A method and system for identifying and warning abnormal vehicle driving

Abstract

The present invention relates to the field of traffic safety management technology, specifically a method and system for identifying and warning abnormal vehicle driving; first, the vehicle position, speed, direction and driver operation data are obtained in real time, then the time series data is corrected by a smoothing denoising method of adaptive segmented clustering, and the information confirmation code is generated through standardization, and then the multi-source data is weighted and fused to extract the key indicators of vehicle speed change rate, lane deviation and driver behavior characteristics, and abnormal driving behavior is identified and graded by setting a threshold, and a comprehensive abnormal score is calculated. According to the comparison result between the abnormal score and the set threshold, the strategy decision module generates a graded intervention strategy for minor, medium and severe abnormalities, and finally combines the automatic intervention inside the vehicle with the global coordinated intervention of the road manager. The present invention generates an intervention strategy based on multi-source data collection, processing, analysis and decision-making, thereby improving vehicle driving safety and road management efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of traffic safety management, and in particular to a method and system for identifying and warning abnormal vehicle driving. Background Art

[0002] With the continuous development of intelligent transportation systems, vehicle driving safety has become an important issue in traffic management. Traditional road monitoring and traffic management methods have been unable to cope with increasingly complex driving behaviors and sudden traffic accidents. Abnormal driving behaviors, such as sudden acceleration, sudden braking, lane departure, etc., are often important causes of traffic accidents. Therefore, real-time identification and warning of abnormal driver behavior is of great significance for reducing traffic accidents and improving road safety.

[0003] The Chinese invention patent with announcement number CN118762526B discloses a method and system for identifying and warning abnormal driving of vehicles on highways, the method comprising: S1: collecting vehicle trajectory information and lane boundary line information; S2 calculating vehicle driving status information; S3 storing the remaining trajectory fragments in a driving behavior set; S4 judging whether the vehicle is within a threshold range of abnormal driving; S5 judging whether the vehicle is driving abnormally; S6 if the current vehicle is identified as abnormal driving, abnormal driving warning information is sent to the abnormal driving vehicle. The invention fully conforms to the current development trend of "vehicle-road-cloud integration", and can assist traffic management personnel in conducting all-weather, uninterrupted traffic safety control on highways, especially key sections, and promote the transformation of highway traffic safety from "passive management after the event" to "active prevention before the event", which helps to reduce the accident rate of highways and improve the level of highway traffic safety management.

[0004] As vehicle-to-everything (V2X) technology continues to mature, intelligent analysis combined with multi-source data can provide more accurate judgments and real-time warnings for abnormal driving behaviors. At the same time, by dynamically adjusting intervention strategies and implementing precise driving behavior interventions, it is possible to improve traffic efficiency while ensuring road safety. However, how to process massive amounts of dynamic data, accurately identify abnormal behaviors, and perform reasonable interventions based on the severity of the behaviors remains a technical challenge. Summary of the invention

[0005] The purpose of the present invention is to address the problems existing in the background technology and to propose a method and system for identifying and warning of abnormal vehicle driving.

[0006] The technical solution of the present invention is a method for identifying and warning abnormal driving of a vehicle, comprising the following specific implementation steps:

[0007] S1. Obtain vehicle dynamic monitoring data in real time through a number of sensors;

[0008] S2. A smoothing denoising method based on adaptive segmented clustering divides the vehicle dynamic monitoring data into multiple dynamic windows according to the time series. In each window, K-means clustering is used to divide the data points into mainstream and abnormal classes, and an adaptive smoothing function is used to correct the noise points. At the same time, the smooth connection of the boundary points of adjacent windows is considered. Finally, the corrected data of all windows are merged into a complete time series, and the data is standardized, and an information confirmation code is generated for the monitoring data of any vehicle;

[0009] S3. Calculate the audit code to verify the rationality and non-interference of the data, and use the weighted fusion method to integrate the data from different data sources into a unified feature vector, and extract key features from the feature vector. Based on the key features and preset thresholds, determine whether the vehicle has abnormal behavior and generate a behavior label. Then, assign different weights according to the severity of each abnormal behavior and calculate the comprehensive anomaly score;

[0010] S4. Based on the comparison results of the comprehensive anomaly score and the preset severity assessment threshold, intelligently classify vehicle driving behaviors and generate corresponding intervention strategies: in case of minor anomalies, the driver is reminded through the in-vehicle prompt system, and other drivers are alerted by adjusting traffic lights or turning on warning lights in high-risk areas; in case of moderate anomalies, the vehicle initiates auxiliary intervention to reduce risks, and road managers use V2X technology to communicate with other vehicles in real time to optimize traffic flow; in case of serious anomalies, the vehicle immediately initiates auxiliary intervention, and road managers activate the regional traffic dispatch system to adjust driving strategies and control traffic flow;

[0011] S5, execute the output intervention strategy, combining the automatic intervention inside the vehicle with the global coordinated intervention of the road manager to respond to each abnormal behavior;

[0012] S6. Record abnormal behavior levels, intervention strategies and final effects, and provide comprehensive operation reports.

