Risk over-speed behavior identification and early warning method based on track space-time matching analysis

Abstract

The invention discloses a risk overspeed behavior identification and early warning method based on track space-time matching analysis. The method comprises the following steps: extracting a risk overspeed event data set based on track space-time matching analysis; constructing a risk over-speed event multi-dimensional feature vector, and obtaining historical data of the risk over-speed event multi-dimensional feature vector corresponding to the risk over-speed event data set; constructing a risk over-speed prediction model, and training the risk over-speed prediction model; and obtaining a risk overspeed behavior prediction result based on the real-time data of the risk overspeed event multi-dimensional feature vector and the trained risk overspeed prediction model, and then executing hierarchical early warning operation. According to the invention, through integrating multi-dimensional data such as the vehicle track data, the static traffic characteristics, the dynamic traffic characteristics and the environmental factors, the risk overspeed behavior is accurately identified, and graded early warning is carried out according to the risk overspeed index, so that the traffic safety management level is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent traffic management, and particularly to a method for identifying and warning risk over-speed behaviors based on trajectory spatio-temporal matching analysis. Background Art

[0002] With the acceleration of the urbanization process, traffic congestion and traffic accidents occur frequently, especially the traffic accidents caused by vehicle over-speed behaviors are increasing day by day. Traditional traffic management methods mainly rely on devices such as traffic monitoring cameras and radar speed measurement, but these methods have problems such as limited coverage and insufficient real-time performance. In recent years, with the development of vehicle networking and big data technologies, the acquisition and analysis of vehicle trajectory data have become possible, providing new ideas for traffic management.

[0003] However, most of the existing technologies only rely on single-dimensional data analysis, lacking the comprehensive analysis and processing of multi-dimensional data such as vehicle trajectories, traffic environments, and road characteristics, resulting in low recognition accuracy of risk over-speed behaviors and unsatisfactory warning effects. Therefore, there is an urgent need for a technical solution that can comprehensively analyze multi-dimensional data, accurately identify risk over-speed behaviors, and give effective warnings. Summary of the Invention

[0004] In view of the above deficiencies in the prior art, the present invention provides a method for identifying and warning risk over-speed behaviors based on trajectory spatio-temporal matching analysis. By comprehensively analyzing multi-dimensional data such as vehicle trajectory data, static traffic characteristics, dynamic traffic characteristics, and environmental factors, it accurately identifies risk over-speed behaviors and conducts hierarchical warnings according to the risk over-speed index, thereby improving the level of traffic safety management.

[0005] In order to achieve the above invention purpose, the technical solution adopted by the present invention is as follows:

[0006] A method for identifying and warning risk over-speed behaviors based on trajectory spatio-temporal matching analysis, comprising the following steps:

[0007] S1. Extract a risk over-speed event data set based on trajectory spatio-temporal matching analysis;

[0008] S2. Construct a multi-dimensional feature vector of risk over-speed events, including static traffic characteristics, dynamic traffic characteristics, event environment impact factors, over-speed trajectory attributes, and behavior characteristics, and obtain historical data of the multi-dimensional feature vector of risk over-speed events corresponding to the risk over-speed event data set;

[0009] S3. Construct a risk over-speed prediction model, train the risk over-speed model based on the risk over-speed event data and the historical data of the multi-dimensional feature vector of risk over-speed events corresponding to the risk over-speed event data set, and obtain a trained risk over-speed prediction model;

[0010] S4. Obtain the real-time data of the multi-dimensional feature vector of the risk over-speed event, and based on the real-time data of the multi-dimensional feature vector of the risk over-speed event and the trained risk over-speed prediction model, obtain the risk over-speed behavior prediction result. The risk over-speed behavior prediction result includes the over-speed determination prediction result and the risk over-speed index prediction result. Perform a hierarchical early warning operation according to the risk over-speed index prediction result.

[0011] Further, step S1 includes the following steps:

[0012] S11. Construct a trajectory database in the neighborhood of the accident location based on trajectory spatio-temporal matching analysis;

[0013] S12. Identify risk over-speed events and calculate the risk over-speed index for the trajectory database in the neighborhood of the accident location to extract the risk over-speed event data set.

