Method for evaluating driving risk of mountainous highway operation area under multi-source dynamic risk field

Abstract

The present application relates to a method for evaluating driving risk in mountainous highway work zones under multi-source dynamic risk fields, belonging to the technical field of traffic safety. The method comprises collecting measured vehicle operation data in mountainous highway work zones; constructing a driving risk field force model integrating vehicle motion state and road environment in work zones; calibrating parameters of the driving risk field force model; determining risk threshold values of the driving risk field force model using collision time reciprocal and deceleration for avoiding collision; based on the calibrated driving risk field force model and the determined risk threshold values, calculating driving risk field force values of vehicles passing through the highway work zones, and performing spatial gridding visualization to identify high-risk areas under different work zones. The present application realizes precise quantification and visualized analysis of driving risk in mountainous highway work zones, thereby providing technical support for traffic safety management and dynamic risk early warning in work zones.

Description

Technical Field

[0001] This invention belongs to the field of traffic safety technology, specifically relating to a method for assessing driving risks in mountainous highway work areas under a multi-source dynamic risk field. Background Technology

[0002] The latest data from the Ministry of Transport shows that the total length of highways in China has reached 5.4904 million kilometers, of which 99.9% are under maintenance. It is estimated that approximately 30,000 kilometers of highways will require expansion and upgrades in the future. While half-width closures are a common method for periodic maintenance and expansion of highways, ensuring traffic flow, the reduction in lanes can lead to frequent lane changes and acceleration / deceleration, resulting in traffic congestion and increased accidents.

[0003] In the identification and assessment of risks in highway work areas, current methods for identifying driving risks can be broadly categorized into three types: identification methods based on historical accident datasets, identification methods based on alternative safety metrics (SSMs), and identification methods based on field theory. Among these, identification methods based on historical accident datasets are significantly limited by the scarcity of accident data in highway work areas. While identification methods based on alternative safety metrics (using time to collision (TTC), time to collision (WTTC), maximum deceleration rate (DRAC) for collision avoidance, and headway (TWH) as alternative safety metrics offer advantages such as computational simplicity and high interpretability, these traditional alternative safety metrics still exhibit the following drawbacks when used to measure driving risks in work areas:

[0004] First, it does not consider the impact of vehicle and road properties on collision risk;

[0005] Secondly, because the required lateral safety distance for vehicles is ignored, many risks associated with complex driving behaviors (such as lane changes) cannot be fully described;

[0006] Third, it is insufficient to measure the driving risk of a vehicle on a highway solely based on relative speed.

[0007] Given the inherent limitations of traditional SSM (Signal-Driven Model), potential field theory has been introduced to describe road driving risks. Its advantage lies in its ability to measure driving risks based on vehicle operation data, reducing reliance on accident data. Furthermore, the three subfields in the field theory model (static potential field, dynamic potential field, and behavioral field) can comprehensively describe the risks caused by people, vehicles, and roads. However, previous field theory models often used interactive objects as field sources. While this method effectively illustrates the distribution of collision risks on roads, its direct application to work area scenarios presents the following problems:

[0008] (1) It is difficult to determine the weight of the field generated by different interactive objects. In addition to vehicles traveling on the road, highway work areas also include other interactive objects, such as rigid guardrails, cones, temporary signs, construction workers and construction equipment. Although some traditional field theory models define the potential field of interactive objects such as cones or people, most models focus on vehicle-to-vehicle interaction and are difficult to fully reflect the complex environment of the work area.

[0009] (2) Many undetermined parameters in traditional field theory models need to be calibrated. Traditional field theory models generally include static potential fields (static obstacles, roads, lane lines), dynamic potential fields (moving objects), and behavioral fields (driving behavior), all of which have a large number of parameters to be calibrated. In order to meet the risk identification requirements of construction scenarios, parameter calibration work often takes a long time.

[0010] (3) When using interactive objects as field sources, the field strength of each interactive object must be calculated separately. In complex road environments such as highway operation zones, interactive objects are complex and diverse, not limited to vehicle-to-vehicle interactions in normal road environments, which leads to a sharp increase in the amount of computation.

[0011] (4) Distortion of risk superposition. Traditional methods obtain the comprehensive field force by superimposing the risk field forces of multiple interactive objects. However, under the condition of multi-vehicle interaction, even if the individual risk is low, it may be misjudged as high risk after superposition, resulting in identification bias.

[0012] While field theory models possess significant advantages in describing traffic risks, there is an urgent need to address some of the shortcomings of traditional field theory models to adapt them to risk assessment in the complex scenarios of highway operation zones. Therefore, overcoming the deficiencies of existing technologies is a pressing issue that needs to be resolved in the field of traffic safety technology. Summary of the Invention

[0013] The purpose of this invention is to address the shortcomings of existing technologies and provide a method for assessing driving risks in mountainous highway work areas under a multi-source dynamic risk field. This method aims to achieve accurate quantification and visualization analysis of driving risks in mountainous highway work areas, thereby providing technical support for traffic safety management and dynamic risk early warning in work areas.

[0014] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0015] The method for assessing driving risks in mountainous highway work areas under multi-source dynamic risk fields includes the following steps:

[0016] S1. Collect measured vehicle operation data in the mountainous highway operation area;

[0017] S2. Construct a driving risk field force model that integrates the vehicle's motion state and the road environment of the work area to quantify the interaction risk between the target vehicle and surrounding interactive objects; the driving risk field force model takes the target vehicle as the only risk field source, and its output risk value is the maximum value of the risk field force between the target vehicle and all interactive objects.

[0018] S3. Perform parameter calibration on the driving risk field force model;

[0019] S4. The risk threshold of the driving risk field force model is determined by using the reciprocal of collision time (TTCI) and the deceleration to avoid collision (DRAC).

[0020] S5. Based on the calibrated driving risk field force model and the determined risk threshold, calculate the driving risk field force value of vehicles passing through the highway operation area, and perform spatial grid visualization to identify high-risk areas under different operation areas.

[0021] Furthermore, the specific method of step S1 is as follows: collect video of vehicle operation in the mountain highway operation area; use video analysis software to extract vehicle trajectory data, obtain vehicle operation data including vehicle number, time, position, speed, acceleration and heading angle, and calibrate, verify and process the data.

[0022] Furthermore, step S1 includes the following steps:

[0023] S1.1 Data Collection Scope and Method: The standard work area of ​​a highway consists of six parts: warning area, upstream transition area, buffer zone, construction area, downstream transition area, and termination area. On a single-lane, two-way mountain highway, video data is collected from the work area, which includes the closed driving lanes or overtaking lanes of straight sections or curves. The weather must be clear during the collection period, and the layout of the work area must meet the specifications.

[0024] S1.2 Data Extraction:

[0025] First, import the collected aerial videos of the construction area into Kinovea or Tracker video analysis software; then, calibrate the scale of the highway lane width measured by a handheld rangefinder at the same location in the aerial video of the video analysis software.

[0026] Secondly, in the video analysis software, the direction of traffic flow is defined as the positive direction of the X-axis. The X-axis is established along the outer lane boundary line of the road. The Y-axis is established perpendicular to the X-axis at the starting point of the upstream transition zone. The positive direction of the Y-axis is along the opposite lane. The intersection of the two axes is the origin O, and a rectangular coordinate system is established.

[0027] Next, tracking points are manually established on the vehicle body and the vehicle's driving trajectory is automatically tracked to obtain the operating data of each vehicle; the vehicle operating data includes vehicle number, time, horizontal coordinate, vertical coordinate, lateral velocity, longitudinal velocity, speed, vehicle heading angle, lateral acceleration, longitudinal acceleration, and acceleration.

[0028] Finally, the lane width extracted by the video analysis software was compared with the highway lane width measured manually using a handheld push-type rangefinder to ensure that the aerial video was consistent with the actual scale. If they were inconsistent, the system was recalibrated. After successful verification, a portion of vehicle speed data acquired by the video analysis software at the same time and location was compared with the actual vehicle speed measured by a manual radar speed gun to verify the accuracy of the speed data extracted by the video analysis software. If the data were inconsistent, the parameters of the video analysis software were adjusted to make its output data consistent with the measured data.

[0029] S1.3 Data Processing: Due to frame skipping at tracking points, missing or outlier values ​​may appear in the data output by the video analysis software. Identify the missing and outlier values ​​in the data; if the number of frames corresponding to consecutive missing values ​​exceeds a preset threshold, they are removed; otherwise, interpolation methods are used to fill in the missing values.

[0030] If the number of frames corresponding to consecutive outliers exceeds a preset threshold, they are removed; otherwise, the mean or median is used to replace the outliers.

[0031] Furthermore, the preset threshold is 5 frames.

