. t Of the four intersection points, when x =± L A / 2, if y ∈[- W A / 2, W A / 2], then these points are retained as potential collision points, when y =± W A / 2, if x ∈[-L A / 2, L A If / 2], then retain it, and intersection points 2 and 3 in the diagram are potential collision points and record t. Based on the above process, calculate the potential collision points of the remaining corner points of car B. Then, the potential collision time of the corner point of car B colliding with the edge of car A can be expressed as { t B1 , t B2 , t B3 …}.

[0022] The second type, such as Figure 3 As shown, the corner of car A touches the edge of car B. This situation is slightly more complex. Before determining whether a corner point is in contact with an edge, we first need to define the expression for the edge. The expression for the edge is related to the vehicle's motion equation. Taking the front edge as an example, we construct the equation for the intersection of the corner point of A and the front edge of B: Substitute the coordinates of the four corner points of car A into the equation and solve. t Since the formula is a nonlinear equation, the bisection method is used to find an approximate solution. t Taking the front left corner of car A as an example ( L A / 2, W A For example, / 2), the solution obtained t Substituting into formula (5), we obtain the coordinates of the front corner of car B. x FL , y FL ), ( x FR , y FR ),because t Since the solution is not exact, its validity cannot be verified by checking whether the corner of car A lies on the edge of car B. Therefore, the criteria are relaxed here. L A / 2∈[ x FL， x FR ]and W A / 2∈[ y FL， y FR ], then t The potential collision time is recorded. Intersection 1 in the diagram is the potential collision point; the other three points are discarded. The potential collision time of a corner of car A colliding with an edge of car B can be represented as { tA1 , t A2 , t A3 …}.

[0023] By merging the potential collision times of the two scenarios above, the minimum collision time is selected as the MCT for the two vehicles. Unlike the TTC obtained through simulation, the above calculation method can provide an accurate solution for the expected collision time of the two vehicles in a two-dimensional plane.

[0024] To address the limitations of traditional TTC indicators, this application also includes an indicator describing overall accident risk, namely time integral. (TI-MACT), the time integral calculation formula is: (15) In the formula, (14), Set the unsafe threshold for MACT. It lasts for 3 seconds. This is for simulating step size; TI-MACT integrates MCT information over a period of time using the above formula, providing a weighted risk over the overall time dimension for a vehicle in a potentially hazardous state, making risk assessment more flexible.

[0025] The advantage of this design lies in the fact that MCT assesses the severity of a conflict risk by defining a certain threshold or range, while TI-MACT provides a more comprehensive description of the conflict process, fully considering the evasive actions a driver might take when approaching a collision, and allowing for adjustments to the risk assessment threshold based on different scenarios and conditions, making risk assessment more flexible. The above calculation method was programmed into Python and applied to the extracted trajectory data of potential conflict vehicles to calculate ACT and MCT for different collision types and initial vehicle speeds. The results are as follows... Figure 4 As shown (this is based on data from subsequent embodiments), MACT's main advantage lies in its ability to more accurately predict and assess conflict risks, especially in complex traffic environments where vehicle behavior is constantly changing. By considering vehicle acceleration and deceleration, MACT can identify potential hazards in advance, allowing for earlier intervention and more effective avoidance measures. In 4c, it can be observed that vehicle 1 briefly accelerated before decelerating to avoid a conflicting vehicle; because it considers vehicle acceleration and deceleration, MACT can capture the hazard signal earlier. As shown in 4b, because ACT does not consider changes in vehicle speed, its conflict risk warning is delayed compared to MACT. In summary, MACT significantly improves the predictive capability of conflict risk warning systems by more comprehensively analyzing real-time data and vehicle dynamics.

