Method and system for road weaving area traffic conflict identification

Abstract

The application discloses a road interlaced area traffic conflict recognition method and system, and particularly relates to the field of traffic management, and is used for solving the problems of low traffic conflict recognition precision and early warning delay in the interlaced area, accurately capturing risk events based on a composite trigger mechanism of a turn signal activation and a transverse displacement rate, and filtering non-intention behavior interference; adaptive interception of an asymmetric time window is realized by matching a behavior characteristic vector with a conflict type template library, and a risk evolution cycle of scenes such as impatient entry and slow lane change is completely covered; a dynamic window expansion mechanism driven by a conflict energy gradient is introduced, and the whole process of risk accumulation is ensured to be captured; a heat map is generated by combining a convolution network to extract multi-scale space-time features, and fine positioning of the risk space-time distribution is realized; a warning signal grading mechanism and a closed-loop feedback optimization system are established, the system adaptability is continuously improved, real-time and accurate decision support is provided for an intelligent traffic management system, secondary accident risks are effectively curbed, and the road network passing efficiency is optimized.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of traffic management, more particularly, the present application relates to a traffic conflict identification method and system for road weaving area. BACKGROUND

[0002] In modern traffic systems, road weaving areas, as complex regions of traffic flow merging, splitting and lane changing, frequently occur interactive behaviors between vehicles, which easily lead to traffic conflict events. Especially in the scenes of highway ramps, roundabout entrances and exits, and urban intersections, vehicles need to complete lane changing, merging or splitting operations in a short time, drivers face high-frequency decision-making and coordination pressure, traffic flow changes dynamically, and the spatiotemporal distribution of risk signals presents asymmetry and uncertainty. Traditional traffic safety monitoring technologies mostly rely on fixed sensors or manual patrols, which are difficult to capture the micro-interaction behaviors between vehicles in real time, and the dynamic characteristics of the risk evolution process are lagging behind, which limits the effectiveness and timeliness of traffic safety management.

[0003] However, the existing traffic conflict identification method for road weaving area generally uses fixed time window or static threshold for data interception and analysis, ignoring the dynamic characteristics of conflict event evolution chain and the forward long tail feature, resulting in the spatiotemporal distribution of risk signals being treated one-sidedly. The traditional method relies on a single trigger point (such as the collision prediction time) to symmetrically intercept data when analyzing the event, which fails to cover the implicit behaviors (such as speed game and lane position offset) in the premonitory stage of conflict events, limiting the accuracy and comprehensiveness of risk assessment. In addition, the existing technology mostly relies on single-scale trajectory data in the feature extraction link, which is difficult to represent both micro-trajectory fluctuations and macro-interaction trends, affecting the precision of conflict identification and the timeliness of early warning. The parameter setting of the existing system is mostly static, lacking adaptive optimization mechanism, which is difficult to cope with the dynamic changes of traffic scenes, further restricting the practicality and reliability of the technology. In order to solve the above problems, a technical solution is provided. SUMMARY

[0004] In order to overcome the above-mentioned defects of the prior art, embodiments of the present application provide a road weaving area traffic conflict identification method and system, which accurately captures risk events based on a composite triggering mechanism of turn signal activation and lateral displacement rate, filters non-intentional behavior interference; adopts a behavior feature vector to match a conflict type template library to realize adaptive interception of an asymmetric time window, completely covers the risk evolution period of scenes such as abrupt entry and slow lane change; introduces a dynamic window expansion mechanism driven by a conflict energy gradient to ensure the whole process capture in the risk accumulation stage; combines a convolution network to extract multi-scale spatiotemporal features to generate a heat map, realizes fine positioning of the spatiotemporal distribution of risks; establishes a warning signal grading mechanism and a closed-loop feedback optimization system, continuously improves the adaptability of the system, provides real-time and accurate decision support for intelligent traffic management systems, effectively suppresses the risk of secondary accidents and optimizes the efficiency of the road network, so as to solve the problems raised in the above background art.

[0005] To achieve the above object, the present application provides the following technical scheme:

[0006] The road weaving area traffic conflict identification method comprises the following steps:

[0007] S1: The turn signal state and lane deviation trajectory are monitored by using a vehicle-mounted camera, the lateral displacement rate is calculated and compared with a dynamic threshold value, and when the turn signal is activated and the lateral displacement rate exceeds the threshold value, a conflict event analysis is triggered;

[0008] S2: A vehicle feature vector is constructed, and a corresponding asymmetric time window template is determined by similarity matching with a preset conflict type template library;

[0009] S3: The time window is optimized based on the conflict energy gradient, if the peak value of the conflict energy gradient in the window does not reach a preset threshold value, the window is gradually expanded until the peak value reaches the threshold value or a termination condition is met, and the time window containing the complete risk evolution chain is determined;

[0010] S4: The vehicle micro trajectory data is collected in the optimized time window, the spatiotemporal features are extracted by using a pyramid convolution network, the spatiotemporal conflict heat map is generated by fusion, and the high-risk spatiotemporal coordinates are marked;

[0011] S5: The hierarchical warning signal is generated according to the peak intensity of the spatiotemporal conflict heat map and the distribution density of the high-risk spatiotemporal coordinates, and the asymmetric time window template parameters in the conflict type template library are dynamically updated through the warning accuracy rate feedback.

[0012] In a preferred embodiment, step S1 comprises the following contents:

[0013] The turning light state and lane deviation trajectory of the vehicle are monitored in real time by the vehicle-mounted camera, the lateral displacement rate of the vehicle is calculated, and a dynamic threshold is generated according to the lane width, the speed of the vehicle and the environmental complexity factor. When the turning light state is activated and the lateral displacement rate exceeds the dynamic threshold, a conflict event analysis process is triggered.

[0014] In a preferred embodiment, step S1 further comprises the following:

[0015] The flashing condition of the turning light is captured by the vehicle-mounted camera to determine whether the turning light state is activated or not activated, and the lateral distance data sequence of the vehicle relative to the lane center line is obtained through image processing technology, and then the lateral displacement rate is calculated. The dynamic threshold is obtained by dividing the lane width by the reference lane changing time, multiplying the ratio of the speed of the vehicle to the reference speed, and multiplying the environmental complexity factor. The trigger condition is that the turning light state is activated and the lateral displacement rate is greater than the dynamic threshold. After triggering, the trigger time point is recorded.

[0016] In a preferred embodiment, step S2 comprises the following:

[0017] A feature vector of the vehicle is constructed, wherein the feature vector includes the turning light activation duration, the lateral acceleration change rate and the lane deviation angle, wherein the turning light activation duration is defined as the time difference from the turning light activation starting time to the trigger time; the lateral acceleration change rate is defined as the difference between the lateral acceleration at the trigger time and the previous frame lateral acceleration divided by the inter-frame time difference; and the lane deviation angle is defined as the inverse tangent function value of the ratio of the lateral displacement rate to the longitudinal speed at the trigger time.

[0018] In a preferred embodiment, step S2 further comprises the following:

[0019] After the construction is completed, the feature vector is matched with a preset conflict type template library, the conflict type template library contains three types of sudden entry, slow lane change and continuous lane change, each type corresponds to a feature vector center value; the matching process determines the similarity by calculating the inverse function form of the Euclidean distance between the feature vector and the feature vector center value of each type, selects the conflict type with the highest matching degree according to the similarity value, and determines the corresponding asymmetric time window template.