[0013] Preferably, the implementation process of the smoothing denoising method based on adaptive segmented clustering is as follows:

[0014] S21, divide the data into dynamic windows according to the time series, each window contains w continuous data points, and the window size is adaptively adjusted according to the fluctuation characteristics of the data. Thus, the time series data is divided into multiple windows: {W1, W2, ..., W i , …, W m};

[0015] Among them, W i represents any window; m represents the number of divided windows;

[0016] Dynamic window size formula:

[0017] ;

[0018] In the formula, w represents the dynamic window size; λ represents the sensitivity factor, which is used to control the dynamic change of the window size; w min Indicates the minimum window size, ranging from 5 to 10 data points; w max Indicates the maximum window size; std(X) indicates the standard deviation of window data X;

[0019] S22, clustering the data points in each window, using K-means clustering, and dividing the data points into two categories: mainstream and abnormal, i.e., main data trends and noise points;

[0020] S23, correcting the data points determined as noise in the cluster, and reconstructing the signal using an adaptive smoothing function, the correction formula is:

[0021] ;

[0022] ;

[0023] In the formula, represents the corrected data point; represents the local smoothing value; represents the data trend value, which is determined by linear fitting; α represents the smoothing coefficient, which is dynamically adjusted by the distance between the noise point and the center of the mainstream class; C1 represents the initial clustering center of the mainstream class; x j Represents any window W i Any data point within; std(W i ) represents any window W i The standard deviation of the data;

[0024] S24, smoothly connect the boundary points of adjacent windows, the connection formula is:

[0025] ;

[0026] In the formula, Indicates the final value after correction of the window boundary point; Indicates the correction value of the previous window boundary point; Indicates the correction value of the current window boundary point;

[0027] S25, merge all window-corrected data into a complete time series X = {X1, X2, ..., X n}.

[0028] Preferably, the clustering process for clustering the data points in each window is as follows:

[0029] S31, cluster center initialization:

[0030] ;

[0031] ;

[0032] Where C1 represents the initial cluster center of the mainstream class; C2 represents the initial cluster center of the abnormal class; median(W i ) represents any window W i The median of the data; mean(W i ) represents any window W i The mean of the data; std(W i ) represents any window W i The standard deviation of the data; k represents the offset factor, which is used to adjust the degree of deviation of the abnormal cluster center;

[0033] S32, clustering process: for each data point x j ∈W i , calculate its distance to the two cluster centers:

[0034] ;

[0035] ;

[0036] S33, Comparison d 1j and d 2j The size of will be assigned to the class with a closer distance, that is, the data point x j The class with the smallest distance to the cluster center;

[0037] S34, determination of mainstream class and noise class: the mainstream class has a large number of data points and is concentrated in the middle of the window, while the noise class has a small number of points and deviates from the mainstream class.

[0038] Preferably, the vehicle dynamic monitoring data includes high-definition camera data, GPS data and V2X data.

[0039] Preferably, the process of generating the information confirmation code is as follows:

[0040] S51, select the element generation code ug for the high-definition camera data 1i ∈Z p , based on which the meta confirmation code uc is calculated 1i :

[0041] ;

[0042] Select the element to generate the code ug for GPS data 2i ∈Z p , based on which the meta confirmation code uc is calculated 2i :

[0043] ;

[0044] Select the meta-generator code ug for V2X data 3i ∈Z p , based on which the meta confirmation code uc is calculated 3i :

[0045] ;

[0046] Among them, g2 is the generator of the additive cyclic group G2; Z p is a cyclic group, Z p ={0,1,p-1}; p is the order of the additive cyclic group G2;

[0047] S52, for any vehicle i Generate temporary information generation code IG i =uc 1i × 2i × 3i ;

[0048] S53, generating a meta information confirmation code IUC for the high-definition camera data 1i :

[0049] ;

[0050] Generate meta information confirmation code IUC for GPS data 2i :

[0051] ;

[0052] Generate meta information confirmation code IUC for V2X data 3i :

[0053] ;

[0054] Among them, m i =bC i ||bL i ||bD i , bC i Represents high-definition camera data C i The binary form of bL i Indicates GPS data L i The binary form of ;bD i Indicates V2X data D i Binary form; H0() is a hash function, H0: {0,1} * →G1, G1 is a q-order additive cyclic group; || represents the string concatenation symbol;

[0055] S54, generate information to be confirmed code ICi =IUC 1i ×IUC 2i ×IUC 3i .

[0056] Preferably, the verification process of calculating the audit code to verify the rationality and non-intrusiveness of the data is as follows:

[0057] S61. Extracting high-definition camera data C i 、GPS data i and V2X data D i , the high-definition camera data C i , GPS data L i and V2X data D i Convert to binary data bC i , bL i and bD i , and splice them together to get m i =bC i ||bL i ||bD i ;

[0058] S62, calculate audit code AC i :

[0059] ;

[0060] Among them, e() is a bilinear mapping, e:G1×G2→G T , G T is a multiplicative cyclic group; IC i represents the information to be confirmed code; g2 represents the generator of the additive cyclic group G2; H0() is the hash function, H0: {0,1} * →G1, G1 is a q-order additive cyclic group; IG i Indicates temporary information generation code;

[0061] S63, if AC i =1, then the high-definition camera data C i 、GPS data i and V2X data D i The audit passed, which means that the high-definition camera data C i 、GPS data i and V2X data D i rationality and non-interference.