[0026] Regarding the second avoidance severity index, although MCT makes vehicle collision risk assessment more accurate and realistic by incorporating vehicle dynamic behavior, traffic participants will take evasive action such as deceleration, acceleration, or changing direction when they are aware of a potential conflict. However, MCT does not consider the avoidance behavior characteristics of non-motorized vehicles, resulting in certain limitations in risk assessment. Therefore, it is necessary to include an avoidance severity index to provide more comprehensive information when measuring the severity of a conflict. Based on this, an index EI is proposed, which considers the vehicle's braking deceleration and lateral avoidance behavior in risk scenarios to measure the severity of the vehicle's avoidance behavior in unsafe conditions. Its calculation formula is as follows: (16) In the formula, , The unsafe threshold for MACT is set to 3 seconds based on experience in this paper; The third indicator, the severity of potential collision consequences, reflects the collision risk of a vehicle in a conflict. However, considering only the collision probability may not fully reflect the overall risk of a vehicle in a conflict scenario. In reality, non-motorized vehicles, due to their structural fragility and the degree of exposure of their occupants, typically suffer more severe injuries in a collision. Therefore, the Expected Delta-V (EDV), a conflict indicator based on a dynamic model, is introduced to supplement the shortcomings of risk indicators in assessing the severity of potential consequences.

[0027] Based on the same assumptions used in calculating MCT, information such as speed, angle, and mass at the time of collision is obtained. In the dynamic model, the vehicle weight is referenced to the average weight of a mid-sized sedan, with a total weight of 1.4 tons including the driver. According to previous research, the weight of the electric bicycle is set at 50 kg. Assuming the standard weight of an adult is 70 kg, the total weight of the electric bicycle including the driver is 0.12 tons. The EDV formula is derived using the dynamic model based on the inelastic collision assumption proposed by Evans. A schematic diagram of the inelastic collision is shown below. Figure 5 As shown.

[0028] The steps for calculating the severity index of potential collision consequences are as follows: Step S331: Reconstruct the coordinate system with the orientation of the first vehicle as the X-axis. Then, according to the law of conservation of momentum, on the X-axis: (17) In the formula, , The masses of the two vehicles in the conflict vehicle group are respectively. The angle between the two vehicles before the collision. It is the angle between the directions of the two vehicles after the collision and the original direction of travel of the first vehicle. , These represent the speeds of the two vehicles in the conflict vehicle group before the collision. v The speeds of the two vehicles in the conflict vehicle group after the collision; Step S332, according to the law of conservation of momentum on the Y-axis: (18) Step S333, combining steps S331 and S332, the speeds of the two vehicles after the collision are expressed as follows: (19) Step S334, the formula for calculating the severity index of potential collision consequences is: (twenty two) In the formula, The velocity vector difference between the first vehicle in a collision group before and after the collision is expressed by the following formula: (20) The formula for calculating the velocity vector difference between the second vehicle in the collision group before and after the collision is: (twenty one) The severity of the entire potential collision consequences can be considered as two... v The largest one.

[0029] Step S4: Based on the conflict measurement index constructed in step S3, establish a cloud model for assessing the severity of traffic conflicts. Determine the digital characteristics of the cloud model through the threshold of the conflict measurement index, and then use the FCG algorithm to obtain the membership degree.

[0030] The specific steps are as follows: Step S41: The cloud model combines a qualitative concept with a quantitative description. It is a model that can naturally transform between the qualitative concept of traffic conflict safety levels and the degree of different safety level assessment indicators, addressing fuzziness and uncertainty distributions. Its core idea is to use three mathematical features—expectation Ex, entropy En, and hyperentropy He—to describe a qualitative concept. The cloud model is established as follows: (twenty three) set up U For the domain of quantitative discourse, i.e., concepts T The set of all possible quantitative values ​​corresponding to it. T It is the domain U The fuzzy qualitative concept above, if the quantitative value x satisfy x U and x It is a qualitative concept.T A random implementation of then x exist T The membership degree in a given number can be represented as a random number with a stable trend. μ ( x ),say( x , μ () represents a cloud droplet in a cloud model; Step S42: Calculate the digital features of the cloud model based on the 3En rule of normal clouds and the threshold values ​​of each parameter. Ex , En and He : (twenty four) In the formula, For the maximum value of the concept, The minimum value of the concept, the hyperentropy of the cloud model. He Choose a suitable constant based on the maximum range of each predictor. k ,set up He ≤0.5, k Take an empirical value of 0.01; Step S43: Generate normally distributed random numbers based on the three numerical characteristics of the normal cloud. ; Step S44: Generate normally distributed random numbers ; Step S45, based on steps S42-S44, according to the formula (25) Calculation is a qualitative concept. T membership degree ,in, i Indicates the first i The index of a cloud droplet i =1,2,..., N ; Step S46: Repeat steps S42-S45 until generated. N Cloud droplets in individual domains ( x i , μ i ); Step S47, calculate the first cloud for each cloud. The first cloud droplet The indicator belongs to the first p Membership of level And then according to X The principle of conditional cloud generator is to obtain the mean membership degree. And use this value as the final membership degree: (26) In the formula, The number of cloud droplets generated by the X-conditional cloud generator. For the first i Membership degree of each cloud droplet j For the number of indicators, j =1,2,…, m , k To determine the number of levels, k =1,2,…, p .

