Method and device for dynamically evaluating degree of impact of highway traffic accident

Abstract

The present application belongs to the technical field of highway traffic control. Disclosed in the present application are a method and device for dynamically evaluating the degree of impact of a highway traffic accident, which method and device are used for solving the existing technical problem of after a traffic accident occurs on a highway, transportation efficiency being prone to being subjected to severe negative impacts caused by it being difficult to dynamically predict the degree of impact of the highway traffic accident, which is not conducive to effectively estimating the severity of the accident. The method comprises: determining a local highway network of an accident site; performing index fusion on preset structural indexes and functional indexes, and performing continuous equalization processing on the severity of a highway traffic accident on the basis of comprehensive indexes obtained after fusion, so as to obtain an accident severity evaluation system for the highway traffic accident; and performing data fusion on information of the natural evolution patterns of traffic flows, information of the severity of the accident at the moment when the accident occurs, and information of abrupt changes in the traffic flows that are induced by the accident, so as to obtain an accident impact factor system of the highway traffic accident.

Description

A method and device for dynamically evaluating influence degree of highway traffic accident

[0001] The present application claims priority to the Chinese patent application No. 202410822196.6, filed on June 25, 2024, and entitled "A method and device for dynamically evaluating influence degree of highway traffic accident", the whole content of which is incorporated herein by reference. TECHNICAL FIELD

[0002] The present application relates to the field of highway traffic control, in particular to a method and device for dynamically evaluating influence degree of highway traffic accident. BACKGROUND

[0003] In recent years, the highway in China has developed rapidly, with the total mileage ranking first in the world, effectively promoting the development of economy and society, and bringing great convenience to people's travel. However, with the increasing demand for travel and the complexity of highway network structure, more and more frequent traffic accidents on highways have followed. When a traffic accident occurs on a highway section, part of the lanes are occupied by accident vehicles, and the traffic capacity of the whole section is sharply reduced, which can easily cause large-scale traffic congestion upstream of the accident site and secondary traffic accidents, and further seriously affect the traffic efficiency of the whole highway network.

[0004] After a traffic accident occurs on a highway, it is difficult to complete the accident responsibility identification and traffic relief in a very short time. Therefore, the influence time of the traffic accident on the highway is relatively longer than that on urban roads. In this case, how to effectively evaluate the medium and long-term impact of the accident on highway transportation has important reference value for the emergency management of traffic police and highway management personnel. SUMMARY

[0005] The embodiments of the present application provide a method and device for dynamically evaluating influence degree of highway traffic accident, which are used to solve the following technical problems: after a traffic accident occurs on an existing highway, it is difficult to dynamically predict the influence degree of the highway traffic accident, which is not conducive to effectively estimating the severity of the accident and can easily cause serious negative impact on traffic efficiency.

[0006] The embodiments of the present application adopt the following technical solutions:

[0007] In one aspect, the embodiments of the present application provide a method for dynamically evaluating the influence degree of a highway traffic accident, comprising: dividing a local sub-network of a highway traffic network according to the location information of a highway traffic accident to determine a local highway network of an accident point; evaluating a functional index based on the time of occurrence of the accident in the local highway network of the accident point and the number of highway gantries; fusing a preset structural index and the functional index, and performing continuous and balanced processing on the severity of the highway traffic accident according to the fused comprehensive index to obtain an accident severity evaluation system of the highway traffic accident; extracting traffic characteristic features from a traffic flow sequence set at a time when the accident did not occur to obtain traffic flow natural evolution law information; fusing the traffic flow natural evolution law information with the severity information of the accident itself at the time of occurrence of the accident and the accident-induced traffic flow mutation information to obtain an accident influence factor system of the highway traffic accident; and dynamically predicting the influence degree of the highway traffic accident according to the accident severity evaluation system and the accident influence factor system to obtain dynamic influence evaluation information based on the highway traffic accident.

[0008] By dividing the local highway network of the accident point and evaluating the functional index, the embodiments of the present application can more accurately evaluate the severity of the highway traffic accident. This helps the relevant departments to take appropriate rescue and response measures in a timely manner, improves the efficiency and safety of accident handling. This helps better understand the causes and effects of the accident, providing a basis for developing improvement measures. It can also more objectively compare the severity of different accidents, providing more valuable information for decision-making. It helps to better understand the occurrence mechanism and influence range of the accident, providing guidance for prevention and reduction of accidents. At the same time, it can timely understand the impact of the accident on traffic and take corresponding traffic management measures to reduce the impact of the accident on traffic and improve the efficiency and safety of traffic operation. It is beneficial to develop reasonable emergency plans, resource allocation and traffic management strategies based on the prediction results to minimize the loss and impact of the accident.

