A vehicle abnormal driving detection method based on multi-target trajectory prediction

Abstract

The present application relates to the field of intelligent transportation systems, and more particularly to a vehicle abnormal driving detection method based on multi-target trajectory prediction. First, based on the obtained target vehicle state data and surrounding vehicle state data, an improved constant acceleration traffic coupling trajectory prediction method is used to obtain the predicted position vector of the target vehicle. Then, based on the predicted position vector of the target vehicle and the actual position vector in the target vehicle state data, an improved normalized trajectory deviation evaluation method is used to obtain the direction-sensitive deviation rate index. Finally, based on the direction-sensitive deviation rate index, the adaptive threshold is determined by the 3σ criterion, the direction-sensitive deviation rate index is compared with the adaptive threshold, and the determination result is obtained. The technical problems of ignoring the influence of surrounding vehicles on the movement of the target vehicle, insufficient recognition ability for abnormal behaviors with significant direction change characteristics, and lack of comparability in abnormal determination between high-speed and low-speed vehicles are solved.

Description

Technical Field

[0001] This invention relates to the field of intelligent transportation systems, and in particular to a method for detecting abnormal vehicle driving based on multi-target trajectory prediction. Background Technology

[0002] With the rapid development of intelligent transportation systems and autonomous driving technologies, the complexity and dynamism of vehicle behavior in road traffic environments have significantly increased. Traditional methods for detecting abnormal vehicle behavior often rely on rule-based judgments or single-target behavior analysis, such as identifying anomalies by setting speed thresholds or lane departure levels. However, when facing complex traffic scenarios, such as multi-vehicle interactions, sudden events, and congested environments, these methods often suffer from insufficient adaptability, lack of robustness, and high false alarm rates. Traditional methods rely heavily on single-vehicle motion models or simple rule-based judgments, making it difficult to accurately identify abnormal behavior in multi-vehicle interactions and high-density traffic flow environments. Especially in urban roads, intersections, and highway congestion scenarios, the coupling relationships between vehicles are not fully modeled, leading to poor stability and high false alarm rates in anomaly detection results. Furthermore, existing methods often lack unified evaluation standards for different vehicle speeds and traffic flow states, making it difficult to achieve consistent detection across scenarios and limiting their effectiveness in practical traffic management. Therefore, a vehicle abnormal driving detection method based on multi-target trajectory prediction is proposed, which can not only improve the accuracy and real-time performance of detection, but also provide a more intelligent technical path for safety early warning and decision support in complex traffic environments, and has broad application prospects. Summary of the Invention

[0003] This invention provides a vehicle anomaly detection method based on multi-target trajectory prediction to address the following issues: ignoring the influence of surrounding vehicles on the target vehicle's motion, lacking the ability to characterize the overall traffic flow behavior, resulting in insufficient prediction accuracy in scenarios such as following, lane changing, and congestion; failing to consider the attenuation relationship of the influence between vehicles with distance changes, lacking modeling of the consistency of vehicle driving direction, and failing to accurately reflect the true effect of neighboring vehicles on the target vehicle's motion trend; not considering vehicle heading change factors, resulting in insufficient ability to identify abnormal behaviors with significant directional change characteristics such as sharp turns, serpentine driving, and abnormal lane changes, easily leading to missed detections; still having scale inconsistencies under different speed conditions, resulting in a lack of comparability in anomaly judgments between high-speed and low-speed vehicles; and using a fixed threshold, which is difficult to adapt to dynamic changes in different traffic densities and road environments, and is prone to thresholds that are too tight or too wide in complex environments, affecting detection sensitivity and stability.