[0031] When assessing the overall risk of a conflict, the importance of each indicator may vary. Therefore, setting appropriate weights for the evaluation indicators to improve accuracy and effectiveness is a fundamental component of the assessment model. In this application, principal component analysis is used to determine the weights of each indicator to objectively reflect its importance. Driver behavior patterns may differ significantly across different collision types. For example, in a rear-end collision, drivers tend to brake to avoid the collision, while in a frontal collision, drivers may be more inclined to steer to avoid the collision with minimal speed loss. Calculating the weights of each indicator separately for different collision types helps to more accurately assess the driver's emergency response behavior. Therefore, this paper estimates the weights for frontal, lateral, and rear-end collisions separately according to the following steps to account for the heterogeneity of conflict indicators across different collision types.

[0032] Step S5: Based on the conflict measurement index data obtained in Step S3, principal component analysis is used to obtain the weight of each conflict measurement index in the comprehensive risk assessment. Specifically, based on step S3, all data for the three indicators are obtained, then converted into the original data matrix, and their mean and standard deviation are calculated to obtain the standardized matrix. Z and its transpose matrix The calculation formula is: (27) In the formula, It is the first l The first collision type j The standardized value of an indicator object, It is the first l The first collision type j The actual value of the object of each indicator ， It is the first l The mean of the index across all collision types. It is the first l Standard deviation of the index in each collision type; Step S52, based on the standardized matrix obtained in step S51 Z and its transpose matrix The covariance matrix is ​​calculated to reflect the correlation between indicators. M The calculation formula is: (28) In the formula, M It is the covariance matrix; Step S53: Solve the characteristic equation to obtain the eigenvalues ​​of the covariance matrix. λ and eigenvectors v : (29) In the formula, λ It is an eigenvalue. u It is the corresponding feature vector; Step S54: Select the components with a cumulative contribution rate of 70% or higher as principal components. Based on the eigenvectors of the calculated principal components, calculate the weights of each indicator in the comprehensive risk assessment. (30) (31) In the formula, Indicates the first r The variance contribution rate of each principal component r =1,2,…, m , Indicates the first r The eigenvalues ​​corresponding to each principal component are obtained by formula (29). t Auxiliary subscripts indicating summation. For the front q The cumulative variance contribution rate of each principal component q Indicates the number of principal components selected. q m ; (32) In the formula, To extract the principal component coefficient matrix that satisfies a cumulative contribution rate of over 70%, It can be obtained from formula (29); Step S55: Calculate the weight of each indicator in the comprehensive risk assessment. (33) In the formula, For the first j The relative importance of each indicator in the principal components For the first The first indicator in the The degree of contribution of each principal component For the t-th indicator at the th... Loading factors in each principal component.

[0033] Step S6: Calculate the weighted membership degree based on the membership degree obtained in step S4 and the weight value calculated in step S5, and obtain the standardized correlation coefficient through the normalization formula.

[0034] The steps to determine the standardized correlation coefficient are as follows: Step S61: Based on the membership degree of each conflict measurement index obtained in step S4 and the weight of each conflict index in the comprehensive risk assessment obtained in step S5, calculate the object... With severity level The normalized correlation coefficient between them The calculation process is as follows: (34) In the formula, ω j As an indicator j The weights; Step S7: Based on the standardized correlation coefficient obtained in step S6, the conflict is classified into levels according to the confidence criterion, and the severity of traffic conflict is determined.