[0009] In a feasible implementation, before the local sub-network of the expressway traffic network is divided according to the location information of the expressway traffic accident, the method further comprises: determining the expressway traffic network based on a set of expressway network nodes and a set of connections between the expressway network nodes, wherein the set of expressway network nodes at least includes expressway toll stations and interchanges; evaluating the similarity of the shortest distance between any two expressway network nodes in the set of connections to obtain the static similarity of the expressway network nodes; identifying all road segment sets connected to a single expressway network node in the set of connections, and collecting the connected road segment flow sequence in the historical time period of all connected road segment sets; according to obtaining the flow sequence feature seq_v i of the expressway network node i; wherein seq_e ik is the connected road segment flow sequence; e ik is all road segments connected to the expressway network node i; K is the number of other expressway network nodes connected to the expressway network node i, and k is the other expressway network node; according to obtaining the dynamic similarity p ij of the expressway network node; wherein cov(seq_v i , seq_v j ) represents the covariance of the flow sequence feature seq_v i of the expressway network node i and the flow sequence feature seq_v j of the expressway network node j; and and respectively represent the standard deviation of the flow sequence feature seq_v i of the expressway network node i and the flow sequence feature seq_v j of the expressway network node j; according to sim ij = a1·d ij + b1·p ij , the node similarity between the expressway network node i and the expressway network node j is obtained; wherein a1 is the weight of the static similarity; b1 is the weight of the dynamic similarity; based on the expressway traffic network, the node similarity matrix is constructed to obtain the node similarity matrix.

[0010] In an embodiment, the highway traffic network is divided into local subnets according to the location information of the highway traffic accident, and a local highway network at the accident point is determined, which specifically includes: based on the point similarity matrix and through a preset Louvain algorithm, the highway traffic network is divided into subnets to determine a plurality of highway subnets; the location information of the highway traffic accident is extracted; according to the location information, the plurality of highway subnets are matched in position to determine the local highway network at the accident point where the highway traffic accident occurs; wherein the local highway network at the accident point includes: a local highway network node set, a road segment set between local highway network nodes, and a local highway network node number.

[0011] In an embodiment, before evaluating the functional index based on the time of accident occurrence and the number of highway gantries in the local highway network at the accident point, the method further includes: obtaining the traffic efficiency E of the local highway network at the accident point local ; wherein d ij represents the number of edges on the shortest path between the i th node and the j th node in the local highway network at the accident point; G is the number of nodes in the local highway network at the accident point; according to the time of accident occurrence of the highway traffic accident, the traffic efficiency is divided into a first traffic efficiency before the time of accident occurrence and a second traffic efficiency after the time of accident occurrence; based on the first traffic efficiency and the second traffic efficiency, the structural index is constructed.

[0012] In an embodiment, the functional index is evaluated based on the time of accident occurrence and the number of highway gantries in the local highway network at the accident point, which specifically includes: obtaining the functional index F local (t) after the time of accident occurrence; wherein M is the number of highway gantries in the local highway network at the accident point; m is a highway gantry; v m_t is the average of the instantaneous speed of all passing vehicles within a time interval centered at time t after the time of accident occurrence for the highway gantry m; v m_0 is the average speed of the highway gantry m before the time of accident occurrence.

[0013] In an embodiment, the preset structural index and the functional index are fused, and the severity of the highway traffic accident is continuously and uniformly processed according to the fused comprehensive index to obtain an accident severity evaluation system for the highway traffic accident, which specifically includes: according to S local (t) = a2·Elocal + β2· F (t) local (t), to obtain the comprehensive index; wherein, E local is the structural index; F local (t) is the functional index; α2 is the weight of the structural index; β2 is the weight of the functional index; according to obtain the accident severity evaluation system L; wherein, T accident is the accident influence time; and, S local (0) = α2· E local_bef + β2· F local (0); and, S local (t) = α2· E local_aft + β2· F local (t); E local_bef is the first traffic efficiency before the accident occurrence moment; E local_aft is the second traffic efficiency after the accident occurrence moment.

[0014] In a feasible implementation, the traffic flow sequence set at the accident non-occurrence moment in the highway traffic accident is subjected to traffic characteristic feature extraction processing to obtain traffic flow natural evolution law information, specifically including: passing through the high-speed gantry and based on the accident non-occurrence moment of the highway traffic accident, a first traffic flow sequence set of the first traffic flow sequence set of the accident point position local highway network connected with the highway traffic accident is obtained; based on the first traffic flow sequence set, a first traffic flow sequence length is determined; through a preset time convolution network, multi-level sequence feature extraction is performed on the first traffic flow sequence length, and traffic flow feature information is output; and the traffic flow feature information is determined as the traffic flow natural evolution law information.

[0015] In a feasible implementation, the traffic flow natural evolution law information is fused with the accident severity information at the time of the accident and the accident-induced traffic flow mutation information to obtain an accident influencing factor system of the highway traffic accident, specifically including: acquiring, by a UAV and based on the time of the accident, a second traffic flow sequence set of a road section connected to the highway traffic accident in a local highway network at the accident point; performing multi-level sequence feature extraction on the second traffic flow sequence set by a preset time convolution network to determine the accident-induced traffic flow mutation information; collecting the accident severity information in the highway traffic accident; the accident severity information includes the number of injured persons, the number of on-site deaths, the number of involved vehicles, and the accident lane occupancy rate; performing information set processing on the traffic flow natural evolution law information, the accident severity information, and the accident-induced traffic flow mutation information to obtain an accident influencing factor set; and constructing the accident influencing factor system based on the accident influencing factor set.

