Road interleaving area mixed driving risk identification method and system, and computer storage medium

Abstract

The embodiment of the invention discloses a mixed driving risk identification method and system for a road interleaving area and a computer storage medium, and the method comprises the steps: respectively quantifying the contribution value of each index to a traffic accident and a traffic conflict for each index in a multi-dimensional risk identification index system; sorting the indexes according to the contribution values to obtain a traffic accident contribution value sorting result and a traffic conflict contribution value sorting result; according to the traffic accident contribution value sorting result and the traffic conflict contribution value sorting result, indexes of which the contribution values to the traffic accidents and the traffic conflicts exceed a preset threshold value are selected and combined to form a high-contribution index system; screening the indexes of which the correlation among the indexes is smaller than a threshold value from the high-contribution index system; training a double-risk identification model based on the screened high-contribution index system; and constructing an accident conflict two-dimensional risk matrix so as to detect real-time traffic flow data by using the trained double-risk identification model and the accident conflict two-dimensional risk matrix.

Description

Technical Field

[0001] The embodiments of this specification relate to the field of road traffic safety technology, and in particular to a method, system, and computer storage medium for identifying mixed traffic risks in weaving areas. Background Art

[0002] Weaving areas are areas where vehicles merge and diverge, such as highway entrances and exits. This type of road section has complex traffic organization patterns and frequent traffic conflicts, and has always been a key area for traffic accident prevention and control. Autonomous driving is the development trend of intelligent road transportation systems. With the development and application of vehicle-road collaboration and autonomous driving technologies, autonomous driving vehicles are gradually transitioning from testing to operation in actual road environments. Mixed traffic flows consisting of manually driven and autonomous driving vehicles will become the new normal for road operations. However, due to the functional design requirements of weaving areas, moving vehicles need to complete weaving lane changes within a limited space. In this environment, conflicts between manually driven vehicles and autonomous driving vehicles, and between vehicles and road facilities will become more prominent, bringing new problems and challenges to the proactive prevention of traffic accidents and traffic safety management.

[0003] Due to the differences in driving behavior, human-driven and autonomous vehicles exhibit significant differences in driving behavior, information perception, and application. Mixed traffic flows exhibit more complex randomness and uncertainty. In particular, in weaving zones, vehicles in different driving modes adopt different lane-changing and following behaviors, resulting in diverse and complex disturbances to road safety. Existing traffic risk identification, early warning, and control methods are primarily designed for human-driven vehicles and conventional road operation scenarios, and are unable to meet the needs of preventing traffic safety risks in weaving zones under mixed traffic flow conditions.

[0004] With the gradual maturity of multi-source data perception and vehicle-road collaboration technologies, road traffic management models are evolving from traditional signal control to vehicle-road-cloud collaborative management and control. Leveraging more accurate real-time interactive data between vehicles and roads will revolutionize traffic risk assessment and early warning and control technologies. Effectively identifying the risks of mixed traffic operations involving human and autonomous driving in weaving areas has become crucial and of great significance for future road traffic safety management. Summary of the Invention

[0005] In view of this, the embodiments of this specification provide a method, system and computer storage medium for identifying mixed traffic risks in road weaving areas, which are used to solve the problems of comprehensive risk identification of traffic operation safety when automatic driving and manually driven vehicles mix in complex road sections such as road weaving areas in the existing technology.

[0006] The embodiments of this specification adopt the following technical solutions:

[0007] The present invention provides a method for identifying mixed traffic risks in a weaving area, and the method includes:

[0008] For each indicator in the constructed multi-dimensional risk identification indicator system, quantify the contribution value of each indicator to traffic accidents and traffic conflicts;

[0009] According to the size of the contribution value, the indicators are sorted respectively to obtain the traffic accident contribution value sorting result and the traffic conflict contribution value sorting result;

[0010] Using the traffic accident contribution value ranking results and the traffic conflict contribution value ranking results, respectively select indicators whose contribution values ​​to traffic accidents and traffic conflicts exceed preset thresholds, and combine them to form a high contribution index system;

[0011] From the high contribution index system, screening indicators whose correlations among the indicators are less than a threshold;

[0012] Training the dual-risk identification model based on the screened high-contribution indicator system, wherein the correlation between the indicators in the screened high-contribution indicator system is less than a threshold;

[0013] A two-dimensional accident conflict risk matrix is ​​constructed to detect real-time traffic flow data using the trained dual-risk identification model and the two-dimensional accident conflict risk matrix, to obtain risk level and risk status identification results corresponding to the real-time traffic flow data, and to use the two-dimensional accident conflict risk matrix as a risk level determination criterion.

[0014] The embodiment of this specification further provides a system for identifying the risk of mixed traffic in weaving areas, the system comprising:

[0015] A quantification module quantifies the contribution of each indicator in the constructed multi-dimensional risk identification indicator system to traffic accidents and traffic conflicts;

[0016] A sorting module, which sorts the indicators according to the contribution values ​​to obtain traffic accident contribution value sorting results and traffic conflict contribution value sorting results;

[0017] a merging module, which uses the traffic accident contribution value ranking results and the traffic conflict contribution value ranking results to select indicators whose contribution values ​​to traffic accidents and traffic conflicts exceed preset thresholds, respectively, and merges them into a high contribution indicator system;

[0018] A screening module, screening indicators whose correlation between indicators is less than a threshold from the high contribution indicator system;

[0019] a training module for training the dual-risk identification model based on a screened high-contribution indicator system, wherein the correlation between the indicators in the screened high-contribution indicator system is less than a threshold;

[0020] A construction module is provided to construct a two-dimensional risk matrix for accident conflicts, so as to detect real-time traffic flow data using the trained dual-risk identification model and the two-dimensional risk matrix for accident conflicts, obtain risk level and risk status identification results corresponding to the real-time traffic flow data, and use the two-dimensional risk matrix for accident conflicts as a risk level determination standard.

[0021] The embodiments of this specification also provide a computer storage medium including a program for use in conjunction with an electronic device, the program being executable by a processor to perform the following steps:

[0022] For each indicator in the constructed multi-dimensional risk identification indicator system, quantify the contribution value of each indicator to traffic accidents and traffic conflicts;

[0023] According to the size of the contribution value, the indicators are sorted respectively to obtain the traffic accident contribution value sorting result and the traffic conflict contribution value sorting result;

[0024] Using the traffic accident contribution value ranking results and the traffic conflict contribution value ranking results, respectively select indicators whose contribution values ​​to traffic accidents and traffic conflicts exceed preset thresholds, and combine them to form a high contribution index system;

[0025] From the high contribution index system, screening indicators whose correlations among the indicators are less than a threshold;

[0026] Training the dual-risk identification model based on the screened high-contribution indicator system, wherein the correlation between the indicators in the screened high-contribution indicator system is less than a threshold;

[0027] A two-dimensional accident conflict risk matrix is ​​constructed to detect real-time traffic flow data using the trained dual-risk identification model and the two-dimensional accident conflict risk matrix, to obtain risk level and risk status identification results corresponding to the real-time traffic flow data, and to use the two-dimensional accident conflict risk matrix as a risk level determination criterion.

[0028] At least one of the above technical solutions adopted in the embodiments of this specification can achieve the following beneficial effects:

[0029] By constructing a dual-risk identification model and a two-dimensional risk matrix for accident conflicts, we can detect real-time traffic flow data and predict the risk level and risk status of mixed traffic in weaving areas, so as to provide accurate early warning of risk factors, reduce the probability of traffic accidents and conflicts, specialize in mixed traffic scenarios in weaving areas, and cover complex conflicts.