[0032] Furthermore, in step S2, the driving risk field force model is as follows:

[0033]

[0034] In the formula: For the target vehicle The greatest driving risk field force, For the target vehicle Facing the interaction object Driving risk field force, For the target vehicle Facing the interaction object The magnitude of the risk field, For the target vehicle Interacting with objects The contribution of interactive risks;

[0035]

[0036] In the formula: For the target vehicle Interacting with objects Interactive risk contribution, For the target vehicle The speed of the car, For interactive objects speed, For the target vehicle Heading angle;

[0037] When the vehicle's heading angle is 0, and the speed difference between the two is 0, When the speed difference is 5.56 m / s, that is, 20 km / h, ;

[0038] The driving risk field model is shown below:

[0039]

[0040] In the formula: For the target vehicle Facing the interaction object The risk field strength; , These are undetermined constants; For the target vehicle Equivalent quality; for Factors affecting road conditions; Lane line filtering factor; For the target vehicle The acceleration; For interactive objects Relative to the target vehicle The relative position vector is used to describe the spatial relative position and interaction direction of the target vehicle and the interactive object within the work area; For vectors Euclidean modulus, used to characterize the target vehicle and interaction objects The standardized spatial distance between them;

[0041] equivalent quality The formula is as follows:

[0042]

[0043] In the formula: For the target vehicle Type parameters; For the target vehicle The quality;

[0044] Road condition influencing factors The formula is as follows:

[0045]

[0046] In the formula: The road surface adhesion coefficient, For road slope, For road curvature, Visibility;

[0047] Target vehicle acceleration The formula is as follows:

[0048]

[0049] In the formula: , The target vehicles Lateral and longitudinal acceleration; , These are acceleration parameters; , They are respectively The angle between the x and y coordinate axes;

[0050] Lane line filtering factor The formula is as follows:

[0051]

[0052] In the formula: Lane line type parameter; For the target vehicle The distance between the current edge of the vehicle body and the lane line it crosses; For the target vehicle The distance between the edge of the vehicle body and the lane lines it crosses when driving along the center line of the current lane.

[0053] further, and The formula is as follows:

[0054]

[0055]

[0056] Where: target vehicle The coordinates are Interaction objects The coordinates are ; These are parameters to be determined. For the target vehicle speed; , For the target vehicle The length and width of the vehicle body;

[0057] When a vehicle has a steering angle while curving or changing lanes, the target vehicle The resulting risk field will also shift; coordinate system transformation is used to describe the overall deflection of the risk field with the vehicle's steering angle; with the target vehicle... Construct a coordinate system with the centroid as the origin when the vehicle is traveling in a straight line. When the vehicle's heading angle is When, its coordinate system is transformed , ( , ) is the interactive object The deflection coordinates of the point are shown below:

[0058]

[0059] at this time, and In the formula and After deflection and When the target vehicle's heading angle When the value is 0, no coordinate rotation is required;

[0060] For the target vehicle The type parameter, when the target vehicle is a passenger car. The value is 1.000; when the target vehicle is a cargo truck, The value is 1.443; when the target vehicle is a motorcycle, It is 0.335;

[0061] This is a lane line type parameter; when the lane line type is dashed, ... The value is 0.700; when the lane line type is a single solid line, It is 0.500.

[0062] Furthermore, step S3 includes the following steps:

[0063] S3.1. Based on the road traffic safety accident dataset, the road surface adhesion coefficient is... Road slope Road curvature ,visibility Lane line type parameters Vehicle type parameters Perform calibration;

[0064] S3.2, For the driving risk field force model , , , , These parameters are calibrated using a genetic algorithm.

[0065] Furthermore, in step S3.1:

[0066] In step S3.1:

[0067] The road surface adhesion coefficient is calibrated using a parameter calibration method based on statistical data. Road slope Road curvature ,visibility Lane line type parameters Vehicle type parameters If the parameter is a continuous variable, such as the road surface adhesion coefficient, then a polynomial is used to fit the relationship between the parameter and the property loss per unit accident, thereby completing the calibration; if it is a discrete variable, such as lane line type parameter, vehicle type parameter, visibility, road curvature, road slope, then the maximum property loss per unit accident is used as the standard value, and the following formula is used to establish a lookup table.

[0068]

[0069] In the formula: Discrete parameters Values The calibration value at that time Discrete parameters for Property losses caused by accidents at the workplace; Discrete parameters The set of accident property losses corresponding to all possible value states; for Losses due to accidents at the time; for Losses due to accidents at the time; for Losses due to accidents at the time;

[0070] In step S3.2: When using the genetic algorithm for calibration, the population size is 200, the number of iterations is 600, the crossover probability is 0.8, the mutation probability is 0.2, and the parameter value corresponding to the minimum root mean square error is taken as the final parameter value.

[0071] Furthermore, step S4 specifically includes:

[0072] The risk threshold of the driving risk field force model is determined using the reciprocal of the collision time (TTCI) and the collision avoidance deceleration (DRAC). The TTCI is the reciprocal of the collision time (TTC), and the conflict thresholds for the TTCI and DRAC are set to 0.2 s and 1.4 m / s, respectively. 2 The formulas are as follows:

[0073]

[0074]

[0075] in, , For the vehicle in front and the target vehicle; , For the vehicle in front and target vehicle exist The position at that moment; and For the vehicle in front and target vehicle exist The speed of time; , For vehicles Vehicle length and vehicle width; For the target vehicle exist The angle between the driving direction and the horizontal axis at any given time, i.e., the vehicle's heading angle;

[0076] If the time to collision (TTCI) of the target vehicle is greater than the conflict threshold of the time to collision (TTCI), and the deceleration rate (DRAC) for avoiding the collision is greater than the conflict threshold of the deceleration rate (DRAC) for avoiding the collision, then the target vehicle is considered a high-risk vehicle sample.

[0077] Then, the maximum driving risk field force value of all high-risk vehicle samples is calculated, and the minimum value among them is taken as the risk threshold of the driving risk field force model.

[0078] When the maximum driving risk field force experienced by any vehicle during operation exceeds the risk threshold of the driving risk field force model, it is determined to be in a high-risk state.

[0079] Furthermore, step S5 specifically includes:

[0080] The road area is divided into grids with a horizontal size of 2.5m and a vertical size of 0.5m. The driving risk field force value of vehicles passing through the highway operation area is calculated. The maximum driving risk field force value of each grid is displayed and reflected intuitively on the visualization map to reflect the spatial distribution of vehicle driving risk in the road environment. The darker the color, the higher the driving risk.

[0081] Video analysis software outputs data with missing or outlier values ​​due to frame skipping at tracking points. To address this, missing and outlier values ​​are identified, and interpolation methods are used to fill in the missing values. Outliers are replaced with the mean or median, and data with too many missing or outliers is discarded. Specifically, for trajectory data with a small number of consecutive missing frames or a limited number of outliers, the above methods are used for filling or replacement. Vehicle trajectory data with more than a threshold of consecutive missing frames or too many outliers is considered unreliable and is discarded.

[0082] This invention employs a data-driven method of "threshold inference based on high-risk samples" to determine the risk threshold of the vehicle risk field force (WTRF). High-risk vehicle samples are initially screened using the thresholds of the time-to-collision inverse (TTCI) and collision avoidance deceleration (DRAC). The WTRF model is then used to calculate the maximum WTRF value for these high-risk vehicle samples, and the minimum value is taken as the critical threshold. This threshold represents the lowest level of danger that the WTRF model can identify among vehicles deemed to have a risk of conflict by the TTCI and DRAC. When the WTRF force experienced by any vehicle during operation exceeds this threshold, it can be determined to be in a high-risk state. This invention constructs a vehicle risk field force model for work areas that uses the target vehicle as the sole field source and comprehensively considers vehicle motion characteristics and environmental factors of the work area, achieving accurate identification and dynamic assessment of vehicle operation risks in mountainous highway work areas.

[0083] Compared with the prior art, the beneficial effects of this invention are as follows:

[0084] 1. Traditional field theory models often use interacting objects as field sources, obtaining a comprehensive risk field by superimposing the potential fields of multiple objects. This results in computational complexity and distortion caused by risk superposition. The Vehicle Risk Field Force Model (WTRF) proposed in this invention uses the target vehicle as the sole field source. Based on the principle that "forces are mutual," it simultaneously calculates the field force interactions between the target vehicle and multiple interacting objects, and uses the maximum field force to characterize the vehicle's risk state. This fundamentally eliminates the misjudgment problem caused by the superposition of multiple fields, significantly reducing model complexity and computational load. Compared to the Time-to-Collision Inverse (TTCI) and Collision Avoidance Deceleration (DRAC), the WTRF model demonstrates superior performance in risk identification, improving identification accuracy by 70.25% and 64.46%, respectively.

[0085] 2. Due to the complexity and diversity of interactive objects in the work area environment, including vehicles, guardrails, cones, signs, construction personnel, and equipment, traditional methods struggle to reasonably determine the weights of each source. This invention avoids the multi-source weight allocation and complex parameter calibration process by modeling the target vehicle as a single source, resulting in a simpler model structure, fewer parameters, and higher computational efficiency.

[0086] 3. This invention introduces a vehicle motion state correction term into the driving risk field force model (WTRF) and superimposes the road environment influence factor of the work area, which can dynamically reflect the comprehensive influence of vehicle speed, acceleration, lane changing behavior and road geometry on risk distribution, thereby achieving accurate characterization of the unique traffic environment of the work area.

[0087] 4. The method of this invention is applicable to various mountainous highway operation zone scenarios, including closed driving lanes on straight sections, closed overtaking lanes on straight sections, closed driving lanes on curves, closed overtaking lanes on curves, and general highway sections. This model features strong versatility, high parameter transferability, and good scalability, and can be widely used in operation zone traffic safety assessment, risk monitoring, and dynamic early warning system design.

[0088] 5. The present invention can realize the spatial visualization of driving risks in the work area by calculating the risk field force distribution, and intuitively reflect the high-risk area and its spatiotemporal evolution characteristics. Attached Figure Description

[0089] Figure 1 This is a flowchart illustrating the driving risk assessment method for mountainous highway operation areas under a multi-source dynamic risk field according to the present invention.