[0062] Preferably, the calculation process of the comprehensive anomaly score is as follows:

[0063] S71, through the weighted fusion method, the data from different data sources are integrated into a unified feature vector Fi , and give each data source a certain weight:

[0064] ;

[0065] Among them, F i Indicates any vehicle i The unified feature vector of data from different data sources; w1, w2 and w3 represent the weights of GPS data, HD camera data and V2X data respectively, w1+w2+w3=1; C i Indicates high-definition camera data; L i Indicates GPS data; D i Indicates V2X data;

[0066] S72. From the unified eigenvector F i In the above, key features related to abnormal driving behavior are extracted, including:

[0067] Vehicle speed change rate: Estimate the acceleration or deceleration behavior of the vehicle based on the change in vehicle speed before and after the moment;

[0068] ;

[0069] In the formula, Indicates the rate of change of velocity; v i Indicates the speed of the vehicle at the current moment; v i-1 Indicates the speed of the vehicle at the last moment; Indicates a time interval;

[0070] Lane deviation: The lane boundary is identified by a high-definition camera, and the position y of the lane centerline is calculated lanec and obtain the vehicle location through GPS data i , and obtain the lateral position y of the vehicle at the current moment i , by comparing the lateral position y of the vehicle i and the position y of the lane centerline lanec , calculate the lane deviation δ i :

[0071] ;

[0072] Driver behavior characteristics: Extract driver behavior data from V2X data to identify whether sudden acceleration and braking occur;

[0073] S73. Determine whether the vehicle has abnormal behavior based on the result of key feature extraction and a set threshold:

[0074] If the speed change rate Exceeds the set threshold T acc, it is judged as a rapid acceleration behavior;

[0075] If the speed change rate Less than the set negative threshold T brack , it is judged as an emergency braking behavior;

[0076] If the lane departure degree δ i Exceeds the set threshold T lane , it is judged as lane departure behavior;

[0077] If the driver behavior data extracted from the V2X data identifies sudden acceleration and braking, it is determined to be an abnormal operation;

[0078] Generate behavior label B based on this i :

[0079] ;

[0080] S74. Based on the severity of each abnormal behavior, a different weight is assigned to each behavior, and a comprehensive abnormality score S is calculated accordingly. i :

[0081] ;

[0082] Wherein, α1, α2, α3 and α4 are the weight coefficients set for sudden acceleration behavior, sudden braking behavior, lane departure behavior and abnormal operation, respectively.

[0083] The technical solution of the present invention is a vehicle abnormal driving identification and warning system, which is used to execute the above-mentioned vehicle abnormal driving identification and warning method, comprising:

[0084] A data acquisition module, which is used to collect dynamic data of vehicles and roads using a variety of sensor devices;

[0085] The data processing module is used to pre-process the dynamic data acquired by the data acquisition module;

[0086] Intelligent analysis module, responsible for identifying abnormal driving behavior and analyzing it;

[0087] A strategy decision module is used to generate an intervention strategy based on the abnormal driving classification results and the current traffic conditions;

[0088] The intervention execution module is responsible for implementing the intervention strategy and influencing driving behavior through various means to ensure traffic safety.

[0089] The user interaction and feedback module is used to visualize intervention strategies and record system operation data and intervention effects.

[0090] Preferably, the data acquisition module is configured with multiple types of sensing devices installed on the monitored road section, including:

[0091] High-definition cameras for collecting road video data;

[0092] GPS sensor, used to obtain the vehicle's location, speed and direction of travel;

[0093] Vehicle-road cooperative equipment uses V2X technology to communicate with vehicles and receive information about the vehicle's internal status and driver operation.

[0094] Compared with the prior art, the above technical solution of the present invention has the following beneficial technical effects:

[0095] The present invention designs a method and system for identifying and warning abnormal driving of vehicles, which utilizes the technical advantages of multi-source data fusion, intelligent analysis and real-time intervention to achieve efficient and accurate abnormal driving behavior identification and warning functions, and provides an innovative solution for improving road traffic safety and driving experience:

[0096] (1) Improving the comprehensiveness and accuracy of data collection: The data collection module combines high-definition cameras, GPS sensors, and vehicle-road cooperative equipment to collect dynamic information about vehicles and road environments in a multi-dimensional and all-round manner, and uses V2X technology to obtain driver operation data, improving data real-time and reliability, and providing comprehensive input for subsequent analysis;

[0097] (2) Achieve high efficiency and accuracy in data processing: By using the smoothing denoising method of adaptive segmented clustering and dynamically adjusting the window size and clustering analysis, it is possible to accurately identify and correct noise points, greatly improving the smoothness and accuracy of time series data. In addition, by using standardized processing and information confirmation code generation, multi-source data is standardized and information confirmation code generated to ensure that the data dimension is unified and interference-free, thus ensuring that the subsequent analysis module can accurately understand and process the data;