[0020] In a preferred embodiment, step S3 comprises the following:

[0021] In the determined asymmetric time window template, the peak value of the conflict energy gradient time sequence is calculated. If the peak value does not reach the preset threshold, the window boundary is expanded step by step, i.e. each time the window boundary is expanded by a fixed time length, until the peak value of the conflict energy gradient in the window reaches the preset threshold or the total time length of the window reaches the preset maximum allowed time length, and finally the optimized time window is determined.

[0022] In a preferred embodiment, step S3 further comprises the following:

[0023] wherein the conflict energy gradient is defined as the difference of the conflict risk indicator between adjacent time points divided by the inter-frame time difference, and the conflict risk indicator is calculated based on the relative distance, relative speed and relative acceleration between the vehicle and the adjacent vehicle; the window expansion is preferentially forward to cover the earlier risk accumulation stage, and the expanded window is used to recalculate the peak value of the conflict energy gradient until the termination condition is met.

[0024] In a preferred embodiment, step S4 comprises the following:

[0025] Collecting the micro-trajectory data of the vehicle in the optimized time window, extracting the spatio-temporal features of the micro-trajectory data using a pyramid convolution network, capturing short-time, medium-time and long-time spatio-temporal features through multi-scale convolution layers respectively, and generating a comprehensive spatio-temporal feature map after aggregation and fusion through a pooling layer; mapping the comprehensive spatio-temporal feature map to a risk intensity value through a fully connected layer, and generating a spatio-temporal conflict heat map through three-dimensional linear interpolation and normalization processing; in the spatio-temporal conflict heat map, marking the spatio-temporal points with risk intensity values exceeding a preset threshold as high-risk spatio-temporal coordinates.

[0026] In a preferred embodiment, step S5 comprises the following:

[0027] Obtaining the spatio-temporal conflict heat map and the high-risk spatio-temporal coordinates, calculating the peak intensity of the spatio-temporal conflict heat map and the distribution density of the high-risk spatio-temporal coordinates, generating a hierarchical warning signal according to the peak intensity and the distribution density, recording the hierarchical warning signal and its corresponding conflict type to a conflict record database, regularly calculating the warning accuracy rate of each conflict type, and adjusting the forward window length or the backward window length for the conflict type with a warning accuracy rate lower than a preset threshold, the adjustment amplitude being a preset fixed value or dynamically determined according to the difference between the warning accuracy rate and the preset accuracy rate threshold, and updating the adjusted asymmetric time window template parameters to the conflict type template library.

[0028] The system for traffic conflict identification in road weaving areas comprises a trigger analysis module, a window matching module, a window optimization module, a feature extraction module and a warning feedback module.

[0029] The trigger analysis module: uses a vehicle-mounted camera to monitor the state of the turn signal and the lane deviation trajectory, calculates the lateral displacement rate and compares it with a dynamic threshold, and triggers conflict event analysis when the turn signal is activated and the lateral displacement rate exceeds the threshold.

[0030] The window matching module: constructs a vehicle feature vector, and determines the corresponding asymmetric time window template through similarity matching with a preset conflict type template library.

[0031] Window optimization module: optimize the time window based on the conflict energy gradient, if the peak value of the conflict energy gradient in the window does not reach the preset threshold, gradually expand the window until the peak value reaches the threshold or meets the termination condition, determine the time window containing the complete risk evolution chain;

[0032] Feature extraction module: collect vehicle micro trajectory data in the optimized time window, extract spatio-temporal features using pyramid convolution network, fuse to generate spatio-temporal conflict heat map and label high-risk spatio-temporal coordinates

[0033] Early warning feedback module: generate hierarchical early warning signals according to the peak intensity of spatio-temporal conflict heat map and the distribution density of high-risk spatio-temporal coordinates, and dynamically update the asymmetric time window template parameters in the conflict type template library through early warning accuracy feedback.

[0034] The technical effects and advantages of the road weaving area traffic conflict identification method and system of the present application are:

[0035] The present application significantly improves the road weaving area traffic conflict identification efficiency through multi-dimensional technology cooperation: based on the composite trigger mechanism of turn signal activation and lateral displacement rate, the risk event is accurately captured, and the non-intention behavior interference is effectively filtered; the adaptive interception of asymmetric time window is realized by matching the behavior feature vector with the conflict type template library, which completely covers the risk evolution period of scenes such as impatient entry and slow lane change; the dynamic window expansion mechanism driven by conflict energy gradient is introduced to ensure the whole process capture in the risk accumulation stage; combined with pyramid convolution network to extract multi-scale spatio-temporal features to generate high-precision heat map, the fine positioning of risk spatio-temporal distribution is realized; the hierarchical warning signal mechanism and closed-loop feedback optimization system are established, the window template parameters are dynamically calibrated through historical data, and the system adaptability is continuously improved. Provide real-time and accurate decision support for intelligent traffic management system, effectively curb the risk of secondary accidents and optimize road network traffic efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0036] Figure 1 The flowchart of the road weaving area traffic conflict identification method of the present application is shown.

[0037] Figure 2 The structure diagram of the road weaving area traffic conflict identification system of the present application is shown. DETAILED DESCRIPTION

[0038] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0039] Embodiment 1: Figure 1 The application provides a traffic conflict recognition method for a road weaving area, which comprises the following steps:

[0040] S1: A vehicle-mounted camera is used to monitor the state of a turn signal and a lane deviation trajectory, a lateral displacement rate is calculated and compared with a dynamic threshold, and when the turn signal is activated and the lateral displacement rate exceeds the threshold, a conflict event analysis is triggered.

[0041] S2: A vehicle feature vector is constructed, and a corresponding asymmetric time window template is determined by similarity matching with a preset conflict type template library.

[0042] S3: The time window is optimized based on the conflict energy gradient, if the peak value of the conflict energy gradient in the window does not reach a preset threshold, the window is gradually expanded until the peak value reaches the threshold or a termination condition is met, and a time window containing a complete risk evolution chain is determined.

[0043] S4: In the optimized time window, vehicle micro trajectory data is collected, spatiotemporal features are extracted by using a pyramid convolution network, a spatiotemporal conflict heat map is generated by fusion, and high-risk spatiotemporal coordinates are marked.

[0044] S5: A hierarchical early warning signal is generated according to the peak intensity of the spatiotemporal conflict heat map and the distribution density of the high-risk spatiotemporal coordinates, and the asymmetric time window template parameters in the conflict type template library are dynamically updated through the early warning accuracy feedback.

[0045] In the modern urban traffic system, road weaving areas (such as highway ramps, island entrances and exits, and urban road intersections) are key areas where traffic accidents frequently occur due to their complex traffic flow directions and dynamic vehicle interaction behaviors. In these areas, vehicles need to frequently perform lane changing, merging or splitting operations, and drivers must make decisions and coordinate with other vehicles within a short time. Due to the intensity of traffic flow and the unpredictability of vehicle behavior, traditional traffic safety analysis methods have significant limitations in capturing the dynamic evolution process of conflict risks. In particular, traditional methods rely on fixed time windows or static thresholds, which are difficult to adapt to the implicit behaviors (such as speed adjustment or lane position deviation) in the pre-conflict phase, resulting in insufficient spatiotemporal accuracy of risk warning and easy lag or misjudgment. The application proposes a dynamic analysis method based on visual recognition and intelligent algorithms to accurately capture the whole process of conflict events and provide technical support for improving traffic safety.