[0004] The present invention provides a method for detecting abnormal vehicle driving based on multi-target trajectory prediction, comprising the following steps:

[0005] S1. Acquire the target vehicle's own state data, including its actual position vector, velocity, heading angle, and acceleration, as well as the surrounding vehicle state data, including the actual position vector, velocity, heading angle, and acceleration of neighboring vehicles. Using an improved constant acceleration traffic coupling trajectory prediction method, calculate the Euclidean distance between the target vehicle and neighboring vehicles based on their actual position vectors. Based on the Euclidean distance and a set influence radius, calculate the distance decay weight using an exponential decay function. Based on the distance decay weight, combine the velocity, heading angle of neighboring vehicles, and heading angle of the target vehicle to obtain a directional consistency coupling factor. After mapping the directional consistency coupling factor, fuse it with the target vehicle's acceleration to obtain the target vehicle's equivalent acceleration. Based on the target vehicle's velocity, equivalent acceleration, and sampling time interval, calculate the theoretical displacement. Based on the theoretical displacement, obtain the predicted position vector of the target vehicle.

[0006] S2. Based on the predicted and actual position vectors of the target vehicle, an improved normalized trajectory deviation assessment method is adopted, combined with theoretical motion displacement, to obtain spatial deviation characteristics; based on the heading angles of the target vehicle at the next and current moments, the standardized heading angle change of the target vehicle is calculated; the absolute value of the standardized heading angle change is added by 1 to obtain the direction change sensitivity factor; based on the spatial deviation characteristics and the direction change sensitivity factor, the direction sensitivity deviation rate index is obtained; based on the direction sensitivity deviation rate index, an adaptive threshold is determined by the 3σ criterion, and the direction sensitivity deviation rate index is compared with the adaptive threshold to obtain the judgment result.

[0007] Preferably, S1 specifically includes:

[0008] In the improved constant acceleration traffic coupling trajectory prediction method, the directional projection is calculated using a cosine function based on the difference between the heading angles of neighboring vehicles and the target vehicle. The directional consistency coupling factor is calculated by summing the products of the speed of neighboring vehicles, the distance attenuation weight, and the directional projection, and then dividing by the sum of the products of the distance attenuation weight and the speed of neighboring vehicles.

[0009] Preferably, S1 specifically includes:

[0010] The hyperbolic tangent function is used to map the directional consistency coupling factor, resulting in the mapped directional consistency coupling factor.

[0011] Preferably, S1 specifically includes:

[0012] The equivalent acceleration of the target vehicle is calculated by multiplying the result of adding 1 to the mapped direction consistency coupling factor, dividing by 2, and then adding 1 again, with the acceleration of the target vehicle.

[0013] Preferably, S1 specifically includes:

[0014] Based on the sine and cosine values ​​of the target vehicle's heading angle, the direction unit vector is calculated; the product of the theoretical displacement and the direction unit vector is added to the actual position vector of the target vehicle to obtain the predicted position vector of the target vehicle.

[0015] Preferably, S2 specifically includes:

[0016] In the improved normalized trajectory deviation assessment method, the positional difference between the predicted and actual position vectors is calculated by subtracting them; the spatial deviation characteristics are calculated by dividing the positional difference by the theoretical motion displacement.

[0017] Preferably, S2 specifically includes:

[0018] Multiply the direction change sensitivity factor by the spatial deviation characteristic to calculate the direction sensitivity deviation rate index of the target vehicle. The calculation formula is as follows: , in, Indicates the first The target vehicle in time Direction-sensitive deviation rate index; Indicates the first The target vehicle in time The actual position vector; Indicates the first The target vehicle in time The predicted position vector; Indicates the first The target vehicle in time speed; Indicates the sampling time interval; Indicates the first The target vehicle in time The equivalent acceleration; Indicates the first The target vehicle from time Time The standardized change in heading angle.

[0019] Preferably, S2 specifically includes:

[0020] Based on the direction-sensitive deviation rate index, a sliding time window statistical mechanism is adopted to continuously record the direction-sensitive deviation rate index within the sampling time to construct a time series sample set; the mean and standard deviation are calculated in the time series sample set, and the adaptive threshold is determined by the 3σ criterion.