[0035] The steps to determine the severity level of a conflict are as follows: Step S71: Assess the severity of the conflict using the confidence criterion, setting the confidence level to 1. * The current level of conflict severity for (35) In the formula, this study uses empirical values ​​based on previous research. This paper uses the cumulative frequency curve method to determine the threshold, and uses the 0-40%, 15-60%, 40-85%, and 60-100% quantile ranges as four severity levels, thus classifying the severity of conflict into four categories: low risk, medium risk, high risk, and extremely high risk. k =1,2,3,4. Example

[0036] Taking six typical signalized intersections in Nanjing as examples, to ensure the quantity and quality of the research samples, a survey method combining UAV aerial photography and ground manual investigation was used to obtain the driving trajectories of motor vehicles and non-motor vehicles at the intersections, the characteristics of non-motor vehicles running red lights, and the basic parameters of the intersections. To obtain the motion trajectory information of traffic participants, this application adopts an image detection and tracking framework combining YOLO-v8 and Deepsort multi-object tracking algorithms for trajectory data extraction. Motor vehicles and non-motor vehicles in the video are identified, and their position coordinates in each frame are obtained. The target detection and tracking model is continuously trained and optimized to improve the detection accuracy to over 90%. For the six intersections studied, trajectory data including ID, target frame number, center X coordinate, center Y coordinate, and bounding box width were successfully output, effectively extracting a total of 13,456,325 vehicle trajectory data points.

[0037] A three-step method—interest trajectory filtering, removal of total error, and elimination of random error—was used to efficiently clean trajectory data, extracting a total of 85,813 valid trajectories, providing basic data for traffic conflict identification.

[0038] The number of trajectories and sample points extracted at each intersection are shown in Table 1.

[0039] Table 1. Trajectories and number of sample points before and after data cleaning

[0040] Based on the potential conflict identification using a safe distance, the reaction time for both motor vehicle and non-motor vehicle drivers is set to 1.5 seconds. g and μ These represent the acceleration due to gravity and the coefficient of friction with the ground, respectively, used to measure the braking acceleration of a vehicle. For practical purposes, the braking acceleration of a typical motor vehicle is set to 7 m / s². 2 The braking acceleration of non-motorized vehicles is set to 5 m / s. 2 The bus travels at 6 meters per second. 2 .

[0041] Traffic conflict determination based on PET marks vehicle pairs that have been in conflict for less than 1.5 seconds as conflicting vehicles.

[0042] Continuing with the above method, the traffic conflict indices TI-MACT, EI, and EDV for each intersection are calculated and the data is then organized. Example data can be found here. Figure 6 .

[0043] The various indicators and the severity of conflicts were categorized into four levels: low risk (Level I), medium risk (Level II), high risk (Level III), and extremely high risk (Level IV). Then, 5000 traffic conflict data points collected from various intersections were grouped at 0.2-second intervals. For ease of calculation, the 15th, 40th, 60th, and 85th percentiles were used as thresholds for the expected values ​​of the four levels. Ex The 0-40%, 15-60%, 40-85%, and 60-100% quantile ranges were defined as the four severity levels. The thresholds for the three conflict indicators were calculated, and the results are shown in Table 2.

[0044] Table 2 Domains of each conflict indicator

[0045] The numerical characteristics of each conflict indicator were calculated based on cloud model theory and the data in Table 2, as shown in Table 3.

[0046] Table 3 Numerical characteristics of each conflict indicator

[0047] The weights of each conflict index in the three types of collisions are then calculated according to the weighting formula, as shown in Table 4.

[0048] Table 4. Weights of each conflict indicator

[0049] Finally, the severity of conflicts at all intersections was assessed, and [the following was taken]: The coefficients and grade marking results are shown in Table 5. Figure 7 The conflict level and its distribution across three indicators are shown in a three-dimensional graph.