[0046] The following is the specific processing technology logic of step S1, which focuses on monitoring vehicle behavior through a vehicle-mounted camera and triggering a conflict event analysis process to ensure clear connection and consistent parameter definition with subsequent steps.

[0047] Step S1 includes the following content:

[0048] The turning light state and lane deviation trajectory of the vehicle are acquired in real time through the vehicle-mounted camera. The monitoring of the turning light state relies on the vehicle-mounted camera to capture the flashing of the turning light, and the image recognition technology is used to determine whether the turning light is in the active state. If the turning light is detected to flash, the turning light state is recorded as active; if no flashing is detected, the turning light state is recorded as inactive. The monitoring of the lane deviation trajectory is achieved through the image processing technology of the vehicle-mounted camera, specifically, the position offset of the vehicle relative to the lane center line is analyzed, and a time-varying lateral distance data sequence is generated. Each item in the lateral distance data sequence represents the offset distance of the vehicle from the lane center line at a certain time, where a positive value indicates that the vehicle is offset to the right, and a negative value indicates that the vehicle is offset to the left.

[0049] Monitoring the turning light state can directly reflect the lane changing or turning intention of the driver, as an explicit signal input, avoiding misjudgment of unintentional slight deviation as lane changing behavior, thereby improving the accuracy of judgment.

[0050] The calculation logic of the lateral displacement rate is based on the lateral distance data sequence of the lane deviation trajectory, which is used to quantify the lateral movement speed of the vehicle in unit time. The specific calculation method is: selecting two consecutive time points, extracting the lateral distance values corresponding to the two time points from the lateral distance data sequence, calculating the difference between the two values, and then dividing the difference by the time interval between the two time points to obtain the lateral displacement rate. The time interval is determined by the frame time difference of data acquisition, for example, if the vehicle-mounted camera acquires data at a frequency of 30 frames per second, the time interval is the time required to acquire one frame. The unit of lateral displacement rate is meters per second, and its numerical value reflects the degree of urgency of the vehicle lane changing behavior.

[0051] The calculation of the lateral displacement rate can dynamically represent the degree of urgency of the vehicle lane changing behavior, which is more accurate than relying on static lateral distance to capture high-risk and urgent behaviors.

[0052] The calculation logic of the dynamic threshold generates an adaptively adjusted reference value by comprehensively considering the lane width, the vehicle speed, and the environmental complexity factor, for judging whether the lateral displacement rate is abnormal. The specific calculation method is as follows: first, taking the lane width as a basic parameter, the lane width is divided by a fixed reference lane-changing time, and the reference lane-changing time is set as a typical time value required for normal lane-changing; then, the vehicle speed is compared with a fixed reference speed to calculate the ratio of the two; next, the division result is multiplied by the ratio of the vehicle speed and the reference speed; finally, the result is multiplied by an environmental complexity factor to obtain the dynamic threshold. The value range of the environmental complexity factor is adjusted according to the complexity of the traffic environment, for example, a higher value is taken in areas with dense traffic flow or complex road types, and a lower value is taken in simple environments. For example, the dynamic threshold is calculated as follows:

[0053] The dynamic threshold comprehensively considers the lane width, the vehicle speed, and the environmental complexity, and the formula is as follows:

[0054]

[0055] Wherein:

[0056] W lane : current lane width, unit: meter, reflecting the physical constraint of lane-changing behavior.

[0057] T ref : reference lane-changing time, fixed as 3 seconds, representing the time required for normal lane-changing.

[0058] V ego : vehicle speed, unit: meter / second, affecting the dynamic characteristics of lane-changing.

[0059] V base : reference speed, fixed as 10 meters / second (about 36 kilometers / hour), used for normalization.

[0060] K env : environmental complexity factor, value range [1, 1.5], adjusted according to traffic flow density or road type (for example, a higher value is taken in weaving areas).

[0061] The introduction of the dynamic threshold avoids the defects of the traditional fixed threshold that cannot adapt to different traffic scenarios, and by combining the lane width, the vehicle speed, and the environmental complexity factor, the judgment standard can be adaptively adjusted according to the actual driving conditions. This adaptability ensures the applicability of the system in wide lanes, high-speed driving, or complex traffic environments, and improves the accuracy and flexibility of abnormal behavior recognition.

[0062] The judgment logic of the trigger condition is based on a comprehensive analysis of the turn signal state and lateral displacement rate, which is used to determine whether to start the conflict event analysis process. The specific judgment process is as follows: first, check if the turn signal state is active, if the turn signal state is not active, the condition is not met; if the turn signal state is active, further compare the lateral displacement rate with the dynamic threshold. If the lateral displacement rate exceeds the dynamic threshold, it is determined that the trigger condition is met, the current time is recorded as the trigger time point, and the subsequent analysis process is started; if the lateral displacement rate does not exceed the dynamic threshold, the condition is not met and is not triggered.

[0063] The dual judgment of the trigger condition combines the dynamic characteristics of the driver's intention and the lane change behavior, ensuring that the analysis is triggered only when the driver explicitly expresses the intention to change lanes and the lane change behavior is unusually urgent. This mechanism can effectively filter out unintentional or normal lateral movement, reduce false triggering, and improve the accuracy and efficiency of the system.

[0064] The execution of step S1 requires a precondition, i.e., the vehicle has entered the road weaving area, the vehicle-mounted camera has completed initialization and can operate normally, and continuously provides image data of the turn signal state and lane deviation trajectory. In the case of meeting the trigger condition, the processing result of step S1 is to record the trigger time point and related data of the turn signal state, lateral displacement rate and lane deviation trajectory, which will be passed as output to the subsequent steps for further constructing the vehicle feature vector and matching the conflict type template.

[0065] The specific processing technology logic of step S1 monitors the turn signal state and lane deviation trajectory through the vehicle-mounted camera, calculates the lateral displacement rate and compares it with the dynamic threshold generated according to the lane width, vehicle speed and environmental complexity factor, and triggers the conflict event analysis when the turn signal state is active and the lateral displacement rate exceeds the dynamic threshold. This technical logic provides a precise trigger mechanism for traffic conflict identification in road weaving areas, and its adaptability and dual condition design significantly improve the accuracy, flexibility and efficiency of the system. The output of the trigger time point and related data lays a foundation for subsequent analysis, ensuring that the whole process of the conflict event can be effectively captured and processed.

[0066] In the traffic conflict identification in the road weaving area, step S1 monitors the turn signal state and lane deviation trajectory of the vehicle in real time through the vehicle-mounted camera, triggers the conflict event analysis process when the turn signal is activated and the lateral displacement rate of the vehicle exceeds the dynamic threshold, and records the trigger time and related data (turn signal activation start time, lateral displacement rate, lane deviation trajectory). Step S2 constructs the vehicle feature vector and matches the conflict type template library based on the data provided by S1, determines the adaptive asymmetric time window template, and provides accurate data basis for window optimization of step S3.

