CN115331425A - Traffic early warning method, device and system - Google Patents

Abstract

The embodiments of the present invention disclose a traffic early warning method, device and system. The method includes: marking a special area, establishing a road network topology map corresponding to the special area; correlating a city road network model, collecting traffic data in the special area; building a route library, analyzing the correlation of routes in the special area combined with the traffic data, and screening key traffic detections convert the data collected by the key traffic detectors into the spatial density of traffic flow, and set the pressure threshold of regional traffic warning according to the spatial density of traffic flow; obtain the congestion information of special areas according to the pressure threshold, and confirm and mark the congestion information, Generate labeled data; train a congestion prediction model based on the labeled data and traffic data, and use the congestion prediction model to warn of traffic conditions in special areas. The solution provided by the present invention can reduce the amount of input samples, reduce the training pressure, and reduce the technical effect of overfitting under the premise of ensuring the accuracy of the model.

Description

A traffic warning method, device and system

technical field

The invention relates to the application field of traffic road network technology, in particular to a traffic early warning method, device and system.

Background technique

With the increase of private car ownership, the contradiction between the supply and demand of urban roads is increasing day by day, the driving environment on the roads is also more complicated, and traffic congestion often occurs, which generally reduces the driving speed of travelers and increases the travel time. Therefore, it is necessary to Traffic congestion early warning technology detects congestion points in a timely manner, and efficiently assigns tasks to staff to quickly deal with congestion, helping travelers plan travel routes more effectively and save travel time.

However, the traditional road congestion warning is mostly aimed at a single node, and the coverage is small. When the monitored intersection is about to be congested, an alarm will be issued, and it is impossible to monitor and advance early warning for the upstream and downstream of the congestion according to the actual road traffic conditions. In some specific time periods When the congestion around buildings with high traffic density in special areas such as hospitals and schools is serious, the traditional traffic early warning system cannot quickly determine the type of traffic pressure and the source of the problem, which leads to the inability to effectively and quickly deal with the dredging of problematic road sections, and it is difficult to quickly and efficiently relieve traffic congestion. It is difficult to guarantee the smoothness of traffic.

In view of the above-mentioned problems that the type of traffic pressure and the source of the problem cannot be quickly judged in the related technology, which leads to the inability to effectively and quickly deal with the dredging of the problem road section, it is difficult to quickly and efficiently alleviate the traffic jam, and it is difficult to ensure the smoothness of the traffic. At present, no effective solution has been proposed. .

Contents of the invention

In order to solve the above technical problems, the present invention expects to provide a traffic early warning method, device and system to at least solve the problem of the inability to quickly determine the type of traffic pressure and the source of the problem in the related art, which in turn leads to the inability to effectively and quickly deal with dredging problem road sections, and it is difficult to quickly Efficiently alleviate traffic jams, and it is difficult to ensure the smoothness of traffic.

Technical scheme of the present invention is realized like this:

In the first aspect, the present invention provides a traffic early warning method, including: marking a special area, establishing a road network topology map corresponding to the special area; associating the urban road network model, collecting traffic data in the special area; constructing a route library, and analyzing traffic data Path correlation in special areas, screening key traffic detectors; converting data collected by key traffic detectors into traffic flow spatial density, and setting pressure thresholds for regional traffic warnings based on traffic flow spatial density; obtaining congestion in special areas based on pressure thresholds information, and confirm and mark the congestion information to generate marked data; train the congestion prediction model based on the marked data and traffic data, and use the congestion prediction model to give early warning of traffic conditions in special areas.

Optionally, marking a special area and establishing a road network topology map corresponding to the special area includes: marking a special area, wherein the special area is a building or facility with a traffic demand greater than a preset threshold in a specific time period and its surrounding preset related areas. The area covered by the road facilities in the adjacent range; filter the central node of the special area, and set the intersection of the central node’s preset adjacent range as the node, take the node as the center, and use all the associated road sections of the node as road sections close to the central node , to establish a road network topology map of a special area.

Further, optionally, collecting the traffic data of the regional road network model in the road network topology map by associating the urban road network model includes: collecting the traffic data of the regional road network model in the road network topology map by associating the urban road network model according to the traffic detectors of each road section The traffic data of the road network model, wherein the traffic data includes: data of the regional road network model and traffic detector data.

Optionally, build a path library, combine traffic data to analyze path correlation in special areas, and filter key traffic detectors including: based on historical traffic detector data, filter vehicles passing around the central node, track vehicle trajectories, and reverse vehicles in special areas Based on the traffic routes in the area, build a route library; based on the historical traffic data, count the traffic data of the route in the specified period to obtain the traffic characteristics of each route; Traffic characteristics; judge the correlation between the path and the central node through the correlation coefficient of the traffic characteristics of the path and the traffic characteristics of the central node, and filter the target path according to the correlation; for the target path corresponding to each correlation, consider the path according to the weight of the corresponding target path The impact of flow on regional congestion, according to the flow direction on the route, select the traffic detector installed in the flow direction in the road section as the key traffic detector.

Further, optionally, the method further includes: for a road segment missing a bayonet device in the route, using a route search algorithm to search for the shortest route and complete the route.