[0014] Further, step S11 includes the following steps:

[0015] S111. Extract the accident time, accident location, and matching vehicle trajectory data from the set of vehicle trajectories in the accident record;

[0016] S112. Determine the adjacent time period of the accident occurrence and the exact accident occurrence location based on the accident time, accident location, and matching vehicle trajectory data;

[0017] S113. Extract the trajectory points within 300 meters before the accident based on the exact accident occurrence location, and calculate the estimated driving speed before the accident based on the adjacent time period of the accident occurrence and the distance between the first and last trajectory points within 300 meters before the accident;

[0018] S114. Traverse the entire set of vehicle trajectories, extract the vehicle trajectory data that spatio-temporally matches the accident, and calculate the relative estimated driving speed to construct the trajectory database in the neighborhood of the accident location.

[0019] Further, in step S12, the calculation of the risk over-speed index for the trajectory database in the neighborhood of the accident location is expressed as:

[0020] indi_v(a) = s(a) * avg_v(a) / (v_accident - avg_v(a))

[0021] Where: indi_v(a) is the risk over-speed index of accident a, s(a) is the severity parameter of accident a, avg_v(a) is the average vehicle speed in the neighborhood of accident a, and v_accident is the estimated driving speed before the accident.

[0022] Further, in step S2, the static traffic features include adjacent road attributes and road alignment geometric parameters, a total of 7 variables; the adjacent road attributes include road type, road grade, number of lanes, and design speed, a total of 4 variables; the adjacent road alignment geometric parameters include slope, radian, and turning radius, a total of 3 variables;

[0023] The dynamic traffic features include traffic flow level and the vehicle types passing by and their corresponding average vehicle speeds, a total of 10 variables; the traffic flow level is 1 aggregated variable; the vehicle types passing by and their corresponding average vehicle speeds include the average vehicle speeds of 9 vehicle types, a total of 9 variables;

[0024] The event environment impact factors include sunny days, cloudy days, rainy days, snowy days, foggy days, strong winds, visibility, temperature, humidity, and road construction conditions, a total of 10 variables;

[0025] The attributes and behavioral characteristics of the overspeed trajectory include the accident vehicle type, the continuous driving duration before the accident, the driver attributes, and the observed vehicle speed list within 300 meters before the accident.

[0026] Further, in step S3, the risk overspeed prediction model includes:

[0027] A structured feature vector processing module, which is used to elevate the 27-dimensional feature vector composed of static traffic features, dynamic traffic features, and event environment impact factors to a 32-dimensional hidden vector;

[0028] A convolution module, which is used to process the attributes and behavioral characteristics of the overspeed trajectory to obtain a 32-dimensional hidden vector with the inter-vehicle interaction relationship;

[0029] A vector splicing module, which is used to splice the 32-dimensional hidden vector output by the structured feature vector processing module and the 32-dimensional hidden vector output by the convolution module to obtain a 64-dimensional feature vector;

[0030] An information compression module, which is used to compress the 64-dimensional feature vector through multiple layers of Dense-dropout units to obtain a 32-dimensional feature latent vector;

[0031] A risk overspeed behavior prediction module, which is used to map the 32-dimensional data to a 2-dimensional probability distribution, judge whether it is overspeed to obtain the overspeed determination prediction result, and directly obtain the risk overspeed index prediction result.

[0032] Further, in the convolution module, 8 convolution kernels with a size of 3×3 and a sliding step of 1 are stacked in three layers to process the attributes and behavioral characteristics of the overspeed trajectory.

[0033] Further, the multi-layer Dense-dropout unit includes 5 Dense layers connected in sequence. The number of neurons in the first Dense layer is set to 128, the number of neurons in the second Dense layer is set to 256, the number of neurons in the third Dense layer is set to 128, the number of neurons in the fourth Dense layer is set to 64, and the number of neurons in the fifth Dense layer is set to 32.