[0067] Step S2 includes the following contents:

[0068] The processing logic of feature vector construction calculates the turn signal activation duration, the lateral acceleration change rate and the lane deviation angle based on the data obtained from step S1, and generates a vehicle behavior feature vector.

[0069] Turn signal activation duration: the time difference between the start time of the turn signal activation monitored by the vehicle-mounted camera and the trigger time recorded in step S1 is calculated. Specifically, the time interval from the start time of the turn signal activation to the trigger time is directly subtracted, and the result is the turn signal activation duration, which is used to represent the urgency of the driver's lane change decision.

[0070] Lateral acceleration change rate: first, the lateral acceleration is calculated according to the lateral displacement rate provided in step S1, which is the difference between the lateral displacement rate of the current frame and the lateral displacement rate of the previous frame, divided by the inter-frame time difference of data collection, to obtain the lateral acceleration; then, the difference between the lateral acceleration at the trigger time and the lateral acceleration of the previous frame is divided by the inter-frame time difference, to obtain the lateral acceleration change rate, which is used to represent the urgency of the lane change behavior.

[0071] For example, the lateral acceleration change rate A lat_rate :

[0072] is defined as the rate of change of the vehicle's lateral acceleration in unit time, representing the urgency of the lane change behavior.

[0073]

[0074] A lat (t trigger ) is the lateral acceleration at the trigger time, which is calculated by differentiating the lateral displacement rate V lat (t trigger ).

[0075] A lat (t trigger -Δt) is the lateral acceleration of the previous frame before the trigger time.

[0076] Δt is the fixed inter-frame time difference of data collection, determined by the sampling frequency of the vehicle-mounted camera.

[0077] V lat (t trigger ) is the lateral displacement rate at the trigger time in step S1.

[0078] Lane deviation angle: the value of the arctangent function is calculated by using the ratio of the lateral displacement rate and the longitudinal speed at the triggering moment, and is converted into an angle. Specifically, the lateral displacement rate is divided by the longitudinal speed, the result of the arctangent function is calculated, and then the result is multiplied by 180 divided by pi to obtain an angle value, which is used to reflect the angle between the vehicle driving direction and the lane line direction.

[0079] For example, the lane deviation angle θ dev :

[0080] defined as the angle between the vehicle driving direction and the lane line direction, reflecting the degree of deviation of the vehicle from the lane line.

[0081]

[0082] V lat (t trigger ): the lateral displacement rate at the triggering moment.

[0083] V long (t trigger ): the longitudinal speed at the triggering moment, i.e. the vehicle's own speed, recorded by the on-board sensor.

[0084] Finally, the steering light activation time, lateral acceleration change rate and lane deviation angle are combined into a vehicle behavior feature vector.

[0085] By constructing the vehicle behavior feature vector, the decision urgency, urgency and deviation degree of the vehicle lane changing behavior can be quantified comprehensively, and a multi-dimensional behavior description is provided.

[0086] The processing logic of the conflict type template library defines three conflict types, namely, urgent entry, slow lane change and continuous lane change, each type corresponding to a feature vector range and an asymmetric time window template. The feature vector range of the urgent entry type is defined as short steering light activation time, high lateral acceleration change rate and large lane deviation angle, and its corresponding asymmetric time window template is designed as long forward window and short backward window, specifically, the forward window covers a longer period of time before the triggering moment, and the backward window covers a shorter period of time after the triggering moment, to capture the risk accumulation stage before the urgent lane change. The feature vector range of the slow lane change type is defined as medium steering light activation time, low lateral acceleration change rate and small lane deviation angle, and its corresponding asymmetric time window template is designed as medium forward window and medium backward window, specifically, the time span of the two is similar, to balance the coverage of the risk evolution process. The feature vector range of the continuous lane change type is defined as long steering light activation time, lateral acceleration change rate fluctuation and frequent lane deviation angle change, and its corresponding asymmetric time window template is designed as long forward window and long backward window, specifically, both cover a long period of time, to include the risk evolution process of multiple lane changing behaviors.

[0087] By predefining the conflict type template library, the vehicle behavior feature vector can be quickly classified according to its characteristics, and the asymmetric time window template matching it can be selected.

[0088] The processing logic of feature vector matching quantifies the similarity between the vehicle behavior feature vector and the center value of each type of feature vector in the conflict type template library by calculating the Euclidean distance. The specific method is as follows: first, calculate the difference between the vehicle behavior feature vector and the center value of each conflict type feature vector in the steering light activation time, the lateral acceleration change rate and the lane deviation angle; then, calculate the square of the three differences respectively, and then add the three square values to get the sum of squares; then, take the square root of the sum of squares to get the Euclidean distance; finally, take the reciprocal of the Euclidean distance after adding 1, get the similarity value, the range of similarity value is between 0 and 1, the larger the value is, the higher the matching degree is. After the above calculation, the conflict type with the highest similarity value is selected, and the asymmetric time window template corresponding to the conflict type is used as the basis for subsequent analysis.

[0089] The logic of the overall processing flow is executed in the following order: first, obtain the trigger time, steering light activation start time, lateral displacement rate and longitudinal velocity from step S1; then, calculate the steering light activation time using the steering light activation start time and the trigger time; then, calculate the lateral acceleration change rate based on the lateral displacement rate, and calculate the lane deviation angle combined with the longitudinal velocity; then, combine the steering light activation time, the lateral acceleration change rate and the lane deviation angle into the vehicle behavior feature vector; then, calculate the similarity between the vehicle behavior feature vector and the center value of each type of feature vector in the conflict type template library; finally, select the conflict type with the highest matching degree according to the similarity value, and determine the corresponding asymmetric time window template.

[0090] The overall processing flow integrates data acquisition, feature calculation and template matching organically to form an adaptive asymmetric time window selection mechanism. This method can quickly respond and provide appropriate time window in different conflict scenarios, providing a solid foundation for subsequent risk evolution analysis.

[0091] The execution of step S2 depends on the data provided by step S1, such as trigger time, steering light activation start time, lateral displacement rate and longitudinal velocity, to ensure the construction of vehicle behavior feature vector has real-time and accuracy. After completing the feature vector matching, the determined asymmetric time window template will be passed to step S3 for optimization and expansion of the time window to support more comprehensive risk assessment process.

[0092] Step S2 realizes the adaptive selection of the asymmetric time window template by constructing the vehicle behavior feature vector and performing similarity matching with the conflict type template library. It can accurately adapt to the needs of different conflict scenarios, providing a flexible and accurate analysis basis for traffic conflict identification in road weaving areas, and significantly improving the pertinence and comprehensiveness of risk assessment.

[0093] Step S2 determines the initial asymmetric time window template by matching the vehicle behavior feature vector with the preset conflict type template library, providing a preliminary time range for risk evolution analysis. However, due to the complexity of traffic scenarios and the dynamic changes of vehicle behavior, the asymmetric time window template determined in step S2 may not fully cover the entire process of the conflict risk evolution chain, especially when the risk signal peak is not fully revealed or the conflict event evolution chain is long. Therefore, step S3 introduces a window optimization and expansion mechanism based on conflict energy gradient, which analyzes the time series peak of conflict energy gradient within the window in real time and dynamically adjusts the window boundary, ensuring that the finally determined time window can fully contain the key stages of the risk evolution chain.