Optionally, converting the data collected by key traffic detectors into traffic flow spatial density includes: using the flow rate or headway distance, calculating an approximate value of the current queue length according to the sum of the standard small car length and the headway distance, and calculating the approximate value of the current queue length Compared with the length of the road section, the first traffic flow space density of the road section is obtained; using the headway or queue length, the time interval collected by two consecutive vehicles passing through the key traffic detector in the fixed point is multiplied by the free flow speed of the road section to obtain Distance interval, compare the distance interval with the length of the road section to obtain the second traffic flow spatial density of the road section; normalize the first traffic flow spatial density and the second traffic flow spatial density, and weight the normalized data On average, the target traffic flow spatial density of the link is obtained.

Further, optionally, setting the pressure threshold of the regional traffic warning according to the traffic flow spatial density includes: classifying the traffic pressure according to the traffic flow spatial density, and setting the pressure threshold of the regional traffic warning; wherein, the traffic pressure classification includes inflow pressure and internal pressure ; There are multiple groups of pressure threshold settings.

Optionally, obtain congestion information in a special area according to the pressure threshold, and confirm and mark the congestion information. Generating marked data includes: obtaining the real-time congestion situation in the area according to the pressure alarm of the pressure threshold and giving an alarm, and confirming the real-time congestion situation manually, Mark the correct perception results and generate marked data; among them, manual confirmation, real-time or offline confirmation through monitoring and traffic index comparison.

Further, optionally, training the congestion prediction model based on the marked data and traffic data, and using the congestion prediction model to warn the traffic situation in a special area includes: sorting the key traffic detector data in the target path by time, performing data Process and obtain the traffic status indicators of the target path, and associate the congestion label of the marked data; input the neural network time series model for training, and obtain the time series congestion early warning model; use the state estimation algorithm to adjust the time series prediction results obtained through the test set, and train Obtain the congestion prediction model of the special area; according to the congestion prediction model, the traffic situation of the special area is given an early warning.

In the second aspect, the present invention provides a traffic early warning device, including: an establishment module, used to mark a special area, and establish a road network topology map corresponding to the special area; a collection module, used for associating the urban road network model, and collecting the road network topology map The traffic data of the road network model in the middle area; the screening module is used to build the path library, and analyzes the correlation of paths in special areas combined with traffic data, and screens key traffic detectors; the conversion module is used to convert the data collected by key traffic detectors into The spatial density of traffic flow, and set the pressure threshold of regional traffic early warning according to the spatial density of traffic flow; the marking module is used to obtain the congestion information of special areas according to the pressure threshold, and confirm and mark the congestion information to generate marked data; the early warning module, It is used to train the congestion prediction model based on the marked data and traffic data, and to warn the traffic conditions in special areas through the congestion prediction model.

In the third aspect, the present invention provides a traffic early warning system, including: a marking module, used to divide and mark special areas; a data collection module, used to collect road network models and traffic data of key road sections; a data conversion module, used to It is used to convert and calculate the collected data; the key path monitoring module is used to monitor the real-time overall traffic status of key paths in the area and display the heat map based on the correlation weight; the real-time congestion perception module is used to give real-time alarms to the road network in special areas , and manually confirm the alarm situation; the congestion early warning module inputs real-time data into the offline model library for congestion prediction and sends out early warning signals;

Among them, the data conversion module includes: a length data conversion module, which is used to calculate the approximate value of the current queue length for the sum of the length of the standard small car and the distance between the fronts, and compare the result with the length of the road section to obtain the first traffic flow space density of the road section; The speed data conversion module is used to compare the time interval between two consecutive vehicles passing a fixed point with the free flow speed of the road section to obtain the distance interval, and compare the distance interval with the length of the road section to obtain the second traffic flow space density of the road section.

The invention provides a traffic warning method, device and system. By marking special areas, establish road network topology maps corresponding to special areas; associate urban road network models, collect traffic data in special areas; build path libraries, analyze path correlations in special areas combined with traffic data, and screen key traffic detectors; The data collected by key traffic detectors is converted into traffic flow spatial density, and the pressure threshold of regional traffic warning is set according to the traffic flow spatial density; the congestion information of special areas is obtained according to the pressure threshold, and the congestion information is confirmed and marked to generate marked data ;According to the marked data and traffic data, train the congestion prediction model, and use the congestion prediction model to warn the traffic situation in special areas, so that the input sample size can be reduced, the training pressure can be reduced, and the over-fitting can be reduced under the premise of ensuring the accuracy of the model. Combined technical effect.

Description of drawings

The accompanying drawings described here are used to provide a further understanding of the present invention and constitute a part of the application. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations to the present invention. In the attached picture:

Fig. 1 is a schematic flow chart of a traffic early warning method provided by Embodiment 1 of the present invention;

2 is a schematic diagram of a road network topology diagram in a traffic early warning method provided by Embodiment 1 of the present invention;

FIG. 3 is a schematic diagram of a traffic warning device provided in Embodiment 2 of the present invention;

FIG. 4 is a schematic diagram of a traffic warning system provided by Embodiment 3 of the present invention.

Detailed ways

In order to enable those skilled in the art to better understand the solutions of the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only It is an embodiment of a part of the present invention, but not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts shall fall within the protection scope of the present invention.

It should be noted that the terms "first" and "second" in the specification, claims and drawings of the present invention are used to distinguish different objects, rather than to limit a specific order.

It should also be noted that the following embodiments of the present invention may be implemented independently, or may be implemented in combination with each other, which is not specifically limited in the embodiments of the present invention.