[0098] (3) Improving the accuracy and reliability of abnormal driving behavior identification: Based on multi-source data fusion and feature extraction methods, a weighted fusion strategy is adopted to integrate GPS, camera, and V2X data into a unified feature vector, from which key features such as vehicle speed change rate, lane deviation, and driver behavior characteristics are extracted, thereby enhancing the ability to identify abnormal driving behaviors. Based on the hierarchical judgment method with set thresholds, through quantitative analysis of key features and combined with set thresholds, it can accurately distinguish abnormal behaviors such as sudden acceleration, sudden braking, and lane departure, effectively reducing the misjudgment rate;

[0099] (4) Intelligent classification and response to abnormal behaviors: The calculation method of the comprehensive abnormality score is based on the weight distribution of behavior severity, which can comprehensively reflect the abnormal degree of driving behavior, and intelligently classify abnormal behaviors, and formulate multi-level intervention strategies from mild to severe, including in-vehicle prompts, automatic intervention, and global collaborative intervention by road managers, ensuring efficient handling of abnormal behaviors;

[0100] (5) Enhance the synergy between vehicle and road management: V2X technology enables real-time communication between vehicles and road managers, allowing road managers to dynamically adjust traffic signals, issue warning information and optimize traffic flow, significantly improving the safety and smoothness of road traffic. In serious abnormal situations, the regional traffic dispatch system can respond quickly, coordinate all vehicles and traffic facilities in the area, reduce accident risks and control traffic flow in a timely manner;

[0101] (6) Improving the intelligence and response speed of the system: Combining the vehicle's internal automatic intervention measures (such as automatic braking and steering) with the road manager's global collaborative intervention (such as traffic light adjustment and broadcast warnings) ensures that each abnormal behavior can be handled in a timely and accurate manner, and can respond quickly in complex traffic scenarios, thereby improving the intelligence and response speed of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0102] Figure 1 This is a system architecture diagram of a vehicle abnormal driving recognition and warning system proposed by the present invention;

[0103] Figure 2 The present invention provides a method flow chart of a method for identifying and warning of abnormal vehicle driving. DETAILED DESCRIPTION

[0104] Embodiment 1, as Figure 1 As shown, a vehicle abnormal driving identification and warning system proposed by the present invention includes: a data acquisition module, a data processing module, an intelligent analysis module, a strategy decision module, an intervention execution module, and a user interaction and feedback module.

[0105] The data collection module is equipped with multiple types of sensing devices installed on the monitored road section, and is used to collect vehicle dynamic monitoring data using multiple sensor devices;

[0106] Among them, multiple types of sensing devices include but are not limited to:

[0107] High-definition camera: collects road video data, including but not limited to road environment, lane markings, surrounding vehicles and pedestrian information;

[0108] GPS sensor: Accurately obtain the vehicle's location, speed and driving direction information;

[0109] The vehicle-road cooperative device uses V2X technology (Vehicle to Everything, wireless communication technology for vehicles) to communicate with the vehicle and receive the vehicle's internal status (including but not limited to braking and throttle opening) and driver operation information;

[0110] The data processing module pre-processes the dynamic data acquired by the data acquisition module;

[0111] The intelligent analysis module is responsible for identifying abnormal driving behavior and analyzing it;

[0112] The strategy decision module is used to generate reasonable and effective intervention measures based on the classification results of abnormal driving and the current traffic conditions;

[0113] The intervention execution module is responsible for the specific implementation of intervention strategies, influencing driving behavior through various means to ensure traffic safety;

[0114] The user interaction and feedback module is used to visualize the intervention strategy and record the system operation data and intervention effects.

[0115] Embodiment 2, as Figure 2 As shown, a vehicle abnormal driving identification and warning method proposed by the present invention is applied to a vehicle abnormal driving identification and warning system proposed in Example 1, and its specific implementation steps are as follows:

[0116] S1. The data acquisition module obtains vehicle dynamic monitoring data in real time through several sensors, including but not limited to:

[0117] High-definition cameras collect road video data in real time, including but not limited to road environment, lane markings, surrounding vehicles and pedestrian information;

[0118] GPS sensor: Get any vehicle i Location i , speed v i and driving direction d i ;

[0119] The vehicle-road cooperative equipment uses V2X technology to obtain driver operation data from inside the vehicle;

[0120] Based on this, the data acquisition model transmits the acquired real-time dynamic information to the data processing module.

[0121] S2, the data processing module receives the real-time dynamic information and pre-processes it. The processing process is as follows:

[0122] S21. A smoothing denoising method based on adaptive segmented clustering is used to divide the time series data into multiple dynamic windows, perform clustering analysis based on data features in each window, identify noise points and use adaptive smoothing functions to correct them, thereby achieving smoothing denoising of time series data. The specific implementation process is as follows:

[0123] S2101, divide the data into dynamic windows according to the time series, each window contains w continuous data points, and the window size is adaptively adjusted according to the fluctuation characteristics of the data, thereby the time series data is divided into multiple windows: {W1, W2, ..., W i , …, W m};

[0124] Among them, W i represents any window; m represents the number of divided windows;

[0125] Dynamic window size formula:

[0126] ;