[0050] Table 5. Conflict Severity Assessment Results

[0051] In summary, compared with the traditional traffic conflict research method that integrates spatiotemporal proximity indicators and risk avoidance behavior indicators, this application introduces indicators that reflect the severity of potential collision consequences, and constructs a conflict measurement index system that includes three dimensions: spatiotemporal proximity, risk avoidance behavior, and the severity of collision consequences. By introducing a third-dimensional indicator, the potential damage level of traffic conflicts can be characterized from the perspective of accident dynamics, providing a more comprehensive portrayal of the risk characteristics of traffic conflicts that are difficult to reflect the severity of potential collision consequences. Compared with traditional indicators, the MCT indicator directly inputs dynamic acceleration and angle factors into the equation of motion, describing vehicle motion as a two-dimensional uniformly accelerated motion process, and actually considering the characteristics such as the vehicle's external dimensions. It can more realistically reflect the speed changes and driving behavior of vehicles during the conflict process, thus showing more obvious advantages in terms of the accuracy of conflict identification, early warning, and adaptability to complex traffic scenarios. By introducing a cloud model, the randomness and fuzziness of the system can be characterized simultaneously within a unified framework. By using numerical features such as expectation, entropy, and hyperentropy to describe the mapping relationship between qualitative concepts and quantitative data, the fuzzy transition characteristics of conflict indicators between different risk levels can be clearly presented. Compared with traditional deterministic threshold classification methods, it has obvious advantages in handling traffic conflict assessment problems that combine randomness and fuzziness, and is more suitable for comprehensive judgment and assessment of complex traffic conflict risks at signalized intersections.

[0052] Those skilled in the art will understand that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. It should also be understood that terms such as those defined in general dictionaries should be understood to have the same meaning as in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless defined as herein.

[0053] The meaning of "and / or" as used in this application includes situations where each exists alone or both exist simultaneously.

[0054] The term "connection" as used in this application can mean a direct connection between components or an indirect connection between components through other components.

[0055] Based on the above-described preferred embodiments of the present invention, and through the foregoing description, those skilled in the art can make various changes and modifications without departing from the inventive concept. The technical scope of this invention is not limited to the contents of the specification, but must be determined according to the scope of the claims.

Claims

1. A traffic conflict risk assessment method integrating multi-dimensional conflict indicators and cloud models, characterized in that: Includes the following steps: Step S1: Obtain the vehicle's original trajectory data and clean the trajectory data; Step S2: Based on the preset safe distance threshold and time threshold, perform effective conflict identification on the trajectory data obtained in step S1. First, determine whether the actual distance of the vehicle group is less than the safe distance threshold. If it is less, determine that the traffic conflict vehicle group has a potential conflict. Then, based on the time threshold, determine whether the traffic conflict vehicle group has a traffic conflict and determine the specific type of the traffic conflict, such as a frontal conflict, a lateral conflict, or a rear-end collision. The time threshold is set to <1.5s. Step S3: Based on the effective traffic conflict trajectory data obtained in step S2, construct a traffic conflict measurement index; The traffic conflict measurement indicators include the expected collision time indicator based on spatiotemporal proximity, the degree of avoidance based on vehicle avoidance behavior, and the severity of potential collision consequences. Step S4: Based on the conflict measurement index constructed in step S3, establish a cloud model for assessing the severity of traffic conflicts, determine the digital features of the cloud model through the threshold of the conflict measurement index, and then use the FCG algorithm to obtain the membership degree. Step S5: Based on the conflict measurement index data obtained in Step S3, principal component analysis is used to obtain the weight of each conflict measurement index in the comprehensive risk assessment. Step S6: Calculate the weighted membership degree based on the membership degree obtained in step S4 and the weight value calculated in step S5, and obtain the standardized correlation coefficient through the normalization formula. Step S7: Based on the standardized correlation coefficient obtained in step S6, the conflict is classified into levels according to the confidence criterion, and the severity of traffic conflict is determined.

2. The traffic conflict risk assessment method integrating multi-dimensional conflict indicators and cloud models according to claim 1, characterized in that: In step S1, the process for cleaning the trajectory data is as follows: Step S11: Remove redundant trajectory data. Methods include ROI filtering, first and last point spacing filtering, and trajectory point count filtering. Step S12: Identify and remove total error. The identification method is to first use the modified KNN nearest neighbor algorithm to identify and remove outliers associated with vehicle ID errors, and then analyze the moving average and moving standard deviation of motion features such as speed to remove outliers caused by vehicle trajectory fluctuations. Step S13: Remove random errors. The method for removing random errors is to use a Kalman filter method that embeds vehicle acceleration and deceleration information to smooth the trajectory data.