[0030] In addition, the risk identification dimension realizes multi-factor coupling, takes into account the impact of multiple factors such as people, vehicles, roads, and environment on risks, takes into account static road conditions, historical accident events, dynamic traffic conflicts and other issues, and proposes a scientific and reasonable comprehensive risk identification method to solve the technical problems of traditional single risk identification based on a single factor. It also solves the problem of the lack of mixed risk identification methods for manual driving and automatic driving in complex sections such as weaving areas, realizes the dual prediction of dynamic and static indicators combined with accident conflicts, forms a "monitoring-prediction-warning-update" closed loop, and supports real-time interaction among multiple terminals. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] The drawings described herein are used to provide a further understanding of the embodiments of this specification and constitute a part of the embodiments of this specification. The illustrative embodiments and descriptions of this specification are used to explain this application and do not constitute an improper limitation on this application. In the drawings:

[0032] Figure 1 A flow chart of a method for identifying mixed traffic risk in a weaving area provided in an embodiment of this specification;

[0033] Figure 2 A logical diagram of the construction of a multi-dimensional risk identification indicator system corresponding to a mixed traffic risk identification method for weaving areas provided in an embodiment of this specification;

[0034] Figure 3 A schematic diagram of a two-dimensional accident conflict risk matrix corresponding to a mixed traffic risk identification method for a weaving area provided in an embodiment of this specification;

[0035] Figure 4 A schematic diagram of a road risk identification process corresponding to a mixed traffic risk identification method for a weaving area provided in an embodiment of this specification;

[0036] Figure 5 A schematic diagram of the structure of a mixed traffic risk identification system for road weaving areas provided in an embodiment of this specification;

[0037] Figure 6 A schematic diagram of the structure of a computer storage medium corresponding to a method for identifying mixed traffic risks in road weaving areas provided in an embodiment of this specification. DETAILED DESCRIPTION

[0038] In the existing technology, the mixed traffic risk identification method in road weaving areas has the following defects:

[0039] Due to scenario limitations, existing risk identification research focuses on conventional roads or single driving modes. The research scenarios for traffic risk identification in mixed traffic of manually driven and automatically driven vehicles are relatively limited. Most of them are manual driving or human-machine co-driving scenarios in the same vehicle. They cannot provide scientific methodological support for the effective formulation of strategies to improve the safety of mixed traffic flows of manually driven and automatically driven vehicles.

[0040] In addition, current risk identification research scenarios are mostly concentrated on conventional road sections, and there is a lack of risk identification methods for complex sections in weaving areas. In particular, there is a relative lack of research on risk identification of road sections when manually driven and autonomous vehicles mix in weaving areas. The data is not real-time enough, and dynamic traffic flow and static environmental factors are not integrated.

[0041] The model lacks adaptability. The existing mixed traffic risk identification model in weaving areas is only applicable to single-vehicle conflict judgment, cannot support section-level comprehensive risk assessment, does not distinguish between the independent contributions of accidents and conflicts, and the static weights are difficult to adapt to the dynamic scenarios in weaving areas.

[0042] The warning lacks accuracy. The existing solution focuses on vehicle collaborative decision-making and has not established a risk tracing mechanism, resulting in the generalization of warning information (such as only prompting "high-risk road sections").

[0043] Therefore, the embodiments of this specification provide a method, system and computer storage medium for identifying mixed traffic risks in road weaving areas. By constructing a dual risk identification model and a two-dimensional risk matrix for accident conflicts, real-time traffic flow data can be detected to predict the risk level and risk status of mixed traffic in weaving areas, so as to provide accurate early warning of risk factors, reduce the probability of traffic accidents and conflicts, specialize in mixed traffic scenarios in weaving areas, and cover complex conflicts.

[0044] In addition, the risk identification dimension realizes multi-factor coupling, takes into account the impact of multiple factors such as people, vehicles, roads, and environment on risks, takes into account static road conditions, historical accident events, dynamic traffic conflicts and other issues, and proposes a scientific and reasonable comprehensive risk identification method to solve the technical problems of traditional single risk identification based on a single factor. It also solves the problem of the lack of mixed risk identification methods for manual driving and automatic driving in complex sections such as weaving areas, realizes the dual prediction of dynamic and static indicators combined with accident conflicts, forms a "monitoring-prediction-warning-update" closed loop, and supports real-time interaction among multiple terminals.

[0045] To make the objectives, technical solutions, and advantages of this application more clear, the technical solutions of this application will be clearly and completely described below in conjunction with the specific embodiments of this specification and the corresponding drawings. Obviously, the embodiments described are only part of the embodiments of this specification, not all of the embodiments. Based on the embodiments in this specification, all other embodiments obtained by ordinary technicians in this field without making any creative efforts are within the scope of protection of this application.

[0046] The technical solutions provided by the embodiments of this specification are described in detail below with reference to the accompanying drawings.

[0047] like Figure 1 The figure shows a flow chart of a method for identifying mixed traffic risks in road weaving areas provided by an embodiment of this specification.

[0048] In the embodiment of this specification, the method for identifying mixed traffic risk in a weaving area may specifically include the following steps:

[0049] S101: quantifying the contribution of each indicator in the constructed multi-dimensional risk identification indicator system to traffic accidents and traffic conflicts;

[0050] S103: sorting the indicators according to the contribution values ​​to obtain traffic accident contribution value sorting results and traffic conflict contribution value sorting results;

[0051] S105: Using the traffic accident contribution value ranking results and the traffic conflict contribution value ranking results, respectively selecting indicators whose contribution values ​​to traffic accidents and traffic conflicts exceed preset thresholds, and merging them to form a high contribution indicator system;

[0052] S107: Screening indicators whose correlations among indicators are less than a threshold value from the high contribution indicator system;

[0053] S109: Training the dual-risk identification model based on the screened high-contribution indicator system, wherein the correlation between the indicators in the screened high-contribution indicator system is less than a threshold;

[0054] S111: Construct a two-dimensional risk matrix for accident conflicts, and use the trained dual-risk identification model and the two-dimensional risk matrix for accident conflicts to detect real-time traffic flow data, obtain risk level and risk status identification results corresponding to the real-time traffic flow data, and use the two-dimensional risk matrix for accident conflicts as a risk level determination standard.

[0055] The embodiments in this specification primarily address the problem of identifying comprehensive traffic safety risks in complex road scenarios such as weaving zones, where autonomous and manually driven vehicles coexist. This method proposes a comprehensive risk level identification method by comprehensively considering the coupling of risk factors across multiple dimensions, including human, vehicle, road, and environmental factors. The method also considers the contribution of indicators to risk impact and the correlation between indicators. It also considers traffic accident and conflict prediction.

[0056] As an application embodiment of this specification, for step S101, building a multi-dimensional risk identification indicator system may specifically include:

[0057] Establishing primary indicators including road condition indicators, traffic flow status indicators, mixed traffic ratio indicators for autonomous vehicles, indicators for illegal human driving behavior, traffic conflict indicators, and environmental factor indicators;

[0058] On the basis of each of the first-level indicators, establish the second-level indicators corresponding to each of the first-level indicators;

[0059] Based on the first-level indicators and the second-level indicators, the multi-dimensional risk identification indicator system is constructed.

[0060] In the embodiments of this specification, according to actual application scenarios, the sources of mixed traffic risks in road weaving areas mainly come from road geometry constraints, traffic flow disorder, driving behavior differences and environmental disturbances. Based on these risk sources, the first-level indicators can be established.

[0061] Among them, the road condition indicators can specifically include six secondary indicators, such as the length of the weaving area, the width of the emergency lane, the number of lanes in the weaving area, the length of the acceleration and deceleration lanes, the length of the entrance and exit markings, the intersection angle of the main line and the ramp markings at the entrance and exit, and the integrity rate of traffic safety facilities, integrating geometric parameters with the status of traffic facilities.

[0062] The traffic flow status indicators can specifically include seven secondary indicators: saturation of the weaving area, traffic volume on the entrance ramp, traffic volume on the exit ramp, main line speed difference at the entrance and exit, proportion of large vehicles in the weaving area, proportion of large vehicles among incoming vehicles, and proportion of large vehicles among outgoing vehicles. These can dynamically capture the characteristics of traffic disorder.

[0063] The autonomous driving vehicle mixed traffic ratio index can specifically use the autonomous driving vehicle mixing rate index to characterize the impact of the autonomous driving vehicle mixed traffic ratio, wherein the autonomous driving vehicle mixing rate is a secondary indicator corresponding to the autonomous driving vehicle mixed traffic ratio index, which can specifically be the ratio of the number of autonomous driving vehicles to the traffic volume per unit time in the weaving area, and can be the core variable that quantifies the degree of mixing.