[0127] In the formula, w represents the dynamic window size; λ represents the sensitivity factor, which is used to control the dynamic change of the window size; w min Indicates the minimum window size, ranging from 5 to 10 data points; w max Indicates the maximum window size; std(X) indicates the standard deviation of window data X;

[0128] S2102. Cluster the data points in each window using K-means clustering to divide the data points into two categories: mainstream (i.e., main data trends) and abnormal (i.e., noise points):

[0129] (1) Cluster center initialization:

[0130] ;

[0131] ;

[0132] Where C1 represents the initial cluster center of the mainstream class; C2 represents the initial cluster center of the abnormal class; median(W i ) represents any window W i The median of the data; mean(W i ) represents any window W i The mean of the data; std(W i ) represents any window W i The standard deviation of the data; k represents the offset factor, which is used to adjust the degree of deviation of the abnormal cluster center;

[0133] (2) Clustering process: For each data point xj ∈W i , calculate its distance to the two cluster centers:

[0134] ;

[0135] ;

[0136] Compare 1j and d 2j The size of will be assigned to the class with closer distance;

[0137] (3) Determination of mainstream and noise classes: The mainstream class has a large number of data points and is concentrated in the middle of the window, while the noise class has a small number of points and deviates from the mainstream class;

[0138] S2103, correcting the data points determined as noise in the cluster, and reconstructing the signal using an adaptive smoothing function, the correction formula is:

[0139] ;

[0140] ;

[0141] In the formula, represents the corrected data point; represents the local smoothing value, and this embodiment uses the mean value of the data in the window; represents the data trend value, which is determined by linear fitting; α represents the smoothing coefficient, which is dynamically adjusted by the distance between the noise point and the center of the mainstream class; C1 represents the initial clustering center of the mainstream class; x j Represents any window W i Any data point within; std(W i ) represents any window W i The standard deviation of the data;

[0142] S2104, smoothly connect the boundary points of adjacent windows, the connection formula is:

[0143] ;

[0144] In the formula, Indicates the final value after correction of the window boundary point; Indicates the correction value of the previous window boundary point; Indicates the correction value of the current window boundary point;

[0145] S2105, merge all the corrected window data into a complete time series X={X1, X2, ..., X n};

[0146] S22. Standardize the data and adjust the data of each dimension to a unified dimension range. The standardization formula is:

[0147] ;

[0148] Where, X norm (t) represents the standardized data; X(t) represents the denoised data; μ and σ represent the mean and standard deviation of the data, respectively;

[0149] S23, in order to ensure the rationality and non-interference of multi-source data during transmission, and to ensure that the intelligent analysis module can obtain the monitoring data of the corresponding vehicle from the data of many vehicles, the data processing module is any vehicle i Monitoring data (i.e., high-definition camera data C i , GPS data L i and V2X data D i )Generate information to be confirmed code IC i , and its generation process is as follows:

[0150] S2301, select the element generation code ug for the high-definition camera data 1i ∈Z p , based on which the meta confirmation code uc is calculated 1i :

[0151] ;

[0152] Select the element to generate the code ug for GPS data 2i ∈Z p , based on which the meta confirmation code uc is calculated 2i :

[0153] ;

[0154] Select the meta-generator code ug for V2X data 3i ∈Z p , based on which the meta confirmation code uc is calculated 3i :

[0155] ;

[0156] Among them, g2 is the generator of the additive cyclic group G2; Z p is a cyclic group, Z p ={0,1,p-1}; p is the order of the additive cyclic group G2;

[0157] S2302, for any vehicle i Generate temporary information generation code IG i =uc 1i × 2i× 3i ;

[0158] S2303, generate meta information confirmation code IUC for high-definition camera data 1i :

[0159] ;

[0160] Generate meta information confirmation code IUC for GPS data 2i :

[0161] ;

[0162] Generate meta information confirmation code IUC for V2X data 3i :

[0163] ;

[0164] Among them, m i =bC i ||bL i ||bD i , bC i Represents high-definition camera data C i The binary form of bL i Indicates GPS data L i The binary form of ;bD i Indicates V2X data D i Binary form; H0() is a hash function, H0: {0,1} * →G1, G1 is a q-order additive cyclic group; || represents the string concatenation symbol;

[0165] S2404, generate information to be confirmed code IC i =IUC 1i ×IUC 2i ×IUC 3i ;

[0166] S24, {IG i ,q,G1,G2,g2,H0} are made public to the intelligent analysis module, and {IC i , C i , L i , D i}Transmitted to the intelligent analysis module.