[0090] Figure 2 This is a schematic diagram of the layout of a standard highway work area;

[0091] Figure 3 A schematic diagram illustrating the process of constructing a field force model for traffic risks in the work area;

[0092] Figure 4 This is a diagram illustrating vehicle coordinate transformation.

[0093] Figure 5 This is a schematic diagram of the interaction risk between other field theory models and the driving risk field force model in the work area; where (a) is the traditional safety potential field model and (b) is the driving risk field force model of this invention.

[0094] Figure 6The driving risk spatial distribution is shown in the field force model of the driving risk in the work area; where (a) driving risk distribution of decelerating straight driving; (b) driving risk distribution of constant speed straight driving; (c) driving risk distribution of accelerating straight driving; (d) driving risk distribution of decelerating turning; (e) driving risk distribution of constant speed turning; (f) driving risk distribution of accelerating turning; (g) driving risk distribution of decelerating lane changing; (h) driving risk distribution of constant speed lane changing; (i) driving risk distribution of accelerating lane changing.

[0095] Figure 7 The risk threshold of the driving risk field force model in the work area is determined; where (a) is the driving risk identification result diagram of TTCI and WTRF, and (b) is the driving risk identification result diagram of DRAC and WTRF.

[0096] Figure 8 The following are comparative analysis charts of the identification of driving risks in the work area by WTRF, TTCI, and DRAC; where (a) is the result chart of TTCI identification of driving risks in the work area; (b) is the result chart of DRAC identification of driving risks in the work area; and (c) is the result chart of WTRF identification of driving risks in the work area.

[0097] Figure 9 The diagrams show the comparison of driving risks with and without work zones under different road conditions; (a) shows the spatial distribution of driving risks when work zones occupy straight road lanes; (b) shows the spatial distribution of driving risks when there are no work zones on straight road lanes; (c) shows the spatial distribution of driving risks when work zones occupy curved road lanes; and (d) shows the spatial distribution of driving risks when there are no work zones on curved road lanes.

[0098] Figure 10 The diagrams show the comparison of driving risks with and without work zones under different road conditions. (a) shows the spatial distribution of driving risks when a work zone occupies the overtaking lane on a straight road; (b) shows the spatial distribution of driving risks when there is no work zone in the overtaking lane on a straight road; (c) shows the spatial distribution of driving risks when a work zone occupies the overtaking lane on a curve; and (d) shows the spatial distribution of driving risks when there is no work zone in the overtaking lane on a curve. Detailed Implementation

[0099] The present invention will now be described in further detail with reference to the embodiments.

[0100] Those skilled in the art will understand that the following embodiments are for illustrative purposes only and should not be construed as limiting the scope of the invention. Where specific techniques or conditions are not specified in the embodiments, they are performed in accordance with the techniques or conditions described in the literature in the field or according to the product instructions. Materials or equipment whose manufacturers are not specified are all conventional products that can be obtained by purchase.

[0101] Example 1

[0102] The method for assessing driving risks in mountainous highway work areas under multi-source dynamic risk fields includes the following steps:

[0103] S1. Collect measured vehicle operation data in the mountainous highway operation area;

[0104] S2. Construct a driving risk field force model that integrates the vehicle's motion state and the road environment of the work area to quantify the interaction risk between the target vehicle and surrounding interactive objects; the driving risk field force model takes the target vehicle as the only risk field source, and its output risk value is the maximum value of the risk field force between the target vehicle and all interactive objects.

[0105] S3. Perform parameter calibration on the driving risk field force model;

[0106] S4. The risk threshold of the driving risk field force model is determined by using the reciprocal of collision time (TTCI) and the deceleration to avoid collision (DRAC).

[0107] S5. Based on the calibrated driving risk field force model and the determined risk threshold, calculate the driving risk field force value of vehicles passing through the highway operation area, and perform spatial grid visualization to identify high-risk areas under different operation areas.

[0108] The specific method of step S1 is as follows: collect video of vehicle operation in the working area of ​​the mountain highway; use video analysis software to extract vehicle trajectory data, obtain vehicle operation data including vehicle number, time, position, speed, acceleration and heading angle, and calibrate, verify and process the data.

[0109] Example 2

[0110] The method for assessing driving risks in mountainous highway work areas under multi-source dynamic risk fields includes the following steps:

[0111] S1. Collect measured vehicle operation data in the mountainous highway operation area;

[0112] S2. Construct a driving risk field force model that integrates the vehicle's motion state and the road environment of the work area to quantify the interaction risk between the target vehicle and surrounding interactive objects; the driving risk field force model takes the target vehicle as the only risk field source, and its output risk value is the maximum value of the risk field force between the target vehicle and all interactive objects.

[0113] S3. Perform parameter calibration on the driving risk field force model;

[0114] S4. The risk threshold of the driving risk field force model is determined by using the reciprocal of collision time (TTCI) and the deceleration to avoid collision (DRAC).

[0115] S5. Based on the calibrated driving risk field force model and the determined risk threshold, calculate the driving risk field force value of vehicles passing through the highway operation area, and perform spatial grid visualization to identify high-risk areas under different operation areas.

[0116] The specific method of step S1 is as follows: collect video of vehicle operation in the working area of ​​the mountain highway; use video analysis software to extract vehicle trajectory data, obtain vehicle operation data including vehicle number, time, position, speed, acceleration and heading angle, and calibrate, verify and process the data.

[0117] Step S1 includes the following steps:

[0118] S1.1 Data Collection Scope and Method: The standard work area of ​​a highway consists of six parts: warning area, upstream transition area, buffer zone, construction area, downstream transition area, and termination area. On a single-lane, two-way mountain highway, video data is collected from the work area, which includes the closed driving lanes or overtaking lanes of straight sections or curves. The weather must be clear during the collection period, and the layout of the work area must meet the specifications.

[0119] S1.2 Data Extraction:

[0120] First, import the collected aerial videos of the construction area into Kinovea or Tracker video analysis software; then, calibrate the scale of the highway lane width measured by a handheld rangefinder at the same location in the aerial video of the video analysis software.

[0121] Secondly, in the video analysis software, the direction of traffic flow is defined as the positive direction of the X-axis. The X-axis is established along the outer lane boundary line of the road. The Y-axis is established perpendicular to the X-axis at the starting point of the upstream transition zone. The positive direction of the Y-axis is along the opposite lane. The intersection of the two axes is the origin O, and a rectangular coordinate system is established.

[0122] Next, tracking points are manually established on the vehicle body and the vehicle's driving trajectory is automatically tracked to obtain the operating data of each vehicle; the vehicle operating data includes vehicle number, time, horizontal coordinate, vertical coordinate, lateral velocity, longitudinal velocity, speed, vehicle heading angle, lateral acceleration, longitudinal acceleration, and acceleration.

[0123] Finally, the lane width extracted by the video analysis software was compared with the highway lane width measured manually using a handheld push-type rangefinder to ensure that the aerial video was consistent with the actual scale. If they were inconsistent, the system was recalibrated. After successful verification, a portion of vehicle speed data acquired by the video analysis software at the same time and location was compared with the actual vehicle speed measured by a manual radar speed gun to verify the accuracy of the speed data extracted by the video analysis software. If the data were inconsistent, the parameters of the video analysis software were adjusted to make its output data consistent with the measured data.

[0124] S1.3 Data Processing: Due to frame skipping at tracking points, missing or outlier values ​​may appear in the data output by the video analysis software. Identify the missing and outlier values ​​in the data; if the number of frames corresponding to consecutive missing values ​​exceeds a preset threshold, they are removed; otherwise, interpolation methods are used to fill in the missing values.

[0125] If the number of frames corresponding to consecutive outliers exceeds a preset threshold, they are removed; otherwise, the mean or median is used to replace the outliers.

[0126] The preset threshold is 5 frames.

[0127] In step S2, the driving risk field force model is as follows:

[0128]

[0129] In the formula: For the target vehicle The greatest driving risk field force, For the target vehicle Facing the interaction object Driving risk field force, For the target vehicle Facing the interaction object The magnitude of the risk field, For the target vehicle Interacting with objects The contribution of interactive risks;

[0130]

[0131] In the formula: For the target vehicle Interacting with objects Interactive risk contribution, For the target vehicle The speed of the car, For interactive objects speed, For the target vehicle Heading angle;

[0132] The driving risk field model is shown below:

[0133]

[0134] In the formula: For the target vehicle Facing the interaction object The risk field strength; , These are undetermined constants; For the target vehicle Equivalent quality; for Factors affecting road conditions; Lane line filtering factor; For the target vehicle The acceleration; For interactive objects Relative to the target vehicle The relative position vector; For vectors Euclidean mold length;

[0135] equivalent quality The formula is as follows:

[0136]

[0137] In the formula: For the target vehicle Type parameters; For the target vehicle The quality;

[0138] Road condition influencing factors The formula is as follows:

[0139]

[0140] In the formula: The road surface adhesion coefficient, For road slope, For road curvature, Visibility;

[0141] Target vehicle acceleration The formula is as follows:

[0142]

[0143] In the formula: , The target vehicles lateral acceleration and longitudinal acceleration; , These are acceleration parameters; , They are respectively The angle between the x and y coordinate axes;

[0144] Lane line filtering factor The formula is as follows:

[0145]

[0146] In the formula: Lane line type parameter; For the target vehicle The distance between the current edge of the vehicle body and the lane line it crosses; For the target vehicle The distance between the edge of the vehicle body and the lane lines it crosses when driving along the center line of the current lane.