[0094] Step S3 is based on the asymmetric time window template determined in step S2, and by calculating the time series peak of conflict energy gradient within the window and comparing it with the preset threshold, it determines whether the risk evolution is fully revealed. If the peak does not reach the preset threshold, the window boundary is expanded step by step until the peak reaches the threshold or the termination condition is met, and the time window containing the complete risk evolution chain is finally determined. This optimized time window will be directly passed to step S4 for subsequent collection and feature extraction of micro-trajectory data.

[0095] Step S3 includes the following:

[0096] The processing logic of conflict energy gradient calculation aims to quantitatively evaluate the trend of traffic conflict risk evolution by quantifying the change rate of conflict risk indicators per unit time. The specific method is as follows:

[0097] First, calculate the conflict risk indicator, which is a quantitative evaluation based on the relative motion parameters of the vehicle and its adjacent vehicles. The calculation process is as follows: multiply the reciprocal of the relative distance between the vehicle and its adjacent vehicles with the weighted sum of the relative speed and relative acceleration to obtain the conflict risk indicator, where the reciprocal of the relative distance is used to reflect the proximity between vehicles, and the relative speed and relative acceleration are used to represent the difference in motion state between vehicles; then, the conflict energy gradient is defined as the difference between the conflict risk indicators at adjacent time points divided by the inter-frame time difference of data collection, and the result is in the unit of risk indicator change rate per second, which is used to represent the dynamic change speed of risk evolution. In the calculation process, the conflict risk indicators at adjacent time points are obtained through continuous collection of vehicle motion data, and the inter-frame time difference is determined by the sampling frequency of the data collection device.

[0098] For example, the calculation of the conflict risk indicator can be as follows:

[0099] Conflict risk indicator R conflict (t) is based on the comprehensive evaluation of the relative motion parameters between vehicles, and the formula is:

[0100]

[0101] d rel (t): the relative distance between the vehicle and the adjacent vehicle at time t, obtained by real-time monitoring of the vehicle-mounted camera.

[0102] v rel (t): the relative speed between the vehicle and the adjacent vehicle at time t, calculated by the difference of trajectory data.

[0103] a rel (t): the relative acceleration between the vehicle and the adjacent vehicle at time t, calculated by the second difference of speed data.

[0104] The calculation of the conflict energy gradient can monitor the intensification or slowing down trend of risk evolution in real time. Compared with relying only on the static conflict risk indicator, this method can more accurately capture the key evolution stage of the traffic conflict event. By dynamically quantifying the risk change speed, it provides a reliable basis for the optimization of the subsequent time window, ensuring that the window adjustment can reflect the actual risk evolution needs.

[0105] The processing logic of peak detection in the time window is based on the asymmetric time window template determined in step S2, which analyzes the time series of the conflict energy gradient in the window, identifies its maximum value and compares it with the preset threshold. The specific method is:

[0106] First, obtain the asymmetric time window template provided in step S2 to determine the initial time window, which covers a specific time period before and after the trigger time; then, in the initial time window, calculate the time series of the conflict energy gradient, which consists of multiple consecutive time points of the conflict energy gradient value, and find the maximum value of the conflict energy gradient from it; then, compare the maximum value with the preset conflict energy gradient threshold, which is a benchmark value determined according to historical traffic conflict data and expert experience, to judge whether the risk evolution is fully manifested in the current window.

[0107] The processing logic of the window expansion mechanism is to expand the window boundary step by step to cover a wider time period when the maximum value of the conflict energy gradient in the time window does not reach the preset conflict energy gradient threshold, ensuring the integrity of the traffic conflict risk evolution chain. The specific method is:

[0108] When the maximum value of the conflict energy gradient within the time window is detected to be less than the preset conflict energy gradient threshold, a window expansion operation is triggered, and the time window boundary is preferentially expanded forward to cover the earlier traffic conflict risk accumulation stage; the length of each expansion is a fixed value, for example, 0.5 seconds is added each time, the conflict energy gradient time series within the new window is recalculated after expansion, and a new maximum value is found; the expansion process continues until the maximum value of the conflict energy gradient within the window reaches or exceeds the preset conflict energy gradient threshold, or the total length of the window reaches the preset maximum allowed length, which is a system-set upper limit value used to avoid excessive window expansion.

[0109] The window expansion mechanism can adapt to the risk evolution length of different traffic conflict scenarios by dynamically adjusting the window boundary, and ensure that the analysis window completely contains the key risk stage. The system's ability to identify long-term traffic conflict events is improved, and by setting the maximum allowed length as the termination condition, the waste of computing resources caused by unlimited expansion is avoided.

[0110] The final time window determination processing logic stops expansion and determines the optimized time window when the maximum value of the conflict energy gradient reaches the preset conflict energy gradient threshold or the total length of the window reaches the maximum allowed length during the window expansion process. The specific method is:

[0111] When the maximum value of the conflict energy gradient within the expanded time window reaches or exceeds the preset conflict energy gradient threshold, it indicates that the traffic conflict risk evolution chain has fully emerged, and the current time window is determined as the final optimized time window; if the total length of the time window reaches the preset maximum allowed length, but the maximum value of the conflict energy gradient has not reached the preset conflict energy gradient threshold, the time window that reaches the maximum allowed length is taken as the final optimized time window to ensure the timeliness of the analysis process.

[0112] The determination of the final time window controls the length of the time window while ensuring the integrity of the traffic conflict risk evolution chain, and balances the real-time performance and accuracy of the system.

[0113] The execution of step S3 depends on the asymmetric time window template provided by step S2 as the starting point of the initial time window, ensuring the pertinence of window optimization. After completing the time window optimization, the final optimized time window is directly passed to step S4 for collecting vehicle micro trajectory data and extracting spatio-temporal features, ensuring the continuity and consistency of the entire analysis process.

[0114] The specific processing technology logic of step S3 ensures that the final optimized time window can completely cover the key stages of the traffic conflict risk evolution chain in the road weaving area through the calculation of the conflict energy gradient and the dynamic expansion mechanism of the time window. It can adapt to the risk evolution length of different traffic conflict scenarios, providing accurate time range support for the collection and spatiotemporal feature extraction of subsequent vehicle micro-trajectory data, significantly improving the accuracy and comprehensiveness of traffic conflict identification.

[0115] Step S3 determines the optimized time window containing the complete risk evolution chain by optimizing the conflict energy gradient and expanding the time window, providing an accurate time range for subsequent feature extraction. However, optimizing the time window alone is not enough to fully represent the spatiotemporal distribution of traffic conflict risk, and key features need to be extracted from micro-trajectory data to capture vehicle interaction behavior and potential conflict points. Therefore, step S4 collects vehicle micro-trajectory data and extracts spatiotemporal features within the optimized time window determined in step S3, generates a spatiotemporal conflict heat map, and marks high-risk spatiotemporal coordinates to provide visual risk assessment basis for the generation of step S5's hierarchical warning signals.

[0116] Step S4 collects vehicle micro-trajectory data, including position, speed, and acceleration, within the optimized time window determined in step S3, extracts spatiotemporal features of these data using a pyramid convolutional network, generates a spatiotemporal conflict heat map, and marks high-risk spatiotemporal coordinates. This step provides a visual risk assessment tool for step S5, ensuring that the spatiotemporal distribution of traffic conflict risk is accurately represented.