Embodiment one

In the first aspect, the embodiment of the present invention provides a traffic early warning method, and Fig. 1 is a schematic flow chart of a traffic early warning method provided by Embodiment 1 of the present invention; as shown in Fig. 1 , the traffic early warning method provided by the embodiment of the present application includes :

Step S101, marking a special area, and establishing a road network topology map corresponding to the special area;

Optionally, marking special areas in step S101, and establishing a road network topology map corresponding to special areas includes: marking special areas, wherein special areas are buildings or facilities and their surroundings that have a traffic demand greater than a preset threshold in a specific time period The area covered by the road facilities in the preset adjacent range; filter the central node of the special area, and set the intersection of the preset adjacent range of the central node as a node, take the node as the center, and use all the associated road sections of the node as the center node For similar road sections, build a road network topology map of a special area.

Among them, the special areas in the embodiment of this application include but are not limited to: shopping malls, schools, hospitals, and scenic spots, where there may be multiple central nodes in the scenic spot;

It should be noted that the road network topology map corresponding to the special area established in the embodiment of the present application is shown in FIG. 2 , which is a schematic diagram of the road network topology map in a traffic early warning method provided by Embodiment 1 of the present invention. In order to meet the subsequent calculation needs, the intersection screening range of the central building needs to include three-level intersections.

Step S102, associating with the urban road network model, and collecting traffic data in special areas;

Specifically, associating the urban road network model in step S102, collecting traffic data in a special area includes: collecting traffic data of the regional road network model in the road network topology map according to the traffic detectors of each road section through associating the urban road network model, wherein, The traffic data includes: the data of the regional road network model and the traffic detector data.

Among them, the data of the regional road network model can include: roads, road sections, intersections, and lane data; the traffic detector data include: traffic flow, occupancy rate, headway, queue length data;

In the embodiment of this application, the traffic data in the regional road network model is collected by traffic detectors such as geomagnetism and checkpoints. The traffic detector takes the vehicle as the detection target, detects the information of the vehicle and the passing status, and calculates or counts various traffic at the same time. Parameters, such as traffic flow, occupancy rate, headway, queue length, etc., are used to provide multi-dimensional traffic information for the control system for signal control.

Among them, the flow rate is the number of vehicles passing the detector per unit time;

Occupancy rate includes time occupancy rate and space occupancy rate. Time occupancy rate refers to the ratio of the cumulative value of vehicle passing time to the total observation time on any section of the road, which is called time occupancy rate; the calculation formula is

The space occupancy rate refers to the ratio of the total length of vehicles to the total length of the road section on a certain section of the road, which is called the space occupancy rate, usually expressed as a percentage, and the calculation formula is:

Among them, R is the occupancy rate, L is the length of the road segment, and T is the accumulated value of the vehicle passing time;

The headway is the time interval between two consecutive vehicles passing a fixed point in the traffic flow;

The queue length is the length of the traffic flow waiting to pass through the intersection;

Among them, in the relationship between the flow rate and the occupancy rate, when the flow rate is small, the corresponding occupancy rate may be relatively small or relatively large. Small traffic flow and small occupancy rate indicate that the traffic pressure on the road section is small, and all passing vehicles can pass through the road section smoothly; small flow rate and large occupancy rate indicate that there is congestion in the road section, and there are few passing vehicles, and there may be vehicles staying in the detector for a long time The situation in the detection area is convenient to remind the staff on duty to deal with it in time.

Step S103, building a route library, analyzing the correlation of routes in a special area in combination with traffic data, and screening key traffic detectors;

Optionally, in step S103, a path library is constructed, and the path correlation in a special area is analyzed in combination with traffic data, and the screening of key traffic detectors includes: based on historical traffic detector data, screening vehicles passing around the central node, tracking vehicle trajectories, Reversing the driving path of the vehicle in the special area, constructing the path library; based on the historical traffic data, counting the traffic data of the path within a specified period to obtain the traffic characteristics of each path; and based on the real-time traffic data, counting the specified The traffic data flowing into the central node within a period obtains the traffic characteristics of the central node; the correlation coefficient between the traffic characteristics of the path and the traffic characteristics of the central node is used to judge the correlation between the path and the central node, Filter target paths according to the correlation; for the target paths corresponding to each correlation, consider the impact of path traffic on regional congestion according to the weight of the corresponding target path, and select road sections to be installed in the selected road section according to the flow direction on the path The traffic detector in the flow direction is used as the key traffic detector.

Further, optionally, the traffic early warning method provided in the embodiment of the present application further includes: for a road segment missing a checkpoint device in the route, using a route search algorithm to search for the shortest route and complete the route.

Optionally, the traffic early warning method provided in the embodiment of the present application further includes: according to the similarity between the historical data of the route and the real-time data of the central building, obtaining a route flow model from the historical data to identify the target route.

Specifically, based on the historical bayonet detector data, the vehicles passing around the central building are screened, the vehicle trajectory is tracked, and the driving path of the vehicle in the special area is reversed. To screen the vehicles passing around the central building, it is necessary to screen vehicles that pass the nearest bayonet of the central building and do not pass the next detector within a certain period of time. Associating the checkpoint detector device with the road network, sorting the checkpoint data of each vehicle in reverse order, and judging whether the checkpoint corresponding to the temporally adjacent data belongs to the connected road section in space. If it belongs to a connected road section, then the road section is included in the path; otherwise, algorithms such as Dijkstra and Floyd (that is, the path search algorithm in the embodiment of the application) are used to search for the shortest path between two checkpoints at adjacent times, and the path is determined. Completion. Integrate the paths obtained by data aggregation to generate a path library in a special area.