[0147] and The formula is as follows:

[0148]

[0149]

[0150] Where: target vehicle The coordinates are Interaction objects The coordinates are ; These are parameters to be determined. , For the target vehicle The length and width of the vehicle body;

[0151] When a vehicle has a steering angle while curving or changing lanes, the target vehicle The resulting risk field will also shift; coordinate system transformation is used to describe the overall deflection of the risk field with the vehicle's steering angle; with the target vehicle... Construct a coordinate system with the centroid as the origin when the vehicle is traveling in a straight line. When the vehicle's heading angle is When, its coordinate system is transformed , ( , ) is the interactive object The deflection coordinates of the point are shown below:

[0152]

[0153] at this time, and In the formula and Using deflection as and When the target vehicle's heading angle When the value is 0, no coordinate rotation is required;

[0154] For the target vehicle The type parameter, when the target vehicle is a passenger car. The value is 1.000; when the target vehicle is a cargo truck, The value is 1.443; when the target vehicle is a motorcycle, It is 0.335;

[0155] This is a lane line type parameter; when the lane line type is dashed, ... The value is 0.700; when the lane line type is a single solid line, It is 0.500.

[0156] Step S3 includes the following steps:

[0157] S3.1. Based on the road traffic safety accident dataset, the road surface adhesion coefficient is... Road slope Road curvature ,visibility Lane line type parameters Vehicle type parameters Perform calibration;

[0158] S3.2, For the driving risk field force model , , , , These parameters are calibrated using a genetic algorithm.

[0159] In step S3.1:

[0160] The road surface adhesion coefficient is calibrated using a parameter calibration method based on statistical data. Road slope Road curvature ,visibility Lane line type parameters Vehicle type parameters If the parameter is a continuous variable, the relationship between the parameter and the property loss per unit accident is fitted using a polynomial to complete the calibration; if it is a discrete variable, the maximum property loss per unit accident is used as the standard value, and the following formula is used to establish a lookup table.

[0161]

[0162] In the formula: Discrete parameters Values The calibration value at that time Discrete parameters for Property losses caused by accidents at the workplace; Discrete parameters The set of accident property losses corresponding to all possible value states; for Losses due to accidents at the time; for Losses due to accidents at the time; for Losses due to accidents at the time;

[0163] In step S3.2: When using the genetic algorithm for calibration, the population size is 200, the number of iterations is 600, the crossover probability is 0.8, the mutation probability is 0.2, and the parameter value corresponding to the minimum root mean square error is taken as the final parameter value.

[0164] Step S4 is as follows:

[0165] The risk threshold of the driving risk field force model is determined using the reciprocal of the collision time (TTCI) and the collision avoidance deceleration (DRAC). The TTCI is the reciprocal of the collision time (TTC), and the conflict thresholds for the TTCI and DRAC are set to 0.2 s and 1.4 m / s, respectively. 2 The formulas are as follows:

[0166]

[0167]

[0168] in, , For the vehicle in front and the target vehicle; , For the vehicle in front and target vehicle exist The position at that moment; and For the vehicle in front and target vehicle exist The speed of time; , For vehicles Vehicle length and vehicle width; For the target vehicle exist The angle between the driving direction and the horizontal axis at any given time, i.e., the vehicle's heading angle;

[0169] If the time to collision (TTCI) of the target vehicle is greater than the conflict threshold of the time to collision (TTCI), and the deceleration rate (DRAC) for avoiding the collision is greater than the conflict threshold of the deceleration rate (DRAC) for avoiding the collision, then the target vehicle is considered a high-risk vehicle sample.

[0170] Then, the maximum driving risk field force value of all high-risk vehicle samples is calculated, and the minimum value among them is taken as the risk threshold of the driving risk field force model.

[0171] When the maximum driving risk field force experienced by any vehicle during operation exceeds the risk threshold of the driving risk field force model, it is determined to be in a high-risk state.

[0172] Step S5 is as follows:

[0173] The road area is divided into grids with a horizontal size of 2.5m and a vertical size of 0.5m. The driving risk field force value of vehicles passing through the highway operation area is calculated. The maximum driving risk field force value of each grid is displayed and reflected intuitively on the visualization map to reflect the spatial distribution of vehicle driving risk in the road environment. The darker the color, the higher the driving risk.

[0174] Example 3

[0175] A method for assessing driving risks in mountainous highway work areas under a multi-source dynamic risk field includes the following steps:

[0176] S1. Collect measured vehicle operation data in typical mountainous highway operation areas, covering construction scenarios with different road sections and different closure methods;

[0177] S1.1 Data Collection Scope and Method. The standard operating area of ​​a highway consists of six parts: warning zone (S), upstream transition zone (Ls), buffer zone (H), construction zone (G), downstream transition zone (Lx), and termination zone (Z). Figure 2As shown, the warning zone and upstream transition zone are high-risk accident areas due to frequent lane changes and acceleration / deceleration. Therefore, in-depth analysis of their driving risks is crucial for improving safety in the work area. Thus, data collection was conducted on a single-lane, two-way highway, covering both straight and curved sections, with the work area including closed driving and overtaking lanes. A continuous data collection method combining drones, cameras, and manual labor was employed, with clear weather and the work area layout conforming to regulations.

[0178] S1.2 Data extraction.

[0179] First, import the collected aerial videos of the construction area into Kinovea or Tracker video analysis software; then, calibrate the scale of the highway lane width measured by a handheld rangefinder at the same location in the aerial video of the video analysis software.

[0180] Secondly, in the video analysis software, the direction of traffic flow is defined as the positive direction of the X-axis. The X-axis is established along the outer lane boundary line of the road. The Y-axis is established perpendicular to the X-axis at the starting point of the upstream transition zone. The positive direction of the Y-axis is along the opposite lane. The intersection of the two axes is the origin O, and a rectangular coordinate system is established.

[0181] Next, tracking points are manually established on the vehicle body and the vehicle's driving trajectory is automatically tracked to obtain the operating data of each vehicle; the vehicle operating data includes vehicle number, time, horizontal coordinate, vertical coordinate, lateral velocity, longitudinal velocity, speed, vehicle heading angle, lateral acceleration, longitudinal acceleration, and acceleration.

[0182] Finally, the lane width extracted by the video analysis software was compared with the highway lane width measured manually using a handheld push-type rangefinder to ensure that the aerial video was consistent with the actual scale. If they were inconsistent, the system was recalibrated. After successful verification, a portion of vehicle speed data acquired by the video analysis software at the same time and location was compared with the actual vehicle speed measured by a manual radar speed gun to verify the accuracy of the speed data extracted by the video analysis software. If the data were inconsistent, the parameters of the video analysis software were adjusted to make its output data consistent with the measured data.

[0183] S1.3 Data Processing. Due to frame skipping at tracking points, missing or outlier values ​​may appear in the data output by the video analysis software. The missing and outlier values ​​in the data are identified. If the number of frames corresponding to consecutive missing values ​​exceeds a preset threshold, the missing values ​​are discarded; otherwise, interpolation methods are used to fill in the missing values.

[0184] If the number of frames corresponding to consecutive outliers exceeds a preset threshold, the outlier is removed; otherwise, the mean or median is used to replace the outlier. The preset threshold is preferably 5 frames.

[0185] S2. Risk Field Force Model Construction: Construct a driving risk field force model that integrates the vehicle's motion state and the road environment of the work area to quantify the interaction risk between the target vehicle and surrounding interactive objects; the driving risk field force model takes the target vehicle as the only risk field source, and its output risk value is the maximum value of the risk field force between the target vehicle and all interactive objects.

[0186] A driving risk field is a model analogous to a "physical field" used to characterize the risk impact of a vehicle on its journey through road traffic environments, influenced by the combined effects of people, vehicles, roads, and the environment. This risk field consists of three sub-fields: a "behavioral field" caused by driving behavior, a "kinetic energy field" formed by the influence of moving objects, and a "potential energy field" caused by static obstacles. Since vehicle motion is essentially the "result" of a driver operating the vehicle, driving risk is primarily influenced by the vehicle-road-environment system. The driving risk field force model for the work area constructed in this invention is essentially a superposition of the vehicle's kinetic energy field and the road environment. Drawing on the concept of "reciprocity of forces," it calculates the field forces between the target vehicle and each interacting object, and uses the maximum field force to represent the driving risk of the target vehicle, thus constructing a driving risk field force model that integrates the vehicle's motion state and the road environment of the work area.

[0187] S2.1 The static risks generated by a vehicle are affected by its inherent attributes (such as vehicle type, mass, and dimensions) and motion state, thus constructing a vehicle static risk field model. ;

[0188]

[0189] In the formula: For the target vehicle when stationary For interactive objects The potential danger level is positively correlated with the target vehicle; The coordinates are Interaction objects The coordinates are ; The direction of the field strength and Consistent, and decays fastest in that direction; , All are undetermined constants; For the target vehicle Equivalent quality; for Factors affecting road conditions; Indicates the target vehicle With surrounding objects The vector distance; This indicates that the relationship between risk magnitude and distance is described using a power function.