[0117] Step S4 includes the following:

[0118] The processing logic of micro-trajectory data collection collects vehicle position, speed, and acceleration data in real time through vehicle-mounted cameras and sensors within the optimized time window determined in step S3.

[0119] The collected data is first preprocessed, including denoising and interpolation. The denoising operation uses a sliding average filtering technique to smooth the data by calculating the average of consecutive data points, eliminating the influence of sensor noise; the interpolation operation uses a linear interpolation technique to complete missing data points between adjacent data points according to the time ratio, generating continuous time series data to ensure data integrity in the subsequent feature extraction process.

[0120] The micro-trajectory data collection process can fully describe the behavior characteristics of vehicles in traffic conflict events by collecting multi-dimensional dynamic data. The denoising and interpolation steps in preprocessing improve the quality and continuity of the data, providing a reliable data foundation for subsequent feature extraction, thereby enhancing the accuracy and robustness of traffic conflict analysis.

[0121] The processing logic of the pyramid convolutional network structure extracts spatiotemporal features from microscopic trajectory data by designing multi-scale convolutional layers. The network structure includes four parts: input layer, convolutional layer, pooling layer, and feature fusion layer. The input layer receives preprocessed microscopic trajectory data, which is organized in the form of vehicle number, frame number within the time window, and feature dimension, which includes position, speed, and acceleration. The convolutional layer uses different sizes of convolutional kernels, such as 1x1, 3x3, and 5x5, to extract short-term, medium-term, and long-term features, respectively. Specifically, smaller convolutional kernels are used to capture local changes in vehicle trajectories, while larger convolutional kernels are used to capture global trends in vehicle trajectories. The pooling layer uses the max-pooling technique to aggregate features by selecting the maximum value in the local region, reducing the data dimension, while retaining the main feature information. The feature fusion layer integrates the outputs of different convolutional layers through concatenation operations to generate a comprehensive spatiotemporal feature map containing multi-scale information, reflecting the interaction behavior between vehicles and potential traffic conflict points.

[0122] The pyramid convolutional network structure can capture both local details and overall trends in vehicle trajectories through multi-scale feature extraction techniques, ensuring that the spatiotemporal features of traffic conflict risk are fully characterized. This network design improves the accuracy and adaptability of feature extraction, providing a solid foundation for subsequent generation of accurate spatiotemporal conflict heat maps.

[0123] The processing logic of the spatiotemporal conflict heat map generation is based on the spatiotemporal features extracted by the pyramid convolutional network, and generates a three-dimensional heat map through three steps of risk intensity mapping, spatiotemporal interpolation, and normalization. First, the spatiotemporal feature map is processed through a fully connected layer to convert the features into risk intensity values, which represent the degree of traffic conflict risk at a specific time and spatial location. The role of the fully connected layer is to weight and combine the input features, outputting numerical values reflecting the risk size. Next, the risk intensity values are processed through three-dimensional linear interpolation to generate continuous spatiotemporal distribution data by calculating intermediate values between adjacent spatiotemporal points in proportion, ensuring the smoothness of the heat map in the time and space dimensions. Finally, the generated heat map data is normalized to the range of 0 to 1 by dividing all risk intensity values by the maximum value, facilitating subsequent risk threshold judgment and evaluation.

[0124] The spatiotemporal conflict heat map generation process converts spatiotemporal features into a visual risk distribution map, intuitively displaying the spatiotemporal variation characteristics of traffic conflict risk. Interpolation and normalization processing ensure the continuity and comparability of the heat map, improving the accuracy of risk identification and practical application value.

[0125] The processing logic of high-risk spatiotemporal coordinate labeling is based on the spatiotemporal conflict heat map, identifying spatiotemporal points with risk intensity exceeding the preset threshold, and recording them as high-risk spatiotemporal coordinates. The specific process is as follows:

[0126] First, a preset risk intensity threshold is needed to be set, which is determined according to historical traffic conflict data statistical analysis and expert experience, and the unit is consistent with the risk intensity value in the heat map. Next, each spatiotemporal point in the spatiotemporal conflict heat map is traversed, and its risk intensity value is compared with the size of the preset threshold one by one; if the risk intensity value of a certain spatiotemporal point exceeds the preset threshold, the coordinates of the spatiotemporal point are recorded as high-risk spatiotemporal coordinates.

[0127] The processing logic of step S4 takes the optimized time window provided in step S3 as the starting point of data collection, ensuring that the analysis process is consistent with the time range of the previous steps. After generating the spatiotemporal conflict heat map and high-risk spatiotemporal coordinates, the processing results are directly passed to step S5 for the generation of graded warning signals, ensuring seamless connection of the entire system analysis process.

[0128] The specific processing technology logic of step S4 collects vehicle micro trajectory data within the optimized time window, extracts spatiotemporal features using a pyramid convolution network, generates a spatiotemporal conflict heat map, and marks high-risk spatiotemporal coordinates. This method can accurately represent the spatiotemporal distribution characteristics of traffic conflict risk in road weaving areas, providing a visual and reliable risk assessment basis for subsequent graded warning signal generation, thereby significantly improving the accuracy and application value of traffic conflict identification.

[0129] Marking high-risk spatiotemporal coordinates provides precise spatiotemporal positioning information for subsequent graded warning signal generation, ensuring that the system can accurately identify and warn the key moments and locations of traffic conflict events. By identifying and recording spatiotemporal points with risk intensity exceeding the preset threshold in the spatiotemporal conflict heat map, the system converts abstract risk assessment into specific time and space coordinates, providing intuitive and operational risk prompts for drivers or traffic management systems. This marking method enhances the relevance and timeliness of warning signals, while providing data support for traffic safety management, helping to prevent and reduce traffic accidents.

[0130] Step S4 collects vehicle micro trajectory data within the optimized time window, extracts spatiotemporal features, generates a spatiotemporal conflict heat map and high-risk spatiotemporal coordinates, providing a visual basis for risk assessment. However, relying solely on the spatiotemporal conflict heat map and marked high-risk points is not enough to achieve real-time traffic safety warning, and further conversion of risk information into graded warning signals is needed, as well as optimization of the asymmetric time window template parameters in step S2 through a feedback mechanism to adapt to dynamic traffic environments. Therefore, step S5 needs to explicitly base the processing logic on the spatiotemporal conflict heat map and high-risk spatiotemporal coordinates to generate operational warning signals and achieve closed-loop optimization.

[0131] Step S5 obtains the spatio-temporal conflict heat map and high-risk spatio-temporal coordinates from step S4, generates a hierarchical warning signal based on the peak intensity and distribution density, and feeds back the warning result to the conflict type template library in step S2 to dynamically update the parameters of the asymmetric time window template, forming a closed-loop optimization mechanism.

[0132] Step S5 includes the following contents:

[0133] The generation of the hierarchical warning signal determines the severity of the traffic conflict risk by analyzing the peak intensity of the spatio-temporal conflict heat map and the distribution density of the high-risk spatio-temporal coordinates, and generates the corresponding warning signal.

