Real-time traffic route dynamic optimization system and method based on multi-source data fusion

Abstract

The invention belongs to the field of intelligent traffic, and discloses a real-time traffic route dynamic optimization system and method based on multi-source data fusion, and the method comprises the steps: collecting the information of N traffic road segments, the information of the traffic road segments comprises the traffic density, the vehicle proportion, the vehicle speed, the road segment width and other data; comprehensively analyzing the traffic road section information, generating a traffic congestion coefficient, and judging whether the corresponding road section is congested or not according to the traffic congestion coefficient; inputting the number of the congested nth road section into a pre-constructed traffic network diagram structure to obtain a road section with a connection relation of the nth road section, and marking the road section as an associated road section; obtaining associated feature data of the corresponding associated road section; the spreading prediction model outputs the spreading time of the congested road section, namely the spreading time of reaching each associated road section; and according to the spreading time of the corresponding associated road section, the passing vehicles are timely reminded to avoid, further deterioration of congestion is avoided, and the traffic efficiency is greatly improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of intelligent transportation, more particularly, the present application relates to a real-time traffic routing dynamic optimization system and method based on multi-source data fusion. BACKGROUND

[0002] The prior art such as the Chinese patent with the authorized announcement number CN111024110B discloses a 5G-based intelligent traffic path planning and navigation system and its working method, the user intelligent terminal is used for sending path request to the navigation server, and receiving path information and displaying dynamic navigation process, the navigation server is used for path calculation and pushing settlement result to the user intelligent terminal, the edge computing service MAP intelligent terminal is used for providing congestion situation, path rectification, flow warning and other information, and providing data support for nearby intelligent terminals, relying on edge computing for intelligent driving route planning, realizing intelligent recommended path and real-time traffic information acquisition, so as to achieve intelligent travel.

[0003] The prior art still has the following problems:

[0004] The prior art mainly describes the static planning problem of traffic path, and the traffic congestion in reality has obvious spreading characteristics, that is, the congestion of a road section will affect the adjacent road sections in time; however, the existing prediction model often lacks the prediction of the "spreading" effect of such congestion, resulting in inaccurate prediction of future traffic conditions, so that the traffic manager cannot well provide suggestions for traffic participants to avoid congested road sections, leading to the "spreading" effect of such congestion, further deterioration, and reduction of traffic efficiency.

[0005] In view of this, the present application provides a real-time traffic routing dynamic optimization system and method based on multi-source data fusion to solve the above problems. SUMMARY

[0006] In order to overcome the above-mentioned defects of the prior art, in order to achieve the above-mentioned purpose, the present application provides the following technical scheme: a real-time traffic routing dynamic optimization method based on multi-source data fusion, comprising:

[0007] Collecting N pieces of traffic section information, N being an integer greater than or equal to 1, the traffic section information including traffic density, vehicle proportion, vehicle speed, road section width and other data;

[0008] Comprehensively analyzing the traffic section information, generating a traffic congestion coefficient, and determining whether the corresponding road section is congested according to the traffic congestion coefficient;

[0009] Inputting the number of the nth congested road section into the pre-constructed traffic network graph structure, obtaining the road sections having a connection relationship with the nth road section and marking them as associated road sections; and obtaining the associated feature data of the corresponding associated road sections.

[0010] Further, the traffic congestion coefficient is composed of the ratio of the weighted sum of the jam-promoting factors and the weighted sum of the jam-relieving factors: the jam-promoting factors are the weighted sum of the normalized traffic density and the vehicle proportion, and the jam-relieving factors are the weighted sum of the normalized vehicle speed and the road width, and the specific formula is as follows:

[0011]

[0012] In the formula, JT n is the traffic congestion coefficient of the nth road segment, n∈N; MD n is the traffic density of the nth road segment; BL n is the vehicle proportion of the nth road segment; CS n is the vehicle speed of the nth road segment; KD n is the road width of the nth road segment; and are preset weights, which are values greater than 0.