[0190] S2.2. An equivalent mass is used to describe the relationship between the inherent attributes of a vehicle and its potential risks. Based on highway accident data from the National Bureau of Statistics, the equivalent mass is fitted to obtain the final equivalent mass. The specific expression is shown in the following formula.

[0191]

[0192] In the formula: For the target vehicle Type parameters; For the target vehicle The quality; For the target vehicle The speed.

[0193] S2.3 The level of driving risk is also related to road conditions; the worse the road conditions, the greater the driving risk. Road conditions include the road surface adhesion coefficient. Road slope Road curvature ,visibility Road condition influencing factors As shown in the following formula.

[0194]

[0195] S2.4 In the above static risk field model for vehicles, the differences in vehicle shape and lateral and longitudinal risks are not considered. The equipotential lines of vehicle operation risk are always distributed in concentric circles, which deviates from the actual situation. Therefore, the vehicle shape and speed are modified to make the risk distribution exhibit lateral and longitudinal anisotropy, forming symmetrical elliptical equipotential lines centered on the vehicle. The vehicle speed is positively correlated with the major axis of the ellipse. The vehicle shape and speed are modified from a distance perspective, using... The lateral and longitudinal anisotropy of a vehicle during motion is described by the following formula:

[0196]

[0197]

[0198] In the formula: For interactive objects Relative to vehicles The relative position vector is used to describe the spatial relative position and interaction direction of the target vehicle and the interactive object within the work area; For vectors Euclidean modulus is used to characterize the standardized spatial distance between the target vehicle and the interactive object; These are parameters to be determined. , For vehicles The length and width of the vehicle body.

[0199] S2.5 To further reflect the impact of longitudinal movements such as turning and lane changing on driving risks, an acceleration parameter is introduced. By correcting the acceleration in both the lateral and longitudinal directions, the risk distribution can be made anisotropic, more accurately characterizing the actual driving risks for the target vehicle. acceleration As shown in the following formula.

[0200]

[0201] In the formula: , vehicles lateral acceleration and longitudinal acceleration; , These are acceleration parameters; , They are respectively The angle between the x and y coordinate axes.

[0202] S2.6. When the vehicle is on a curve or changing lanes, it has a steering angle. The resulting risk field will also shift accordingly. Coordinate transformation is used to describe the overall deflection of the risk field with respect to the vehicle's steering angle, with the vehicle's center of gravity as the reference point. Construct a coordinate system with the origin for the vehicle traveling in a straight line. When the vehicle's heading angle is When (clockwise direction is positive), its coordinate system is transformed to , ( , )for The deflection coordinates of the point are obtained from the following formula based on geometric analysis.

[0203]

[0204] at this time, and In the formula and After deflection and When the target vehicle's heading angle When the value is 0, no coordinate rotation is required;

[0205] S2.7 When the target vehicle When there is no intention to change lanes, The field strength of the driving risk generated in the lane needs to be multiplied by the lane line filtering factor. This prevents it from mistakenly identifying vehicles traveling in adjacent lanes as high-risk. This filtering effect is related to the lane type and the target vehicle's movement intention. (Lane filtering factor) As shown in the following formula:

[0206]

[0207] In the formula: Lane line type parameter; For the target vehicle The distance between the current edge of the vehicle body and the lane line it crosses; For the target vehicle The distance between the edge of the vehicle and the lane lines it crosses when driving along the center line of the current lane. When there is no intention to change lanes, the boundary line between the original lane and the target lane will lose its alignment. The resulting risk filtering effect.

[0208] S2.8 The final driving risk field model is shown in the following formula:

[0209]

[0210] In the formula: For vehicles Facing the interaction object The risk field strength; , These are undetermined constants; For vehicles Equivalent quality; for Factors affecting road conditions; Lane line filtering factor; For acceleration; For interactive objects Relative to vehicles The relative position vector is used to describe the spatial relative position and interaction direction of the target vehicle and the interactive object within the work area; For vectors Euclidean modulus is used to characterize the standardized spatial distance between the target vehicle and the interactive object.

[0211] S2.9, Further referencing the force law of a point charge in an electric field. A mathematical and physical analogy model of the driving risk field forces in the work area was established. This model is based on the target vehicle. As the field source, the field strength has already been described above. Detailed analysis, using Describe the interaction risk contribution between the target vehicle and the interacting object. Constructed based on relative speed. , The greater the positive velocity difference between the target vehicle and the interacting object, the higher the probability of a collision and the higher the risk contribution. When the vehicle's heading angle is 0, when the target vehicle... Interacting with objects When the speed difference is 0, When the speed difference is 5.56 m / s (i.e., 20 km / h), .

[0212]

[0213] In the formula: For the target vehicle Interacting with objects Interactive risk contribution, For the target vehicle The speed of the car, For interactive objects speed, This refers to the vehicle's heading angle.

[0214] S2.10, The Driving Risk Field Force Model (WTRF) is used to characterize the interaction risk between the target vehicle and the interacting object. Unlike traditional models that often use the interacting object as the field source and superimpose their risk field forces to obtain the comprehensive field force, the Driving Risk Field Force Model (WTRF) uses the target vehicle as the field source. As the sole source of the field, drawing on the concept of "reciprocity of forces," the field forces between the target vehicle and each interacting object are calculated, and the driving risk of the target vehicle is represented by the maximum field force, as shown in the following formula:

[0215]

[0216] In the formula: For the target vehicle The greatest driving risk field force, For the target vehicle Facing the interaction object Driving risk field force, For the target vehicle Facing the interaction object The magnitude of the risk field, For the target vehicle Interacting with objects The contribution of interactive risks.

[0217] This design eliminates the need to determine the field source weights of interactive objects, significantly reducing model parameters and computational load, while avoiding high-risk misjudgments caused by the superposition of low-risk factors.

[0218] S3, Calibration of parameters for the driving risk field force model in the work area;

[0219] S3.1 Due to complex environmental influencing factors (road surface adhesion coefficient) Road curvature Road slope ,visibility Lane line type and vehicle type Due to the difficulty in accurate measurement, complex environmental impact factors were calibrated using a parameter calibration method based on statistical data, based on the 2010-2016 China Road Traffic Safety Accident Dataset. If the parameter is a continuous variable, such as the road surface adhesion coefficient, a polynomial fitting was used to find the relationship between the parameter and the property loss per unit accident, thus completing the calibration. If it is a discrete variable, such as lane line type parameters, vehicle type parameters, visibility, road curvature, and road slope, the maximum property loss per unit accident was used as the standard value, and a lookup table was established using the following formula.

[0220]

[0221] In the formula: Discrete parameters Values The calibration value at that time Discrete parameters for Property losses caused by accidents at the workplace; Discrete parameters The set of accident property losses corresponding to all possible value states; for Losses due to accidents at the time; for Losses due to accidents at the time; for Losses due to accidents at the time;

[0222] S3.2. The remaining parameters in the driving risk field force model are calibrated using a genetic algorithm. The population size is 200, with 600 iterations, a crossover probability of 0.8, a mutation probability of 0.2, and the minimum root mean square error is taken as the parameter value. The calibration parameters include... , , , , .

[0223] S4. The risk threshold of the driving risk field force model in the work area is determined by using the reciprocal time of collision (TTCI) and the deceleration to avoid collision (DRAC).

[0224] S4.1 The Time-to-Collision Inverse (TTCI) and Deceleration-to-Avoidance (DRAC) calculated based on relative speed and relative distance are widely used for risk identification and assessment in highway work areas due to their high interpretability. The TTCI is the reciprocal of the Time-to-Collision (TTC), and their conflict thresholds are related to the driver's reaction time and the vehicle's maximum deceleration. Therefore, this invention uses TTCI and DRAC to determine the risk thresholds of the driving risk field force model. The conflict thresholds for TTCI and DRAC are set to 0.2 s and 1.4 m / s, respectively. 2 The formulas are as follows:

[0225]

[0226]

[0227] in, , For the vehicle in front and the target vehicle; , For the vehicle in front and target vehicle exist The position at that moment; and For the vehicle in front and target vehicle exist The speed of time; , For vehicles Vehicle length and vehicle width; For the target vehicle exist The angle between the driving direction and the horizontal axis at any given time, i.e., the vehicle's heading angle;

[0228] If the time to collision (TTCI) of the target vehicle is greater than the conflict threshold of the time to collision (TTCI), and the deceleration rate (DRAC) for avoiding the collision is greater than the conflict threshold of the deceleration rate (DRAC) for avoiding the collision, then the target vehicle is considered a high-risk vehicle sample.

[0229] Then, the maximum driving risk field force value of all high-risk vehicle samples is calculated, and the minimum value among them is taken as the risk threshold of the driving risk field force model.

[0230] When the maximum driving risk field force experienced by any vehicle during operation exceeds the risk threshold of the driving risk field force model, it is determined to be in a high-risk state.

[0231] S5. Based on the results of the field force model of driving risk in the work area, the spatial distribution of vehicle operation risk is visualized, and high-risk areas under different construction work areas are identified.

[0232] Specifically, the road area is divided into grids, with each grid measuring 2.5m horizontally and 0.5m vertically. The Travel Risk Field Force (WTRF) of vehicles passing through the highway work area is calculated. The maximum WTRF value for each grid is displayed, visually representing the spatial distribution of vehicle travel risk in the road environment. Darker colors indicate higher travel risk.