[0167] S3, the intelligent analysis module extracts effective features from multi-source data, integrates multi-modal information, identifies whether the driver has abnormal behavior, and classifies the behavior according to the severity. The specific implementation process is as follows:

[0168] S31. Audit the received high-definition camera data Ci , GPS data L i and V2X data D i The audit process is as follows:

[0169] S3101, extract high-definition camera data C i 、GPS data i and V2X data D i , the high-definition camera data C i , GPS data L i and V2X data D i Convert to binary data bC i , bL i and bD i , and splice them together to get m i =bC i ||bL i ||bD i ;

[0170] S3102, calculate audit code AC i :

[0171] ;

[0172] Among them, e() is a bilinear mapping, e:G1×G2→G T , G T is a multiplicative cyclic group; IC i represents the information to be confirmed code; g2 represents the generator of the additive cyclic group G2; H0() is the hash function, H0: {0,1} * →G1, G1 is a q-order additive cyclic group; IG i Indicates temporary information generation code;

[0173] S3103, if AC i =1, then the high-definition camera data C i 、GPS data i and V2X data D i The audit passed, which means that the high-definition camera data C i 、GPS data i and V2X data D i reasonableness and non-intrusiveness;

[0174] S32, through the weighted fusion method, the data from different data sources are integrated into a unified feature vector F i , and give each data source a certain weight:

[0175] ;

[0176] Among them, Fi Indicates any vehicle i The unified feature vector of data from different data sources; w1, w2 and w3 represent the weights of GPS data, high-definition camera data (i.e., road video data) and V2X data, respectively, w1+w2+w3=1; C i Indicates high-definition camera data; L i Indicates GPS data; D i Indicates V2X data;

[0177] S33, from the unified eigenvector F i In the above example, we extract the key features related to abnormal driving behavior:

[0178] Vehicle speed change rate: Estimate the acceleration or deceleration behavior of the vehicle based on the change in vehicle speed before and after the moment;

[0179] ;

[0180] In the formula, Indicates the rate of change of velocity; v i Indicates the speed of the vehicle at the current moment; v i-1 Indicates the speed of the vehicle at the last moment; Indicates a time interval;

[0181] Lane deviation: The lane boundary is identified by a high-definition camera, and the position y of the lane centerline is calculated lanec and obtain the vehicle location through GPS data i , and obtain the lateral position y of the vehicle at the current moment i , by comparing the lateral position y of the vehicle i and the position y of the lane centerline lanec , calculate the lane deviation δ i :

[0182] ;

[0183] Driver behavior characteristics: Extract driver behavior data from V2X data, including but not limited to the intensity and frequency of acceleration, braking, and steering behaviors, and identify whether sudden acceleration and braking behaviors occur;

[0184] S34. Determine whether the vehicle has abnormal behavior based on the result of key feature extraction and the set threshold:

[0185] If the speed change rate Exceeds the set threshold T acc , it is judged as a rapid acceleration behavior;

[0186] If the speed change rate Less than the set negative threshold Tbrack , it is judged as an emergency braking behavior;

[0187] If the lane departure degree δ i Exceeds the set threshold T lane , it is judged as lane departure behavior;

[0188] If the driver behavior data extracted from the V2X data identifies sudden acceleration and braking, it is determined to be an abnormal operation;

[0189] Generate behavior label B based on this i :

[0190] ;

[0191] S35. Based on the severity of each abnormal behavior, a different weight is assigned to each behavior, and the comprehensive abnormality score S is calculated accordingly. i :

[0192] ;

[0193] Where α1, α2, α3 and α4 are the weight coefficients set for sudden acceleration behavior, sudden braking behavior, lane departure behavior and abnormal operation respectively;

[0194] S36, the comprehensive anomaly score S i Transmitted to the policy decision module.

[0195] S4, the strategy decision module analyzes the comprehensive anomaly score S i The comparison result with the set severity assessment threshold is used to grade the vehicle's driving behavior and intelligently generate intervention strategies:

[0196] Intervention strategy 1: If S i <T low (set level 1 severity threshold), i.e., a minor anomaly, generates an intervention strategy: the driver is reminded to pay attention to the driving behavior only through the vehicle's internal prompt system (including but not limited to voice reminders or vibration feedback). If the abnormal behavior occurs in a high-risk traffic area (including but not limited to near an intersection), the road manager adjusts the traffic lights or turns on the warning lights to increase the alertness of other drivers in the area;

[0197] Intervention strategy 2: If T low ≤S i ≤T high(set secondary severity threshold), i.e. medium abnormality, generates intervention strategy: the vehicle initiates auxiliary intervention (including but not limited to automatic braking or automatic steering) to mitigate potential risks, and the road manager communicates with other nearby vehicles in real time through V2X to instruct other vehicles to slow down, or issue warnings on the road ahead, coordinate changes in traffic lights in the area, optimize traffic flow, and reduce traffic congestion;

[0198] Intervention strategy three: If S i >T high (The set three-level severity threshold), i.e., serious abnormality, generates an intervention strategy: the vehicle initiates auxiliary intervention (including but not limited to automatic braking or automatic steering), and the road manager will activate the regional traffic dispatch system, adjust the driving strategies of surrounding vehicles, and control the surrounding traffic flow in a timely manner;

[0199] The intervention strategy is then transmitted to the intervention execution module.

[0200] S5. The intervention execution module combines the automatic intervention inside the vehicle with the global coordinated intervention of the road manager based on the intervention strategy, ensuring that each abnormal behavior can be responded to in a timely manner, including the vehicle's intervention measures (including but not limited to: automatic braking, steering or acceleration restrictions) and the coordinated intervention of the road manager (including but not limited to traffic light adjustment, warning information release or emergency response).

[0201] S6. The user interaction and feedback module records the detailed process of intervention, including but not limited to the abnormal behavior level, intervention strategy and final effect, provides a comprehensive operation report, analyzes the overall performance and road management effect, and helps managers formulate long-term traffic planning.