[0013] Further, the method for constructing a traffic network graph structure comprises:

[0014] obtaining road network basic data and the connection relationship of each road segment, the road network basic data including the number, length and endpoints of each road segment, the endpoints including the starting point and the ending point; the connection relationship of each road segment is directly obtained from map data;

[0015] the nodes in the traffic network graph structure are the endpoints of each road segment; the edges in the traffic network graph structure are the connection relationship between road segments, the road segment length and the road segment width.

[0016] Further, the training method of the spread prediction model comprises:

[0017] pre-collecting j groups of spread analysis data, j being an integer greater than 1, the spread analysis data including spread feature data and spread time corresponding to the spread feature data; the spread feature data including the k time series traffic congestion coefficients and the associated feature data of the corresponding associated road segment;

[0018] converting the spread analysis data into a corresponding group of third feature vectors; taking the third feature vectors as the input of the spread analysis data, taking the spread time corresponding to each group of spread feature data as the output, and taking the actual spread time corresponding to each group of spread feature data as the prediction target; taking the minimization of the sum of the prediction errors of all spread feature data as the training target; training the spread prediction model until the sum of the prediction errors reaches convergence, and then stopping the training; the spread prediction model is a deep neural network model.

[0019] Further, the method for determining whether the corresponding road segment is congested according to the traffic congestion coefficient comprises:

[0020] The traffic congestion coefficient is compared with a preset congestion coefficient threshold value, if the traffic congestion coefficient is greater than or equal to the preset congestion coefficient threshold value, it is determined that the corresponding road section is congested, if the traffic congestion coefficient is less than the preset congestion coefficient threshold value, it is determined that the corresponding road section is normal.

[0021] Further, the traffic density acquisition method comprises:

[0022] Collecting pictures in the monitoring area of the road section monitoring camera, and marking as traffic density pictures;

[0023] Inputting the traffic density pictures into the trained vehicle recognition model, and outputting the number of vehicles and the number of large vehicles of the corresponding road section; taking the ratio of the number of vehicles to the actual area of the monitoring area of the corresponding road section monitoring camera as the traffic density of the corresponding road section.

[0024] Further, the vehicle proportion is the ratio of the number of large vehicles to the number of vehicles; the vehicle speed is the average value of the vehicle speed collected by the road section speedometer in a unit time.

[0025] Further, the correlation feature data comprises the average value and the standard deviation of the road section length, the road section width and the traffic congestion coefficient.

[0026] The average value and the standard deviation of the traffic congestion coefficient are acquired by the method comprising:

[0027] Collecting k time series traffic congestion coefficients of the corresponding correlation road section at a preset interval time from the starting point, establishing a congestion coefficient analysis set, and calculating the average value and the standard deviation of the congestion coefficient analysis set.

[0028] Further, when the nth road section is congested, the time series traffic congestion coefficients collected in the predetermined time of the nth road section and the green light interval time of the traffic signal at the intersection node of the nth road section are input into the pre-trained correlation analysis model, and the correlation result of the current congestion and the green light interval time is output; the correlation result comprises correlation and non-correlation; if the correlation result is non-correlation, the correlation road section is acquired.

[0029] If the correlation result is correlation, the time series traffic congestion coefficients collected in the predetermined time of the corresponding road section and the green light interval time of the traffic signal at the intersection node of the corresponding road section are input into the pre-trained signal lamp adjustment model, and the adjusted green light interval time is output.

[0030] The real-time traffic routing dynamic optimization system based on multi-source data fusion is used to realize the real-time traffic routing dynamic optimization method based on multi-source data fusion, and the system comprises:

[0031] The acquisition module is used to acquire N traffic road section information, N is an integer greater than or equal to 1, and the traffic road section information comprises traffic density, vehicle proportion, vehicle speed and road section width.