[0233] Application Examples

[0234] like Figure 1 As shown, this invention provides a method for assessing driving risks in mountainous highway work areas under a multi-source dynamic risk field. The specific steps are as follows:

[0235] S1. Collect measured vehicle operation data in typical mountainous highway work areas, covering various closed construction scenarios;

[0236] S1.1, Data collection scope and collection method. For example... Figure 2 As shown, the standard operating area (SIO) of a highway consists of six parts: warning zone (S), upstream transition zone (Ls), buffer zone (H), operating zone (G), downstream transition zone (Lx), and termination zone (Z). Because vehicles frequently change lanes and accelerate / decelerate in the warning zone and upstream transition zone, these areas are high-risk accident zones. Therefore, in-depth analysis of their driving risks is crucial for improving the safety of the operating area. Thus, on a single-lane, two-way highway, covering both straight and curved sections, the operating area included the closed driving lane and overtaking lane for data collection. Continuous data collection was conducted using a combination of drones, cameras, and manual labor. The weather was clear during the data collection period, and the operating area layout met the required specifications.

[0237] S1.2 Data Extraction. First, import the collected aerial videos of the construction area into Kinovea or Tracker video analysis software; then, calibrate the scale of the highway lane width measured manually with a handheld rangefinder at the same location in the aerial video of the video analysis software.

[0238] Secondly, in the video analysis software, the direction of traffic flow is defined as the positive direction of the X-axis. The X-axis is established along the outer lane boundary line of the road. The Y-axis is established perpendicular to the X-axis at the starting point of the upstream transition zone. The positive direction of the Y-axis is along the opposite lane. The intersection of the two axes is the origin O, and a rectangular coordinate system is established.

[0239] Next, tracking points are manually established on the vehicle body and the vehicle's driving trajectory is automatically tracked to obtain the operating data of each vehicle; the vehicle operating data includes vehicle number, time, horizontal coordinate, vertical coordinate, lateral velocity, longitudinal velocity, speed, vehicle heading angle, lateral acceleration, longitudinal acceleration, and acceleration.

[0240] Finally, the lane width extracted by the video analysis software was compared with the highway lane width measured manually using a handheld push-type rangefinder to ensure that the aerial video was consistent with the actual scale. If they were inconsistent, the system was recalibrated. After successful verification, a portion of vehicle speed data acquired by the video analysis software at the same time and location was compared with the actual vehicle speed measured by a manual radar speed gun to verify the accuracy of the speed data extracted by the video analysis software. If the data were inconsistent, the parameters of the video analysis software were adjusted to make its output data consistent with the measured data.

[0241] Table 1. Vehicle operating parameters and data for highway work areas (partial data illustration)

[0242]

[0243] S1.3 Data Processing. Due to frame skipping at tracking points, missing or outlier values ​​may appear in the data output by the video analysis software. The missing and outlier values ​​in the data are identified. If the number of frames corresponding to consecutive missing values ​​exceeds a preset threshold, the missing values ​​are discarded; otherwise, interpolation methods are used to fill in the missing values.

[0244] If the number of frames corresponding to consecutive outliers exceeds a preset threshold, the outlier is removed; otherwise, the mean or median is used to replace the outlier. The preset threshold is 5 frames.

[0245] S2. Construct a field force model for driving risk in the work area that integrates vehicle motion state and road environment to describe the interaction risk between the target vehicle and the interacting object, such as... Figure 3 The diagram shows the flowchart illustrating the construction process of the driving risk field force model in the highway operation area.

[0246] S2.1 The static risks generated by vehicles are affected by their inherent attributes (such as vehicle type, mass, and dimensions) and motion state. Therefore, a static risk field model of vehicles is constructed as shown in the following formula. On this basis, a dynamic operation state correction term of the vehicle is introduced, and the road environment factors of the work area are superimposed to construct a work area driving risk field model that is more in line with the actual working conditions.

[0247]

[0248] In the formula: For the target vehicle when stationary For interactive objects The potential danger level is positively correlated with the target vehicle; The coordinates are Interaction objects The coordinates are ; The direction of the field strength and Consistent, and decays fastest in that direction; , All are undetermined constants; For the target vehicle Equivalent quality; for Factors affecting road conditions; Indicates the target vehicle With surrounding objects The vector distance; This indicates that the relationship between risk magnitude and distance is described using a power function.

[0249] S2.2. An equivalent mass is used to describe the relationship between the inherent attributes of a vehicle and its potential risks. Based on highway accident data from the National Bureau of Statistics, the equivalent mass is fitted to obtain the final equivalent mass. The specific expression is shown in the following formula.

[0250]

[0251] In the formula: For vehicles Type; For vehicles The quality; For vehicles The speed.

[0252] S2.3 The level of driving risk is also related to road conditions; the worse the road conditions, the greater the driving risk. Road conditions include the road surface adhesion coefficient. Road curvature Road slope ,visibility Road condition influencing factors As shown in the following formula.

[0253]

[0254] S2.4 In the above static risk field model for vehicles, the differences in vehicle shape and lateral and longitudinal risks are not considered. The equipotential lines of vehicle operation risk are always distributed in concentric circles, which deviates from the actual situation. Therefore, the vehicle shape and speed are modified to make the risk distribution exhibit lateral and longitudinal anisotropy, forming symmetrical elliptical equipotential lines centered on the vehicle. The vehicle speed is positively correlated with the major axis of the ellipse. The vehicle shape and speed are modified from a distance perspective, using... The lateral and longitudinal anisotropy of a vehicle during travel is described by the following formula.

[0255]

[0256]

[0257] In the formula: For interactive objects Relative to the target vehicle The relative position vector is used to describe the spatial relative position and interaction direction of the target vehicle and the interactive object within the work area; For vectors The Euclidean modulus is used to characterize the standardized spatial distance between the target vehicle and the interactive object; the target vehicle The coordinates are The coordinates of the interactive object are ; These are parameters to be determined. Target vehicle speed; , For the target vehicle The length and width of the vehicle body;

[0258] S2.5 To further reflect the impact of longitudinal movements such as turning and lane changing on driving risks, an acceleration parameter is introduced. By correcting the acceleration in both the lateral and longitudinal directions, the risk distribution can be made anisotropic, more accurately characterizing the actual driving risks for the target vehicle. acceleration As shown in the following formula.

[0259]

[0260] In the formula: , The target vehicles Lateral and longitudinal acceleration; , These are acceleration parameters; , They are respectively The angle between the x and y coordinate axes.

[0261] S2.6 When the vehicle has a steering angle while curving or changing lanes, the target vehicle The resulting risk field will also shift. Coordinate system transformation can be used to describe the overall shift of the risk field with the vehicle's steering angle, such as... Figure 4 As shown, with the target vehicle's center of mass... Construct a coordinate system for the vehicle traveling in a straight line, with the origin as the coordinate system. When the vehicle's heading angle is When (clockwise direction is positive), its coordinate system is transformed to , ( , )for The deflection coordinates of the point are obtained from the following formula based on geometric analysis.

[0262]

[0263] at this time, and In the formula and After deflection and When the target vehicle's heading angle When the value is 0, no coordinate rotation is required;

[0264] S2.7 When the target vehicle When there is no intention to change lanes, The field strength of the driving risk generated in the lane needs to be multiplied by the lane line filtering factor. This prevents it from mistakenly identifying vehicles traveling in adjacent lanes as high-risk. This filtering effect is related to the lane type and the target vehicle's movement intention. (Lane filtering factor) As shown in the following formula.

[0265]

[0266] In the formula: Lane line type parameter; For the target vehicle The distance between the current edge of the vehicle body and the lane line it crosses; For the target vehicle The distance between the edge of the vehicle and the lane lines it crosses when driving along the center line of the current lane. When there is no intention to change lanes, the boundary line between the original lane and the target lane will lose its alignment. The resulting risk filtering effect.

[0267] S2.8 The final driving risk field model is shown in the following formula.

[0268]

[0269] In the formula: For the target vehicle Facing the interaction object The risk field strength; , These are undetermined constants; For the target vehicle Equivalent quality; for Factors affecting road conditions; Lane line filtering factor; For the target vehicle The acceleration; For interactive objects Relative to the target vehicle The relative position vector is used to describe the spatial relative position and interaction direction of the target vehicle and the interactive object within the work area; For vectors Euclidean modulus, used to characterize the target vehicle and interaction objects The standardized spatial distance between them;

[0270] S2.9, Further referencing the force law of a point charge in an electric field. A mathematical and physical analogy model of the driving risk field forces in the work area was established. This model is based on the target vehicle. As the field source, the field strength has already been described above. Detailed analysis, using Describe the interaction risk contribution between the target vehicle and the interacting object. Constructed based on relative speed. , The risk contribution is calculated based on the interaction risk between the target vehicle and the interacting object. A larger positive speed difference between the target vehicle and the interacting object increases the likelihood of a collision and thus the higher the risk contribution. When the speed difference is zero,... When the speed difference is 5.56 m / s (20 km / h), .

[0271]

[0272] In the formula: For the target vehicle Interacting with objects Interactive risk contribution, For the target vehicle The speed of the car, For interactive objects speed, This refers to the vehicle's heading angle.