[0021] Preferably, S2 specifically includes:

[0022] The direction-sensitive deviation rate is compared with an adaptive threshold. If the direction-sensitive deviation rate exceeds the adaptive threshold, it is judged as an abnormal event; otherwise, it is judged as normal.

[0023] The beneficial effects of the technical solution of the present invention are:

[0024] 1. By integrating the target vehicle's own state and the state information of surrounding vehicles, an improved constant acceleration traffic coupling trajectory prediction method is constructed, which realizes high-precision prediction of the short-term motion trend of vehicles, thereby improving the physical consistency and environmental adaptability of the trajectory prediction results.

[0025] 2. By introducing distance attenuation weights and directional consistency coupling factors, and using a bounded nonlinear mapping function to adjust the directional consistency coupling factors, effective modeling of the movement trend of surrounding vehicle groups is achieved. This enables the prediction process to adaptively reflect the consistency or interference of traffic flow directions, thereby improving the adaptability and stability of multi-vehicle interaction scenarios.

[0026] 3. By constructing an improved normalized trajectory deviation assessment method based on the positional difference between the predicted position vector and the actual position vector, and combining the change in heading angle to calculate the direction sensitivity enhancement factor, a joint characterization of abnormal longitudinal motion and abnormal lateral control of the vehicle is achieved, thereby significantly improving the ability to identify complex abnormal behaviors such as sharp turns, abnormal lane changes and serpentine driving.

[0027] 4. By introducing a sliding time window statistical mechanism and combining it with the 3σ criterion to construct an adaptive threshold, the anomaly judgment criteria are dynamically adjusted according to the traffic environment, thereby avoiding the misjudgment problem caused by fixed thresholds, improving the robustness and generalization ability of anomaly detection, and meeting the application requirements of real-time monitoring and early warning of abnormal vehicle driving in intelligent transportation systems. Attached Figure Description

[0028] Figure 1 This is a flowchart of a vehicle abnormal driving detection method based on multi-target trajectory prediction according to the present invention. Detailed Implementation

[0029] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0031] The following description, in conjunction with the accompanying drawings, details a specific scheme for a vehicle abnormal driving detection method based on multi-target trajectory prediction provided by the present invention.

[0032] See attached document Figure 1 The diagram illustrates a flowchart of a vehicle abnormal driving detection method based on multi-target trajectory prediction, provided by an embodiment of the present invention. The method includes the following steps:

[0033] S1. Acquire the target vehicle's own state data, including its actual position vector, velocity, heading angle, and acceleration, as well as the surrounding vehicle state data, including the actual position vector, velocity, heading angle, and acceleration of neighboring vehicles. Using an improved constant acceleration traffic coupling trajectory prediction method, calculate the Euclidean distance between the target vehicle and neighboring vehicles based on their actual position vectors. Based on the Euclidean distance and a set influence radius, calculate the distance decay weight using an exponential decay function. Based on the distance decay weight, combine the velocity, heading angle of neighboring vehicles, and heading angle of the target vehicle to obtain a directional consistency coupling factor. After mapping the directional consistency coupling factor, fuse it with the target vehicle's acceleration to obtain the target vehicle's equivalent acceleration. Based on the target vehicle's velocity, equivalent acceleration, and sampling time interval, calculate the theoretical displacement. Based on the theoretical displacement, obtain the predicted position vector of the target vehicle.