[0032] The first analysis module is used for comprehensive analysis on the traffic section information, generates a traffic congestion coefficient, and judges whether the corresponding section is congested according to the traffic congestion coefficient;

[0033] The fourth analysis module inputs the collected time series traffic congestion coefficient of the nth section within a predetermined time and the green light interval duration of the traffic signal at the intersection node of the nth section into a pre-trained correlation analysis model when the nth section is congested, and outputs the correlation result of the current congestion and the green light interval duration; the correlation result includes correlation and non-correlation; if the correlation result is non-correlation, the correlation section is obtained;

[0034] The fifth analysis module inputs the collected time series traffic congestion coefficient of the corresponding section within a predetermined time and the green light interval duration of the traffic signal at the intersection node of the corresponding section into a pre-trained signal light adjustment model if the correlation result is correlation, and outputs the adjusted green light interval duration.

[0035] The second analysis module inputs the nth section number of the congestion into a pre-constructed traffic network graph structure if the correlation is non-correlation, obtains the sections having a connection relationship with the nth section, and marks them as correlation sections; and obtains the correlation feature data of the corresponding correlation sections;

[0036] The third analysis module inputs the k time series traffic congestion coefficients collected at a preset interval duration and the correlation feature data of the corresponding correlation sections into a pre-trained spread prediction model, and obtains the spread time of the corresponding correlation sections, which is the time required for the congestion of the nth section to spread to the corresponding correlation sections.

[0037] The real-time traffic routing dynamic optimization system and method based on multi-source data fusion have the following technical effects and advantages:

[0038] The present application comprehensively analyzes the specific numerical values of the collected traffic section information, generates a traffic congestion coefficient to judge whether the corresponding section is congested, and objectively shows that the traffic density, vehicle proportion, vehicle speed and section width directly affect road congestion; by analyzing the traffic section information and generating the traffic congestion coefficient, the influence of traffic factors on road congestion is intuitively presented, which helps to accurately judge the congestion situation.

[0039] The spread prediction model is constructed based on a deep network, the correlation sections having a connection relationship with the congested section, i.e., adjacent sections, are found out, the spread time of the congested section to each correlation section is predicted through the spread prediction model; the spread prediction model outputs the spread propagation time of the congested section, i.e., the spread time to each correlation section; according to the spread time of the corresponding correlation section, the passing vehicles are timely reminded to avoid, so as to avoid further deterioration of the congestion and greatly improve the traffic efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0040] Figure 1 The schematic diagram of the real-time traffic routing dynamic optimization system based on multi-source data fusion of the embodiment 1 of the present application;

[0041] Figure 2 The schematic diagram of the real-time traffic routing dynamic optimization system based on multi-source data fusion of the embodiment 2 of the present application;

[0042] Figure 3 The flow chart of the real-time traffic routing dynamic optimization method based on multi-source data fusion of the embodiment 2 of the present application;

[0043] Figure 4 The flow chart of the real-time traffic routing dynamic optimization method based on multi-source data fusion of the embodiment 3 of the present application. DETAILED DESCRIPTION

[0044] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all the other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the present application.

[0045] Embodiment 1

[0046] Please refer to Figure 1 As shown in the figure, the real-time traffic routing dynamic optimization system based on multi-source data fusion described in the embodiment includes a collection module, a first analysis module, a second analysis module and a third analysis module. Each module is connected through wired and / or wireless mode to realize data transmission.

[0047] The collection module is used to collect N pieces of traffic section information, where N is an integer greater than or equal to 1, and the traffic section information includes traffic density, vehicle proportion, vehicle speed, section width and other data. The method for obtaining traffic density includes:

[0048] The collection module collects pictures in the monitoring area demarcated by the section monitoring camera and marks them as traffic density pictures. It should be noted that the section monitoring camera is a section monitoring camera with a length of r from the intersection node.

[0049] The traffic density pictures are input into the trained vehicle recognition model to output the number of vehicles and the number of large vehicles corresponding to the section. The ratio of the number of vehicles to the actual area of the monitoring area demarcated by the corresponding section monitoring camera is taken as the traffic density of the corresponding section. The higher the traffic density, the more vehicles, and the smaller the distance between vehicles, which can easily lead to congestion. When the vehicles are too dense, the driver needs to reduce the speed to maintain a safe distance, resulting in a decrease in overall vehicle speed.