[0016] In a feasible implementation, the accident influencing degree of the highway traffic accident is dynamically predicted based on the accident severity evaluation system and the accident influencing factor system to obtain dynamic influence evaluation information based on the highway traffic accident, specifically including: constructing a traffic accident influencing degree dynamic prediction model based on a neural network technology and a channel attention model; the input of the traffic accident influencing degree dynamic prediction model is the accident influencing factor system, and the output is the accident severity evaluation system; model training of the traffic accident influencing degree dynamic prediction model is completed by using a historical accident influencing factor set in the accident influencing factor system and a historical accident severity evaluation set in the accident severity evaluation system; and current highway traffic accident information is input into the traffic accident influencing degree dynamic prediction model to obtain the dynamic influence evaluation information.

[0017] In another aspect, the embodiments of the present application also provide a highway traffic accident influencing degree dynamic evaluation device, which includes at least one processor and a memory in communication connection with the at least one processor; the memory stores instructions executable by the at least one processor to enable the at least one processor to execute the highway traffic accident influencing degree dynamic evaluation method described in any of the above embodiments.

[0018] The embodiments of the present application provide a highway traffic accident influencing degree dynamic evaluation method and device, and compared with the prior art, the embodiments of the present application have the following beneficial technical effects:

[0019] 1. Accident severity evaluation: By dividing the local highway network at the accident point and evaluating the functional indicators, the severity of the highway traffic accident can be more accurately evaluated. This helps relevant departments to take appropriate rescue and response measures in a timely manner, improving the efficiency and safety of accident handling.

[0020] 2. Comprehensive index fusion: The fusion of structural indicators and functional indicators can comprehensively consider the physical structure and traffic operation of the highway, providing a more comprehensive evaluation system. This helps better understand the causes and effects of accidents and provides a basis for developing improvement measures.

[0021] 3. Continuous equalization processing: Continuous equalization processing of accident severity can reduce the dispersion of evaluation results, improve the accuracy and reliability of evaluation. This helps more objectively compare the severity of different accidents and provides more valuable information for decision-making.

[0022] 4. Accident impact factor analysis: By extracting the information of traffic flow natural evolution law and fusing it with the information of accident severity and traffic flow mutation induced by accident, the impact factors of highway traffic accident can be more comprehensively analyzed. This helps to understand the occurrence mechanism and influence range of the accident, and provides guidance for prevention and reduction of accidents.

[0023] 5. Dynamic prediction: Based on the accident severity evaluation system and the accident impact factor system, the dynamic prediction of the impact of traffic accidents can timely understand the impact of accidents on traffic and take corresponding traffic management measures to reduce the impact of accidents on traffic and improve the efficiency and safety of traffic operation.

[0024] 6. Decision support: The dynamic impact evaluation information provided can provide strong support for traffic management departments, rescue agencies and relevant decision-makers. They can develop reasonable emergency plans, resource allocation and traffic management strategies based on the prediction results to minimize the loss and impact of accidents. BRIEF DESCRIPTION OF DRAWINGS

[0025] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiment or prior art description will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments described in the present application, and those skilled in the art can obtain other drawings according to these drawings without creative labor. In the drawings:

[0026] Figure 1 is a flow chart of a highway traffic accident impact dynamic evaluation method provided by the present application;

[0027] Fig. 2 is an input end of a traffic accident influence degree dynamic prediction model provided by an embodiment of the present application;

[0028] Fig. 3 is an output end of a traffic accident influence degree dynamic prediction model provided by an embodiment of the present application;

[0029] Fig. 4 is a structural schematic diagram of a highway traffic accident influence degree dynamic evaluation device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0030] In order to enable persons skilled in the art to better understand the technical solutions in the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. Based on the embodiments of the present application, all other embodiments obtained by persons skilled in the art without creative labor should fall within the scope of protection of the present application.

[0031] It should be noted that the current highway traffic accident severity analysis and prediction method all take the accident itself attribute as the analysis basis, and some researchers take the traffic state before the accident into the analysis basis. However, the occurrence of the accident will lead to the mutation of the highway traffic state, and it is difficult to effectively depict the change of the future traffic state and the change of the transport efficiency based on only the traffic accident time and the traffic information before the traffic accident.

[0032] In addition, most of the existing researchers use limited discrete grades such as minor accident, general accident, serious accident to measure the traffic accident severity, but in the actual traffic scene, most of the highway traffic accidents are minor traffic accidents, which leads to the problem of serious traffic accident sample imbalance when developing traffic accident severity analysis, which is not conducive to effectively developing accident severity prediction modeling.