3. The traffic conflict risk assessment method integrating multi-dimensional conflict indicators and cloud models according to claim 1, characterized in that: In step S2, the steps for effective traffic conflict identification are as follows: Step S21: Based on the trajectory data cleaned in step S1, obtain the vehicle's position and speed in a continuous time series; Step S22: Calculate the Euclidean distance between each vehicle based on the current operating status of the vehicles. Calculate the safe distance by combining the driver's reaction time and braking performance. ; Step S23, if Euclidean distance <Safe distance If so, then the potential conflict is identified in the traffic conflict vehicle group; Step S24: Calculate the intrusion time for the screened potential conflict vehicle group data. If the intrusion time is later If the time threshold is exceeded, it is determined that there is a traffic conflict in the vehicle group, and the operating status and conflict characteristic information of the corresponding vehicle group are recorded. The formula for calculating the safe distance is as follows: (1) In the formula, For reaction distance, Braking distance, For vehicle speed, For reaction time, It is the acceleration due to gravity. The coefficient of friction of the ground; The formula for calculating the post-intrusion time is: (2) In the formula, For the time of subsequent intrusion, The time when the first vehicle left the conflict zone. This refers to the time when the second vehicle entered the conflict zone.

4. The traffic conflict risk assessment method integrating multi-dimensional conflict indicators and cloud models according to claim 1, characterized in that: In step S3, the calculation steps for the expected collision time index based on spatiotemporal proximity are as follows: Step S311, the motion parameters of the vehicle pair at any given time ( x, y, v x , v y , a x , a y , W, L Using this as input, a reference frame is set, and the expression for the vehicle's center point is obtained based on the input information. The formula for the center point of vehicle B is: （3） The formula for the center point of car A is: （4） Step S312, obtain the corner motion expressions. The formulas for all corners of car B are: （5） （6） （7） （8） The formula for all corner points of car A is: （9） （10） （11） （12） Step S313: Solve for t when the corner point coincides with the edge, which is the potential collision time; Step S314, select t The minimum value is the expected collision time, i.e. (13) In the formula, The potential collision time is when the point of collision is the corner of vehicle A. The potential collision time is when the collision point is the corner of vehicle B.

5. The traffic conflict risk assessment method integrating multi-dimensional conflict indicators and cloud models according to claim 4, characterized in that: Including time integration The calculation formula is as follows: (15) In the formula, (14), Set the unsafe threshold for MACT. It lasts for 3 seconds. This is the simulated step size.

6. The traffic conflict risk assessment method integrating multi-dimensional conflict indicators and cloud models according to claim 1, characterized in that: In step S3, the formula for calculating the avoidance degree index based on vehicle avoidance behavior is as follows: (16) In the formula, a lon For the longitudinal acceleration of the vehicle, a lat This refers to the vehicle's lateral acceleration.

