CN115687709A

CN115687709A - Traffic dynamic control method based on traffic data dimension reduction reconstruction and multidimensional analysis

Info

Publication number: CN115687709A
Application number: CN202211270619.5A
Authority: CN
Inventors: 韩朝晖; 曲媛媛; 张中凯; 吴力刚; 刘丙庆; 陶鹏; 秦志亮; 秦宇; 李莹莹
Original assignee: Shandong New Beiyang Information Technology Co Ltd; Weihai Beiyang Electric Group Co Ltd
Current assignee: Shandong New Beiyang Information Technology Co Ltd; Weihai Beiyang Electric Group Co Ltd
Priority date: 2022-10-16
Filing date: 2022-10-16
Publication date: 2023-02-03

Abstract

The invention relates to the technical field of traffic data processing, in particular to a traffic dynamic control method based on traffic data dimension reduction reconstruction and multidimensional analysis, which is characterized by comprising the following steps of: constructing a traffic intersection region model, planning a passing path of the region model, carrying out information layered extraction on original data of each region, and carrying out dimension reduction reconstruction by using a matrix and coordinates, specifically comprising data acquisition, road condition dimension reduction reconstruction, vehicle condition dimension reduction reconstruction and dimension reduction information storage, extracting layered dimension reduction reconstruction data of each region to form a new coordinate curve or matrix data model, carrying out multi-dimensional characteristic analysis, carrying out multi-dimensional characteristic data mixed calculation, and then carrying out traffic dynamic control; compared with the prior art, the storage capacity of traffic data is greatly reduced, the quantifiable technical indexes are obtained through dimension reduction reconstruction, the data rule can be accurately obtained, and reliable technical support is provided for intelligent traffic control, vehicle-road cooperation and automatic driving.

Description

Traffic dynamic control method based on traffic data dimension reduction reconstruction and multidimensional analysis

The technical field is as follows:

the invention relates to the technical field of traffic data processing, in particular to a traffic dynamic control method based on traffic data dimension reduction reconstruction and multidimensional analysis, which can perform dimension reduction reconstruction, efficient and accurate visual multidimensional index analysis and intelligent output control decision on traffic data so as to effectively improve traffic control capacity.

The background art comprises the following steps:

with the increase of the number of motor vehicles, the problems of traffic jam, traffic accidents and the like are increased gradually, and the problems of large traffic data storage pressure, low vehicle passing efficiency, high energy consumption, high pollution, untimely traffic accident discovery and the like are increasingly prominent because of the laggard traffic data processing technology, so that the traffic management work bears great pressure.

The main method for storing traffic data in the prior art is to store road traffic video and image data in multiple points, but the storage cost is quite high. The existing method for relieving the traffic pressure mainly comprises road traffic expansion, traffic volume reduction control (number limit traffic and the like), public transport and bicycle travel encouragement, tidal lanes, peak traffic shifting and the like, but the effect is not obvious. At present, technologies for improving management and control experience and strategies by analyzing historical traffic data exist, and a 'green wave passing' technology is typical, but practice proves that traffic congestion is often caused frequently in emergencies such as temporary parking, road occupation, traffic accidents and the like, so that an optimal solution can not be obtained only by analyzing historical data, and 'green wave passing' is only suitable for a specific passing direction and a passing speed, and is poor in practicability.

At present, a small amount of artificial intelligence algorithm technologies are adopted, and the traffic control capacity is improved by collecting and analyzing traffic data in real time and optimizing traffic light timing, but practice proves that the methods are incomplete in data characteristic mining, low in data utilization efficiency and relatively comprehensive in technical method, and are too dependent on artificial intelligence black box operation, so that the contents of data analysis and operation cannot be quantitatively explained, the problem of AI uncertainty cannot be avoided, and if the technologies are widely applied to public safety fields such as traffic control and the like, potential safety hazards cannot be avoided.