[0033] In view of the above two problems, the present application provides a highway traffic accident influence degree dynamic evaluation method:

[0034] (1) The present application discards the traditional accident severity grading method, but uses a new traffic accident influence degree evaluation index to form a continuous traffic accident severity value. The sample imbalance problem in the accident severity analysis is effectively overcome.

[0035] (2) The application considers that the severity of the traffic accident includes two levels. The first level is the severity of the traffic accident itself, such as the number of injured persons, the number of deaths, the number of occupied lanes and the like. The second level is the influence degree on the traffic, but the traffic state after the traffic accident is affected by two aspects in nature, which are the natural evolution law of the traffic flow and the traffic flow mutation induced by the accident. In the application, the “severity information of the traffic accident itself”, the “natural evolution law information of the traffic flow” and the “traffic flow mutation information induced by the accident” are fused, and the fusion result is taken as the influence factor of the severity of the traffic accident.

[0036] Specifically, in the application, the natural evolution law information of the traffic flow is extracted based on the traffic flow sequence before the traffic accident occurs, and this result is not affected by the accident. The traffic flow mutation information induced by the accident is extracted through the traffic flow in a short period after the traffic accident occurs.

[0037] It should be noted that the traffic flow in a short period after the traffic accident occurs is obtained based on the highway inspection unmanned aerial vehicle. After the highway traffic accident occurs, the unmanned aerial vehicle can reach the surrounding of the accident location in a short time by using its flexibility, carry out the road image information collection of the accident section and the related sections, and realize the vehicle detection and flow analysis by using the computer vision algorithm, which provides a new data source for the medium and long term evaluation of the severity of the traffic accident.

[0038] The embodiment of the application provides a highway traffic accident influence degree dynamic evaluation method, as shown in FIG. 1, the highway traffic accident influence degree dynamic evaluation method specifically includes steps S101-S106:

[0039] S101, according to the position information of the highway traffic accident, the local subnetwork of the highway traffic network is divided, and the local highway network of the accident point is determined.

[0040] Firstly, the highway traffic network is determined based on the set of highway network nodes and the set of connection lines between the highway network nodes. The set of highway network nodes at least includes: a highway toll station and an interchange.

[0041] In one embodiment, a complex network topology structure of the highway network is constructed, that is, the highway traffic network: road_net=(V,E). Wherein, V represents the set of highway network nodes (including toll stations, interchanges and the like), and H is the number of highway network nodes; E represents the set of connection lines between the highway network nodes, if there is a road section between the highway network node i and the highway network node j, then e ij =1, otherwise, eij =0.

[0042] Furthermore, the similarity between the shortest distances between any two highway network nodes in the connection set is evaluated to obtain the static similarity of the highway network nodes. That is, for nodes i and j, the shortest distance d between them is calculated. ij As a measure of static similarity.

[0043] Furthermore, the set of all road segments connected to a single highway network node in the connection set is identified, and the traffic flow sequence of the connected road segments in the historical time period of all connected road segment sets is collected.

[0044] Furthermore, according to Obtain the traffic sequence features seq_v of node i in the highway network. i Among them, seq_e ik For the flow sequence of connected road segments; e ik Let represent all road segments connected to node i in the highway network; K represents the number of other highway network nodes connected to node i, and k represents the number of other highway network nodes. In other words, the mean of the traffic flow sequence of all road segments connected to node i needs to be used as the traffic flow sequence feature seq_v of node i. i ∈R T .

[0045] In one embodiment, for node i, the set of other road segments connected to that node in road_net is: Where e ik This represents the road segment between node i and the remaining nodes k. Furthermore, within the time interval 0, 1, ..., T-1, e... ik The flow sequence of connected road segments on a road segment is seq_e ik ∈R T .

[0046] Furthermore, according to The dynamic similarity ρ of highway network nodes is obtained. ij Among them, cov(seq_v) i ,seq_v j ) represents the traffic sequence feature seq_v of node i in the highway network. i The flow sequence features seq_v of node j in the highway network j covariance; and Let seq_v represent the traffic sequence features of node i in the highway network. i The flow sequence features seq_v of node j in the highway network j The standard deviation.

[0047] Further, according to the sim ij = a1 d ij + b1 p ij , to obtain the node similarity between the highway network node i and the highway network node j. Wherein, a1 is the weight of static similarity; b1 is the weight of dynamic similarity; based on the highway traffic network, the node similarity is constructed into a matrix to obtain the node similarity matrix.

[0048] Further, based on the node similarity matrix, and through the preset Louvain algorithm, the highway traffic network is divided into subnets to determine a plurality of highway subnets.

[0049] Further, the location information of the highway traffic accident is extracted. Then, according to the location information, the local highway network of the accident point is determined by matching the location of the plurality of highway subnets. Wherein, the local highway network of the accident point includes: a local highway network node set, a road section set between the local highway network nodes, and a local highway network node quantity.