[0034] Further, in step S4, a hierarchical early warning operation is performed according to the risk over-speed index prediction result. Specifically: judge the magnitude of the risk over-speed index prediction result; if the risk over-speed index is in the latter 50%, a yellow early warning is sent; if the risk over-speed index is between the top 50% and the top 20%, an orange early warning is sent and neighboring vehicles are informed; if the risk over-speed index is in the top 20%, a red early warning is sent and the management department is reported for intervention.

[0035] The present invention has the following beneficial effects:

[0036] (1) By integrating multi-dimensional data such as vehicle trajectory data, static traffic characteristics, dynamic traffic characteristics, and environmental factors, the present invention can more accurately identify risk over-speed behaviors;

[0037] (2) By calculating the risk over-speed index, the present invention can quantify the risk level of over-speed behaviors and provide a scientific basis for early warning management;

[0038] (3) The present invention can analyze vehicle trajectory data in real time and perform hierarchical early warning according to the risk over-speed index, effectively improving the level of traffic safety management;

[0039] (4) By constructing a risk over-speed prediction model including a structured feature vector processing module, a convolutional module, a vector splicing module, an information compression module, and a risk over-speed behavior prediction module, the present invention can automatically extract complex features in vehicle trajectories and improve the accuracy and efficiency of risk over-speed behavior recognition. Description of the Drawings

[0040] Figure 1 It is a schematic flow diagram of a method for identifying and warning risk over-speed behaviors based on trajectory spatio-temporal matching analysis;

[0041] Figure 2 It is a schematic structural diagram of a risk over-speed prediction model. Detailed Embodiments

[0042] The following describes the specific embodiments of the present invention to facilitate those skilled in the art of this technology to understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific embodiments. For those of ordinary skill in the art of this technology, as long as various changes are within the spirit and scope of the present invention defined and determined by the appended claims, these changes are obvious, and all inventions created using the concept of the present invention are within the scope of protection.

[0043] As Figure 1 shown, the risk over-speed behavior recognition and warning method based on trajectory spatio-temporal matching analysis includes steps S1 - S4, specifically as follows:

[0044] S1. Extract the risk over-speed event data set based on trajectory spatio-temporal matching analysis.

[0045] In an optional embodiment of the present invention, step S1 includes the following steps:

[0046] S11. Construct a neighborhood trajectory database of the accident location based on trajectory spatio-temporal matching analysis.

[0047] Step S11 includes the following steps:

[0048] S111. Extract the accident time, accident location, and matching vehicle trajectory data from the vehicle trajectory set of the accident record.

[0049] The vehicle trajectory set of the accident record is A_Cars(A, C, G), where A is the accident record list, C is the accident matching vehicle list, and G is the set of matching vehicle trajectory data. The present invention traverses the accident records a in the accident record list, within a range of 5 minutes before and after the accident record time, matches the vehicle c from the accident record a, and then extracts the trajectory data of the matching vehicle c, denoted as g(a, c).

[0050] S112. Determine the adjacent time period of the accident occurrence and the exact accident occurrence location based on the accident time, accident location, and matching vehicle trajectory data.

[0051] Based on the accident time, accident location, and matching vehicle trajectory data, sequentially traverse each trajectory point in the trajectory data g(a, c) of the matching vehicle c, and obtain the zero-speed moment t_accident_real closest to the accident record occurrence time as the exact accident occurrence moment. Five minutes before the exact accident occurrence moment is the corresponding adjacent time period t_seg of the accident occurrence, and the longitude and latitude coordinates c_accident_real corresponding to the exact accident occurrence moment are used as the exact accident occurrence location.

[0052] S113. Extract the trajectory points within 300 meters before the accident based on the exact accident occurrence location, and calculate the estimated driving speed before the accident based on the adjacent time period of the accident occurrence and the distance between the first and last trajectory points within 300 meters before the accident.

[0053] Extract the list of trajectory points Tr_pre(a, c) within 300 meters before the trajectory reaches the exact accident occurrence location based on the exact accident occurrence location, and then, based on the adjacent time period of the accident occurrence, calculate the ratio of the distance between the head and tail trajectory points of this list to the time to calculate the estimated driving speed v_accident before the accident.