[0034] To acquire the target vehicle's own status data, specifically, to obtain the target vehicle's real-time position data in a planar coordinate system through an onboard satellite positioning system. ,in, Indicates the first The target vehicle in time The actual position vector, Indicates the first The target vehicle in time The horizontal coordinate, Indicates the first The target vehicle in time The longitudinal coordinates are obtained; the longitudinal and lateral velocity components of the target vehicle are acquired through the vehicle control bus or navigation speed measurement module, and the speed of the target vehicle at the current moment is calculated accordingly. , Indicates the first The target vehicle in time The transverse velocity component, Indicates the first The target vehicle in time The longitudinal velocity component; the target vehicle's current heading angle is calculated using the longitudinal and lateral velocity components. It is used to describe the direction of motion of the target vehicle; the acceleration of the target vehicle is calculated by the velocity change at two consecutive sampling times. ,in, Indicates the first The target vehicle in time Acceleration, used to represent the trend of change in the target vehicle's velocity, is an important manifestation of the vehicle's dynamic characteristics. Indicates the first The target vehicle in time speed, This represents the sampling time interval, set to 0.1 seconds.

[0035] After acquiring the target vehicle's own status data, the status data of surrounding vehicles is acquired. The position data of nearby vehicles in a planar coordinate system are obtained using roadside millimeter-wave radar. ,in, Indicates the first Nearby vehicles at time The actual position vector, Indicates the first Nearby vehicles at time The horizontal coordinate, Indicates the first Nearby vehicles at time The longitudinal coordinates are used; the velocity components of surrounding vehicles in the plane are calculated by the position difference method at two consecutive sampling times, and the velocities of neighboring vehicles are obtained from this. and heading angle .

[0036] Based on the lateral and longitudinal coordinates of the target vehicle and its neighboring vehicles, the Euclidean distance between the target vehicle and each neighboring vehicle is calculated. A set of neighboring vehicles whose Euclidean distance is less than or equal to the influence radius is selected, forming a neighborhood set. The influence radius can be determined using roadside millimeter-wave radar; for example, a 77GHz millimeter-wave radar has a range of 80-150 meters.

[0037] Based on the target vehicle's own state data and the state data of surrounding vehicles, an improved constant acceleration traffic coupling trajectory prediction method is used to predict the target vehicle's position vector at the next moment.

[0038] To reflect the objective law that the impact of traffic flow weakens with increasing distance, a distance attenuation weight is introduced. The distance attenuation weight decreases exponentially with increasing distance between vehicles.

[0039] A directional consistency coupling factor is constructed, comprehensively considering the angular relationship between the speeds and headings of neighboring vehicles and the heading of the target vehicle. This factor characterizes the degree of consistency between the overall direction of traffic flow and the direction of the target vehicle. When neighboring vehicles are traveling in the same direction as the target vehicle, the directional consistency coupling factor is positive; otherwise, it tends towards zero or negative values. If the neighborhood set is empty at a certain moment, the directional consistency coupling factor is directly set to zero, indicating that the current environment has no significant directional influence on the target vehicle. To avoid numerical instability caused by an excessively large directional consistency coupling factor, a bounded nonlinear mapping function is introduced to restrict the directional consistency coupling factor to a finite interval. The mapped directional consistency coupling factor serves as a moderating factor for the dynamic trend of the target vehicle caused by the traffic environment.

[0040] The mapped direction consistency coupling factor is applied proportionally to the current acceleration of the target vehicle to form an equivalent acceleration, which represents the trend of vehicle dynamics change after considering the influence of traffic flow environment. When the traffic flow direction is consistent, the equivalent acceleration is slightly enhanced, and when there is interference or reverse trend in the traffic flow direction, the equivalent acceleration is suppressed.

[0041] The formula for calculating equivalent acceleration is: , in, Indicates the first The target vehicle in time The equivalent acceleration; This represents the hyperbolic tangent function, a bounded nonlinear mapping function; This represents the directional consistency coupling factor after mapping, which is the adjustment factor of the traffic environment on the dynamic trend of the target vehicle. Indicates the directional consistency coupling factor; Indicates the first Nearby vehicles at time For the The distance decay weight for each target vehicle is expressed by an exponential decay function, and the calculation formula is as follows: , Indicates time No. The target vehicle and the first Euclidean distance between adjacent vehicles Indicates the radius of influence; Indicates the first The neighborhood set of the target vehicles; Indicates the first Nearby vehicles at time speed; Indicates the first Nearby vehicles at time The heading angle; Indicates the first The target vehicle in time The heading angle; Represents a very small positive number, used to avoid the denominator being zero, and takes the value of ; The direction projection is used to represent the angular relationship between the headings of adjacent vehicles and the heading of the target vehicle, reflecting the degree of directional consistency.