[0064] The manual illegal driving behavior indicators may specifically include four secondary indicators, namely, an overspeed rate, an overloading rate, an illegal lane change rate, and an illegal parking rate, which may be associated with the behavioral uncertainty of manually driven vehicles.

[0065] The environmental factor indicators may specifically include secondary indicators such as the number of bad weather days per year and ambient illumination, and may couple the effects of meteorological and optical disturbances.

[0066] Among them, for ambient illumination (unit: lux), if the ambient illumination during the day is greater than 10,000 lux, this may cause a strong glare risk, and it is necessary to link the autonomous driving vehicle sensor to reduce the confidence level; if the ambient illumination at night is less than 10 lux, this may cause insufficient visibility, and it is necessary to trigger the LED board to strengthen the warning.

[0067] The traffic conflict index can specifically use the traffic conflict rate index to characterize the impact of traffic conflicts, which can replace the real-time risk proxy index of accidents. The traffic conflict rate index is a secondary index corresponding to the traffic conflict index, and is calculated as described in the following formula (1):

[0068]

[0069] Where f is the traffic conflict rate, in units of times / (vehicle km); TC is the number of conflicts, in units of times; Q is the traffic volume, in units of vehicles / h; L is the road section length, in units of km.

[0070] For example, in actual applications, the traffic conflict rate indicator can warn of high-risk conditions 20 minutes earlier than accident data.

[0071] By building a multi-dimensional risk identification indicator system, and at the technical level, layered processing of dynamic and static indicators (static calibration combined with dynamic early warning) solves the complex system modeling challenges in weaving areas. At the engineering level, it supports low-cost roadside equipment to collect full indicators, promoting the large-scale implementation of proactive safety control measures.

[0072] As an application embodiment of this specification, in step S101, for each indicator in the constructed multi-dimensional risk identification indicator system, the contribution value of each indicator to traffic accidents and traffic conflicts is quantified, including:

[0073] SHAP value analysis was performed on each of the aforementioned indicators;

[0074] The contribution of the indicators to traffic accidents and traffic conflicts is quantified based on the SHAP value analysis results.

[0075] In the embodiments of this specification, for the same indicator A, the contribution values ​​of indicator A to traffic accidents and traffic conflicts are quantified separately. That is, one indicator corresponds to two different contribution values. This is because the same indicator has a greater impact on traffic accidents, but not necessarily a greater impact on traffic conflicts, so SHAP value analysis needs to be performed separately.

[0076] In a specific application scenario, the calculation formula of the SHAP value is as follows:

[0077]

[0078] Among them, Φ j is the SHAP value of the jth traffic parameter, N is the set of all parameters in the training set, its dimension is M, S is the subset extracted from N, its dimension is |S|, f x (S) is the sample mean calculated using the parameter set S, fx (S∪{i}) is the sample average calculated again based on the parameter set S plus the parameter i, is the weight item.

[0079] In the embodiments of this specification, after constructing the accident traffic parameter combination and the conflict traffic parameter combination, it is also necessary to clarify the degree of influence of each traffic parameter on the accident (conflict) during actual application. The SHAP value method uses the classic Shapley value from game theory and its related extensions to link the optimal credit allocation with local interpretation. It is a method to explain individual predictions based on the optimal Shapley value in game theory, and can quantify the contribution of each indicator (traffic parameter) to the risk identification process.

[0080] When identifying traffic risk states in mixed traffic areas, the first step is to determine the contribution of each traffic parameter to the identification results. SHAP value analysis allows for a visual output of the degree of influence (contribution) of each traffic parameter on the risk identification results. On this basis, the traffic parameters are ranked according to their impact on traffic accidents (conflicts). If the contribution is high, the impact of this factor on the identification results must be considered during risk identification. If the parameter's contribution to the identification results is low, excluding it during identification can reduce model complexity, increase model training speed, and enhance the model's generalization ability. Secondly, the correlation between the selected traffic parameters is low, so each traffic parameter can independently reflect the risk characteristics of a certain dimension, avoiding repeated calculations.

[0081] As an application embodiment of this specification, in step S103, the indicators are sorted according to the contribution values ​​to obtain traffic accident contribution value sorting results and traffic conflict contribution value sorting results, which may specifically include:

[0082] Sorting the indicators according to their contribution to traffic accidents to obtain a ranking result of their contribution to traffic accidents;

[0083] According to the contribution value of each indicator to the traffic conflict, the indicators are sorted to obtain the traffic conflict contribution value sorting result.

[0084] In the embodiment of this specification, since the same indicator has different contribution values ​​for traffic accidents and traffic conflicts, in the process of sorting the contribution values ​​of the indicators, it is necessary to sort them according to the contribution values ​​for traffic accidents and traffic conflicts respectively, and obtain two different sorting results.

[0085] As an application embodiment of the present specification, in step S105, the traffic accident contribution value ranking results and the traffic conflict contribution value ranking results are used to select indicators whose contribution values ​​to traffic accidents and traffic conflicts exceed preset thresholds, respectively, and combine them to form a high contribution index system, which may specifically include:

[0086] Determining, based on the traffic accident contribution value ranking result, a first indicator set whose contribution value to the traffic accident exceeds a preset threshold;

[0087] Determining, based on the traffic conflict contribution value ranking result, a second indicator set whose contribution value to the traffic conflict exceeds a preset threshold;

[0088] The first indicator set and the second indicator set are combined to form the high contribution indicator system.

[0089] In the embodiments of this specification, in actual applications, for indicators with smaller contribution values, the impact on traffic accidents and traffic conflicts can be ignored. In this case, there is no need to perform additional calculations on these indicators. Instead, a first step of screening is required to select a set of indicators that have a greater impact on traffic accidents and traffic conflicts. Finally, the two indicator sets are combined to form a high-contribution indicator system. That is, in the high-contribution indicator system, the contribution values ​​corresponding to all indicators exceed the preset threshold.

[0090] The specific value of the preset threshold can be pre-set according to actual application and is not specifically limited here.

[0091] For example, for indicator A, if the contribution value of indicator A to traffic accidents is greater than the preset threshold, then indicator A can be selected into the first indicator set, and if the contribution value of indicator A to traffic conflicts is less than the preset threshold, then indicator A will be screened out from the second indicator set.

[0092] Furthermore, in step S107, screening indicators whose correlations between indicators are less than a threshold from the high contribution indicator system may specifically include:

[0093] Calculating the correlation between the various indicators in the high contribution index system;

[0094] If the correlation between the two indicators exceeds the threshold, one indicator is selected from the two indicators and the other indicator is screened out, so that in the high contribution indicator system after screening, the correlation between each indicator is less than the threshold.

[0095] In the embodiments of this specification, for different indicators in the high contribution index system, there may be some indicators that are highly correlated. In this case, there is no need to perform repeated calculations during the calculation process. In this case, a second screening is required to select indicators with lower correlation between indicators to avoid repeated calculations.

[0096] For example, in the high contribution index system, there are five indicators that have relatively large contribution values ​​to traffic accidents, but the correlation between these five indicators is also relatively large. In this case, only one indicator needs to be selected from these five indicators to represent the contribution of these five indicators to traffic accidents. In this way, the calculation amount of the system can be reduced.

[0097] Furthermore, calculating the correlation between the indicators in the high contribution index system may specifically include:

[0098] One indicator in the high contribution index system is used as the dependent variable, and the other indicators are used as independent variables, and a cross correlation analysis is performed;

[0099] According to the correlation analysis results, obtain the indicator combination with smaller correlation.

[0100] In the examples presented in this specification, the traffic environment in intersecting roads is complex, and identifying risk states in mixed traffic requires comprehensive consideration of multiple traffic factors. At the same time, the model constructed to address this problem must be computationally simple, accurate, and provide clear conclusions. A tree-based decision tree model, however, can extract a series of rules from known historical traffic flow data to classify newly input traffic flow data. Its relatively simple structure and strong interpretability align with the project's risk identification requirements.

[0101] Furthermore, when calculating a decision tree model, if too many input parameters are used, the model is prone to excessive computation time and the risk of overfitting. Therefore, before training the decision tree model, it is necessary to pre-prune the mixed traffic flow operational characteristic representation system to select parameters that are clearly relevant to the model from multiple traffic parameters.