[0273] S2.10, the Vehicle Risk Field Force Model (WTRF) is used to characterize the interaction risk between the target vehicle and the interacting object. Unlike traditional safety potential field models, which often use the interacting object as the field source and superimpose risk field forces to obtain a comprehensive field force, [the WTRF model is not directly related to the previous sentence and can be omitted]. Figure 5 As shown in (a), WTRF targets the vehicle. As the sole source of the field, drawing on the concept of "reciprocity of forces," the field forces between the target vehicle and each interacting object are calculated, and the driving risk of the target vehicle is represented by the maximum field force. Figure 5 As shown in (b), and in the following equation. This design eliminates the need to determine the field source weights of the interactive objects, significantly reducing model parameters and computational load, while avoiding high-risk misjudgments caused by the superposition of low-risk factors.

[0274]

[0275] In the formula: For the target vehicle The greatest driving risk field force, For the target vehicle Facing the interaction object Driving risk field force, For the target vehicle Facing the interaction object The magnitude of the risk field, For the target vehicle Interacting with objects The contribution of interactive risks.

[0276] S3, Calibration of parameters for the driving risk field force model in the work area;

[0277] S3.1 Due to complex environmental influencing factors (road surface adhesion coefficient) Road slope Road curvature ,visibility Lane line type and vehicle type Due to the difficulty in accurate measurement, complex environmental impact factors were calibrated using a parameter calibration method based on statistical data, based on the 2010-2016 Chinese road traffic safety accident dataset. If the parameter is a continuous variable, such as the road surface adhesion coefficient, a polynomial fitting was used to find the relationship between the parameter and the property loss per unit accident, thus completing the calibration. If it is a discrete variable, such as lane line type, vehicle type, visibility, road curvature, or road slope, the maximum property loss per unit accident was used as the standard value, and a lookup table was established using the following formula.

[0278]

[0279] In the formula: Discrete parameters Values The calibration value at that time Discrete parameters for Property losses caused by accidents at the workplace; Discrete parameters The set of accident property losses corresponding to all possible value states; for Losses due to accidents at the time; for Losses due to accidents at the time; for Losses due to accidents at the time;

[0280] Table 2. Risk Value Criteria for Complex Environmental Impact Factors in the Work Area

[0281]

[0282] S3.2. For the remaining parameters in the driving risk field force model, a genetic algorithm is used for calibration. The population size is 200, with 600 iterations, a crossover probability of 0.8, and a mutation probability of 0.2. The parameter values ​​corresponding to the minimum root mean square error are taken as the final parameter values. The calibration parameters include... , , , , .

[0283] Table 3. Calibration Results of Driving Risk Field Force Model Parameters

[0284]

[0285] Based on the above analysis, the spatial distribution of driving risks under nine different conditions (vehicles decelerating, maintaining a constant speed, and accelerating in three scenarios: straight driving, curving, and lane changing) is presented, such as... Figure 6 As shown in the figure, the three columns are schematic diagrams illustrating the risks of deceleration, constant speed, and acceleration of a vehicle in three scenarios: driving straight, driving on a curve, and changing lanes.

[0286] S4. The risk threshold of the driving risk field force model in the work area is determined by using the reciprocal time of collision (TTCI) and the deceleration to avoid collision (DRAC).

[0287] S4.1 The Time-to-Collision Inverse (TTCI) and Deceleration-to-Avoidance (DRAC) calculated based on relative speed and relative distance are widely used for risk identification and assessment in highway work areas due to their high interpretability. Their conflict thresholds are related to the driver's reaction time and the vehicle's maximum deceleration. Therefore, this invention uses the Time-to-Collision Inverse (TTCI) and Deceleration-to-Avoidance (DRAC) to determine the risk threshold of the driving risk field force model. TTCI is the reciprocal of TTC, and the conflict thresholds for TTCI and DRAC are set to 0.2 s and 1.4 m / s, respectively. 2 The formulas are as follows:

[0288]

[0289]

[0290] in, , For the vehicle in front and the target vehicle; , For the vehicle in front and target vehicle exist The position at that moment; and For the vehicle in front and target vehicle exist The speed of time; , For vehicles Vehicle length and vehicle width; For the target vehicle exist The angle between the driving direction and the horizontal axis at any given time, i.e., the vehicle's heading angle;

[0291] Figure 7 This paper illustrates the use of TTCI, DRAC, and WTRF to identify high-risk vehicles, with each point representing one vehicle. TTCI and DRAC identified 21 and 25 high-risk vehicles, respectively, for a total of 28 high-risk vehicles. When the WTRF risk threshold was set to 3.2, WTRF identified 62 high-risk vehicles. This demonstrates that compared to TTCI and DRAC, WTRF has a significant advantage in identifying multi-factor coupled risks and can effectively capture interactive risks that traditional time-distance indicators fail to represent.

[0292] To further analyze the heterogeneity of WTRF, TTC, and DRAC in identifying vehicle risks, 300 target vehicles were randomly selected from four types of highway work area scenarios, and risk identification was performed using each of the three models. During the identification process, the interaction type (“CT” indicates that the target vehicle is a car and the interacting vehicle is a truck) and the target vehicle's steering angle were comprehensively analyzed. ,speed acceleration and interactive vehicle steering angle Speed ​​difference between the target vehicle and the interacting vehicle Acceleration difference and horizontal and vertical distance , Key indicators such as... Figure 8 Three parallel index plots illustrate the different results of WTRF, TTC, and DRAC in identifying high-risk vehicles. Each line represents a pair of interacting vehicles, with red lines indicating high risk and green lines indicating low risk.

[0293] right Figure 8 (a) Analysis shows that the vehicle pairs represented by the yellow line can be identified as high-risk by TTCI and WTRF, but DRAC cannot identify them. Their main characteristics are low relative speed and small distance between vehicles, indicating that DRAC exhibits a low risk identification rate for these types of interacting vehicles. Similarly, in Figure 8 (b) The vehicle pairs indicated by the yellow line can be identified using DRAC and WTRF, but not by TTCI. When the vehicle pairs are far apart and have high relative speeds, TTCI exhibits a low risk recognition rate. Furthermore, when the following vehicle's speed is lower than the preceding vehicle's, TTCI shows low risk, as it ignores the potential risk of a traffic accident caused by the following vehicle braking suddenly and the following vehicle not being able to brake in time due to the close following distance.

[0294] The results showed that TTCI identified 36 high-risk vehicles, DRAC identified 43, while WTRF identified 121, demonstrating a higher risk identification rate. WTRF corrects for vehicle shape, speed, acceleration, turning angle, and longitudinal distance, and incorporates road attribute parameters, making it more sensitive and comprehensive in identifying potential vehicle risks in complex road environments such as construction zones. This overcomes the limitations of traditional SSM in terms of adaptability and accuracy. Compared to TTCI and DRAC, WTRF outperforms TTCI and DRAC in risk identification, with identification accuracies improved by 70.25% and 64.46%, respectively.

[0295] S5. Based on the results of the field force model of driving risk in the work area, the spatial distribution of vehicle operation risk is visualized, and high-risk areas under different work areas are identified.

[0296] Specifically, this involves analyzing vehicle operation data collected from four typical mountainous highway work areas. All four work areas are located on single-lane, two-way highways, encompassing both straight and curved sections. The construction areas include closed driving and overtaking lanes. The road area is divided into grids, with each grid measuring 2.5m horizontally and 0.5m vertically. The Travel Risk Field Force (WTRF) values ​​of vehicles passing through the highway work areas are calculated. The maximum WTRF value for each grid is displayed, visually representing the spatial distribution of vehicle travel risk in the road environment. Darker colors indicate higher travel risk.

[0297] from Figure 9 and Figure 10 It can be seen that there is a significant difference in driving risk on the same road segment with and without a work zone. Regardless of whether the work zone is located on a straight or curved driving lane or overtaking lane, it significantly increases the overall driving risk of the road segment. Compared to the situation without a work zone, the driving risk of a closed overtaking lane is slightly higher than that of a closed driving lane, possibly related to the difference in speed limits; and once a work zone is set up, the lanes merging into it usually face greater driving risks.

[0298] Overall, the risk levels corresponding to the four work zone settings are as follows: work zone occupying the overtaking lane on a curve > overtaking lane on a curve > overtaking lane on a straight road > straight road. The differences in risk levels are likely due to limited visibility and speed differences. Especially in curve scenarios, the reduced visibility combined with centrifugal force significantly increases the risk of lateral deviation compared to straight sections. Closing the overtaking lane involves emergency braking of high-speed traffic, resulting in a significantly higher risk spike than closing the driving lane. Therefore, the risks posed by setting the work zone in the overtaking lane are generally higher than those in the driving lane.

[0299] The increase in the proportion of high-risk vehicles is closely related to lane-changing behavior. Among high-risk vehicles, 75.64% were lane-changing vehicles or vehicles that interacted with lane-changing vehicles, indicating that lane changing is a key factor causing driving risks in the work area. In addition, due to the frequent deceleration of vehicles in the work area, driving risks can also spread laterally, affecting the traffic safety of adjacent lanes.