[0134] The specific method is:

[0135] First, the peak intensity is extracted from the spatio-temporal conflict heat map generated in step S4, which is defined as the maximum value of the risk intensity in the spatio-temporal conflict heat map, used to reflect the highest point of traffic conflict risk; second, the distribution density of high-risk spatio-temporal coordinates is calculated, which is obtained by dividing the number of high-risk spatio-temporal coordinates by the product of the spatial area of the road weaving area and the time window length, used to reflect the concentration of high-risk points in space-time; then, combined with the peak intensity and the distribution density, a preset threshold matrix is used for hierarchical judgment, the hierarchical rules are: when the peak intensity is lower than the preset low threshold or the distribution density is lower than the preset low threshold, the warning signal is classified as low level; when the peak intensity is between the preset low threshold and the high threshold and the distribution density is between the preset low threshold and the high threshold, the warning signal is classified as medium level; when the peak intensity is higher than the preset high threshold and the distribution density is higher than the preset high threshold, the warning signal is classified as high level.

[0136] The hierarchical warning signal generation process can comprehensively evaluate the severity of traffic conflicts by comprehensively analyzing the peak intensity and distribution density, providing intuitive and operable warning information. Avoiding the misjudgment caused by a single indicator, improving the accuracy and practicality of the warning signal, providing a basis for timely response to traffic risks for drivers or traffic management systems, thereby reducing the possibility of traffic accidents.

[0137] The feedback mechanism and parameter update record the hierarchical warning signal and its corresponding conflict type, regularly analyze the warning accuracy, and adjust the parameters of the asymmetric time window template in step S2 according to the analysis results. The specific method is:

[0138] First, the generated hierarchical early warning signal and its corresponding conflict type are stored in the conflict record database to form historical data accumulation; then, data is extracted from the conflict record database at regular intervals to calculate the early warning accuracy rate of each conflict type, which is obtained by counting the number of times the early warning signal is consistent with the actual conflict event, and then dividing the total number of generated early warning signals for this conflict type; next, for the conflict type with an early warning accuracy rate lower than the preset accuracy threshold, the corresponding asymmetric time window template parameters in step S2 are adjusted, and the adjustment method is to increase the forward window length or the backward window length of the asymmetric time window template, and the adjustment amplitude can be a pre-set fixed value or dynamically determined according to the difference between the early warning accuracy rate and the preset accuracy threshold; finally, the adjusted asymmetric time window template parameters are updated to the conflict type template library in step S2, and the subsequent changes in early warning accuracy rate are continuously monitored.

[0139] The feedback mechanism and parameter updating process ensure that the asymmetric time window template can adapt to the actual changes of the traffic environment through analysis of historical data and dynamic adjustment of parameters. This method realizes the adaptive optimization of the system, improves the accuracy and stability of traffic conflict recognition, and provides reliable support for long-term operation of traffic safety management.

[0140] The closed-loop optimization mechanism processes by continuously monitoring the early warning accuracy rate and dynamically adjusting the parameters of the asymmetric time window template, forming an adaptive optimization cycle. The specific method is:

[0141] After updating the parameters of the asymmetric time window template, the subsequent generated hierarchical early warning signal and its corresponding conflict type data are continuously collected to calculate the early warning accuracy rate; if the calculated early warning accuracy rate is still lower than the preset accuracy threshold, the forward window length or the backward window length of the asymmetric time window template is further adjusted, and the adjustment process is repeated until the early warning accuracy rate meets the requirements of the preset accuracy threshold. This mechanism ensures that the system can continuously optimize its analysis parameters to adapt to complex changes in the traffic environment.

[0142] The closed-loop optimization mechanism realizes the self-improvement of the traffic conflict recognition system through continuous monitoring and adjustment, ensuring its accuracy and timeliness in different traffic scenarios. It enhances the robustness and reliability of the system, providing continuous improvement of technical support for traffic safety management.

[0143] The processing logic of step S5 takes the spatiotemporal conflict heat map and high-risk spatiotemporal coordinates provided in step S4 as input data, ensuring consistency of the analysis process with the data of the previous steps. After completing the hierarchical warning signal generation and asymmetric time window template parameter adjustment, the updated asymmetric time window template parameters are directly fed back to the conflict type template library of step S2 for subsequent conflict type identification and analysis. This design ensures that the entire system forms a complete adaptive analysis cycle after a traffic conflict event is triggered.

[0144] The specific processing technology logic of step S5 generates hierarchical warning signals by analyzing the peak intensity of the spatiotemporal conflict heat map and the distribution density of the high-risk spatiotemporal coordinates, and dynamically adjusts the parameters of the asymmetric time window template based on the warning accuracy, forming a closed-loop optimization mechanism. This method realizes real-time warning and adaptive optimization of traffic conflicts in road weaving areas, improves the efficiency and accuracy of traffic safety management, and provides technical support for the intelligent development of the transportation system.

[0145] Embodiment 2: Figure 2 The present application provides a road weaving area traffic conflict identification system, which comprises:

[0146] a trigger analysis module, a window matching module, a window optimization module, a feature extraction module, and a warning feedback module;

[0147] Trigger analysis module: use the vehicle-mounted camera to monitor the state of the turn signal and the lane deviation trajectory, calculate the lateral displacement rate and compare it with the dynamic threshold, and when the turn signal is activated and the lateral displacement rate exceeds the standard, trigger the conflict event analysis;

[0148] Window matching module: construct a vehicle feature vector, and determine the corresponding asymmetric time window template by similarity matching with the preset conflict type template library;

[0149] Window optimization module: optimize the time window based on the conflict energy gradient, if the peak value of the conflict energy gradient in the window does not reach the preset threshold, gradually expand the window until the peak value reaches the threshold or meets the termination condition, and determine the time window containing the complete risk evolution chain;

[0150] Feature extraction module: collect vehicle micro-trajectory data in the optimized time window, extract spatiotemporal features using a pyramid convolutional network, and generate a spatiotemporal conflict heat map and mark high-risk spatiotemporal coordinates

[0151] Warning feedback module: generate hierarchical warning signals according to the peak intensity of the spatiotemporal conflict heat map and the distribution density of the high-risk spatiotemporal coordinates, and dynamically update the asymmetric time window template parameters in the conflict type template library through the warning accuracy feedback.

[0152] The above formulas are all de-dimensioned to calculate the numerical values, the formulas are obtained by collecting a large amount of data to simulate a formula of the most recent real situation, and the preset parameters in the formulas are set by the person skilled in the art according to the actual situation.

[0153] It should be noted that the system of the present application can be deployed in the device itself to realize embedded application, or can be run on PC or other terminal with user interface, so as to meet various hardware environments and use requirements.

[0154] The above only describes some exemplary embodiments of the present application by way of illustration, and it is needless to say that the described embodiments can be modified in various ways without departing from the spirit and scope of the present application for those skilled in the art. Therefore, the above figures and descriptions are illustrative in nature and should not be understood as limiting the scope of protection of the claims of the present application.

[0155] It should be noted that in this document, relational terms such as first and second and the like can merely be used to distinguish one entity or action from another, without necessarily requiring or implying that there is any such actual relationship or order between or among the entities or actions. Moreover, the terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus. An element proceeded by "comprises a... " does not, without more constraints, exclude the existence of additional identical elements in the process, method, article, or apparatus that comprises the element.