Among them, the historical data is aggregated by path at a certain granularity, and the data of a certain time window before the time of congestion identification is taken as the historical traffic characteristics of different paths at the current moment. At the same time, according to the same aggregation granularity and time window, the data of the detectors around the central building before the same time are collected, and aggregated into the flow characteristics of the central building. Among them, the detectors around the central building refer to the detectors on the road section that directly flows into the central building. Typically a detector on a road segment surrounding a building. The correlation coefficient between the path flow characteristics and the central building flow is used to judge the correlation between the path and the central building traffic conditions, and the target path can be dynamically screened according to the correlation. The use of time windows to select part of the data is to consider that the use of data close to the moment can better characterize the traffic characteristics of the moment.

For target paths with different correlations, the correlation coefficient is used as the weight to consider the impact of different path traffic on regional congestion. According to the flow direction on the route, the traffic detector installed in the flow direction in the road section is selected as the key traffic detector to calculate the traffic index of the route.

In the embodiment of this application, the traffic correlation is calculated as follows:

Aggregate 3 months of historical data by path at a granularity of 10 minutes, and sort by time series to obtain the historical traffic characteristics of each path within a day. Select the time T that needs to be identified for congestion, such as 15:10, and take the time window data one hour before T, that is, the historical traffic characteristics between 14:10 and 15:10, as the historical traffic change characteristics at this time point, record for X. At the same time, the real-time data is also aggregated at a granularity of 10 minutes, and the data of the same time window is taken as the current flow characteristic of the central building, which is recorded as Y. Computes the correlation between X and Y.

In the embodiment of this application, considering the non-linearity and continuity of the data, after filtering and removing abnormal outliers, the Pearson correlation coefficient calculation method is used,

Among them, cov(X, Y) is the covariance of X and Y, σ _X is the standard deviation of X, and σ _Y is the standard deviation of Y. The correlation coefficient is between -1 and 1. In this embodiment, setting the coefficient less than 0 to 0 means that the flow characteristics of the path do not conform to the flow characteristics of the central building, and the possibility of transporting traffic pressure to the central building is low.

In the embodiment of this application, the idea and process of using the Dijkstra algorithm to search for the shortest path are as follows:

If a directed graph with N nodes and weights is given, first set an array D to save the shortest distance from the starting point to each node, and then define a set T to save all nodes that have found the shortest path from the starting point to this point . Under the initial conditions, the distance from the starting point to the starting point is 0, the distance between the starting point and the nodes connected to it is the corresponding weight, and the distance between nodes without direct connections is defined as ∞ (infinity), and then the distance value to the starting point is selected The smallest node, add this point to the set T, and take the node with the smallest distance from the starting point as the starting point, and update the distance value from the starting point in turn, that is, if the sum of the distance value of this point and the distance value from this point to other points is less than the starting point directly If the distance value reached, replace the direct arrival value with this value, otherwise, do not change, repeat the above process until T contains all the nodes in the graph, and the algorithm ends;

In the embodiment of this application, the idea and process of using the Floyd algorithm to search for the shortest path are as follows:

Assuming there is a directed or undirected graph with N nodes, the Floyd algorithm needs to define two N×N matrices, denoted as D and P, and the element a[i][j] in the D matrix represents the The distance to vertex j, and the element b[i][j] in the P matrix represents the vertex passing through from vertex i to vertex j. In the initial state; the matrix D represents the distance from each vertex to other vertices, which is usually a weight in a computer. If there is no direct connection between two points, the distance value is ∞, and the initial state of the P matrix is all b [i] The j value of the [j] element. Note that the number of updates is K (a total of N updates are required), if (a[i][k-1]+a[k-1][j])<a[i][j], replace it with the former value At the same time, b[i][j]=b[i][k-1] in the P matrix will be updated sequentially until the number of updates is N, then the algorithm ends, and the combination of the D matrix and the P matrix can be obtained from The shortest path from any node to all other nodes;

In the embodiment of the present application, by counting the vehicle driving paths with the central building as the end point of the trip, a library of traffic paths from different nodes to the central building can be constructed, and the paths with high correlation between the flow around the central building and the historical flow of the path can be selected to obtain special areas important path in .

Based on the bayonet detector, the passage path between the node and the central building is established. For the road section missing the bayonet, the Dijkstra and Floyd algorithm (that is, the path search algorithm in the embodiment of the present application) is used to search for the shortest path and establish a passage path library. The traffic flow on the path is marked at each node, and the key lanes on the path are obtained. Correlate the lane-level traffic detectors installed in the key lanes to get the key detectors that need to be focused on.

In the embodiment of this application, the purpose of screening the detectors that need to be focused on is to eliminate the impact of irrelevant traffic flow on the congestion perception of special areas. Highly correlated road sections include traffic flows in different directions. According to the detectors that need to be focused on, the most Traffic pressure that may flow into the central building, enabling accurate pressure analysis.

Step S104, converting the data collected by key traffic detectors into traffic flow spatial density, and setting the pressure threshold of regional traffic early warning according to the traffic flow spatial density;

Optionally, converting the data collected by key traffic detectors into traffic flow spatial density in step S104 includes: using the flow rate or headway distance, calculating an approximate value of the current queue length according to the sum of the length of a standard small car and the headway distance, and calculating the current queue length The approximate value of the length is compared with the length of the road section to obtain the first traffic flow space density of the road section; using the headway or queue length, the time interval collected by two consecutive vehicles passing through the key traffic detector in the fixed point is compared with the free flow speed of the road section By multiplying the distance interval, the distance interval is compared with the length of the road section to obtain the second traffic flow spatial density of the road section; the first traffic flow spatial density and the second traffic flow spatial density are normalized, and the normalized The data are weighted and averaged to obtain the target traffic flow space density of the road section.