[0300] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A method for assessing driving risks in mountainous highway work areas under a multi-source dynamic risk field, characterized in that, Includes the following steps: S1. Collect measured vehicle operation data in the mountainous highway operation area; S2. Construct a driving risk field force model that integrates the vehicle's motion state and the road environment of the work area to quantify the interaction risk between the target vehicle and surrounding interactive objects; the driving risk field force model takes the target vehicle as the only risk field source, and its output risk value is the maximum value of the risk field force between the target vehicle and all interactive objects. S3. Perform parameter calibration on the driving risk field force model; S4. The risk threshold of the driving risk field force model is determined by using the reciprocal of collision time (TTCI) and the deceleration to avoid collision (DRAC). S5. Based on the calibrated driving risk field force model and the determined risk threshold, calculate the driving risk field force value of vehicles passing through the highway operation area and perform spatial grid visualization to identify high-risk areas under different operation areas. In step S2, the driving risk field force model is as follows: In the formula: For the target vehicle The greatest driving risk field force, For the target vehicle Facing the interaction object Driving risk field force, For the target vehicle Facing the interaction object The magnitude of the risk field, For the target vehicle Interacting with objects The contribution of interactive risks; In the formula: For the target vehicle Interacting with objects Interactive risk contribution, For the target vehicle The speed of the car, For interactive objects speed, For the target vehicle Heading angle; The driving risk field model is shown below: In the formula: For the target vehicle Facing the interaction object The risk field strength; , These are undetermined constants; For the target vehicle Equivalent quality; for Factors affecting road conditions; Lane line filtering factor; For the target vehicle The acceleration; For interactive objects Relative to the target vehicle The relative position vector; For vectors Euclidean mold length; equivalent quality The formula is as follows: In the formula: For the target vehicle Type parameters; For the target vehicle The quality; Road condition influencing factors The formula is as follows: In the formula: The road surface adhesion coefficient, For road slope, For road curvature, Visibility; Target vehicle acceleration The formula is as follows: In the formula: , The target vehicles Lateral and longitudinal acceleration; , These are acceleration parameters; , They are respectively The angle between the x and y coordinate axes; Lane line filtering factor The formula is as follows: In the formula: Lane line type parameter; For the target vehicle The distance between the current edge of the vehicle body and the lane line it crosses; For the target vehicle The distance between the edge of the vehicle body and the lane lines it crosses when driving along the center line of the current lane.

2. The method for assessing driving risks in mountainous highway work areas under a multi-source dynamic risk field as described in claim 1, characterized in that, The specific method of step S1 is as follows: collect video of vehicle operation in the working area of ​​the mountain highway; use video analysis software to extract vehicle trajectory data, obtain vehicle operation data including vehicle number, time, position, speed, acceleration and heading angle, and calibrate, verify and process the data.

3. The method for assessing driving risks in mountainous highway work areas under a multi-source dynamic risk field according to claim 2, characterized in that, Step S1 includes the following steps: S1.1 Data Collection Scope and Method: The standard work area of ​​a highway consists of six parts: warning area, upstream transition area, buffer zone, construction area, downstream transition area, and termination area. On a single-lane, two-way mountain highway, video data is collected from the work area, which includes the closed driving lanes or overtaking lanes of straight sections or curves. The weather must be clear during the collection period, and the layout of the work area must meet the specifications. S1.2 Data Extraction: First, import the collected aerial videos of the construction area into Kinovea or Tracker video analysis software; then, calibrate the scale of the highway lane width measured by a handheld rangefinder at the same location in the aerial video of the video analysis software. Secondly, in the video analysis software, the direction of traffic flow is defined as the positive direction of the X-axis. The X-axis is established along the outer lane boundary line of the road. The Y-axis is established perpendicular to the X-axis at the starting point of the upstream transition zone. The positive direction of the Y-axis is along the opposite lane. The intersection of the two axes is the origin O, and a rectangular coordinate system is established. Next, tracking points are manually established on the vehicle body and the vehicle's driving trajectory is automatically tracked to obtain the operating data of each vehicle; the vehicle operating data includes vehicle number, time, horizontal coordinate, vertical coordinate, lateral velocity, longitudinal velocity, speed, vehicle heading angle, lateral acceleration, longitudinal acceleration, and acceleration. Finally, the lane width extracted by the video analysis software was compared with the highway lane width measured manually using a handheld push-type rangefinder to ensure that the aerial video was consistent with the actual scale. If they were inconsistent, the system was recalibrated. After successful verification, a portion of vehicle speed data acquired by the video analysis software at the same time and location was compared with the actual vehicle speed measured by a manual radar speed gun to verify the accuracy of the speed data extracted by the video analysis software. If the data were inconsistent, the parameters of the video analysis software were adjusted to make its output data consistent with the measured data. S1.3 Data Processing: Due to frame skipping at tracking points, missing or outlier values ​​may appear in the data output by the video analysis software. Identify the missing and outlier values ​​in the data; if the number of frames corresponding to consecutive missing values ​​exceeds a preset threshold, they are removed; otherwise, interpolation methods are used to fill in the missing values. If the number of frames corresponding to consecutive outliers exceeds a preset threshold, they are removed; otherwise, the mean or median is used to replace the outliers.

4. The method for assessing driving risks in mountainous highway work areas under a multi-source dynamic risk field according to claim 3, characterized in that, The preset threshold is 5 frames.

5. The method for assessing driving risks in mountainous highway work areas under a multi-source dynamic risk field according to claim 1, characterized in that, and The formula is as follows: Where: target vehicle The coordinates are Interaction objects The coordinates are ; These are parameters to be determined. , For the target vehicle The length and width of the vehicle body; When a vehicle has a steering angle while curving or changing lanes, the target vehicle The resulting risk field will also shift; coordinate system transformation is used to describe the overall deflection of the risk field with the vehicle's steering angle; with the target vehicle... Construct a coordinate system with the centroid as the origin when the vehicle is traveling in a straight line. When the vehicle's heading angle is When, its coordinate system is transformed , ( , ) is the interactive object The deflection coordinates of the point are shown below: at this time, and In the formula and After deflection and When the target vehicle's heading angle When the value is 0, no coordinate rotation is required; For the target vehicle The type parameter, when the target vehicle is a passenger car. The value is 1.000; when the target vehicle is a cargo truck, The value is 1.443; when the target vehicle is a motorcycle, It is 0.335; This is a lane line type parameter; when the lane line type is dashed, ... The value is 0.700; when the lane line type is a single solid line, It is 0.

500.

6. The method for assessing driving risks in mountainous highway work areas under a multi-source dynamic risk field according to claim 1, characterized in that, Step S3 includes the following steps: S3.

1. Based on the road traffic safety accident dataset, the road surface adhesion coefficient is... Road slope Road curvature ,visibility Lane line type parameters Vehicle type parameters Perform calibration; S3.2, For the driving risk field force model , , , , These parameters are calibrated using a genetic algorithm.

7. The method for assessing driving risks in mountainous highway work areas under a multi-source dynamic risk field according to claim 6, characterized in that, In step S3.1: The road surface adhesion coefficient is calibrated using a parameter calibration method based on statistical data. Road slope Road curvature ,visibility Lane line type parameters Vehicle type parameters If the parameter is a continuous variable, the relationship between the parameter and the property loss per unit accident is fitted using a polynomial to complete the calibration; if it is a discrete variable, the maximum property loss per unit accident is used as the standard value, and the following formula is used to establish a lookup table. In the formula: Discrete parameters Values The calibration value at that time Discrete parameters for Property losses caused by accidents at the workplace; Discrete parameters The set of accident property losses corresponding to all possible value states; for Losses due to accidents at the time; for Losses due to accidents at the time; for The accident loss at that time; In step S3.2: When using the genetic algorithm for calibration, the population size is 200, the number of iterations is 600, the crossover probability is 0.8, the mutation probability is 0.2, and the parameter value corresponding to the minimum root mean square error is taken as the final parameter value.

8. The method for assessing driving risks in mountainous highway work areas under a multi-source dynamic risk field according to claim 1, characterized in that, Step S4 is as follows: The risk threshold of the driving risk field force model is determined using the reciprocal of the collision time (TTCI) and the collision avoidance deceleration (DRAC). The TTCI is the reciprocal of the collision time (TTC), and the conflict thresholds for the TTCI and DRAC are set to 0.2 s and 1.4 m / s, respectively. 2 The formulas are as follows: in, , For the vehicle in front and the target vehicle; , For the vehicle in front and target vehicle exist The position at that moment; and For the vehicle in front and target vehicle exist The speed of time; , For vehicles Vehicle length and vehicle width; For the target vehicle exist The angle between the driving direction and the horizontal axis at any given time, i.e., the vehicle's heading angle; If the time to collision (TTCI) of the target vehicle is greater than the conflict threshold of the time to collision (TTCI), and the deceleration rate (DRAC) for avoiding the collision is greater than the conflict threshold of the deceleration rate (DRAC) for avoiding the collision, then the target vehicle is considered a high-risk vehicle sample. Then, the maximum driving risk field force value of all high-risk vehicle samples is calculated, and the minimum value among them is taken as the risk threshold of the driving risk field force model. When the maximum driving risk field force experienced by any vehicle during operation exceeds the risk threshold of the driving risk field force model, it is determined to be in a high-risk state.

9. The method for assessing driving risks in mountainous highway work areas under a multi-source dynamic risk field according to claim 1, characterized in that, Step S5 is as follows: The road area is divided into grids with a horizontal size of 2.5m and a vertical size of 0.5m. The driving risk field force value of vehicles passing through the highway operation area is calculated. The maximum driving risk field force value of each grid is displayed and reflected intuitively on the visualization map to reflect the spatial distribution of vehicle driving risk in the road environment. The darker the color, the higher the driving risk.

Images