[0156] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of changes or replacements within the technical range disclosed in the present application, which should be covered in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for road weaving zone traffic conflict identification, characterized in that, The method comprises the steps of: S1: using a vehicle-mounted camera to monitor the state of the turn signal and the lane deviation trajectory, calculating the lateral displacement rate and comparing it with the dynamic threshold, triggering the conflict event analysis when the turn signal is activated and the lateral displacement rate exceeds the threshold; S2: constructing a vehicle feature vector, determining the corresponding asymmetric time window template by similarity matching with the preset conflict type template library; S3: optimizing the time window based on the conflict energy gradient, if the peak value of the conflict energy gradient in the window does not reach the preset threshold, gradually expanding the window until the peak value reaches the threshold or the termination condition is met, and determining the time window containing the complete risk evolution chain; S4: collecting vehicle micro-trajectory data in the optimized time window, extracting spatio-temporal features using a pyramid convolutional network, generating a spatio-temporal conflict heat map and marking high-risk spatio-temporal coordinates; S5: generating a hierarchical warning signal according to the peak intensity of the spatio-temporal conflict heat map and the distribution density of the high-risk spatio-temporal coordinates, and dynamically updating the asymmetric time window template parameters in the conflict type template library through the warning accuracy feedback.

2. The method for traffic conflict identification in weaving area of road according to claim 1, characterized in that, Step S1 includes the following content: The vehicle-mounted camera monitors the state of the turn signal and the lane deviation trajectory in real time, calculates the lateral displacement rate of the vehicle, and generates a dynamic threshold according to the lane width, vehicle speed and environmental complexity factor. When the turn signal is activated and the lateral displacement rate exceeds the dynamic threshold, the conflict event analysis process is triggered.

3. The method for traffic conflict identification in weaving area of road according to claim 2, characterized in that, Step S1 also includes the following content: The vehicle-mounted camera captures the flashing of the turn signal to determine whether the turn signal is activated or not, and obtains the lateral distance data sequence of the vehicle relative to the lane center line through image processing technology, and then calculates the lateral displacement rate. The dynamic threshold is obtained by dividing the lane width by the reference lane-changing time, multiplying the ratio of the vehicle speed to the reference speed, and multiplying the environmental complexity factor. The trigger condition is that the turn signal is activated and the lateral displacement rate is greater than the dynamic threshold. After triggering, the trigger time point is recorded.

4. The method for traffic conflict identification in weaving area of road according to claim 3, characterized in that, Step S2 includes the following content: A feature vector of the vehicle is constructed, which includes the turn signal activation time, the lateral acceleration change rate and the lane deviation angle. The turn signal activation time is defined as the time difference from the start time of the turn signal activation to the trigger time. The lateral acceleration change rate is defined as the difference between the lateral acceleration at the trigger time and the previous frame lateral acceleration divided by the frame time difference. The lane deviation angle is defined as the inverse tangent function value of the ratio of the lateral displacement rate to the longitudinal speed at the trigger time.

5. The method for traffic conflict identification in weaving area of road according to claim 4, characterized in that, Step S2 also includes the following content: After the construction is completed, the feature vector is matched with the preset conflict type template library. The conflict type template library includes three types: sudden entry, slow lane change and continuous lane change, each type corresponding to a feature vector center value. The matching process determines the similarity by calculating the inverse function form of the Euclidean distance between the feature vector and the feature vector center value of each type. According to the similarity value, the conflict type with the highest matching degree is selected to determine its corresponding asymmetric time window template.

6. The method for traffic conflict identification in weaving area of road according to claim 5, characterized in that, Step S3 includes the following content: In the determined asymmetric time window template, the peak value of the conflict energy gradient time series is calculated, if the peak value does not reach the preset threshold, the window boundary is expanded step by step, that is, each time a fixed time length is expanded, until the peak value of the conflict energy gradient in the window reaches the preset threshold or the total time length of the window reaches the preset maximum allowed time length, and finally the optimized time window is determined.

7. The method for traffic conflict identification in weaving area of road according to claim 6, characterized in that, Step S3 also includes the following: Wherein, the conflict energy gradient is defined as the difference between the conflict risk indicators at adjacent time points divided by the inter-frame time difference, and the conflict risk indicator is calculated based on the relative distance, relative speed and relative acceleration of the vehicle and the adjacent vehicle; The window is expanded preferentially forward to cover the earlier risk accumulation stage, and the expanded window is used to recalculate the peak value of the conflict energy gradient until the termination condition is met.

8. The method for traffic conflict identification in weaving area of road according to claim 7, characterized in that, Step S4 includes the following: Collect micro trajectory data of the vehicle in the optimized time window, extract spatio-temporal features of the micro trajectory data using a pyramid convolution network, capture short, medium and long time spatio-temporal features through multi-scale convolution layers, and generate a comprehensive spatio-temporal feature map after aggregation and fusion through a pooling layer; map the comprehensive spatio-temporal feature map to a risk intensity value through a fully connected layer, and generate a spatio-temporal conflict heat map through three-dimensional linear interpolation and normalization processing; In the spatio-temporal conflict heat map, mark the spatio-temporal points with risk intensity values exceeding the preset threshold as high-risk spatio-temporal coordinates.

9. The method for traffic conflict identification in weaving area of road according to claim 8, characterized in that, Step S5 includes the following: Obtain the spatio-temporal conflict heat map and high-risk spatio-temporal coordinates, calculate the peak intensity of the spatio-temporal conflict heat map and the distribution density of the high-risk spatio-temporal coordinates, generate a hierarchical warning signal according to the peak intensity and the distribution density, record the hierarchical warning signal and its corresponding conflict type to the conflict record database, regularly calculate the warning accuracy rate of each conflict type, and for the conflict type with a warning accuracy rate lower than the preset threshold, adjust by increasing the forward window length or the backward window length, the adjustment amplitude is a pre-set fixed value or dynamically determined according to the difference between the warning accuracy rate and the preset accuracy rate threshold, and update the adjusted asymmetric time window template parameters to the conflict type template library.

10. A system for road weaving zone traffic conflict identification for implementing the method for road weaving zone traffic conflict identification according to any one of claims 1 to 9, characterized in that, includes: trigger analysis module, window matching module, window optimization module, feature extraction module and warning feedback module; Trigger analysis module: use the vehicle-mounted camera to monitor the turn signal state and lane deviation trajectory, calculate the lateral displacement rate and compare it with the dynamic threshold, and trigger the conflict event analysis when the turn signal is activated and the lateral displacement rate exceeds the standard; Window matching module: construct a vehicle feature vector, and determine the corresponding asymmetric time window template by similarity matching with the preset conflict type template library; Window optimization module: optimize the time window based on the conflict energy gradient, if the peak value of the conflict energy gradient in the window does not reach the preset threshold, expand the window step by step until the peak value reaches the threshold or the termination condition is met, and determine the time window containing the complete risk evolution chain; Feature extraction module: collect vehicle micro trajectory data in the optimized time window, extract spatio-temporal features using a pyramid convolution network, generate a spatio-temporal conflict heat map and mark high-risk spatio-temporal coordinates Early warning feedback module: according to the peak intensity of space-time conflict heat map and the distribution density of high-risk space-time coordinates, generate graded early warning signals, and dynamically update the asymmetric time window template parameters in the conflict type template library through early warning accuracy feedback.

Images