Further, optionally, setting the pressure threshold of the regional traffic warning according to the traffic flow spatial density includes: classifying the traffic pressure according to the traffic flow spatial density, and setting the pressure threshold of the regional traffic warning; wherein, the traffic pressure classification includes inflow pressure and internal pressure ; There are multiple groups of pressure threshold settings.

Specifically, when the traffic flow space density reaches the quantile threshold on the road section that transmits traffic flow to the area on the node, it will have an impact on the traffic situation inside the area. For the inflow pressure, the traffic flow space density of the road section adjacent to the building reaches Threshold, indicating that the traffic pressure around the building is greater, which is the internal pressure.

In the embodiment of the present application, the pressure threshold is set with multiple groups of levels, and different alarm levels correspond to a certain node or multiple nodes in a special area reaching the congested situation, which can be divided into unblocked, slow, congested, and severe congested according to the actual situation And so on multiple levels.

In the embodiment of this application, the pressure threshold for the pressure alarm is based on the lane-level data and the intersection facility model, and the historical traffic rules are calculated to obtain the historical traffic density data sets of different road sections, and a group of quantiles is selected to distinguish the road sections The congestion and unimpeded conditions of different road sections are different. Using the quantile as the pressure threshold can be targeted to get the early warning standard based on the historical traffic law of the road section itself. At the same time, the adjustment of the quantile can be adjusted according to manpower and actual alarm needs When the quantile value is set larger, the number of problem areas is small, but there are usually traffic problems that need to be solved urgently; Situation flexible adjustment design.

It should be noted that in the embodiment of this application, the alarm sensitivity can be adjusted according to user requirements. For scenes with high sensitivity requirements, an alarm will be issued as long as a threshold is reached; The internal pressure corresponds to the pressure situation. Specifically, the above example is only used as a preferred example of the traffic early warning method provided by the embodiment of the present application, and is subject to the implementation of the traffic early warning method provided by the embodiment of the present application, and is not specifically limited. Step S105, obtaining congestion information in a special area according to the pressure threshold, and confirming and marking the congestion information to generate marked data;

Optionally, in step S105, the congestion information of the special area is obtained according to the pressure threshold, and the congestion information is confirmed and marked, and the generation of marked data includes: obtaining the real-time congestion situation of the area according to the pressure alarm of the pressure threshold and giving an alarm, and manually confirming the real-time Congestion, mark the correct perception results, and generate marked data; among them, manual confirmation, real-time or offline confirmation through monitoring and traffic index comparison.

Specifically, based on the weight of the correlation coefficient in S103, a weight threshold between 0 and 1 is defined according to the congestion alarm business requirements, and the target path that meets the weight requirements is considered to be relatively similar to the traffic characteristics of the central building at the moment when congestion identification is required. This path is more likely to transmit traffic pressure to the central building. When the traffic flow space density of the road section on the above-mentioned path reaches the warning threshold, it means that the path has input more traffic into the area, which will soon affect the traffic conditions inside the area. In order to reduce the inflow pressure, a congestion alarm can be issued to control the peripheral traffic flow and outflow. The regional traffic flow can be guided to guide the signal timing personnel to locate specific problems, and can cooperate with the implementation of measures such as increasing downstream release efforts to quickly relieve traffic pressure. At the same time, the target path is considered as a whole, and priority warnings are given for areas with multiple path congestion.

In the embodiment of the present application, the manual confirmation and marking of the sensing results is through real-time monitoring of regional traffic parameters, recording the alarm conditions that meet the pressure threshold, obtaining real-time congestion alarms, and confirming the data alarms through manual observation of intersection monitoring and other methods. Then get the congestion alarm accuracy label.

Step S106, according to the marked data and traffic data, train the congestion prediction model, and use the congestion prediction model to give an early warning of the traffic situation in the special area.

Optionally, in step S106, according to the marked data and traffic data, training the congestion prediction model, and using the congestion prediction model to warn the traffic situation in the special area includes: sorting the key traffic detector data in the target path by time, performing data Process and obtain the traffic status indicators of the target path, and associate the congestion label of the marked data; input the neural network time series model for training, and obtain the time series congestion early warning model; use the state estimation algorithm to adjust the time series prediction results obtained through the test set, and train Obtain the congestion prediction model of the special area; according to the congestion prediction model, the traffic situation of the special area is given an early warning.

Specifically, the number of road sections included in different paths is different, and the number of key detectors in different road sections on the path is different. Therefore, it is necessary to aggregate and calculate the detector data, and calculate the flow sum, average occupancy rate, and average traffic flow space density in units of road sections and other indicators, and then calculate the average value of traffic indicators in units of routes as the route indicators. Input the above path index data in order of time, build a congestion model in a special area, mark the congestion label, and use LSTM, RNN, GRU and other neural network time series models to perform cyclic calculations to obtain the data of the traffic monitor before the congestion occurs. Threshold, the threshold data record label, that is, the congestion prediction label of the special area is obtained. Using state estimation methods such as Kalman time update equation and least squares method, the time series prediction results obtained through the test set are adjusted to obtain the prediction results of the mixed model.

In the embodiment of this application, the key detector data on the key path is used for learning and training, and the detectors that have little influence on the regional traffic conditions are excluded. On the premise of ensuring the accuracy of the model, the input sample size can be reduced, and the training pressure can be reduced. Reduce overfitting.