[0042] Under the principle of constant acceleration kinematics, the predicted displacement of the target vehicle at the next moment, i.e., the theoretical displacement, is calculated using the target vehicle's current velocity, equivalent acceleration, and sampling time interval. Combined with the current direction unit vector, the predicted position vector is obtained. The current direction unit vector is a two-dimensional vector determined by the target vehicle's heading angle at the current moment, representing the target vehicle's direction of travel in the planar coordinate system.

[0043] The formula for calculating the predicted location vector is: , in, Indicates the first The target vehicle in time The predicted position vector; Indicates the first The target vehicle in time The actual position vector; Indicates the first The target vehicle in time speed; This represents the sampling time interval, set to 0.1 seconds. Indicates the first The target vehicle in time The direction of the unit vector, ; Indicates the first The predicted displacement of the target vehicle at the next moment, i.e., the theoretical displacement.

[0044] Based on the traditional constant acceleration trajectory prediction model, a traffic flow direction consistency coupling mechanism is introduced. By applying the environmental influence to the vehicle acceleration through a bounded proportional correction method, the physical rationality and environmental adaptability of the trajectory prediction are improved.

[0045] S2. Based on the predicted and actual position vectors of the target vehicle, an improved normalized trajectory deviation evaluation method is adopted, combined with theoretical motion displacement, to obtain spatial deviation characteristics; based on the heading angles of the target vehicle at the next and current moments, the standardized heading angle change of the target vehicle is calculated; the absolute value of the standardized heading angle change is added by 1 to obtain the direction change sensitivity factor; based on the spatial deviation characteristics and the direction change sensitivity factor, the direction sensitivity deviation rate index is obtained; based on the direction sensitivity deviation rate index, an adaptive threshold is determined by the 3σ criterion, and the direction sensitivity deviation rate index is compared with the adaptive threshold to obtain the judgment result.

[0046] Based on the predicted and actual position vectors, an improved normalized trajectory deviation assessment method is employed to calculate the direction sensitivity deviation rate index. Traditional normalized trajectory deviation methods calculate the spatial error between the predicted and actual trajectories and normalize it using theoretical displacement to eliminate the influence of vehicle speed and time scales on the error assessment results. However, these methods do not consider the impact of vehicle heading changes on trajectory deviation, limiting their effectiveness in identifying complex driving behaviors such as sharp turns, serpentine driving, and abnormal lane changes. This paper improves the traditional normalized trajectory deviation assessment method by introducing a direction sensitivity enhancement factor based on the change in heading angle. This improvement comprehensively reflects the vehicle's longitudinal motion stability and lateral handling characteristics, significantly enhancing the accuracy and robustness of abnormal driving detection.

[0047] By calculating the position difference between the predicted position vector and the actual position vector, the degree to which the target vehicle's movement deviates from the theoretically predicted trajectory is reflected. If the vehicle exhibits abnormal behavior, such as sudden steering, emergency lane changing, rapid acceleration, or rapid deceleration, the prediction model will have difficulty accurately matching the movement trajectory, and the position difference will increase significantly.

[0048] To avoid inconsistencies in deviation scale under different speed conditions, a theoretical displacement is introduced for normalization. The theoretical displacement is determined by both the current speed and the equivalent acceleration, reflecting the distance the target vehicle should theoretically travel within the time interval. By proportionalizing the positional difference to the theoretical displacement, the scale effect caused by vehicle speed differences can be eliminated, thus ensuring comparability between vehicles at different speeds.