[0054] S114. Traverse the entire vehicle trajectory set, extract the vehicle trajectory data that matches the accident in terms of time and space, and calculate the relative estimated driving speed to construct a trajectory database in the neighborhood of the accident location.

[0055] For the entire vehicle trajectory set T_Cars(C_total, G_total), where C_total is the list of all vehicles and G_total is the list of corresponding trajectory data for all vehicles, traverse each vehicle c in C_total, and for its trajectory set g(c), obtain the subset of trajectory points g_S(c, A_m) that is consistent with the time range of A_T_seg, where A_m is the subset of the corresponding accident record list for vehicle c that is consistent with the time range of A_T_seg. For example, if an accident a_m in A_T_seg occurs during the time period t1 - t2 and vehicle c has a set of trajectory records g'(c) during this time period, then add this set to g_S(c, A_m) and add the accident a_m record to A_m;

[0056] If g_S(c, A_m) is an empty set, skip to the next vehicle. Otherwise, traverse each accident record a_m in A_m, search within a range of 500 meters adjacent to the exact occurrence location c_accident_real_m of a_m. If there is trajectory data with the same driving direction (the included angle of the forward direction is less than 90 degrees) in the corresponding vehicle trajectory record set g'(c) within this range, it can be considered that the trajectory data set g(c, a_m) of the vehicle within the corresponding range has a spatio-temporal matching relationship with the accident record a_m. At the same time, calculate the ratio of the distance between the head and tail trajectory points in g(c, a_m) to the time to calculate the estimated driving speed v_related(c, a_m) before the accident a_m.

[0057] After traversing T_Cars(C_total, G_total), the set of spatio-temporally matching vehicles M_C, the matching trajectory set M_G, and the set of estimated driving speeds before the accident M_V for all accidents can be obtained. Construct the set of spatio-temporally matching vehicles M_C, the matching trajectory set M_G, and the set of estimated driving speeds before the accident M_V for all accidents together as a trajectory database in the neighborhood of the accident location.

[0058] S12. Identify risk over-speed events and calculate the risk over-speed index for the trajectory database in the neighborhood of the accident location to extract the risk over-speed event data set.

[0059] The present invention calculates a risk over - speed index for the trajectory database in the neighborhood of the accident location, expressed as:

[0060] indi_v(a) = s(a)*avg_v(a) / (v_accident - avg_v(a))

[0061] Where: indi_v(a) is the risk over - speed index of accident a, s(a) is the severity parameter of accident a, avg_v(a) is the average vehicle speed in the neighborhood of accident a, and v_accident is the estimated driving speed before the accident.

[0062] S2. Construct a multi - dimensional feature vector for risk over - speed events, including static traffic features, dynamic traffic features, event environment impact factors, over - speed trajectory attributes, and behavior features, and obtain historical data of the multi - dimensional feature vector of risk over - speed events corresponding to the risk over - speed event dataset.

[0063] In an alternative embodiment of the present invention, the static traffic features include adjacent road attributes and road alignment geometric parameters, a total of 7 variables; the adjacent road attributes include road type (such as bridge, tunnel, highway, urban road), road grade, number of lanes, and design speed, a total of 4 variables; the adjacent road alignment geometric parameters include slope, radian, and turning radius, a total of 3 variables;

[0064] The dynamic traffic features include traffic flow level and vehicle types passing by and their corresponding average vehicle speeds, a total of 10 variables; the traffic flow level is 1 aggregated variable; the vehicle types passing by and their corresponding average vehicle speeds include the average vehicle speeds of 9 vehicle types (mini - car, small sedan, medium sedan, SUV, large bus, small truck, medium truck, large truck, special vehicle), a total of 9 variables;

[0065] The event environment impact factors include sunny, cloudy, rainy, snowy, foggy, windy, visibility, temperature, humidity, and road construction conditions, a total of 10 variables;

[0066] The over - speed trajectory attributes and behavior features include accident vehicle type, continuous driving duration before the accident, driver attributes, and a list of observed vehicle speeds within 300 meters before the accident.