7. The traffic conflict risk assessment method integrating multi-dimensional conflict indicators and cloud models according to claim 1, characterized in that: In step S3, the calculation steps for the severity index of potential collision consequences are as follows: Step S331: Reconstruct the coordinate system with the orientation of the first vehicle as the X-axis. Then, according to the law of conservation of momentum, on the X-axis: (17) In the formula, , The masses of the two vehicles in the conflict vehicle group are respectively. The angle between the two vehicles before the collision. It is the angle between the directions of the two vehicles after the collision and the original direction of travel of the first vehicle. , These represent the speeds of the two vehicles in the conflict vehicle group before the collision. v The speeds of the two vehicles in the conflict vehicle group after the collision; Step S332, according to the law of conservation of momentum on the Y-axis: (18) Step S333, combining steps S331 and S332, the speeds of the two vehicles after the collision are expressed as follows: (19) Step S334, the formula for calculating the severity index of potential collision consequences is: (22) In the formula, The velocity vector difference between the first vehicle in a collision group before and after the collision is expressed by the following formula: (20) The formula for calculating the velocity vector difference between the second vehicle in the collision group before and after the collision is: (21) 8. The traffic conflict risk assessment method integrating multi-dimensional conflict indicators and cloud models according to claim 1, characterized in that: The specific steps of step S4 are as follows: Step S41, establish the cloud model as follows: (23) set up U For the domain of quantitative discourse, i.e., concepts T The set of all possible quantitative values ​​corresponding to it. T It is the domain U The fuzzy qualitative concept above, if the quantitative value x satisfy x U and x It is a qualitative concept. T A random implementation of then x exist T The membership degree in a given number can be represented as a random number with a stable trend. μ ( x ),say( x , μ () represents a cloud droplet in a cloud model; Step S42: Calculate the digital features of the cloud model based on the 3En rule of normal clouds and the threshold values ​​of each parameter. Ex , En and He : (24) In the formula, For the maximum value of the concept, The minimum value of the concept, the hyperentropy of the cloud model. He Choose a suitable constant based on the maximum range of each predictor. k ,set up He ≤0.5, k Take an empirical value of 0.01; Step S43: Generate normally distributed random numbers based on the three numerical characteristics of the normal cloud. ; Step S44: Generate normally distributed random numbers ; Step S45, based on steps S42-S44, according to the formula (25) Calculation is a qualitative concept. T membership degree ,in, i Indicates the first i The index of a cloud droplet i =1,2,..., N ; Step S46: Repeat steps S42-S45 until generated. N Cloud droplets in individual domains ( x i , μ i ); Step S47, calculate the first cloud for each cloud. The first cloud droplet The indicator belongs to the first p Membership of level And then according to X The principle of conditional cloud generator is to obtain the mean membership degree. And use this value as the final membership degree: (26) In the formula, The number of cloud droplets generated by the X-conditional cloud generator. For the first i Membership degree of each cloud droplet j For the number of indicators, j =1,2,…, m , k To determine the number of levels, k =1,2,…, p .

9. The traffic conflict risk assessment method integrating multi-dimensional conflict indicators and cloud models according to claim 1, characterized in that: The specific steps of step S5 are as follows: Step S51: Based on the data obtained in step S3 for the three indicators, convert them into the original data matrix, calculate their mean and standard deviation, and obtain the standardized matrix. Z and its transpose matrix The calculation formula is: (27) In the formula, It is the first l The first collision type j The standardized value of an indicator object, It is the first l The first collision type j The actual value of the object of each indicator ， It is the first l The mean of the index across all collision types. It is the first l Standard deviation of the index in each collision type; Step S52, based on the standardized matrix obtained in step S51 Z and its transpose matrix The covariance matrix is ​​calculated to reflect the correlation between indicators. M The calculation formula is: (28) In the formula, M It is the covariance matrix; Step S53: Solve the characteristic equation to obtain the eigenvalues ​​of the covariance matrix. λ and eigenvectors v : (29) In the formula, λ It is an eigenvalue. u It is the corresponding feature vector; Step S54: Select the components with a cumulative contribution rate of 70% or higher as principal components. Based on the eigenvectors of the calculated principal components, calculate the weights of each indicator in the comprehensive risk assessment. (30) (31) In the formula, Indicates the first r The variance contribution rate of each principal component r =1,2,…, m , Indicates the first r The eigenvalues ​​corresponding to each principal component are obtained by formula (29). t Auxiliary subscripts indicating summation. For the front q The cumulative variance contribution rate of each principal component q Indicates the number of principal components selected. q m ; (32) In the formula, To extract the principal component coefficient matrix that satisfies a cumulative contribution rate of over 70%, It can be obtained from formula (29); Step S55: Calculate the weight of each indicator in the comprehensive risk assessment. (33) In the formula, For the first j The relative importance of each indicator in the principal components For the first The first indicator in the The degree of contribution of each principal component For the t-th indicator at the th... Loading factors in each principal component.

10. The traffic conflict risk assessment method integrating multi-dimensional conflict indicators and cloud models according to claim 1, characterized in that: The steps in step S6 to determine the severity classification are as follows: Step S61: Based on the membership degree of each conflict measurement index obtained in step S4 and the weight of each conflict index in the comprehensive risk assessment obtained in step S5, calculate the object... With severity level The normalized correlation coefficient between them The calculation process is as follows: (34) In the formula, ω j As an indicator j The weights; The step in step S7 to classify the severity of the conflict is as follows: Step S71: Assess the severity of the conflict using the confidence criterion, setting the confidence level to 1. * The current level of conflict severity for (35) In the formula, Take the empirical value of 0.6.

Images