[0050] The vehicle proportion is a ratio of the number of large vehicles, such as trucks, buses and the like, or vehicles defined as exceeding preset length and width, to the number of vehicles, and the large vehicles occupy more road section space, start and accelerate slowly, and to some extent, reduce the road section passing efficiency and increase the congestion risk.

[0051] The vehicle speed is an average value of vehicle speed collected by a road section speedometer in a unit time, and the road section speedometer is a road section speedometer installed at a distance of r length from a road intersection node to ensure the consistency of data obtained by each road section.

[0052] The road section width is directly obtained through road construction data, and the narrower the road section width, the smaller the lateral distance between vehicles, the more cautious the driver needs to be, the driving speed will be reduced, and the speed will be reduced due to insecurity, resulting in a higher risk of congestion on the road section.

[0053] The first analysis module is configured to comprehensively analyze the traffic section information, generate a traffic congestion coefficient, and determine whether the corresponding road section is congested according to the traffic congestion coefficient; the traffic congestion coefficient is composed of a ratio of a weighted sum of congestion promoting factors to a weighted sum of congestion alleviating factors: the congestion promoting factors are a weighted sum of normalized traffic density and vehicle proportion, and the congestion alleviating factors are a weighted sum of normalized vehicle speed and road section width, and the specific formula is as follows:

[0054]

[0055] In the formula, JT n is the traffic congestion coefficient of the nth road section, n∈N; MD n is the traffic density of the nth road section; BL n is the vehicle proportion of the nth road section; CS n is the vehicle speed of the nth road section; KD n is the road section width of the nth road section; and are preset weights, which are values greater than 0, and the larger the traffic congestion coefficient, the more congested the corresponding road section, and vice versa.

[0056] The method for determining whether the corresponding road section is congested according to the traffic congestion coefficient comprises the following steps:

[0057] The traffic congestion coefficient is compared with a preset congestion coefficient threshold value, if the traffic congestion coefficient is greater than or equal to the preset congestion coefficient threshold value, it is determined that the corresponding road section is congested; if the traffic congestion coefficient is less than the preset congestion coefficient threshold value, it is determined that the corresponding road section is normal.

[0058] The second analysis module inputs the congested nth road section number into the pre-constructed traffic network graph structure, obtains road sections having a connection relationship with the nth road section, and marks the road sections as associated road sections. The number of associated road sections can be multiple, and is not limited herein. The associated feature data of the corresponding associated road sections is obtained.

[0059] The third analysis module inputs the k time series traffic congestion coefficients collected at the preset interval time length and the associated feature data of the corresponding associated road sections into the pre-trained spread prediction model, obtains the spread time of the corresponding associated road sections, and determines the time when the nth road section is congested as the starting point. The spread time is the time required for the congestion of the nth road section to spread to the corresponding associated road sections. When there are multiple associated road sections, the spread prediction model is inputted multiple times in sequence to obtain the spread time of the corresponding associated road sections.

[0060] The method for constructing the traffic network graph structure comprises the following steps:

[0061] The connection relationship of each road section is obtained from the road network basic data. The road network basic data comprises the number, length and endpoints of each road section. The endpoints comprise the starting point and the ending point. It should be noted that the road section in the embodiment refers to each section of the one-way road. For example, a road is a two-way road, and each section of the road on both sides (a section refers to a road section between two intersections) needs to be numbered, that is, the number of the road section. The connection relationship of each road section can be directly obtained from the map data (GIS data).

[0062] The nodes in the traffic network graph structure are the endpoints of each road section. The edges (lines) in the traffic network graph structure are the connection relationship between the road sections, the length of the road section and the width of the road section. The nodes and edges in the traffic network graph structure are shown as follows:

[0063] The starting point of the A road section is A1, and the ending point is A2. The starting point of the B road section is B1, and the ending point is B2. A1, A2, B1 and B2 are nodes in the traffic network graph structure. The ending point A2 of the A road section is connected with the starting point B1 of the B road section (that is, the vehicle of the A road section can pass to the B road section, but the vehicle of the B road section cannot pass to the A road section), that is, the connection relationship between the road sections.