[0067] Specifically, the present invention includes four features: static traffic features, dynamic traffic features, event environment influencing factors, and over-speed trajectory attributes and behavior characteristics. All four features are stored in matrix form. The method of converting to a matrix is as follows: First, taking the accident vehicle as the center, divide the space within 300 m before the accident (in the tail direction) into grids of 1 m each. With a lane width of 3.75 m and 4 lanes, for example, the entire lane is 300 m long and 15 m wide, forming a matrix A of 300 * 15. The present invention fills the corresponding vehicle IDs into matrix A according to the trajectory data; in the second step, the corresponding attribute values of the vehicle IDs are filled into the corresponding positions of the matrix to form an attribute tensor T. If there are 4 attributes, a tensor of 4 * 300 * 15 is formed.

[0068] S3. Construct a risk over-speed prediction model, and train the risk over-speed model based on the risk over-speed event data and the historical data of the risk over-speed event multi-dimensional feature vectors corresponding to the risk over-speed event dataset to obtain a trained risk over-speed prediction model.

[0069] In an alternative embodiment of the present invention, the risk over-speed prediction model includes a structured feature vector processing module, a convolution module, a vector splicing module, an information compression module, and a risk over-speed behavior prediction module, as Figure 2 shown.

[0070] The structured feature vector processing module is used to elevate the 27-dimensional feature vector composed of static traffic features, dynamic traffic features, and event environment influencing factors to a 32-dimensional hidden vector.

[0071] The convolution module is used to process the over-speed trajectory attributes and behavior characteristics to obtain a 32-dimensional hidden vector with the mutual influence relationship between vehicles. In the convolution module, 8 convolution kernels of size 3×3 and a sliding step of 1 (8, 3, 3, 1) are stacked in three layers to process the over-speed trajectory attributes and behavior characteristics.

[0072] The vector splicing module is used to splice the 32-dimensional hidden vector output by the structured feature vector processing module and the 32-dimensional hidden vector output by the convolution module to obtain a 64-dimensional feature vector.

[0073] The information compression module is used to compress the 64-dimensional feature vector through multiple layers of Dense-dropout units to obtain a 32-dimensional feature latent vector. The multiple layers of Dense-dropout units include 5 sequentially connected Dense layers. The number of neurons in the first Dense layer is set to 128, the number of neurons in the second Dense layer is set to 256, the number of neurons in the third Dense layer is set to 128, the number of neurons in the fourth Dense layer is set to 64, and the number of neurons in the fifth Dense layer is set to 32.

[0074] A risk over-speed behavior prediction module is used to map 32-dimensional data to a 2D probability distribution, determine whether it is over-speed to obtain an over-speed determination prediction result, and directly obtain a risk over-speed index prediction result. Specifically, in the risk over-speed behavior prediction module, a dense layer with sigmoid as the activation function is used to map 32-dimensional latent variables to a 1D probability, and this probability value is used as the prediction result for determining whether it is over-speed; in the risk over-speed behavior prediction module, a 1D dense layer with ReLU as the activation function is used to obtain the risk over-speed index prediction result.

[0075] Specifically, during the training process of the risk over-speed model in the present invention, the risk over-speed model simultaneously generates a binary classification result of whether it is over-speed and a regression numerical result of the over-speed index. Therefore, two loss functions are used for training; for the classification prediction head, binary cross-entropy is used as the loss function; for the regression prediction head, root mean square error is used as the loss function.

[0076] S4. Obtain the real-time data of the multi-dimensional feature vector of the risk over-speed event, and based on the real-time data of the multi-dimensional feature vector of the risk over-speed event and the trained risk over-speed prediction model, obtain the risk over-speed behavior prediction result. The risk over-speed behavior prediction result includes an over-speed determination prediction result and a risk over-speed index prediction result, and perform a hierarchical early warning operation according to the risk over-speed index prediction result.