In the embodiment of this application, the time-series model is based on a sequence of times at a series of moments, by reading the variables of the traffic flow data monitored in the traffic detector at each moment, and integrating the acquired discrete data to obtain a time-series Model;

In the embodiment of this application, splitting the data set belongs to the existing technology in the AI training and learning system, including splitting the acquired data set into a test set and a training set and performing tests and training on the samples. The split test The set and the training set are mutually exclusive, that is, the test samples should not appear in the training set as much as possible, and have not been used in the training set to ensure the accuracy of the time series congestion early warning model;

In the embodiment of this application, in predicting the hybrid model according to the Kalman time update equation, set X(k) as the system state at time k, U(k) as the control amount of the system at time k, and A and B as System parameters, A is the traffic flow spatial density change parameter, B is the traffic flow control parameter, for the mixed model, they are a matrix, Z(k) is the measured value at time k, H is the parameter of the measurement system, W(k ) and V(k) represent the noise of the process and measurement respectively, and the covariance is represented by Q and R respectively, and Q and R are set not to change with the system state, and a set of linear stochastic differential equations are introduced:

X(k)=AX(k-1)+BU(k)+W(k)

System measurement value Z(k)=HX(k)+V(k);

First, predict the next state according to the process model of the system; set the system state at a certain moment as k, and predict the state at a certain moment according to the previous state of the system model:

X(k|k-1)=AX(k-1|k-1)+BU(k) (1)

X(k|k-1) is the result predicted by the previous state, X(k-1|k-1) is the optimal result of the previous state, U(k) is the control amount of the state at a certain moment, if there is no control amount, it can be 0;

Up to a certain moment, the system result has been updated, but the covariance corresponding to X(k|k-1) has not been updated, denoted by P;

P(k|k-1)=AP(k-1|k-1)+A'+Q (2)

P(k|k-1) is the covariance corresponding to X(k|k-1), A' represents the transposition matrix of A, and Q is the covariance of the system process. The comprehensive calculation formulas (1) and (2) can be According to X(k|k-1), the predicted result of the hybrid model is obtained and compared with the preset threshold. If the calculation result exceeds the threshold, it means that the predicted result will be congested in the next period, and a predicted congestion alarm will be issued at this time.

The traffic early warning method provided by the embodiment of the present application can realize the real-time monitoring of the traffic flow spatial density data of the surrounding road sections with a special area as the center, and can quickly determine the type of traffic pressure and the problematic road section when congestion occurs, and realize the detection of excessive traffic flow space density. Real-time and timely congestion awareness and fast processing and relief of problem road sections to ensure smooth traffic around special areas. And it can independently adjust the congestion perception threshold around the special area, which is helpful to observe and control the traffic situation in the whole area, and can calculate and warn the traffic congestion in time through the data of the traffic detector, guide the signal timing and personnel positioning processing in advance, and can Cooperate with the implementation of measures such as increasing the intensity of downstream release, and quickly ease the traffic pressure on the whole. Through real-time data input, real-time alarms can be realized for road networks in special areas, and congestion can be predicted in advance based on real-time data and data matching and calculation of the offline model library, which facilitates the command center to grasp the traffic conditions in special areas in real time, and to respond and command in advance , Do traffic diversion to the congestion point, reduce the occurrence of congestion, and ensure the smooth flow of traffic. And, in the time series forecasting model, there may be some errors between the forecasted series and the real time series. By re-estimating and optimizing the sequence with errors and re-adjusting the forecast sequence dynamically, the prediction accuracy can be improved.

An embodiment of the present invention provides a traffic early warning method. By marking special areas, establish road network topology maps corresponding to special areas; associate urban road network models, collect traffic data in special areas; build path libraries, analyze path correlations in special areas combined with traffic data, and screen key traffic detectors; The data collected by key traffic detectors is converted into traffic flow spatial density, and the pressure threshold of regional traffic warning is set according to the traffic flow spatial density; the congestion information of special areas is obtained according to the pressure threshold, and the congestion information is confirmed and marked to generate marked data ;According to the marked data and traffic data, train the congestion prediction model, and use the congestion prediction model to warn the traffic situation in special areas, so that the input sample size can be reduced, the training pressure can be reduced, and the over-fitting can be reduced under the premise of ensuring the accuracy of the model. Combined technical effect.

Embodiment two

In a second aspect, an embodiment of the present invention provides a traffic warning device. FIG. 3 is a schematic diagram of a traffic warning device provided in Embodiment 2 of the present invention; as shown in FIG. 3 , the traffic warning device provided in the embodiment of the present application includes: Build module 31, be used for marking special area, establish the road network topological map corresponding to special area; Acquisition module 32, be used for associating urban road network model, collect the traffic data of special area; Screening module 33, be used for by constructing path library, Combining traffic data to analyze path correlation in a special area, and screening key traffic detectors; conversion module 34, used to convert data collected by key traffic detectors into traffic flow spatial density, and set the pressure of regional traffic early warning according to traffic flow spatial density Threshold; Marking module 35, is used for obtaining the congestion information of special area according to pressure threshold value, and confirms and marks the congestion information, generates label data; Early warning module 36, is used for training congestion prediction model according to label data and traffic data, and Early warning of traffic conditions in special areas through the congestion prediction model.