The invention content is as follows:

aiming at the defects and shortcomings in the prior art, the invention provides a traffic dynamic control method based on traffic data dimension reduction reconstruction and multidimensional analysis, which can reduce the dimension of traffic data in real time and reconstruct the traffic data, then convert the data into a matrix and coordinate data model in multiple dimensions, analyze and process the matrix and coordinate data model by an artificial intelligence algorithm, and further efficiently and accurately output a control strategy, thereby effectively improving the traffic control capacity.

The invention is achieved by the following measures:

a traffic dynamic control method based on traffic data dimension reduction reconstruction and multidimensional analysis is characterized by comprising the following steps:

step 1: constructing a traffic intersection region model, setting a central crossing region of the traffic intersection as a control region and setting road extension regions in multiple directions as monitoring regions by taking a stop line and an extension line of the stop line as boundaries; setting the area of data flowing from the monitoring area to the control area as an out area and setting the area of data flowing from the control area to the monitoring area as an in area in each monitoring area;

step 2: planning a traffic path of the area model: taking each out area of the monitoring area as a starting point, communicating with the corresponding in area according to three directions of left-going, straight-going and right-going, and constructing a communication relation graph of the out areas and the in areas;

and step 3: performing information layering extraction on the original data of each region, and performing dimension reduction reconstruction by using a matrix and coordinates, wherein the method specifically comprises data acquisition, road condition dimension reduction reconstruction, vehicle condition dimension reduction reconstruction and dimension reduction information storage;

and 4, step 4: extracting the layered dimensionality reduction reconstruction data of each area to form a new coordinate curve or matrix data model, and performing multi-dimensional characteristic analysis, wherein the method specifically comprises the following steps:

(1) Establishing an OUT area to-be-released vehicle monitoring data model: monitoring the vehicle condition matrix in each OUT area, calculating and extracting the number of vehicles to be released at a certain time in the matrix in real time to form a new matrix, then establishing a coordinate system according to the number of time, and generating a monitoring curve of the number of vehicles to be released in the OUT area for characteristic analysis;

(2) Establishing an OUT area on-road traffic flow monitoring data model: monitoring a vehicle condition matrix in each OUT area, calculating the number and the position of the on-road traffic flow in real time and forming a new matrix, then establishing a coordinate system according to the number of time, and generating an on-road traffic flow monitoring curve of the OUT area for characteristic analysis;

(3) Establishing a passing time fitting data model of the vehicles to be released in the OUT area: monitoring a vehicle condition matrix in each OUT area, calculating the number and the passing time of vehicles to be released in a certain batch in the matrix, then establishing a coordinate system according to the number and the time, recording and fitting multiple batches of data, and obtaining a fitted data curve of the passing time of the vehicles to be released in the OUT area for characteristic analysis;

(4) Establishing a vehicle outflow data model in an OUT region (the model principle can be used for analyzing vehicle inflow data in a control region): monitoring a vehicle condition matrix in each OUT area, setting statistical duration t, calculating the vehicle outflow quantity in a time-sharing or accumulated mode by taking t as a unit, forming a new matrix, establishing a coordinate system by time quantity, and forming a vehicle outflow data curve in the OUT area for characteristic analysis;

(5) Establishing an IN area vehicle inflow data model (the model principle can be used for analyzing vehicle outflow data IN a control area): monitoring the vehicle condition matrix of each IN area, setting a statistical time length t, calculating the vehicle inflow quantity IN a time-sharing mode or accumulating mode by taking t as a unit, forming a new matrix, then establishing a coordinate system by time quantity, and forming a vehicle inflow data curve of the IN area for characteristic analysis;

(6) Establishing a vehicle speed tracking data model of each area: monitoring a vehicle condition matrix in each region, extracting speed data of each vehicle, calculating the average speed of the vehicles, forming a new matrix, establishing a coordinate system by time speed, and generating a vehicle speed data curve in real time for characteristic analysis;