[0077] In an optional embodiment of the present invention, the road network space is divided into grids of 500m×500m, the vehicle trajectory data in each grid is dynamically collected, and real-time analysis is carried out in combination with static traffic characteristics and event environment influence factors. Then, for any vehicle, every time it passes through a grid, the over-speed trajectory attributes and behavior characteristics in its previous grid are dynamically analyzed, and in combination with the static traffic characteristics, dynamic traffic characteristics and environmental factor characteristics of the current area, the trained risk over-speed prediction model is used to obtain the risk over-speed behavior prediction result.

[0078] The present invention performs a hierarchical early warning operation according to the risk over-speed index prediction result, specifically: judge the size of the risk over-speed index prediction result; if the risk over-speed index is in the latter 50%, send a yellow early warning; if the risk over-speed index is between the former 50% and the former 20%, send an orange early warning and inform adjacent vehicles; if the risk over-speed index is in the former 20%, send a red early warning and report to the management department for intervention.

[0079] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to embodiments of the present invention. It should be understood that each flow and / or block in the flowchart and / or block diagram, as well as the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate a means for implementing the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or a means for implementing the functions specified in multiple blocks.

[0080] These computer program instructions can 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, such that the instructions stored in the computer-readable memory generate a manufactured article including an instruction means that implements the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or a means for implementing the functions specified in multiple blocks.

[0081] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operating steps are executed on the computer or other programmable device to generate a computer-implemented process, so that the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or a means for implementing the functions specified in multiple blocks.

[0082] Specific embodiments are applied in the present invention to elaborate on the principles and implementation manners of the present invention. The descriptions of the above embodiments are only used to help understand the method of the present invention and its core idea; at the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation manners and application scopes. In summary, the content of this specification should not be construed as a limitation to the present invention.

[0083] Those of ordinary skill in the art will realize that the embodiments described herein are for helping readers understand the principles of the present invention, and it should be understood that the protection scope of the present invention is not limited to such specific statements and embodiments. Those of ordinary skill in the art can make various other specific deformations and combinations that do not depart from the essence of the present invention based on the technical revelations disclosed in the present invention, and these deformations and combinations are still within the protection scope of the present invention.

Claims

1. A risk overspeeding behavior identification and early warning method based on trajectory spatiotemporal matching analysis is characterized by: The following steps are involved: S1, extract risk overspeeding event data set based on trajectory spatiotemporal matching analysis; S2. Construct a multidimensional feature vector of risky speeding events, including static traffic characteristics, dynamic traffic characteristics, event environment influencing factors, speeding trajectory attributes and behavior characteristics, and obtain historical data of the multidimensional feature vector of risky speeding events corresponding to the risky speeding event data set; S3, constructing a risk overspeed prediction model, training the risk overspeed model based on the risk overspeed event data and the historical data of the multi-dimensional feature vector of the risk overspeed event corresponding to the risk overspeed event data set, and obtaining the trained risk overspeed prediction model; S4. Obtain real-time data of the multidimensional feature vector of risky speeding events, and obtain risky speeding behavior prediction results based on the real-time data of the multidimensional feature vector of risky speeding events and the trained risky speeding prediction model. The risky speeding behavior prediction results include speeding determination prediction results and risky speeding index prediction results. Perform graded warning operations based on the risky speeding index prediction results.

2. The risk overspeeding behavior identification and early warning method based on trajectory spatiotemporal matching analysis according to claim 1 is characterized in that: Step S1 includes the following steps: S11. Constructing the neighborhood trajectory database of the accident location based on the spatiotemporal matching analysis of trajectories; S12. Identify risky speeding events and calculate risky speeding indexes on the accident location neighborhood trajectory database to extract a risky speeding event data set.

3. The risk overspeeding behavior identification and early warning method based on trajectory spatiotemporal matching analysis according to claim 2 is characterized in that: Step S11 includes the following steps: S111, extracting accident time, accident location and matching vehicle trajectory data from the vehicle trajectory set of the accident record; S112, determining the time period near the accident and the precise location of the accident based on the accident time, accident location and matching vehicle trajectory data; S113, extracting trajectory points within 300 meters before the accident based on the precise location of the accident, and calculating the estimated driving speed before the accident based on the time period near the accident and the distance between the first and last trajectory points within 300 meters before the accident; S114, traversing all vehicle trajectory sets, extracting vehicle trajectory data that matches the accident time and space, and calculating the relative estimated driving speed to construct an accident location neighborhood trajectory database.