[0064] The associated feature data comprises the length of the road section, the width of the road section, the mean value and the standard deviation of the traffic congestion coefficient.

[0065] The method for obtaining the mean value and the standard deviation of the traffic congestion coefficient comprises the following steps:

[0066] The k time series traffic congestion coefficients of the corresponding associated road sections are collected at the preset interval time length from the starting point, a congestion coefficient analysis set is established, the mean value and the standard deviation of the congestion coefficient analysis set are calculated, and the mean value and the standard deviation can reflect the dispersion degree of the traffic condition of the corresponding associated road sections with the change of time.

[0067] The training method of the spread prediction model comprises:

[0068] j sets of spread analysis data are collected in advance, j is an integer greater than 1, the spread analysis data comprises spread characteristic data and a spread time corresponding to the spread characteristic data, and the spread characteristic data comprises the k time-series traffic congestion coefficients and the characteristic data of the corresponding associated road section.

[0069] The spread analysis data is converted into a corresponding set of third feature vectors, the third feature vectors are taken as the input of the spread analysis data, the spread analysis data takes the spread time corresponding to each set of spread characteristic data as the output, takes the actual spread time corresponding to each set of spread characteristic data as the prediction target, takes the minimization of the sum of the prediction errors of all the spread characteristic data as the training target, and the calculation formula of the prediction error is wherein ν j is the prediction error of the jth set of spread analysis data, is the predicted spread time corresponding to the jth set of spread characteristic data, β j is the actual spread time corresponding to the jth set of spread characteristic data; the spread prediction model is trained until the sum of the prediction errors reaches convergence, and the training is stopped; and the spread prediction model is a deep neural network model.

[0070] In this embodiment, the collected specific traffic road section information is comprehensively analyzed to generate a traffic congestion coefficient to determine whether the corresponding road section is congested, and the direct influence of the road section on road congestion due to traffic density, vehicle proportion, vehicle speed and road section width is objectively presented. By analyzing the traffic road section information and generating the traffic congestion coefficient, the influence of traffic factors on road congestion is intuitively presented, which helps to accurately determine the congestion situation.

[0071] A spread prediction model is constructed based on a deep learning network, the associated road sections having a connection relationship with the congested road section, i.e. adjacent road sections, are found out, the spread time of the congested road section to each associated road section is predicted through the spread prediction model, the spread prediction model outputs the spread propagation time of the congested road section, i.e. the spread time to each corresponding associated road section, and the passing vehicles are timely reminded to avoid, so as to avoid further deterioration of the congestion and greatly improve the traffic efficiency.

[0072] Embodiment 2

[0073] Please refer to Figure 2 and Figure 3 The real-time traffic routing dynamic optimization system based on multi-source data fusion described in this embodiment is used to further analyze whether the cause of the congestion is associated with the green light interval length before performing the spread time analysis when the road section is congested, if not associated, the spread time analysis is performed, if associated, the green light interval length adjustment is performed; the system further comprises a fourth analysis module and a fifth analysis module.

[0074] The fourth analysis module inputs the collected time series traffic congestion coefficient (which can reflect the law of change of traffic congestion coefficient over time) of the nth road section within the predetermined time and the green interval duration of the traffic signal at the intersection node of the nth road section into the pre-trained correlation analysis model when the nth road section is congested, and outputs the correlation result of this congestion and the green interval duration; the correlation result includes correlation and non-correlation. If the correlation result is non-correlation, return to the operation of the second analysis module.