[0049] Furthermore, a direction change sensitivity factor is introduced. Abnormal vehicle behavior is often accompanied by significant changes in heading angle, such as serpentine driving, sudden lane changes, and reverse steering. Relying solely on positional differences may be insufficient to promptly identify short-term but high-risk abrupt changes in direction. By incorporating the change in heading angle between adjacent sampling times into the calculation of the direction sensitivity deviation rate index, abrupt changes in direction can enhance the index. The change in heading angle reflects the degree of steering abruptness of the vehicle within the current sampling period, and the error caused by heading angle period jumps is eliminated through angle normalization difference. Furthermore, coupling and fusing with spatial deviation features enables simultaneous sensitive detection of both trajectory deviation anomalies and directional control anomalies, thereby improving the accuracy and stability of abnormal driving behavior identification in complex traffic scenarios.

[0050] The formula for calculating the direction sensitivity deviation rate index is: , in, Indicates the first The target vehicle in time Direction-sensitive deviation rate index; Indicates the first The target vehicle in time The actual position vector; Indicates the first The target vehicle in time The positional difference between the actual position vector and the predicted position vector; Indicates the first The target vehicles from time Time The standardized change in heading angle is used to characterize the actual steering amplitude and steering direction of the target vehicle within a sampling time interval. The calculation formula is: , Indicates the first The target vehicle in time The heading angle; This represents the sampling time interval, set to 0.1 seconds. Indicates the sensitivity factor to changes in direction; Indicates spatial deviation characteristics; This represents the theoretical displacement.

[0051] After calculating the direction-sensitive deviation rate index, the adaptive threshold determination stage begins. To avoid the problem of poor adaptability of a fixed threshold under different traffic environments, a sliding time window statistical mechanism is adopted to continuously record the direction-sensitive deviation rate index over several recent sampling periods, constructing a time-series sample set. The average value is calculated from the time-series sample set to reflect the baseline level of the vehicle's normal driving state under the current environment, and the standard deviation is calculated to characterize the natural fluctuation range of the direction-sensitive deviation rate index.

[0052] An adaptive threshold is determined based on the 3σ criterion in statistical process control theory. If the current directional sensitive deviation rate exceeds the adaptive threshold, it can be judged as an abnormal event; otherwise, it can be judged as normal. The adaptive threshold determination stage does not require manual setting of a fixed threshold. It can automatically adjust the judgment criteria according to changes in the road environment. For example, under complex road conditions or high-density traffic conditions, the natural fluctuation range of the indicator increases, and the adaptive threshold automatically increases. Under stable road conditions, the adaptive threshold automatically tightens, thereby improving detection sensitivity.

[0053] In summary, a method for detecting abnormal vehicle driving based on multi-target trajectory prediction has been developed.

[0054] The order of the embodiments is for illustrative purposes only and does not represent the superiority or inferiority of the embodiments. The processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0055] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0056] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for detecting abnormal vehicle driving based on multi-object trajectory prediction, characterized in that, Includes the following steps: S1. Acquire the target vehicle's own state data, including the target vehicle's actual position vector, velocity, heading angle, and acceleration, as well as the surrounding vehicle state data, including the actual position vector, velocity, heading angle, and acceleration of neighboring vehicles; Using an improved constant acceleration traffic coupling trajectory prediction method, calculate the Euclidean distance between the target vehicle and neighboring vehicles based on their actual position vectors. Based on the Euclidean distance between the target vehicle and neighboring vehicles and the set influence radius, the distance decay weight is calculated using an exponential decay function; Based on the distance attenuation weight, the directional consistency coupling factor is obtained by combining the speed of nearby vehicles, the heading angle of nearby vehicles and the heading angle of the target vehicle. After mapping the directional consistency coupling factor, the result of adding 1 to the mapped directional consistency coupling factor, dividing by 2, and then adding 1 again is multiplied with the acceleration of the target vehicle and fused to obtain the equivalent acceleration of the target vehicle. The theoretical displacement is calculated based on the target vehicle's speed, equivalent acceleration, and sampling time interval. Based on the theoretical motion displacement, the predicted position vector of the target vehicle is obtained; S2. Based on the predicted and actual position vectors of the target vehicle, an improved normalized trajectory deviation evaluation method is adopted, combined with theoretical motion displacement, to obtain spatial deviation characteristics; based on the heading angles of the target vehicle at the next and current moments, the standardized heading angle change of the target vehicle is calculated. Add 1 to the absolute value of the standardized change in heading angle to obtain the direction change sensitivity factor; Based on spatial deviation characteristics and directional change sensitivity factors, a directional sensitivity deviation rate index is obtained. Based on the directional sensitivity deviation rate index, an adaptive threshold is determined by the 3σ criterion. The directional sensitivity deviation rate index is compared with the adaptive threshold to obtain the judgment result.