[0050] In one embodiment, based on the SIM node similarity matrix, the Louvain algorithm is used to divide the local road network to form a plurality of highway subnets, and the subnet where the traffic accident occurs is called the local highway network of the accident point, denoted as local_road_net = (local_V, local_E), wherein, local_V represents the set of local highway network nodes of the accident point (including toll stations, interchanges, etc.), and G is the number of local highway network nodes of the accident point; E is the road section between the local highway network nodes of the accident point, if there is a road section between node i and node j, then e ij = 1, otherwise, e ij = 0; and the road section e accident of the traffic accident is in local_E.

[0051] S102, based on the time of the accident at the accident point and the number of highway gantries in the local highway network of the accident point, the functional index is evaluated.

[0052] First, according to the traffic efficiency E local of the local highway network of the accident point is obtained. Wherein, d ij represents the number of edges on the shortest path between the i-th node and the j-th node in the local highway network of the accident point; G is the number of nodes in the local highway network of the accident point.

[0053] Further, according to the accident occurrence time of the expressway traffic accident, the traffic efficiency is divided into a first traffic efficiency before the accident occurrence time and a second traffic efficiency after the accident occurrence time. Based on the first traffic efficiency and the second traffic efficiency, a structural index is constructed.

[0054] Further, if there are M gantries in the local expressway network of the accident point, the average speed of the M gantries before the accident occurrence time is After the accident occurrence time t, the average speed detected by the M gantries is Then, at the time t: according to the functional index F after the accident occurrence time is obtained local (t). Wherein, M is the number of high-speed gantries in the local expressway network of the accident point. m is a high-speed gantry. v m_t is the average speed of all passing vehicles within the time interval centered at the time t after the accident occurrence time of the high-speed gantry m; v m_0 is the average speed of the high-speed gantry m before the accident occurrence time.

[0055] S103, the preset structural index and the functional index are fused, and according to the fused comprehensive index, the severity of the expressway traffic accident is continuously and evenly processed to obtain an accident severity evaluation system of the expressway traffic accident.

[0056] First, according to S local (t)=α2·E local +β2·F local (t), the comprehensive index is obtained. Wherein, E local is the structural index; F local (t) is the functional index; α2 is the weight of the structural index; β2 is the weight of the functional index.

[0057] Further, then according to the accident severity evaluation system L is obtained. Wherein, T accident is the accident influence time; and, S local (0)=α2·E local_bef +β2·F local (0); and, S local (t)=α2·E local_aft +β2·F local (t); E local_bef is the first traffic efficiency before the accident occurrence time; E local_aft is the second traffic efficiency after the accident occurrence time.

[0058] As a feasible implementation manner, the transformation rate of the comprehensive index S local (t) and the accident influence time Taccident The cumulative product of these factors is used as a measure of the severity of highway traffic accidents, thus constructing a severity evaluation system for highway traffic accidents.

[0059] S104. Extract traffic characteristic features from the traffic flow sequence set at the time before the accident occurred in the highway traffic accident to obtain information on the natural evolution law of traffic flow.

[0060] First, by using a high-speed gantry and based on the time before the highway traffic accident occurred, the first set of traffic flow sequences for the road segments connected to the highway traffic accident in the local highway network at the accident location is obtained.

[0061] Furthermore, the length of the first traffic flow sequence is determined based on the first traffic flow sequence set.

[0062] Furthermore, using a pre-defined temporal convolutional network, multi-level sequence feature extraction is performed on the length of the first traffic flow sequence to output traffic flow feature information. This traffic flow feature information is then identified as information reflecting the natural evolution of traffic flow. The multi-level process includes an input layer, hidden layers, and an output layer.

[0063] In one embodiment, a set of the first traffic flow sequences of G local highway segments within N_bef minutes prior to a traffic accident is collected based on highway roadside equipment (such as ETC gantry), denoted as... in This represents the traffic flow sequence of the g-th road segment, where the length of the first traffic flow sequence is N. tf_bef This value satisfies Where seg_len_bef is the time interval for traffic flow statistics.

[0064] In one embodiment, this application uses a TCN (Temporal Convolutional Network) to extract traffic characteristics at the time of the accident. TCNs have the capability to extract features from time-series data. Taking the g-th road segment as an example, ... If the sequence is used as input to the TCN, then the output of the TCN is the information on the natural evolution of traffic flow, denoted as inf_bef. g ∈R 1×1 Therefore, for the local highway network (local_road_net) at the accident site, the feature set of G road segments that can form information on natural evolution patterns is as follows:

[0065] S105. By fusing information on the natural evolution of traffic flow with information on the severity of the accident itself at the time of the accident and information on sudden changes in traffic flow induced by the accident, a system of accident influencing factors for highway traffic accidents is obtained.

[0066] It is necessary to use drones to obtain a second set of traffic flow sequences for the road segments connected to the highway accident in the local highway network at the accident site, based on the time of the accident.

[0067] Furthermore, a pre-defined temporal convolutional network is used to extract multi-level sequence features from the second traffic flow sequence set to determine the traffic flow mutation information induced by the accident.