[0075] It should be noted that the green interval duration directly affects the variable data change in the traffic road section information, such as the change of traffic density and vehicle speed. Taking the time series traffic congestion coefficient within the predetermined time and the green interval duration as the input data of the correlation analysis model can make the model understand the change of the time series traffic congestion coefficient within the predetermined time and the green interval duration. If the time series traffic congestion coefficient within the predetermined time presents a large fluctuation, the congestion caused this time may be caused by unreasonable green interval duration setting; otherwise, it is not correlated, and the congestion caused this time is caused by other factors, such as too large traffic flow, traffic accident, etc. Through the synchronous analysis of traffic congestion coefficient and green interval duration, it can be analyzed whether the cause of congestion is related to the setting of green interval duration.

[0076] The training method of the correlation analysis model includes:

[0077] Pre-collect b sets of correlation analysis data, b is an integer greater than 1, the correlation analysis data includes correlation feature data and correlation result corresponding to the correlation feature data, the correlation is marked as "1", and the non-correlation is marked as "0"; the correlation feature data includes time series traffic congestion coefficient within a unit time and green interval duration.

[0078] Convert the correlation feature data into a corresponding set of second feature vectors; take the second feature vector as the input of the correlation analysis model, the correlation analysis model takes the correlation result corresponding to each set of correlation feature data as the output, and takes the actual correlation result corresponding to each set of correlation feature data as the prediction target; take the sum of the prediction errors of all correlation feature data as the training target; wherein the calculation formula of the prediction error is Wherein v b is the prediction error of the bth set of correlation analysis data, is the predicted correlation result corresponding to the bth set of correlation feature data, β b is the actual correlation result corresponding to the bth set of correlation feature data; train the correlation analysis model until the sum of the prediction errors reaches convergence to stop training; the correlation analysis model is specifically a deep neural network model.

[0079] The fifth analysis module inputs the collected time series traffic congestion coefficients of the corresponding road section within the predetermined time and the green light interval length of the traffic signal at the intersection node of the corresponding road section into the pre-trained signal light adjustment model, and outputs the adjusted green light interval length.

[0080] The signal light adjustment model is trained in the same way as the correlation analysis model. It should be noted that before training the signal light adjustment model, the adjusted green light interval length is adjusted by the traffic management personnel when the road section is congested. When the road section is not congested, the adjusted green light interval length is used as the corresponding label of the time series traffic congestion coefficient of the road section within the predetermined time, i.e., the adjusted green light interval length. The training data of the signal light adjustment model is collected in the same way as above. The green light interval length in the input data of the signal light adjustment model reflects the normal traffic flow change at the intersection node, i.e., the green light interval length represents the road conditions of different road sections. The green light interval length as a basic restriction avoids the output of the adjusted green light interval length deviating from the actual situation, so it is used as the input data, which can better enable the signal light adjustment model to understand the relationship between the adjusted green light interval length and the green light interval length in the input data, avoid abnormal data output, and improve the prediction accuracy of the signal light adjustment model.

[0081] In this embodiment, the traffic congestion coefficient when the road section is congested and the green light interval length of the traffic signal are first input into the correlation analysis model based on deep network, and the correlation result of road congestion and the green light interval length of the traffic signal is output to determine whether the road congestion is correlated with the green light interval length of the traffic signal. If the correlation exists, the green light interval length of the traffic signal is adjusted through a model similar to the correlation analysis model. The correlation analysis model can effectively identify the correlation between congestion and green light interval, and optimize the setting of the traffic signal according to the result. The signal light adjustment model is used to dynamically adjust the traffic signal at the intersection node of the congested road section in real time, so as to reduce the probability of road congestion. The source of congestion is found in time, and the congestion is solved in time to avoid further spread of the congestion.

[0082] Embodiment 3

[0083] Please refer to Figure 4 The real-time traffic routing dynamic optimization method based on multi-source data fusion described in this embodiment is as follows:

[0084] N pieces of traffic section information are collected, N is an integer greater than or equal to 1, and the traffic section information includes traffic density, vehicle proportion, vehicle speed, road section width, etc.

[0085] The traffic section information is comprehensively analyzed to generate a traffic congestion coefficient, and whether the corresponding road section is congested is determined according to the traffic congestion coefficient.