(7) Establishing a vehicle density distribution data model of each area: monitoring the vehicle condition matrix of each area, calculating the vehicle coverage rate in the matrix in real time and forming a new matrix, then establishing a coordinate system according to the time-coverage rate, and generating a vehicle density distribution curve in real time for characteristic analysis;

and 5: and (3) multi-dimensional feature data hybrid calculation: the traffic dynamic control method comprises the steps of calculation and acquisition of vehicles on the way IN an OUT area and vehicles to be released, accident monitoring of each area, and congestion state monitoring of an IN area and a control area, wherein calculation and prediction results provide information data support for traffic dynamic control:

step 5-1: and (3) acquiring or predicting the number of vehicles to be released and the number of large traffic streams in the OUT area in real time by using the data models in the step (1) and the step (2) in the step (4):

the number of the vehicles waiting to be released in the lane is N through calculation _await State z of vehicle to be released _await ：

The number of the vehicles in the lane in the on-the-way traffic flow is N through calculation _onway The on-road vehicle mass flow state is z _onway ，

Step 5-2: and (4) acquiring vehicle speed characteristics and vehicle density information which change along with time in the area by using the data models in the steps (6) and (7), setting a density threshold value as D and a duration threshold value as P, and monitoring accidents in a monitoring area and a control area by monitoring whether the vehicle speed drops to 0 steeply and the duration of the vehicle density greater than D exceeds P or not. When the density in the area exceeds D and the duration exceeds P, or the number of vehicles with the vehicle speed steeply reduced to 0 in the area is more than or equal to 1, the traffic accident is determined to occur, otherwise, the traffic accident does not occur. Traffic accident status AD is as follows:

step 5-3: acquiring characteristic information of inflow and outflow volumes of vehicles, vehicle speeds and vehicle densities IN an IN area and a control area along with time change by using the data models (4), (5), (6) and (7) IN the step 4, and calculating congestion states of the IN area and the control area by using a machine learning algorithm, wherein the data models (4), (5), (6) and (7) IN the step 4 are specifically used for the IN area, the data models (5), (6) and (7) IN the step 4 are specifically used for the IN area, and the congestion state calculation methods of the control area and the IN area are the same;

step 6: traffic dynamic management and control, including signal lamp management and control:

step 6-1-1: setting an initial state: setting all red lights (red light = 1) in the initial state of the signal lamp;

step 6-1-2: determining a passable path: determining a combination of straight lanes and left-turn lanes which can be released simultaneously (red light = 0) according to the step 2, and then determining whether the right-turn lanes can pass according to the traffic data of the combination;

step 6-1-3: calculating a single lane release index: calculating a single-lane release index according to the numerical value or state output in the steps 5-1, 5-2 and 5-3, and determining whether to release and the priority of release according to the numerical value of the single-lane release index, wherein the single-lane release index is calculated according to the following formula:

let x _n The total number of traffic flows to be calculated for a certain lane in the monitoring area is calculated as follows:

x _n ＝z _await *N _await +z _onway *N _onway

let x _spec Is special inThe state of the vehicle (the special service vehicle),

let x _in In order to be in the in-zone congestion state,

let x _cross In order to manage the congestion status of the area,

let x _ad The current abnormal state of the lane is detected,

and if the single lane release index is y, the expression for evaluating y is as follows:

y＝f(x _spec ,x _in ,x _cross ,x _ad ,x _n )

＝C _spec x _spec +C _in x _in +C _cross x _cross +C _ad x _ad

+(1-x _spec )(1-x _in )(1-x _cross )(1-x _ad )*x _n

wherein, C _spec ,C _in ,C _cross ,C _ad Is a constant number, C _spec ＞＞x _n ,C _in ＞＞x _n ,C _cross ＞＞x _n ,C _ad ＞＞x _n 。