[0068] In one embodiment, after the accident occurs, i.e., based on the time of the highway traffic accident, a set of traffic flow sequences for G local highway segments, totaling τ minutes (T, T+1, ..., T+τ), is collected using highway roadside equipment (such as ETC gantry), denoted as […]. in Let N represent the traffic flow sequence of the g-th road segment, and let N be the set of the second traffic flow sequences. tf_aft , can be represented as And satisfy Where seg_len_bef is the time interval for traffic flow statistics.

[0069] In one implementation, this application uses a Time-Series Network (TCN) to extract traffic characteristics at the time of the accident. The TCN has the capability to extract features from time-series data. Taking the g-th road segment as an example, it will... If the sequence is used as input to the TCN, then the output of the TCN is the traffic flow mutation information induced by the accident, denoted as inf_aft. g At this point, for the local highway network (local_road_net) at the accident site, the G road segments can form a feature set of information on traffic flow changes induced by the accident.

[0070] Furthermore, information on the severity of highway traffic accidents is collected. This information includes: the number of injured, the number of people who died at the scene, the number of vehicles involved, and the lane occupancy rate.

[0071] In one embodiment, this application uses the following metrics: accident-induced traffic flow mutation N_injured, number of on-site fatalities N_dead, number of vehicles involved N_veh, and lane occup ratio R_occup. The formula for calculating R_occup is: Here, Num_occupancy_lane refers to the number of lanes occupied by traffic accidents, and Num_lane refers to the total number of lanes on the road.

[0072] Further, the traffic flow natural evolution law information, the accident itself severity information and the accident-induced traffic flow mutation information are collected to obtain an accident influence factor set. Based on the accident influence factor set, an accident influence factor system is constructed.

[0073] In one embodiment, the "accident itself severity information", "traffic flow natural evolution law information" and "accident-induced traffic flow mutation information" are fused to form an accident influence factor set X of the highway traffic accident:

[0074] {inf_bef1,...,inf_bef G ,N_injured,N_dead,N_veh,R_occup,inf_aft1,...,inf_aft G}∈R 2×G+4 Then, based on the accident influence factor set X, an accident influence factor system is constructed.

[0075] S106, according to the accident severity evaluation system and the accident influence factor system, the traffic accident influence degree of the highway traffic accident is dynamically predicted to obtain dynamic influence evaluation information based on the highway traffic accident.

[0076] The neural network technology and the channel attention model are needed to construct the traffic accident influence degree dynamic prediction model. The input of the traffic accident influence degree dynamic prediction model is the accident influence factor system, and the output is the accident severity evaluation system.

[0077] Further, the historical accident influence factor set in the accident influence factor system and the historical accident severity evaluation set in the accident severity evaluation system are used to complete the model training of the traffic accident influence degree dynamic prediction model.

[0078] Further, the current highway traffic accident information is input into the traffic accident influence degree dynamic prediction model to obtain the dynamic influence evaluation information.

[0079] In one embodiment, Fig. 2 is an input end of a traffic accident influence degree dynamic prediction model provided by an embodiment of the application, and Fig. 3 is an output end of a traffic accident influence degree dynamic prediction model provided by an embodiment of the application. The CNN+channel attention model is used to construct the traffic accident influence degree dynamic prediction model Model. The basic framework of the Model model is shown in Figs. 2 and 3. The input of the model is the historical accident influence factor set X in the accident influence factor system, and the output of the model is the historical accident influence factor set L in the accident severity evaluation system. The purpose of the attention channel is to give different weights to different channels (i.e. different influence factors).

[0080] In one embodiment, the training of the traffic accident impact degree dynamic prediction model: first collect the historical traffic accident data of the expressway, assume that a total of Q traffic accidents are collected For the qth traffic accident his_accident q According to the specific implementation steps described above, the historical accident impact factor set X of the historical traffic accident is obtained q And the historical accident impact factor set L q Finally, the training sample set of the model is obtained Based on train_set, the traffic accident impact degree dynamic prediction model Model can be trained. Then, for the current expressway traffic accident accident c According to the specific implementation steps described above, the accident impact factor set X of the expressway traffic accident can be obtained c Using the trained Model, the accident impact factor set L can be inferred c Further, the dynamic impact evaluation information reflecting the dynamic evaluation result of the impact degree of the expressway traffic accident is formed.

[0081] In addition, the embodiment of the application also provides a dynamic evaluation device for the impact degree of the expressway traffic accident, as shown in Figure 4, the dynamic evaluation device for the impact degree of the expressway traffic accident 400 specifically comprises:

[0082] At least one processor 401; and a memory 402 communicatively connected with the at least one processor 401; wherein the memory 402 stores instructions executable by the at least one processor 401, so that the at least one processor 401 can execute:

[0083] According to the location information of the expressway traffic accident, the local subnetwork of the expressway traffic network is divided, and the local expressway network of the accident point is determined;

[0084] Based on the time of the accident occurrence in the local expressway network of the accident point and the number of expressway gantries, the functional index is evaluated;

[0085] The preset structural index and the functional index are fused, and the severity of the expressway traffic accident is continuously and uniformly processed according to the fused comprehensive index, so that the accident severity evaluation system of the expressway traffic accident is obtained;

[0086] The traffic flow sequence set of the expressway traffic accident at the time when the accident does not occur is extracted for traffic characteristic features, and the traffic flow natural evolution law information is obtained;

[0087] The traffic flow natural evolution rule information is fused with the accident severity information at the accident occurrence time and the accident-induced traffic flow mutation information to obtain an accident influence factor system of the highway traffic accident.