[0102] In specific application scenarios, the Jarque-Bera test method can be used to screen the correlation between traffic parameters. The specific process is as follows:

[0103] Step 1) Conduct hypothesis test H0: a certain traffic parameter is correlated with other traffic parameters;

[0104] Conduct hypothesis test H1: a certain traffic parameter is not correlated with other traffic parameters;

[0105] Step 2) Assume that a traffic parameter in the mixed traffic flow operation characteristic representation system is Traffic parameters The kurtosis is K, the traffic parameter The skewness of is S, then construct the test statistic:

[0106]

[0107] Among them, χ 2 (2) indicates that the sample obeys the chi-square distribution with 2 degrees of freedom; n indicates the sample size;

[0108] Step 3) Use JB to check the chi-square distribution table and obtain the corresponding significance p-value;

[0109] Step 4) extracting traffic parameters with significance less than 0.05 from the test results, and constructing traffic parameter combinations with correlation analysis below the threshold based on the traffic parameters with significance less than 0.05.

[0110] As an application embodiment of this specification, in step S109, the dual risk identification model is trained based on the screened high contribution index system, including:

[0111] Constructing a dual-risk identification model, wherein the dual-risk identification model includes an accident prediction model and a conflict prediction model;

[0112] Collect historical traffic flow data based on the screened high-contribution indicator system;

[0113] performing a pre-cleaning operation on the historical traffic flow data according to the traffic accident contribution value ranking results and the traffic conflict contribution value ranking results to obtain training sample data;

[0114] Dividing the training sample data into a training set and a test set;

[0115] Using the training set to train the dual-risk identification model, and using the test set to test the dual-risk identification model;

[0116] Cross-validation is performed based on the training results and the test results, and the parameters of the dual-risk identification model are optimized to obtain a trained dual-risk identification model.

[0117] In the embodiment of this specification, the accident prediction model can output the probability of future accidents, and the conflict prediction model can output the frequency of short-term conflicts.

[0118] For mixed traffic weaving areas, the risk identification model constructed in the embodiments of this specification can be specifically abstracted as a model in which both input and output are sequences (sequence to sequence, seq2seq). The input sequence is a set of characteristic indicators of mixed traffic flow operation in the weaving area, and the output sequence is the risk level of the weaving area. Obviously, the difficulty in model construction is that the input and output sequences are of unequal length.

[0119] The Transformer model is a deep learning architecture for processing sequential data. Compared to traditional recurrent neural networks (RNNs) and long short-term memory (LSTM) networks, the Transformer model introduces a self-attention mechanism to handle dependencies at any position in the input sequence, thereby better capturing long-range dependencies and improving model performance. More importantly, the Transformer model consists of two parts: an encoder and a decoder. When addressing the problem of identifying safety risks in mixed traffic areas, this encoder-decoder structure effectively addresses the issue of unequal input and output sequence lengths, making it highly applicable.

[0120] Therefore, the embodiment of this specification is based on the Transformer architecture to construct an accident prediction model and a conflict prediction model for mixed traffic weaving areas. The data flow is: historical data cleaning → training set / test set division → cross-validation optimization → model accuracy verification.

[0121] The historical traffic flow data is pre-cleaned to obtain training sample data, which may specifically include:

[0122] The large amount of traffic flow data detected by the detector is aggregated, and the detector data within a certain time window is statistically analyzed to obtain the traffic parameters at each weaving area. A multi-dimensional traffic parameter sequence is constructed, and then the data detected by the detector is calibrated with accident and conflict labels.

[0123] The data of each section is processed to appropriately reduce abnormal fluctuations in the data and perform noise reduction on the overall data.

[0124] Furthermore, the processed data is divided into training set and test set.

[0125] Cross-validation optimization can include:

[0126] The training set is input into the Transformer accident prediction model and conflict prediction model for training, and the parameters are adjusted using the cross-validation method to establish the optimal model.

[0127] Model accuracy verification can include:

[0128] The test set data is input into the Transformer accident prediction model and conflict prediction model to obtain the predicted values ​​of the number of accidents and conflicts. The predicted values ​​are compared with the true values ​​to obtain the prediction or recognition accuracy of the model, which is used as a standard to test the recognition ability of the model.

[0129] Furthermore, after obtaining the trained dual-risk identification model, the dual-risk identification model may be tested and verified, which may specifically include:

[0130] The traffic flow data detected by the detector and the road environment conditions in the weaving area are input into the Transformer accident prediction model and the Transformer conflict prediction model to predict the number of accidents and conflicts within the observation time interval in the weaving area.

[0131] After obtaining the trained dual-risk identification model, the risk level judgment criteria for the weaving area can also be determined, and the relationship between the number of accidents and conflicts in the weaving area and the traffic risk identification in the weaving area can be further clarified.

[0132] As an application embodiment of this specification, for step S111, constructing a two-dimensional accident conflict risk matrix and using the two-dimensional accident conflict risk matrix as a risk level determination standard may specifically include:

[0133] The trained dual-risk identification model is used to predict the number of accidents and conflicts on the inspection section within a preset time.

[0134] performing parameter calibration on the number of accidents and the number of conflicts;

[0135] The Delphi method is used to determine the relationship between the number of accidents, the number of conflicts and the risk level, and to construct a two-dimensional risk matrix of accidents and conflicts.

[0136] In the embodiments of this specification, an accident prediction model and a conflict prediction model for mixed traffic weaving areas are constructed based on the Transformer architecture. When defining the risk level determination criteria for weaving areas, a semi-open expert opinion method, the Delphi method, is combined: after training the Transformer accident prediction model and the Transformer conflict prediction model using data such as historical traffic flow data, road environmental conditions, road accident occurrences, and the number of conflicts, the semi-open expert opinion method, the Delphi method, is used to define the risk level determination criteria for weaving areas, taking the number of accidents and conflicts as influencing factors, and constructing a two-dimensional risk matrix for accidents and conflicts.

[0137] As an application embodiment of this specification, for step S111, after obtaining the risk level and risk status identification results corresponding to the real-time traffic flow data, the method further includes:

[0138] If the risk level is high, the SHAP value method is used to calculate the contribution value corresponding to each risk factor;

[0139] According to the contribution value of each risk factor, targeted warnings are issued for major risk factors with larger contribution values.

[0140] In the embodiment of this specification, the risk factor may specifically be the traffic parameter related to accident calibration or conflict calibration selected in the embodiment of the above specification.

[0141] Different traffic risk scenarios involve different risk factors, and the degree of influence of each risk factor is also different. By using the SHAP value method to calculate the contribution value corresponding to each risk factor and quantifying the degree of influence of each risk factor, we can determine the risk factors that have a greater impact on the high-risk traffic scenarios that may occur. In this way, we can issue targeted warnings to vehicle drivers on the road and reduce the probability of traffic accidents.

[0142] Furthermore, the method may further include:

[0143] Storing real-time traffic flow data to regularly update the historical database corresponding to the multi-dimensional risk identification indicator system;

[0144] The parameters of the trained dual-risk identification model are updated according to the updated historical database.

[0145] In the embodiment of this specification, by storing real-time traffic flow data and regularly updating the historical database, the parameters of the dual risk identification model are further updated to achieve iterative updating of data and maintain high accuracy of risk identification by the model.

[0146] The embodiments of this specification provide a method for identifying mixed traffic risks in weaving areas. By constructing a dual risk identification model and a two-dimensional risk matrix for accident conflicts, the method can detect real-time traffic flow data and predict the risk level and risk status of mixed traffic in weaving areas, so as to provide accurate early warning of risk factors, reduce the probability of traffic accidents and conflicts, and specialize in mixed traffic scenarios in weaving areas to cover complex conflicts.

[0147] In addition, the risk identification dimension realizes multi-factor coupling, takes into account the impact of multiple factors such as people, vehicles, roads, and environment on risks, takes into account static road conditions, historical accident events, dynamic traffic conflicts and other issues, and proposes a scientific and reasonable comprehensive risk identification method to solve the technical problems of traditional single risk identification based on a single factor. It also solves the problem of the lack of mixed risk identification methods for manual driving and automatic driving in complex sections such as weaving areas, realizes the dual prediction of dynamic and static indicators combined with accident conflicts, forms a "monitoring-prediction-warning-update" closed loop, and supports real-time interaction among multiple terminals.