[0202] The embodiments of the present invention are described in detail above with reference to the accompanying drawings, but the present invention is not limited thereto, and various changes can be made within the knowledge scope of technicians in the relevant technical field without departing from the purpose of the present invention.

Claims

1. A method for identifying and warning abnormal vehicle driving, characterized in that: The specific implementation steps include the following: S1. Obtain vehicle dynamic monitoring data in real time; S2. A smoothing denoising method based on adaptive segmented clustering divides the vehicle dynamic monitoring data into multiple dynamic windows according to the time series. In each window, K-means clustering is used to divide the data points into mainstream and abnormal classes, and an adaptive smoothing function is used to correct the noise points. At the same time, the smooth connection of the boundary points of adjacent windows is considered. Finally, the corrected data of all windows are merged into a complete time series, and the data is standardized, and an information confirmation code is generated for the monitoring data of any vehicle; The process of generating the information confirmation code is as follows: S51, select the element generation code ug for the high-definition camera data 1i ∈Z p , calculate the meta confirmation code uc 1i : ; Select the element to generate the code ug for GPS data 2i ∈Z p , calculate the meta confirmation code uc 2i : ; Select the meta-generator code ug for V2X data 3i ∈Z p , calculate the meta confirmation code uc 3i : ; Among them, g2 is the generator of the additive cyclic group G2; Z p is a cyclic group, Z p ={0,1,p-1}; p is the order of the additive cyclic group G2; S52, for any vehicle i Generate temporary information generation code IG i =uc 1i × 2i × 3i ; S53, generating a meta information confirmation code IUC for the high-definition camera data 1i : ; Generate meta information confirmation code IUC for GPS data 2i : ; Generate meta information confirmation code IUC for V2X data 3i : ; Among them, m i =bC i ||bL i ||bD i , bC i , bL i , bD i Represents high-definition camera data C i Binary form, GPS data L i Binary form, V2X data D i Binary form; H0() is a hash function, H0: {0,1} * →G1, G1 is a q-order additive cyclic group; || represents the string concatenation symbol; S54, generate information to be confirmed code IC i =IUC 1i ×IUC 2i ×IUC 3i ; S3. Calculate the audit code to verify the rationality and non-interference of the data, and use the weighted fusion method to integrate the data from different data sources into a unified feature vector, and extract key features from the feature vector. Based on the key features and preset thresholds, determine whether the vehicle has abnormal behavior and generate a behavior label. Then, assign different weights according to the severity of each abnormal behavior and calculate the comprehensive anomaly score; The verification process of calculating the audit code to verify the rationality and non-intrusiveness of the data is as follows: S61, receiving the high-definition camera data C i , GPS data L i and V2X data D i Convert to binary data bC i , bL i and bD i , and splice them together to get m i =bC i ||bL i ||bD i ; S62, calculate audit code AC i : ; Among them, e() is a bilinear mapping, e:G1×G2→G T , G T is a multiplicative cyclic group; S63, if AC i =1, it means that the high-definition camera data C i 、GPS data i and V2X data D i reasonableness and non-intrusiveness; S4. Based on the comparison results of the comprehensive anomaly score and the preset severity assessment threshold, intelligently classify vehicle driving behaviors and generate corresponding intervention strategies: in case of minor anomalies, the driver is reminded through the in-vehicle prompt system, and other drivers are alerted by adjusting traffic lights or turning on warning lights in high-risk areas; in case of moderate anomalies, the vehicle initiates auxiliary intervention to reduce risks, and road managers use V2X technology to communicate with other vehicles in real time to optimize traffic flow; in case of serious anomalies, the vehicle immediately initiates auxiliary intervention, and road managers activate the regional traffic dispatch system to adjust driving strategies and control traffic flow; S5, execute the output intervention strategy, combining the automatic intervention inside the vehicle with the global coordinated intervention of the road manager to respond to each abnormal behavior; S6. Record abnormal behavior levels, intervention strategies and final effects, and provide comprehensive operation reports.

2. The method for identifying and warning abnormal vehicle driving according to claim 1, characterized in that: The implementation process of the smoothing denoising method based on adaptive segmented clustering is as follows: S21, divide the data into dynamic windows according to the time series, each window contains w continuous data points, and the window size is adaptively adjusted according to the fluctuation characteristics of the data. Thus, the time series data is divided into multiple windows: {W1, W2, ..., W i , …, W m }; Among them, W i represents any window; m represents the number of divided windows; Dynamic window size formula: ; In the formula, w represents the dynamic window size; λ represents the sensitivity factor, which is used to control the dynamic change of the window size; w min Indicates the minimum window size, ranging from 5 to 10 data points; w max Indicates the maximum window size; std(X) indicates the standard deviation of window data X; S22, clustering the data points in each window, using K-means clustering, and dividing the data points into two categories: mainstream and abnormal, i.e., main data trends and noise points; S23, correcting the data points determined as noise in the cluster, and reconstructing the signal using an adaptive smoothing function, the correction formula is: ； ； ； In the formula, represents the corrected data point; represents the local smoothing value; represents the data trend value, which is determined by linear fitting; α represents the smoothing coefficient, which is dynamically adjusted by the distance between the noise point and the center of the mainstream class; C1 represents the initial clustering center of the mainstream class; x j Represents any window W i For any data point x j ∈W i ; std(W i ) represents any window W i The standard deviation of the data; median(W i ) represents any window W i The median of the data; S24, smoothly connect the boundary points of adjacent windows, the connection formula is: ; In the formula, Indicates the final value after correction of the window boundary point; Indicates the correction value of the previous window boundary point; Indicates the correction value of the current window boundary point; S25, merge all window-corrected data into a complete time series X = {X1, X2, ..., X n }.