[0088] According to the accident severity evaluation system and the accident influence factor system, the traffic accident influence degree of the highway traffic accident is dynamically predicted to obtain dynamic influence evaluation information based on the highway traffic accident.

[0089] The embodiments of the present application can more accurately evaluate the severity of the highway traffic accident by dividing the local highway network of the accident point and evaluating the functional indicators. This helps the relevant departments to take appropriate rescue and response measures in a timely manner, improves the efficiency and safety of accident handling. This helps better understand the causes and effects of the accident, provides a basis for developing improvement measures. It can also more objectively compare the severity of different accidents to provide more valuable information for decision-making. It helps to better understand the mechanisms and impact of accidents, providing guidance for prevention and reduction of accidents. At the same time, it can timely understand the impact of the accident on traffic and take corresponding traffic management measures to reduce the impact of the accident on traffic and improve the efficiency and safety of traffic operation. It is beneficial to develop reasonable emergency plans, resource allocation and traffic management strategies based on the prediction results to minimize the loss and impact of accidents.

[0090] Each of the embodiments in the present application is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments. In particular, for the device embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the part of the method embodiments.

[0091] The above describes specific embodiments of the present application. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be performed in an order different than the order in the embodiments and still achieve the desired result. In addition, the processes depicted in the figures do not necessarily require the particular order shown or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing can be advantageous or required.

[0092] The above only describes the embodiments of the present application and does not limit the present application. The embodiments of the present application can have various changes and variations for those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the embodiments of the present application shall be included in the scope of the claims of the present application.

Claims

1. A dynamic evaluation method for the impact of highway traffic accidents, wherein, The method includes: Based on the location information of highway traffic accidents, the highway network is divided into local sub-networks to determine the local highway network where the accident occurred. Before determining the local highway network location based on the location information of the highway traffic accident by dividing the highway network into local subnets, the method further includes: The highway network is determined based on the set of highway network nodes and the set of connections between highway network nodes; wherein the set of highway network nodes includes at least: highway toll stations and interchanges; The similarity between the shortest distances between any two highway network nodes in the connection set is evaluated to obtain the static similarity of the highway network nodes. Identify all road segments in the connection set that are connected to a single highway network node, and collect the traffic flow sequence of the connected road segments in the historical time period of all connected road segment sets; according to Obtain the traffic sequence features seq_v of node i in the highway network. i ; where seq_e ik The flow sequence of the connected road segments; e ik K represents all road segments connected to the highway network node i; K represents the number of other highway network nodes connected to the highway network node i, and k represents the other highway network nodes. according to The dynamic similarity ρ of the highway network nodes is obtained. ij ; where, cov(seq_v i ,seq_v j ) represents the traffic flow sequence feature seq_v of node i in the highway network. i The flow sequence features seq_v of node j in the highway network j covariance; and The traffic flow sequence features seq_v of node i in the highway network are respectively represented. i and the traffic sequence feature seq_v of the highway network node j j Standard deviation; According to sim ij =α1·d ij +β1·ρ ij The node similarity between highway network node i and highway network node j is obtained; where α1 is the weight of the static similarity and β1 is the weight of the dynamic similarity. Based on the highway traffic network, a node similarity matrix is ​​constructed to obtain the node similarity matrix; Based on the location information of highway traffic accidents, the highway network is divided into local subnetworks to determine the local highway network at the accident location. Specifically, this includes: Based on the point similarity matrix, and using the preset Louvain algorithm, the highway traffic network is divided into subnetworks to determine several highway subnetworks. Extract the location information of the highway traffic accident; Based on the location information, location matching is performed on several of the expressway subnetworks to determine the local expressway network where the expressway traffic accident occurred; wherein, the local expressway network at the accident location includes: a set of local expressway network nodes, a set of road segments between local expressway network nodes, and the number of local expressway network nodes. Based on the time of the accident and the number of highway gantries in the local highway network at the accident site, functional indicators were evaluated. Before assessing functional indicators based on the time of the accident and the number of highway gantries in the local highway network at the accident location, the method further includes: according to The traffic efficiency E of the local highway network at the accident site was obtained. local ; where d ij G represents the number of edges on the shortest path between the i-th node and the j-th node in the local highway network at the accident site; G is the number of nodes in the local highway network at the accident site. Based on the time of the highway traffic accident, the traffic efficiency is divided into a first traffic efficiency before the time of the accident and a second traffic efficiency after the time of the accident. Based on the first traffic efficiency and the second traffic efficiency, structural indicators are constructed; Among them, functional indicators were evaluated based on the time of the accident and the number of highway gantries in the local highway network at the accident site, specifically including: according to The functional index F obtained after the time of the accident occurred local (t); where M is the number of high-speed gantries in the local highway network at the accident site; m is the number of high-speed gantries; v m_t v is the average instantaneous speed of all passing vehicles within a time interval centered at time t after the occurrence of the accident at the high-speed gantry m; m_0 The average speed of the high-speed gantry m before the time of the accident; The preset structural indicators and the functional indicators are fused together, and the severity of the highway traffic accident is continuously balanced based on the fused comprehensive indicators to obtain the highway traffic accident severity evaluation system. Specifically, the pre-set structural indicators and the functional indicators are fused together, and the severity of highway traffic accidents is continuously balanced based on the fused comprehensive indicators to obtain an accident severity evaluation system for highway traffic accidents, which includes: According to S local (t)=α2·E local +β2·F local (t), to obtain the comprehensive index; where, E local For the aforementioned structural index; F local (t) represents the functional index; α2 represents the weight of the structural index; β2 represents the weight of the functional index; according to The accident severity evaluation system L is obtained; where T accident The duration of the accident; and, S local (0)=α2·E local_bef +β2·F local (0); and, S local (t)=α2·E local_aft +β2·F local (t); E local_bef The first traffic efficiency before the time of the accident; E local_aft The second traffic efficiency after the moment the accident occurred; The traffic flow sequence set of the highway traffic accident at the time before the accident occurred is processed by extracting traffic characteristic features to obtain information on the natural evolution law of traffic flow. By fusing the information on the natural evolution of traffic flow with the information on the severity of the accident itself at the time of the accident and the information on sudden changes in traffic flow induced by the accident, the accident influencing factor system of the highway traffic accident is obtained. Based on the accident severity evaluation system and the accident influencing factor system, the impact of the highway traffic accident is dynamically predicted to obtain dynamic impact evaluation information based on the highway traffic accident.