An embodiment of the invention provides a traffic warning device. By marking special areas, establish road network topology maps corresponding to special areas; associate urban road network models, collect traffic data in special areas; build path libraries, analyze path correlations in special areas combined with traffic data, and screen key traffic detectors; The data collected by key traffic detectors is converted into traffic flow spatial density, and the pressure threshold of regional traffic warning is set according to the traffic flow spatial density; the congestion information of special areas is obtained according to the pressure threshold, and the congestion information is confirmed and marked to generate marked data ;According to the marked data and traffic data, train the congestion prediction model, and use the congestion prediction model to warn the traffic situation in special areas, so that the input sample size can be reduced, the training pressure can be reduced, and the over-fitting can be reduced under the premise of ensuring the accuracy of the model. Combined technical effect.

Embodiment three

In a third aspect, an embodiment of the present invention provides a traffic early warning system. FIG. 4 is a schematic diagram of a traffic early warning system provided by Embodiment 3 of the present invention; as shown in FIG. 4 , the traffic early warning system provided by the embodiment of the present application includes: Marking module 41 is used to divide and mark special areas; data collection module 42 is used to collect road network models and traffic data of key road sections; data conversion module 43 is used to convert and calculate collected data; key path monitoring Module 44 is used to monitor the real-time overall traffic status of key routes in the area and display the heat map based on the correlation weight; the real-time congestion perception module 45 is used to issue real-time alarms to the road network in special areas and manually confirm the alarm situation; congestion The early warning module 46 inputs real-time data into the offline model library for congestion prediction and sends out early warning signals.

Optionally, the data conversion module 43 includes: a length data conversion module, which is used to calculate the approximate value of the current queuing length for the sum of the length of the standard small car and the distance between the fronts, and compare the result with the length of the road section to obtain the first traffic flow of the road section Spatial density; speed data conversion module, used to compare the time interval of two consecutive vehicles passing through a fixed point with the free flow speed of the road section to obtain the distance interval, and compare the distance interval with the length of the road section to obtain the second traffic flow spatial density of the road section .

Specifically, as shown in Figure 4, the marking module 41 is recorded as a marking module, the data acquisition module 42 is recorded as a data acquisition module, the data conversion module 43 is recorded as a data conversion module, and the key path monitoring module 44 is marked as a key path monitoring module. The congestion sensing module 45 is referred to as a real-time congestion sensing module, and the congestion warning module 46 is referred to as a congestion warning module.

The traffic early warning system provided by the embodiment of the present application first confirms the special area, searches for the shortest path and divides the key road sections of the area with the special area as the center, and then further collects roads, road sections, intersections, lanes, traffic, occupancy, headway, queuing Length data, and then calculate and integrate the data through normalization and weighted average to obtain the overall traffic flow space density of the road section, and then match the data with the traffic parameters inside the traffic detector, where the node is transmitted to the area When the traffic flow space density reaches the early warning threshold in the road section of traffic flow, it means that the upstream road section has input more traffic into the area, which will soon affect the traffic conditions inside the area. For the inflow pressure, early warning can be carried out to guide the peripheral traffic flow in advance ;

Through real-time data input, real-time alarms can be realized for road networks in special areas, and congestion can be predicted in advance based on real-time data and data matching and calculation of the offline model library, which is convenient for the command center to respond and command in advance, and to conduct traffic relief for congestion points in advance , greatly reducing the occurrence of congestion and ensuring smooth traffic;

At the same time, when the traffic flow space density of the road section adjacent to the special area reaches the threshold value, it means that the traffic pressure around the special area is high. The detector can be used to warn the traffic problem, and the signal timing personnel can be guided to locate the specific problem, and It can cooperate with the implementation of measures such as increasing the intensity of downstream release to quickly relieve traffic pressure.

The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It should be understood that each procedure and/or block in the flowchart and/or block diagram, and a combination of procedures and/or blocks in the flowchart and/or block diagram can be realized by computer program instructions. These computer program instructions may be provided to a general purpose computer, special purpose computer, embedded processor, or processor of other programmable data processing equipment to produce a machine such that the instructions executed by the processor of the computer or other programmable data processing equipment produce a An apparatus for realizing the functions specified in one or more procedures of the flowchart and/or one or more blocks of the block diagram.

These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing apparatus to operate in a specific manner, such that the instructions stored in the computer-readable memory produce an article of manufacture comprising instruction means, the instructions The device realizes the function specified in one or more procedures of the flowchart and/or one or more blocks of the block diagram.

These computer program instructions can also be loaded onto a computer or other programmable data processing device, causing a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process, thereby The instructions provide steps for implementing the functions specified in the flow chart or blocks of the flowchart and/or the block or blocks of the block diagrams.

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the protection scope of the present invention.

Claims (10)

1. A traffic early warning method is characterized by comprising the following steps:

marking a special area, and establishing a road network topological graph corresponding to the special area;

associating the urban road network model, and acquiring traffic data of the special area;

constructing a path library, analyzing the path correlation in the special area by combining the traffic data, and screening a key traffic detector;

converting data collected by the key traffic detector into traffic flow space density, and setting a pressure threshold value of regional traffic early warning according to the traffic flow space density;

acquiring congestion information of the special area according to the pressure threshold, confirming and marking the congestion information, and generating marking data;

and training a congestion prediction model according to the marking data and the traffic data, and early warning the traffic condition of the special area through the congestion prediction model.