3. The method for identifying and warning abnormal vehicle driving according to claim 2, characterized in that: The clustering process for clustering the data points in each window is as follows: S31, cluster center initialization: ； ； Where C1 represents the initial cluster center of the mainstream class; C2 represents the initial cluster center of the abnormal class; median(W i ) represents any window W i The median of the data; mean(W i ) represents any window W i The mean of the data; std(W i ) represents any window W i The standard deviation of the data; k represents the offset factor, which is used to adjust the degree of deviation of the abnormal cluster center; S32, clustering process: for each data point x j ∈W i , calculate its distance to the two cluster centers: ； ； S33, Comparison d 1j and d 2j The size of will be assigned to the class with a closer distance, that is, the data point x j The class with the smallest distance to the cluster center; S34, determination of mainstream class and noise class: the mainstream class has a large number of data points and is concentrated in the middle of the window, while the noise class has a small number of points and deviates from the mainstream class.

4. The method for identifying and warning abnormal vehicle driving according to claim 1, characterized in that: Vehicle dynamic monitoring data includes high-definition camera data, GPS data and V2X data.

5. The method for identifying and warning abnormal vehicle driving according to claim 1, characterized in that: The calculation process of the comprehensive anomaly score is as follows: S71, through the weighted fusion method, the data from different data sources are integrated into a unified feature vector F i , and give each data source a certain weight: ; Among them, F i Indicates any vehicle i The unified feature vector of data from different data sources; w1, w2 and w3 represent the weights of GPS data, HD camera data and V2X data respectively, w1+w2+w3=1; C i Indicates high-definition camera data; L i Indicates GPS data; D i Indicates V2X data; S72. From the unified eigenvector F i In the above, key features related to abnormal driving behavior are extracted, including: Vehicle speed change rate: Estimate the acceleration or deceleration behavior of the vehicle based on the change in vehicle speed before and after the moment; ； In the formula, Indicates the rate of change of velocity; v i Indicates the speed of the vehicle at the current moment; v i-1 Indicates the speed of the vehicle at the last moment; Indicates a time interval; Lane deviation: The lane boundary is identified by a high-definition camera, and the position y of the lane centerline is calculated lanec and obtain the vehicle location through GPS data i , and obtain the lateral position y of the vehicle at the current moment i , by comparing the lateral position y of the vehicle i and the position y of the lane centerline lanec , calculate the lane deviation δ i : ; Driver behavior characteristics: Extract driver behavior data from V2X data to identify whether sudden acceleration and braking occur; S73. Determine whether the vehicle has abnormal behavior based on the result of key feature extraction and a set threshold: If the speed change rate Exceeds the set threshold T acc , it is judged as a rapid acceleration behavior; If the speed change rate Less than the set negative threshold T brack , it is judged as an emergency braking behavior; If the lane departure degree δ i Exceeds the set threshold T lane , it is judged as lane departure behavior; If the driver behavior data extracted from the V2X data identifies sudden acceleration and braking, it is determined to be an abnormal operation; Generate behavior label B based on this i : ； S74. Based on the severity of each abnormal behavior, a different weight is assigned to each behavior, and a comprehensive abnormality score S is calculated accordingly. i : ; Wherein, α1, α2, α3 and α4 are the weight coefficients set for sudden acceleration behavior, sudden braking behavior, lane departure behavior and abnormal operation, respectively.

6. A vehicle abnormal driving recognition and warning system, which is used to execute a vehicle abnormal driving recognition and warning method according to any one of claims 1 to 5, characterized in that: include: A data acquisition module, which is used to collect dynamic data of vehicles and roads using a variety of sensor devices; The data processing module is used to pre-process the dynamic data acquired by the data acquisition module; Intelligent analysis module, responsible for identifying abnormal driving behavior and analyzing it; A strategy decision module is used to generate an intervention strategy based on the abnormal driving classification results and the current traffic conditions; The intervention execution module is responsible for implementing the intervention strategy and influencing driving behavior through various means to ensure traffic safety. The user interaction and feedback module is used to visualize intervention strategies and record system operation data and intervention effects.

7. The vehicle abnormal driving recognition and warning system according to claim 6, characterized in that: The data collection module is equipped with multiple types of sensing devices installed on the monitored road sections, including: High-definition cameras for collecting road video data; GPS sensor, used to obtain the vehicle's location, speed and direction of travel; Vehicle-road cooperative equipment uses V2X technology to communicate with vehicles and receive information about the vehicle's internal status and driver operation.

Images