2. The method for dynamically evaluating the impact of highway traffic accidents according to claim 1, wherein, The traffic flow sequence set of the aforementioned highway traffic accident at the time before the accident occurred is processed by extracting traffic characteristic features to obtain information on the natural evolution law of traffic flow, specifically including: Using the high-speed gantry and based on the time before the highway traffic accident occurred, a first set of traffic flow sequences for the road segments connected to the highway traffic accident in the local highway network at the accident location is obtained. Based on the first set of traffic flow sequences, the length of the first traffic flow sequence is determined; Using a pre-defined temporal convolutional network, multi-level sequence feature extraction is performed on the length of the first traffic flow sequence to output traffic flow feature information; and the traffic flow feature information is determined as the natural evolution law information of the traffic flow.

3. The method for dynamically evaluating the impact of highway traffic accidents according to claim 1, wherein, By fusing the information on the natural evolution of traffic flow with the information on the severity of the accident at the time of its occurrence and the information on sudden changes in traffic flow induced by the accident, a system of accident influencing factors for highway traffic accidents is obtained, specifically including: Using drones, and based on the time of the highway traffic accident, a second set of traffic flow sequences is obtained for the road segments connected to the highway traffic accident in the local highway network at the accident site. By using a pre-defined temporal convolutional network, multi-level sequence feature extraction is performed on the second traffic flow sequence set to determine the traffic flow mutation information induced by the accident. Collect information on the severity of the traffic accidents on the highway; wherein, the information on the severity of the accidents includes: the number of people injured in the accident, the number of people who died at the scene of the accident, the number of vehicles involved, and the lane occupancy rate of the accident lane; The information on the natural evolution of traffic flow, the severity of the accident itself, and the information on sudden changes in traffic flow induced by the accident are combined and processed to obtain a set of factors affecting the accident. Based on the set of factors influencing the accident, the system of factors influencing the accident is constructed.

4. The method for dynamically evaluating the impact of highway traffic accidents according to claim 1, wherein, Based on the aforementioned accident severity evaluation system and accident influencing factor system, the impact of the highway traffic accident is dynamically predicted to obtain dynamic impact evaluation information based on the highway traffic accident, specifically including: Based on neural network technology and channel attention model, a dynamic prediction model for the impact of traffic accidents is constructed; wherein, the input of the dynamic prediction model for the impact of traffic accidents is the accident influencing factor system, and the output is the accident severity evaluation system; The model training for the dynamic prediction model of the impact of traffic accidents is completed by using the set of historical accident influencing factors in the accident influencing factor system and the set of historical accident severity evaluations in the accident severity evaluation system. The current highway traffic accident information is input into the dynamic prediction model of the degree of impact of the traffic accident to obtain the dynamic impact evaluation information.

5. A dynamic evaluation device for the impact of highway traffic accidents, wherein, The device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, enabling the at least one processor to execute a dynamic evaluation method for the impact of highway traffic accidents according to any one of claims 1-4.

Images