[0148] It should be noted that the above-mentioned specific method for identifying mixed traffic risks in road weaving areas is only a specific application embodiment and does not limit the scope of the embodiments of this specification. It can also include other specific embodiments, which will not be described one by one here.

[0149] Based on the same inventive concept, the embodiments of this specification also provide specific application embodiments of the above-mentioned method for identifying mixed traffic risks in weaving areas.

[0150] like Figure 2 The figure shows a logic diagram of constructing a multi-dimensional risk identification index system corresponding to a mixed traffic risk identification method in a road weaving area provided by an embodiment of this specification.

[0151] In the embodiments of this specification, a road condition index can be established based on road geometry constraints, a traffic flow state index can be established based on traffic flow disorder, an autonomous driving vehicle mixed traffic ratio index and a manual illegal driving behavior index can be established based on driving behavior differences, and an environmental factor index can be established based on environmental disturbances.

[0152] Specifically, the road condition index is used to quantify the fundamental constraints of the physical space of the weaving area on safety.

[0153] The collection method and safety impact logic of some secondary indicators corresponding to road condition indicators are as follows:

[0154] The length of the weaving area, specifically the distance from the merging end to the diverging end, can be collected through high-precision map extraction or on-site surveys. The safety impact logic is that insufficient weaving area length leads to compressed lane change space, which in turn leads to increased traffic conflicts.

[0155] The width of the emergency lane, specifically the width of the road shoulder (m), can be collected through high-precision maps or road design documents, drone aerial photography combined with image recognition, or on-site surveys and measurements. The safety impact logic is that if the emergency lane width is less than 3m, then the faulty vehicle is occupying the road, which in turn blocks main-line traffic.

[0156] The intersection angle of the entrance and exit markings, specifically the angle between the ramp and the main road marking (°), can be collected through HD maps or road design documents, drone aerial photography combined with image recognition, or on-site survey and measurement. The safety impact logic is that if the intersection angle of the entrance and exit markings is greater than 30°, the lateral speed of the incoming vehicle will suddenly change.

[0157] The length of the acceleration / deceleration lane can be specifically the length of the ramp acceleration / deceleration section (m). The collection method can be extraction from high-precision maps or road design documents, drone aerial photography combined with image recognition, or on-site survey and measurement. The safety impact logic is that the acceleration / deceleration lane length is insufficient → the vehicle fails to reach the synchronous speed → rear-end collision risk.

[0158] It should be noted that the above-mentioned method of collecting secondary indicators is a specific application embodiment, and there may be other collection methods, which are not specifically limited here.

[0159] Furthermore, traffic flow state indicators can be used to capture the dynamic disorder characteristics of traffic flow.

[0160] Among them, the secondary indicator main line speed difference at the inlet and outlet can be calculated using the following formula:

[0161] Δv=|v main-in -v main-out ∣ Formula (4)

[0162] In the formula, Δv is the main line speed difference at the inlet and outlet, v main-in is the main line speed at the inlet, v main-out is the main line speed at the exit.

[0163] The specific measurement method can use a radar-based integrated device to detect the average velocity of the main line entrance / exit section, which is not specifically limited here.

[0164] For example, it has been experimentally verified that when Δv>20km / h, the probability of traffic conflict increases by 40%.

[0165] Secondary indicators such as the proportion of large vehicles in incoming and outgoing traffic can be measured using AI video classification, such as the YOLOv7 vehicle recognition model, but are not specifically limited here. For example, experimental verification has shown that when the proportion of large vehicles exceeds 15%, the lane change conflict rate increases nonlinearly.

[0166] Furthermore, the secondary indicator corresponding to the AV mixed traffic ratio indicator, the AV mixed traffic rate, can quantify the degree of mixed traffic. Specifically, it can be measured in the following two ways:

[0167] Vehicle-Infrastructure Collaboration: Autonomous vehicles upload V2X status messages (e.g., BSM messages);

[0168] Visual assistance: Roadside cameras combined with an autonomous vehicle sign recognition model (e.g., ResNet50 model).

[0169] In specific application scenarios, other methods may also be used for measurement, which are not specifically limited here.

[0170] Among them, the risk mechanism of the mixing rate of autonomous driving vehicles is:

[0171] If the autonomous vehicle penetration rate is less than 10%, it indicates a low penetration rate of autonomous vehicles. In this case, manually driven vehicles are dominant and their behavior is highly random.

[0172] If the autonomous vehicle penetration rate is between 10% and 30%, it indicates low to medium penetration of autonomous vehicles. In this case, the interaction conflict between human-driven vehicles and autonomous vehicles is prominent.

[0173] If the blending rate of autonomous vehicles is greater than 30%, it indicates that autonomous vehicles have low penetration and high penetration. In this case, autonomous vehicles work together to reduce conflicts, but the risk of system failure increases.

[0174] Furthermore, the manual illegal driving behavior index can be used to quantify the uncertainty risk of drivers of manually driven vehicles.

[0175] Among them, the secondary indicator of overspeed can be detected by real-time detection technology using millimeter-wave radar speed measurement combined with license plate recognition;

[0176] The illegal lane change rate can capture the behavioral specificity of manually driven vehicles. Specifically, it can adopt a real-time detection technology that combines video trajectory tracking with Hough transform line marking detection.

[0177] In specific application scenarios, other methods may also be used for measurement, which are not specifically limited here.

[0178] For example, test verification shows that when the mixing rate of autonomous driving vehicles is greater than 25%, the contribution of illegal lane change rate to risk increases to 32%, while in the purely manually driven vehicle scenario it is only 18%.

[0179] The multi-dimensional risk identification index system constructed in the embodiment of this specification adopts the coordination of dynamic and static indicators, covering four-dimensional dynamic coupling indicators of people, vehicles, roads and environment, lightweight measurement, and edge computing compatibility. Most of the secondary indicators can be directly calculated in real time by the roadside edge computing nodes (delay <100ms).

[0180] Among them, static factors: road geometry conditions (length of weaving area, lane width, etc.) can be used for basic risk calibration.

[0181] Quasi-dynamic factors, environment (weather, illumination), can be used to monitor medium- and short-term risk fluctuations.

[0182] Dynamic factors include traffic flow status (saturation, speed difference, proportion of large vehicles), illegal manual driving behavior (speeding / illegal lane change rate), autonomous vehicle intrusion rate, and traffic conflict rate (TC / (vehicle·km)), which can be used to trigger real-time conflict warnings.

[0183] According to the above embodiments of the specification, a specific application embodiment is described below by taking a certain highway exit interweaving area as an example.

[0184] Measured values ​​of secondary indicators:

[0185] Road environment: The length of the weaving area is, and the ramp intersection angle is 25°;

[0186] Traffic flow status: mainline speed difference is 18km / h, and the proportion of large vehicles is 22%;

[0187] Mixed traffic and behavior: The rate of autonomous vehicles mixing in was 28%, and the illegal lane change rate was 15 times per hour;

[0188] Conflict and environment: The traffic conflict rate is 52 times / (1,000 vehicles / km), and the illumination at noon is 80,000 lux.

[0189] After risk contribution analysis, the three main risk factors in this application scenario are:

[0190] The mix of autonomous vehicles (31% risk contribution) may lead to behavioral mismatches in the mid-to-high penetration range;

[0191] Illegal lane change rate (risk contribution 28%), which may lead to aggressive lane changes by drivers of manually driven vehicles;

[0192] Large vehicle proportion (risk contribution 24%), which may lead to vehicle dynamic performance limitations.

[0193] In this case, the following two warning outputs can be performed:

[0194] The main line LED prompts or sends control strategy information to the vehicle: "There is a weaving area 500m ahead, there are many large vehicles, please pay attention to merging control for autonomous vehicles."

[0195] The navigation app pushes the message: "It has been detected that you have changed lanes three times across the solid line. Please obey traffic regulations. The next exit will be 2 km away."