[0086] The number of the nth congested road section is input into the pre-constructed traffic network graph structure, the road sections having a connection relationship with the nth road section are obtained and marked as associated road sections, and the associated feature data of the corresponding associated road sections is obtained;

[0087] The time when the nth road section is determined to be congested is taken as a starting point, and k time series traffic congestion coefficients and the associated feature data of the corresponding associated road sections are collected at a preset interval length, which are input into the pre-trained spread prediction model to obtain the spread time of the corresponding associated road sections, the spread time being the time required for the congestion of the nth road section to spread to the corresponding associated road sections.

[0088] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited thereto, any person skilled in the art can easily think of changes or replacements within the technical range disclosed by the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0089] Finally: the above is only a preferred embodiment of the present application and is not used to limit the present application, any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application should be included within the protection scope of the present application.

Claims

1. A real-time traffic routing dynamic optimization method based on multi-source data fusion, characterized in that, The method comprises the following steps: Collect N pieces of traffic section information, N is an integer greater than or equal to 1, and the traffic section information includes traffic density, vehicle proportion, vehicle speed, and road section width; Comprehensively analyze the traffic section information to generate a traffic congestion coefficient, and determine whether the corresponding road section is congested according to the traffic congestion coefficient; Input the number of the nth congested road section into a pre-constructed traffic network graph structure to obtain road sections having a connection relationship with the nth road section, which are marked as associated road sections; Obtain associated feature data of the corresponding associated road sections; Collect k time series of traffic congestion coefficients and associated feature data of the corresponding associated road sections at a preset interval time length from the time when the nth road section is determined to be congested as the starting point, input the collected data into a pre-trained spread prediction model, and obtain the spread time of the corresponding associated road sections, wherein the spread time is the time required for the congestion of the nth road section to spread to the corresponding associated road sections. 2.The real-time traffic routing dynamic optimization method based on multi-source data fusion of claim 1, wherein, The traffic congestion coefficient is composed of the ratio of the weighted sum of congestion-promoting factors to the weighted sum of congestion-relieving factors: the congestion-promoting factors are the weighted sum of the normalized traffic density and the vehicle proportion, and the congestion-relieving factors are the weighted sum of the normalized vehicle speed and the road section width. 3.The real-time traffic routing dynamic optimization method based on multi-source data fusion of claim 1, wherein, The method for constructing the traffic network graph structure comprises the following steps: Obtain road network basic data and the connection relationship of each road section, wherein the road network basic data includes the number, length, and endpoints of each road section, the endpoints include the starting point and the ending point, and the connection relationship of each road section is directly obtained from map data; The nodes in the traffic network graph structure are the endpoints of each road section, and the edges in the traffic network graph structure are the connection relationship between road sections, the length of road sections, and the width of road sections. 4.The real-time traffic routing dynamic optimization method based on multi-source data fusion of claim 1, wherein, The training method of the spread prediction model comprises the following steps: Pre-collect j sets of spread analysis data, j is an integer greater than 1, the spread analysis data includes spread feature data and the spread time corresponding to the spread feature data; the spread feature data includes the k time series of traffic congestion coefficients and the associated feature data of the corresponding associated road sections; Convert the spread analysis data into a corresponding set of third feature vectors; take the third feature vectors as the input of the spread analysis data, take the spread time corresponding to each set of spread feature data as the output, take the actual spread time corresponding to each set of spread feature data as the prediction target; take minimizing the sum of prediction errors of all spread feature data as the training target; train the spread prediction model until the sum of prediction errors reaches convergence to stop training; the spread prediction model is a deep neural network model.