4. The risk overspeeding behavior identification and early warning method based on trajectory spatiotemporal matching analysis according to claim 2 is characterized in that: In step S12, the risk overspeed index is calculated for the accident location neighborhood trajectory database, which is expressed as: indi_v(a)=s(a)*avg_v(a) / (v_accident-avg_v(a)) Where indi_v(a) is the risk speeding index of accident a, s(a) is the severity parameter of accident a, avg_v(a) is the average speed of the neighborhood of accident a, and v_accident is the estimated speed before the accident.

5. The risk overspeeding behavior identification and early warning method based on trajectory spatiotemporal matching analysis according to claim 1 is characterized in that: In step S2, the static traffic characteristics include adjacent road attributes and road linear geometry parameters, a total of 7 variables; adjacent road attributes include road type, road grade, number of lanes, design speed, a total of 4 variables; adjacent road linear geometry parameters include slope, curvature, turning radius, a total of 3 variables; Dynamic traffic characteristics include traffic flow level and passing vehicle types and corresponding average speeds, a total of 10 variables; traffic flow level is a cluster variable; passing vehicle types and corresponding average speeds include the average speeds of 9 vehicle types, a total of 9 variables; The factors affecting the event environment include sunny days, cloudy days, rainy days, snowy days, foggy days, strong winds, visibility, temperature, humidity and road construction conditions, a total of 10 variables; The speeding trajectory attributes and behavior characteristics include the accident vehicle type, continuous driving time before the accident, driver attributes, and a list of observed vehicle speeds within 300 meters before the accident.

6. The risk overspeeding behavior identification and early warning method based on trajectory spatiotemporal matching analysis according to claim 1 is characterized in that: In step S3, the risk overspeed prediction model include: The structured feature vector processing module is used to upgrade the 27-dimensional feature vector composed of static traffic characteristics, dynamic traffic characteristics and event environment influencing factors to a 32-dimensional latent vector; Convolution module, used to process the speeding trajectory attributes and behavior characteristics to obtain a 32-dimensional latent vector with the mutual influence relationship between vehicles; A vector concatenation module is used to concatenate the 32-dimensional latent vector output by the structured feature vector processing module and the 32-dimensional latent vector output by the convolution module to obtain a 64-dimensional feature vector; The information compression module is used to compress the 64-dimensional feature vector through multi-layer Dense-dropout units to obtain a 32-dimensional feature latent vector; The risk overspeeding behavior prediction module is used to map 32-dimensional data to a 2-dimensional probability distribution, determine whether it is overspeeding to obtain the overspeeding judgment prediction result, and directly obtain the risk overspeeding index prediction result.

7. The risk overspeeding behavior identification and early warning method based on trajectory spatiotemporal matching analysis according to claim 6 is characterized in that: In the convolution module, eight convolution kernels of size 3×3 and sliding step size 1 are stacked in three layers to process the attributes and behavioral characteristics of the speeding trajectory.

8. The method for identifying and warning risky speeding behaviors based on trajectory spatiotemporal matching analysis according to claim 6 is characterized in that: The multi-layer Dense-dropout unit includes 5 Dense layers connected in sequence, the number of neurons in the first Dense layer is set to 128, the number of neurons in the second Dense layer is set to 256, the number of neurons in the third Dense layer is set to 128, the number of neurons in the fourth Dense layer is set to 64, and the number of neurons in the fifth Dense layer is set to 32.

9. The risk overspeeding behavior identification and early warning method based on trajectory time-space matching analysis according to claim 1 is characterized in that: In step S4, a graded warning operation is performed according to the risk speeding index prediction result, specifically: the size of the risk speeding index prediction result is determined; if the risk speeding index is in the bottom 50%, a yellow warning is sent; if the risk speeding index is between the top 50% and the top 20%, an orange warning is sent and neighboring vehicles are informed; if the risk speeding index is in the top 20%, a red warning is sent and reported to the management department for intervention.

Images