[0196] In this way, by identifying mixed traffic risks in road weaving areas in advance and using appropriate methods to warn drivers and self-driving vehicles, traffic risks can be predicted in advance and the occurrence of traffic accidents and conflicts can be reduced.

[0197] like Figure 3 , which is a schematic diagram of a two-dimensional accident conflict risk matrix corresponding to a method for identifying mixed traffic risks in road weaving areas provided in an embodiment of this specification.

[0198] In the embodiment of this specification, the method for defining the comprehensive risk level threshold of the interweaving area is as follows:

[0199] Step 1: Preprocess the traffic flow data detected by the detector to determine parameters such as road geometry, traffic flow volume, autonomous vehicle penetration rate, and large vehicle penetration rate;

[0200] Step 2: Extract the number of traffic conflicts and the accident conditions in each detector section, and then assign the number of accidents and conflicts as labels to the corresponding data;

[0201] Step 3: After the parameter calibration is completed, use the Delphi method to clarify the relationship between the two and the risk level, and determine the boundaries of the risk level.

[0202] The calibration range for traffic accident counts is monthly, and the calibration range for conflict counts is hourly. For example, if a test location has 60 traffic conflicts within an hour and there has been one traffic accident on that road section within a month, the accident risk level for that test location is determined to be medium.

[0203] In this way, in actual application, after the road detector detects real-time traffic parameter data, the real-time data can be input into the grassroots server of the mixed traffic weaving area. The grassroots server uses the Transformer accident prediction model and the Transformer conflict prediction model to process the real-time data, predict the current number of traffic accidents and conflicts in the weaving area, and further use the risk level judgment standard given by the Delphi method to determine the road traffic risk level of the mixed traffic weaving area, and output the results in a visual way to facilitate real-time grasp of the road risk status.

[0204] like Figure 4 1 is a flow chart of road risk identification corresponding to a method for identifying mixed traffic risks in a weaving area provided by an embodiment of this specification.

[0205] In the embodiment of this specification, the specific steps of comprehensive risk level identification are:

[0206] Step 1: Extract historical traffic flow status data, driver behavior data, road environment conditions, autonomous vehicle penetration rate, traffic accident records, traffic conflict frequency records, and other data from the historical traffic flow database and process them to form a traffic flow risk characterization indicator system for mixed traffic areas.

[0207] Step 2: Aggregate all data from different weaving areas by time interval, and assign accident records and traffic conflict count labels to the data at each time interval;

[0208] Step 3: Use correlation analysis and SHAP value analysis to select parameters with high contribution to accidents and conflicts and low correlation between indicators from the traffic flow risk characterization index system;

[0209] Step 4: Input the filtered parameters into the Transformer trainer and perform Transformer training on the data to obtain the Transformer accident prediction model and Transformer conflict prediction model that can estimate the number of accidents and conflicts in the weaving area based on traffic parameters;

[0210] Step 5: Using the Delphi method, with the number of accidents and conflicts as characteristic parameters, define the risk level determination criteria for weaving zones. The Transformer accident prediction model, Transformer conflict prediction model, and risk level determination criteria are distributed to each grassroots server as the criteria for identifying accident risk status in mixed traffic weaving zones.

[0211] Step 6: After the detector detects real-time traffic flow data within the jurisdiction of the grassroots server in the weaving area, it transmits the real-time data to the grassroots server. The grassroots server first uses the Transformer accident prediction model and the Transformer conflict prediction model to determine the number of accidents and conflicts corresponding to the traffic flow data. Then, the grassroots server identifies the traffic accident risk status of the traffic flow data according to the risk level judgment criteria. The risk level corresponding to the traffic flow data and the traffic accident risk status identification results are then visualized and output to guide engineering practice.

[0212] Step 7: Store the real-time data. When a certain time or data volume reaches a certain scale, update the historical database and update the Transformer accident prediction model and Transformer conflict prediction model.

[0213] Furthermore, based on the risk assessment results, a comprehensive risk identification method can be used to determine if traffic in the weaving zone is at high risk. Based on the contribution of these risk factors (the SHAP value method can be used to calculate the contribution), early warnings can be issued for the major risk factors with the highest contribution. These warnings can be implemented through channels such as LED information boards along the route, map navigation prompts, and autonomous vehicle systems, though these are not specifically limited here. In practical application scenarios, typical early warning scenarios can be shown in Table 1 below.

[0214] Table 1 Typical scenarios for risk warning

[0215] Serial number Risk warning scenarios 1 Emergency traffic incident risk warning 2 Road hazard warning 3 Main line intersection saturation risk warning 4 Import / exit ramp high traffic risk warning 5 Warning of excessive risk of speed difference between mainline vehicles 6 Risk warning for excessively high proportion of large vehicles in weaving areas 7 Risk warning for excessively high proportion of large vehicles in / out 8 Traffic violation risk warning for manually driven vehicles 9 Traffic conflict risk warning 10 High-impact severe weather warning 11 Low environmental visibility risk warning

[0216] The embodiment of this specification provides a mixed traffic risk identification method for road weaving areas, which can provide traffic operation risk status assessment results and risk warning scenarios for traffic safety management departments such as public security traffic management, and support risk warning, risk control and other work.

[0217] Targeted governance can also be carried out for high-risk intersection sections in the road network, combined with risk indicators, to address problems such as the lack of targeted traffic safety governance in the past and unclear risk points and sections under key control.

[0218] In addition, it takes into account the impact of multiple factors such as people, vehicles, roads, and the environment on risks, as well as issues such as static road conditions, historical accidents, and dynamic traffic conflicts, and proposes a scientific and reasonable comprehensive risk identification method to solve the technical problems of traditional single risk identification based on a single factor. It also solves the problem of the lack of risk identification methods for mixed manual and automatic driving in complex sections such as interweaving areas.

[0219] The specific implementation process of the embodiments of this specification can refer to the corresponding implementation steps of the above embodiments, which will not be repeated here.

[0220] Based on the same inventive concept, the embodiment of this specification also provides a mixed traffic risk identification system for road weaving areas. Figure 5 The figure shows a schematic diagram of the structure of a mixed traffic risk identification system for road weaving areas provided by an embodiment of this specification.

[0221] The mixed traffic risk identification system in the weaving area may specifically include:

[0222] Quantification module 501 quantifies the contribution of each indicator in the constructed multi-dimensional risk identification indicator system to traffic accidents and traffic conflicts;

[0223] A sorting module 502 sorts the indicators according to the contribution values ​​to obtain a traffic accident contribution value sorting result and a traffic conflict contribution value sorting result;

[0224] A merging module 503 selects indicators whose contribution values ​​to traffic accidents and traffic conflicts exceed preset thresholds based on the traffic accident contribution value ranking results and the traffic conflict contribution value ranking results, and merges them into a high contribution index system;

[0225] A screening module 504 is configured to screen indicators with correlations between indicators less than a threshold value from the high contribution indicator system;

[0226] A training module 505 is configured to train the dual-risk identification model based on a screened high-contribution indicator system, wherein the correlation between the indicators in the screened high-contribution indicator system is less than a threshold value;

[0227] Construction module 506 constructs a two-dimensional accident conflict risk matrix, and uses the trained dual-risk identification model and the two-dimensional accident conflict risk matrix to detect real-time traffic flow data, obtain risk level and risk status identification results corresponding to the real-time traffic flow data, and use the two-dimensional accident conflict risk matrix as a risk level determination standard.

[0228] based on Figure 5The present specification also provides some specific implementation plans of the system, which are described below.

[0229] Furthermore, a multi-dimensional risk identification indicator system is constructed, including:

[0230] Establishing primary indicators including road condition indicators, traffic flow status indicators, mixed traffic ratio indicators for autonomous vehicles, indicators for illegal human driving behavior, traffic conflict indicators, and environmental factor indicators;

[0231] On the basis of each of the first-level indicators, establish the second-level indicators corresponding to each of the first-level indicators;

[0232] Based on the first-level indicators and the second-level indicators, the multi-dimensional risk identification indicator system is constructed.