5. The real-time traffic routing dynamic optimization method based on multi-source data fusion according to claim 1, characterized in that, The method for determining whether the corresponding road section is congested according to the traffic congestion coefficient comprises the following steps: Compare the traffic congestion coefficient with a preset congestion coefficient threshold value; if the traffic congestion coefficient is greater than or equal to the preset congestion coefficient threshold value, it is determined that the corresponding road section is congested; if the traffic congestion coefficient is less than the preset congestion coefficient threshold value, it is determined that the corresponding road section is normal. 6.The real-time traffic routing dynamic optimization method based on multi-source data fusion of claim 1, wherein, The method for obtaining the traffic density comprises the following steps: Collect pictures in the monitoring area marked by the road section monitoring camera, and mark them as traffic density pictures; Input the traffic density pictures into the trained vehicle recognition model to output the number of vehicles and the number of large vehicles of the corresponding road section; take the ratio of the number of vehicles to the actual area of the monitoring area of the corresponding road section monitoring camera as the traffic density of the corresponding road section. 7.The real-time traffic routing dynamic optimization method based on multi-source data fusion of claim 1, wherein, The vehicle ratio is a ratio of the number of large vehicles to the number of vehicles; and the vehicle speed is an average value of vehicle speeds collected by a road section speed detector in a unit time. 8.The real-time traffic routing dynamic optimization method based on multi-source data fusion of claim 1, wherein, The correlation feature data includes a road section length, a road section width, a mean value and a standard deviation of a traffic congestion coefficient; The mean value and the standard deviation of the traffic congestion coefficient are obtained by: Collecting k time series traffic congestion coefficients of the corresponding correlation road section at a preset interval time from the starting point, establishing a congestion coefficient analysis set, and calculating the mean value and the standard deviation of the congestion coefficient analysis set. 9.The real-time traffic routing dynamic optimization method based on multi-source data fusion of claim 1, wherein, When the nth road section is congested, the time series traffic congestion coefficients collected in a predetermined time of the nth road section and the green light interval time of the traffic signal at the intersection node of the nth road section are input into the pre-trained correlation analysis model, and the correlation result of the congestion and the green light interval time is output; the correlation result includes correlation and non-correlation; if the correlation result is non-correlation, the correlation road section is obtained; If the correlation result is correlation, the time series traffic congestion coefficients collected in a predetermined time of the corresponding road section and the green light interval time of the traffic signal at the intersection node of the corresponding road section are input into the pre-trained signal light adjustment model, and the adjusted green light interval time is output.

10. A real-time traffic routing dynamic optimization system based on multi-source data fusion, characterized in that, The system for implementing the real-time traffic routing dynamic optimization method based on multi-source data fusion of any one of claims 1-9, the system comprising: A collection module for collecting information of N traffic road sections, N being an integer greater than or equal to 1, the traffic road section information including traffic density, vehicle ratio, vehicle speed, and road section width; A first analysis module for comprehensively analyzing the traffic road section information, generating a traffic congestion coefficient, and determining whether the corresponding road section is congested according to the traffic congestion coefficient; A fourth analysis module for, when the nth road section is congested, inputting the time series traffic congestion coefficients collected in a predetermined time of the nth road section and the green light interval time of the traffic signal at the intersection node of the nth road section into a pre-trained correlation analysis model, and outputting the correlation result of the congestion and the green light interval time; the correlation result includes correlation and non-correlation; A fifth analysis module for, if the correlation result is correlation, inputting the time series traffic congestion coefficients collected in a predetermined time of the corresponding road section and the green light interval time of the traffic signal at the intersection node of the corresponding road section into a pre-trained signal light adjustment model, and outputting the adjusted green light interval time; A second analysis module for, if the correlation result is non-correlation, inputting the nth road section number of the congestion into a pre-constructed traffic network graph structure, obtaining road sections having a connection relationship with the nth road section and marking them as correlation road sections; and obtaining correlation feature data of the corresponding correlation road sections; A third analysis module for inputting the k time series traffic congestion coefficients collected at a preset interval time from the time when the nth road section is determined to be congested and the correlation feature data of the corresponding correlation road sections into a pre-trained spread prediction model, and obtaining a spread time of the corresponding correlation road sections, the spread time being a time required for the congestion of the nth road section to spread to the corresponding correlation road sections.

Images