2. The vehicle abnormal driving detection method based on multi-target trajectory prediction according to claim 1, characterized in that, S1 specifically includes: In the improved constant acceleration traffic coupling trajectory prediction method, the directional projection is calculated using a cosine function based on the difference between the heading angles of neighboring vehicles and the target vehicle. The directional consistency coupling factor is calculated by summing the products of the speed of neighboring vehicles, the distance attenuation weight, and the directional projection, and then dividing by the sum of the products of the distance attenuation weight and the speed of neighboring vehicles.

3. The vehicle abnormal driving detection method based on multi-target trajectory prediction according to claim 2, characterized in that, S1 specifically includes: The hyperbolic tangent function is used to map the directional consistency coupling factor, resulting in the mapped directional consistency coupling factor.

4. The vehicle abnormal driving detection method based on multi-target trajectory prediction according to claim 1, characterized in that, S1 specifically includes: Based on the sine and cosine values ​​of the target vehicle's heading angle, the direction unit vector is calculated; the product of the theoretical displacement and the direction unit vector is added to the actual position vector of the target vehicle to obtain the predicted position vector of the target vehicle.

5. The vehicle abnormal driving detection method based on multi-target trajectory prediction according to claim 1, characterized in that, S2 specifically includes: In the improved normalized trajectory deviation assessment method, the positional difference between the predicted and actual position vectors is calculated by subtracting them; the spatial deviation characteristics are calculated by dividing the positional difference by the theoretical motion displacement.

6. The vehicle abnormal driving detection method based on multi-target trajectory prediction according to claim 5, characterized in that, S2 specifically includes: Multiply the direction change sensitivity factor by the spatial deviation characteristic to calculate the direction sensitivity deviation rate index of the target vehicle. The calculation formula is as follows: in, Indicates the first The target vehicle in time Direction-sensitive deviation rate index; Indicates the first The target vehicle in time The actual position vector; Indicates the first The target vehicle in time The predicted position vector; Indicates the first The target vehicle in time speed; Indicates the sampling time interval; Indicates the first The target vehicle in time The equivalent acceleration; Indicates the first The target vehicles from time Time The standardized change in heading angle.

7. The vehicle abnormal driving detection method based on multi-target trajectory prediction according to claim 6, characterized in that, S2 specifically includes: Based on the direction-sensitive deviation rate index, a sliding time window statistical mechanism is adopted to continuously record the direction-sensitive deviation rate index within the sampling time to construct a time series sample set; the mean and standard deviation are calculated in the time series sample set, and the adaptive threshold is determined by the 3σ criterion.

8. The vehicle abnormal driving detection method based on multi-target trajectory prediction according to claim 7, characterized in that, S2 specifically includes: The direction-sensitive deviation rate is compared with an adaptive threshold. If the direction-sensitive deviation rate exceeds the adaptive threshold, it is judged as an abnormal event; otherwise, it is judged as normal.

Images