[0233] Furthermore, for each indicator in the constructed multi-dimensional risk identification indicator system, the contribution value of each indicator to traffic accidents and traffic conflicts is quantified, including:

[0234] SHAP value analysis was performed on each of the aforementioned indicators;

[0235] The contribution of the indicators to traffic accidents and traffic conflicts is quantified based on the SHAP value analysis results.

[0236] Furthermore, the indicators are sorted according to the contribution values ​​to obtain traffic accident contribution value sorting results and traffic conflict contribution value sorting results, including:

[0237] Sorting the indicators according to their contribution to traffic accidents to obtain a ranking result of their contribution to traffic accidents;

[0238] According to the contribution value of each indicator to the traffic conflict, the indicators are sorted to obtain the traffic conflict contribution value sorting result.

[0239] Furthermore, the calculation formula of the SHAP value is as follows:

[0240]

[0241] Among them, Φ j is the SHAP value of the jth traffic parameter, N is the set of all parameters in the training set, its dimension is M, S is the subset extracted from N, its dimension is |S|, f x (S) is the sample mean calculated using the parameter set S, f x (S∪{i}) is the sample average calculated again based on the parameter set S plus the parameter i, is the weight item.

[0242] Furthermore, using the traffic accident contribution value ranking results and the traffic conflict contribution value ranking results, indicators whose contribution values ​​to traffic accidents and traffic conflicts exceed preset thresholds are selected respectively, and combined to form a high contribution index system, including:

[0243] Determining, based on the traffic accident contribution value ranking result, a first indicator set whose contribution value to the traffic accident exceeds a preset threshold;

[0244] Determining, based on the traffic conflict contribution value ranking result, a second indicator set whose contribution value to the traffic conflict exceeds a preset threshold;

[0245] The first indicator set and the second indicator set are combined to form the high contribution indicator system.

[0246] Furthermore, from the high contribution index system, the indexes whose correlations between the indexes are less than a threshold are screened, including:

[0247] Calculating the correlation between the various indicators in the high contribution index system;

[0248] If the correlation between the two indicators exceeds the threshold, one indicator is selected from the two indicators and the other indicator is screened out, so that in the high contribution indicator system after screening, the correlation between each indicator is less than the threshold.

[0249] Furthermore, based on the screened high-contribution indicator system, the dual-risk identification model is trained, including:

[0250] Constructing a dual-risk identification model, wherein the dual-risk identification model includes an accident prediction model and a conflict prediction model;

[0251] Collect historical traffic flow data based on the screened high-contribution indicator system;

[0252] performing a pre-cleaning operation on the historical traffic flow data according to the traffic accident contribution value ranking results and the traffic conflict contribution value ranking results to obtain training sample data;

[0253] Dividing the training sample data into a training set and a test set;

[0254] Using the training set to train the dual-risk identification model, and using the test set to test the dual-risk identification model;

[0255] Cross-validation is performed based on the training results and the test results, and the parameters of the dual-risk identification model are optimized to obtain a trained dual-risk identification model.

[0256] Furthermore, a two-dimensional accident conflict risk matrix is ​​constructed and used as a risk level determination standard, including:

[0257] The trained dual-risk identification model is used to predict the number of accidents and conflicts on the inspection section within a preset time.

[0258] performing parameter calibration on the number of accidents and the number of conflicts;

[0259] The Delphi method is used to determine the relationship between the number of accidents, the number of conflicts and the risk level, and to construct a two-dimensional risk matrix of accidents and conflicts.

[0260] Furthermore, after obtaining the risk level and risk status identification results corresponding to the real-time traffic flow data, the system further includes:

[0261] If the risk level is high, the SHAP value method is used to calculate the contribution value corresponding to each risk factor;

[0262] According to the contribution value of each risk factor, targeted warnings are issued for major risk factors with larger contribution values.

[0263] The embodiments of this specification provide a mixed traffic risk identification system for weaving areas. By constructing a dual risk identification model and a two-dimensional risk matrix for accident conflicts, the system can detect real-time traffic flow data and predict the risk level and risk status of mixed traffic in weaving areas, so as to provide accurate early warning of risk factors, reduce the probability of traffic accidents and conflicts, specialize in mixed traffic scenarios in weaving areas, and cover complex conflicts.

[0264] In addition, the risk identification dimension realizes multi-factor coupling, takes into account the impact of multiple factors such as people, vehicles, roads, and environment on risks, takes into account static road conditions, historical accident events, dynamic traffic conflicts and other issues, and proposes a scientific and reasonable comprehensive risk identification method to solve the technical problems of traditional single risk identification based on a single factor. It also solves the problem of the lack of mixed risk identification methods for manual driving and automatic driving in complex sections such as weaving areas, realizes the dual prediction of dynamic and static indicators combined with accident conflicts, forms a "monitoring-prediction-warning-update" closed loop, and supports real-time interaction among multiple terminals.

[0265] Based on the same inventive concept, an embodiment of this specification further provides an electronic device, including at least one processor and a memory, wherein the memory stores a program and is configured to execute the following steps by the at least one processor:

[0266] For each indicator in the constructed multi-dimensional risk identification indicator system, quantify the contribution value of each indicator to traffic accidents and traffic conflicts;

[0267] According to the size of the contribution value, the indicators are sorted respectively to obtain the traffic accident contribution value sorting result and the traffic conflict contribution value sorting result;

[0268] Using the traffic accident contribution value ranking results and the traffic conflict contribution value ranking results, respectively select indicators whose contribution values ​​to traffic accidents and traffic conflicts exceed preset thresholds, and combine them to form a high contribution index system;

[0269] From the high contribution index system, screening indicators whose correlations among the indicators are less than a threshold;

[0270] Training the dual-risk identification model based on the screened high-contribution indicator system, wherein the correlation between the indicators in the screened high-contribution indicator system is less than a threshold;

[0271] A two-dimensional accident conflict risk matrix is ​​constructed to detect real-time traffic flow data using the trained dual-risk identification model and the two-dimensional accident conflict risk matrix, to obtain risk level and risk status identification results corresponding to the real-time traffic flow data, and to use the two-dimensional accident conflict risk matrix as a risk level determination criterion.

[0272] Among them, other functions of the processor can also refer to the contents recorded in the above embodiments, which will not be repeated here.

[0273] Based on the same inventive concept, an embodiment of this specification further provides a computer-readable storage medium, including a program for use in conjunction with an electronic device, which can be executed by a processor to perform the following steps:

[0274] For each indicator in the constructed multi-dimensional risk identification indicator system, quantify the contribution value of each indicator to traffic accidents and traffic conflicts;

[0275] According to the size of the contribution value, the indicators are sorted respectively to obtain the traffic accident contribution value sorting result and the traffic conflict contribution value sorting result;

[0276] Using the traffic accident contribution value ranking results and the traffic conflict contribution value ranking results, respectively select indicators whose contribution values ​​to traffic accidents and traffic conflicts exceed preset thresholds, and combine them to form a high contribution index system;

[0277] From the high contribution index system, screening indicators whose correlations among the indicators are less than a threshold;

[0278] Training the dual-risk identification model based on the screened high-contribution indicator system, wherein the correlation between the indicators in the screened high-contribution indicator system is less than a threshold;

[0279] A two-dimensional accident conflict risk matrix is ​​constructed to detect real-time traffic flow data using the trained dual-risk identification model and the two-dimensional accident conflict risk matrix, to obtain risk level and risk status identification results corresponding to the real-time traffic flow data, and to use the two-dimensional accident conflict risk matrix as a risk level determination criterion.

[0280] Among them, other functions of the processor can also refer to the contents recorded in the above embodiments, which will not be repeated here.

[0281] like Figure 6 As shown, the embodiment of this specification also provides a structural schematic diagram of a computer storage medium.

[0282] The systems, devices, modules, or units described in the above embodiments may be implemented by computer chips or entities, or by products having certain functions. A typical implementation device is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smartphone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0283] It will be understood by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0284] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0285] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0286] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0287] In a typical configuration, a computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0288] Memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

[0289] Computer-readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. The information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic tape storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include temporary computer-readable media (transitory media), such as modulated data signals and carrier waves.

[0290] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

[0291] The present application may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communications network. In a distributed computing environment, program modules may be located in local and remote computer storage media, including storage devices.