2. The traffic early warning method according to claim 1, wherein the marking of the special area and the establishing of the road network topological graph corresponding to the special area comprise:

marking the special area, wherein the special area is an area covered by buildings or facilities with traffic demands larger than a preset threshold value in a specific time period and road facilities in a preset adjacent range around the buildings or facilities;

and screening the central nodes of the special area, setting intersections in a preset adjacent range of the central nodes as nodes, taking the nodes as centers, taking all associated road sections of the nodes as road sections close to the central nodes, and establishing the road network topological graph of the special area.

3. The traffic early warning method according to claim 2, wherein the constructing a route library, analyzing the relevance of the route in the special area in combination with the traffic data, and screening a key traffic detector comprises:

screening passing vehicles around the central node based on historical traffic detector data, tracking vehicle tracks, reversely deducing driving paths of the vehicles in the special area, and constructing a path library;

on the basis of historical traffic data, flow data of the paths in a specified period are counted to obtain flow characteristics of each path; counting flow data flowing into the central node in a specified period based on real-time traffic data to obtain the flow characteristics of the central node;

judging the correlation between the path and the central node according to the correlation coefficient of the flow characteristics of the path and the flow characteristics of the central node, and screening a target path according to the correlation;

and considering the influence of the path flow on the regional congestion for the target paths corresponding to the correlations according to the weights corresponding to the target paths, and selecting the traffic detector installed on the flow direction in the road section as the key traffic detector according to the flow direction on the paths.

4. The traffic warning method according to claim 3, further comprising:

and for the road section lacking the bayonet device in the path, searching the shortest path by adopting a path searching algorithm and completing the path.

5. The traffic-warning method of claim 3, wherein converting the data collected by the critical traffic detector into a spatial density of traffic flow comprises:

calculating an approximate value of the current queuing length according to the sum of the length of the standard minicar and the distance between the heads by using the flow or the distance between the heads, and comparing the approximate value of the current queuing length with the length of the road section to obtain the first traffic flow space density of the road section;

multiplying the time interval acquired by the key traffic detector in the fixed point by the road section free flow speed by using the headway distance or the queue length to obtain a distance interval, and comparing the distance interval with the road section length to obtain a second traffic flow space density of the road section;

and carrying out normalization processing on the first traffic flow space density and the second traffic flow space density, and carrying out weighted average on data after normalization processing to obtain the target traffic flow space density of the road section.

6. The traffic early warning method according to claim 5, wherein the setting of the pressure threshold of the regional traffic early warning according to the spatial density of the traffic flow comprises:

classifying traffic pressure according to the traffic flow space density, and setting a pressure threshold value of the regional traffic early warning;

wherein the traffic pressure classification includes an inflow pressure and an internal pressure; the pressure threshold values are provided in a plurality of groups.

7. The traffic early warning method according to claim 6, wherein the acquiring congestion information of the special area according to the pressure threshold, and confirming and marking the congestion information, and the generating marking data comprises:

according to the pressure alarm of the pressure threshold value, acquiring a real-time congestion situation of an area, giving an alarm, manually confirming the real-time congestion situation, marking a correct sensing result, and generating marking data;

and the manual confirmation is carried out in real time or off-line confirmation through monitoring and traffic index comparison.

8. The traffic early warning method according to claim 7, wherein the training of a congestion prediction model according to the labeled data and the traffic data and the early warning of the traffic condition in the special area through the congestion prediction model comprises:

sequencing the key traffic detector data in the target path according to time, performing data processing to obtain a traffic state index of the target path, and associating a congestion tag of the marked data;

inputting a neural network time sequence model for training to obtain a time sequence congestion early warning model;

adjusting the time series prediction result obtained through the test set by using a state estimation algorithm, and training to obtain a congestion prediction model of the special area;

and early warning the traffic condition of the special area according to the congestion prediction model.

9. A traffic early warning device, comprising:

the system comprises an establishing module, a judging module and a judging module, wherein the establishing module is used for marking a special area and establishing a road network topological graph corresponding to the special area;

the acquisition module is used for associating the urban road network model and acquiring traffic data of the regional road network model in the road network topological graph;

the screening module is used for constructing a path library, analyzing the path correlation in the special area by combining the traffic data and screening a key traffic detector;

the conversion module is used for converting the data acquired by the key traffic detector into traffic flow space density and setting a pressure threshold value of regional traffic early warning according to the traffic flow space density;

the marking module is used for acquiring the congestion information of the special area according to the pressure threshold, confirming and marking the congestion information and generating marking data;

and the early warning module is used for training a congestion prediction model according to the marking data and the traffic data, and early warning the traffic condition of the special area through the congestion prediction model.

10. A traffic warning system, comprising:

the marking module is used for marking the special area in a dividing way;

the data acquisition module is used for acquiring the road network model of the key road section and traffic data;

the data conversion module is used for carrying out conversion calculation on the acquired data;

the key path monitoring module is used for monitoring the real-time overall traffic state of the key paths in the area and displaying the thermodynamic diagram based on the relevance weight;

the real-time congestion sensing module is used for giving an alarm to a road network in a special area in real time and confirming the alarm condition manually;

the congestion early warning module is used for inputting the real-time data into an offline model library to predict congestion and sending out an early warning signal;

wherein, the data conversion module comprises:

the length data conversion module is used for calculating an approximate value of the current queuing length according to the sum of the length of the standard compact car and the distance between heads of the compact car, and comparing the result with the length of the road section to obtain the first traffic flow space density of the road section;

and the speed data conversion module is used for comparing the time interval of two continuous vehicles passing through the fixed point with the road section free flow speed to obtain a distance interval, and comparing the distance interval with the road section length to obtain a second traffic flow space density of the road section.

Images