[0292] The various embodiments in this specification are described in a progressive manner. Similar parts between the various embodiments can be referred to in conjunction with each other. Each embodiment focuses on the differences between the other embodiments. In particular, the system embodiments are generally similar to the method embodiments, so the description is relatively simple. For relevant parts, refer to the description of the method embodiments.

[0293] The foregoing is merely an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should all be included within the scope of the claims of the present application.

Claims

1. A method for identifying mixed traffic risk in road weaving areas, characterized in that: The method for identifying mixed traffic risk in a weaving area includes: For each indicator in the constructed multi-dimensional risk identification indicator system, quantify the contribution value of each indicator to traffic accidents and traffic conflicts; According to the size of the contribution value, the indicators are sorted respectively to obtain the traffic accident contribution value sorting result and the traffic conflict contribution value sorting result; Using the traffic accident contribution value ranking results and the traffic conflict contribution value ranking results, respectively select indicators whose contribution values ​​to traffic accidents and traffic conflicts exceed preset thresholds, and combine them to form a high contribution index system; From the high contribution index system, screening indicators whose correlations among the indicators are less than a threshold; Training the dual-risk identification model based on the screened high-contribution indicator system, wherein the correlation between the indicators in the screened high-contribution indicator system is less than a threshold; A two-dimensional accident conflict risk matrix is ​​constructed to detect real-time traffic flow data using the trained dual-risk identification model and the two-dimensional accident conflict risk matrix, to obtain risk level and risk status identification results corresponding to the real-time traffic flow data, and to use the two-dimensional accident conflict risk matrix as a risk level determination criterion.

2. The method according to claim 1, wherein Construct a multi-dimensional risk identification indicator system, including: Establishing primary indicators including road condition indicators, traffic flow status indicators, mixed traffic ratio indicators for autonomous vehicles, indicators for illegal human driving behavior, traffic conflict indicators, and environmental factor indicators; On the basis of each of the first-level indicators, establish the second-level indicators corresponding to each of the first-level indicators; Based on the first-level indicators and the second-level indicators, the multi-dimensional risk identification indicator system is constructed.

3. The method according to claim 2, wherein For each indicator in the constructed multi-dimensional risk identification indicator system, the contribution value of each indicator to traffic accidents and traffic conflicts is quantified, including: SHAP value analysis was performed on each of the aforementioned indicators; The contribution of the indicators to traffic accidents and traffic conflicts is quantified based on the SHAP value analysis results.

4. The method according to claim 2, wherein According to the size of the contribution value, the indicators are sorted respectively to obtain the traffic accident contribution value sorting result and the traffic conflict contribution value sorting result, including: Sorting the indicators according to their contribution to traffic accidents to obtain a ranking result of their contribution to traffic accidents; According to the contribution value of each indicator to the traffic conflict, the indicators are sorted to obtain the traffic conflict contribution value sorting result.

5. The method according to claim 3, wherein The calculation formula of the SHAP value is as follows: Among them, Φ j is the SHAP value of the jth traffic parameter, N is the set of all parameters in the training set, its dimension is M, S is the subset extracted from N, its dimension is |S|, f x (S) is the sample mean calculated using the parameter set S, f x (S∪{i}) is the sample average calculated again based on the parameter set S plus the parameter i, is the weight item.

6. The method according to claim 1, wherein Using the traffic accident contribution value ranking results and traffic conflict contribution value ranking results, indicators whose contribution values ​​to traffic accidents and traffic conflicts exceed preset thresholds are selected respectively, and combined to form a high contribution index system, including: Determining, based on the traffic accident contribution value ranking result, a first indicator set whose contribution value to the traffic accident exceeds a preset threshold; Determining, based on the traffic conflict contribution value ranking result, a second indicator set whose contribution value to the traffic conflict exceeds a preset threshold; The first indicator set and the second indicator set are combined to form the high contribution indicator system.

7. The method according to claim 6, wherein From the high contribution index system, the indexes whose correlations between the indexes are less than the threshold are screened, including: Calculating the correlation between the various indicators in the high contribution index system; If the correlation between the two indicators exceeds the threshold, one indicator is selected from the two indicators and the other indicator is screened out, so that in the high contribution indicator system after screening, the correlation between each indicator is less than the threshold.

8. The method according to claim 1, wherein Based on the screened high-contribution indicator system, the dual-risk identification model is trained, including: Constructing a dual-risk identification model, wherein the dual-risk identification model includes an accident prediction model and a conflict prediction model; Collect historical traffic flow data based on the screened high-contribution indicator system; performing a pre-cleaning operation on the historical traffic flow data according to the traffic accident contribution value ranking results and the traffic conflict contribution value ranking results to obtain training sample data; Dividing the training sample data into a training set and a test set; Using the training set to train the dual-risk identification model, and using the test set to test the dual-risk identification model; Cross-validation is performed based on the training results and the test results, and the parameters of the dual-risk identification model are optimized to obtain a trained dual-risk identification model.

9. The method according to claim 1, wherein Constructing a two-dimensional accident conflict risk matrix, and using the two-dimensional accident conflict risk matrix as a risk level determination standard, including: The trained dual-risk identification model is used to predict the number of accidents and conflicts on the inspection section within a preset time. performing parameter calibration on the number of accidents and the number of conflicts; The Delphi method is used to determine the relationship between the number of accidents, the number of conflicts and the risk level, and to construct a two-dimensional risk matrix of accidents and conflicts.

10. The method according to claim 1, wherein After obtaining the risk level and risk status identification results corresponding to the real-time traffic flow data, the method further includes: If the risk level is high, the SHAP value method is used to calculate the contribution value corresponding to each risk factor; According to the contribution value of each risk factor, targeted warnings are issued for major risk factors with larger contribution values.

11. A mixed traffic risk identification system in a road weaving area, characterized in that: The mixed traffic risk identification system in the road weaving area includes: A quantification module quantifies the contribution of each indicator in the constructed multi-dimensional risk identification indicator system to traffic accidents and traffic conflicts; A sorting module, which sorts the indicators according to the contribution values ​​to obtain traffic accident contribution value sorting results and traffic conflict contribution value sorting results; a merging module, which uses the traffic accident contribution value ranking results and the traffic conflict contribution value ranking results to select indicators whose contribution values ​​to traffic accidents and traffic conflicts exceed preset thresholds, respectively, and merges them into a high contribution indicator system; A screening module, screening indicators whose correlation between indicators is less than a threshold from the high contribution indicator system; a training module for training the dual-risk identification model based on a screened high-contribution indicator system, wherein the correlation between the indicators in the screened high-contribution indicator system is less than a threshold; A construction module is provided to construct a two-dimensional risk matrix for accident conflicts, so as to detect real-time traffic flow data using the trained dual-risk identification model and the two-dimensional risk matrix for accident conflicts, obtain risk level and risk status identification results corresponding to the real-time traffic flow data, and use the two-dimensional risk matrix for accident conflicts as a risk level determination standard.

12. A computer storage medium comprising a program for use in conjunction with an electronic device, the program being executable by a processor to perform the following steps: For each indicator in the constructed multi-dimensional risk identification indicator system, quantify the contribution value of each indicator to traffic accidents and traffic conflicts; According to the size of the contribution value, the indicators are sorted respectively to obtain the traffic accident contribution value sorting result and the traffic conflict contribution value sorting result; Using the traffic accident contribution value ranking results and the traffic conflict contribution value ranking results, respectively select indicators whose contribution values ​​to traffic accidents and traffic conflicts exceed preset thresholds, and combine them to form a high contribution index system; From the high contribution index system, screening indicators whose correlations among the indicators are less than a threshold; Training the dual-risk identification model based on the screened high-contribution indicator system, wherein the correlation between the indicators in the screened high-contribution indicator system is less than a threshold; A two-dimensional accident conflict risk matrix is ​​constructed to detect real-time traffic flow data using the trained dual-risk identification model and the two-dimensional accident conflict risk matrix, to obtain risk level and risk status identification results corresponding to the real-time traffic flow data, and to use the two-dimensional accident conflict risk matrix as a risk level determination criterion.

Images