Step 3 of the invention is specifically realized by the following steps:

step 3-1: data acquisition: in a monitoring area and a control area, a map, a road engineering drawing and a camera are utilized to collect road condition information in a shunting section, wherein the road condition information comprises lane length and width, lane line position, sidewalk position, road side access, turnout, road barrier, road bending degree and gradient information; the method comprises the following steps of collecting video, pictures and radar signal information of vehicle characteristics, traffic flow and vehicle motion states at a branch section;

step 3-2: road condition dimension reduction reconstruction: (1) Firstly, establishing a rectangular coordinate system and setting a scale to ensure that the actual road condition and the coordinate system form an exact corresponding relation; (2) Setting a line segment in a coordinate system to correspond to a stop line of a traffic intersection, and marking and reconstructing road condition length, width size, lane line position, sidewalk position, road side entrance and exit, turnout, road barrier, road bending degree and gradient information of an out area in the coordinate system by taking the line segment as a starting point; (3) Dividing the road marked and reconstructed in the coordinate system into K areas for m _i *n _i Gridding and dividing to obtain a value range m _i >0、n _i >0、K≥i>0, then storing the information in the divided grids by using a matrix;

step 3-3: vehicle condition dimensionality reduction reconstruction: (1) Collecting video, picture and radar signal information of vehicle conditions in real time according to K regions, detecting and tracking vehicle characteristics, vehicle positions and vehicle motion conditions by using image identification and signal identification artificial intelligence algorithm, extracting characteristic information indexes, and then obtaining the vehicle characteristics, the vehicle positions and the vehicle motion conditions according to m _i *n _i The matrix is used for data storage, so that a vehicle condition coordinate graph can be generated in a road condition graph in real time after data superposition, and visual display of road condition and vehicle condition information is realized;

step 3-4: and D, storing dimension reduction information: and storing the acquired road condition and vehicle condition multidimensional matrix data, and if the effective information in the matrix is less, compressing and storing the sparse matrix, thereby further saving the storage space.

The calculation steps of the congestion states of the control area and the IN area IN the step 5-3 of the invention are as follows:

step 5-3-1: constructing an original input data set: acquiring a vehicle position characteristic matrix, a vehicle inflow and outflow quantity characteristic curve and a vehicle speed characteristic matrix by using the data models of (4), (5), (6) and (7) in the step 4, and then defining a congestion state corresponding to the characteristics as a label;

step 5-3-2: characteristic pretreatment: arranging the vehicle position characteristic matrixes according to a time sequence to form a group of serialized matrix blocks, and then performing serialized tiling on each matrix block to obtain linear characteristics; carrying out normalization processing on the characteristic curve of the inflow and outflow quantity of the vehicle; and taking the preprocessed three-part features as input features of the model.

Step 5-3-3: building a deep learning model: (1) Building a self-attention mechanism network, inputting the preprocessed vehicle position characteristic matrix into the network, and outputting a linear characteristic layer result; (2) Building a convolutional neural network of a residual error structure, finally outputting a linear layer, inputting a vehicle speed characteristic matrix into the network, and outputting a linear characteristic layer result; (3) Splicing and merging the characteristic curve of the inflow and outflow quantity of the vehicle after the normalization processing and the characteristics of the two linear layers, inputting the merged characteristic curve into a full-connection network, activating the merged characteristic curve, and predicting the congestion state;

step 5-3-4: model training: sending the data set into the deep learning model in the step 5-3-3 for training, calculating the loss of the network model by using a cross entropy loss function, and finishing the model training and storing the calculation parameters in the model if the loss reduction difference values of the front n (n is more than or equal to 1) epochs and the back n epochs are less than a fixed value (the fixed value is more than 0) in the training process;

step 5-3-5: and (3) calculating the congestion state: acquiring a vehicle position characteristic matrix, a vehicle inflow and outflow characteristic curve and a vehicle speed characteristic matrix in real time, inputting the obtained data into a network after preprocessing in the step 5-3-2 to obtain a calculation result PC, wherein the calculation result represents the congestion probability, and a threshold value PC is set _th (ii) a Similarly, when the congestion state is calculated IN the IN area, the calculation result is PI, and the threshold value is set to be PI _th 。

Step 6 of the method also comprises the control of overflow prevention, namely controlling the in-zone vehicle to flow in, wherein the index of the in-zone is obtained through the calculation in step 6-1, and when x exists, the index is used _in And if the traffic congestion information data is not greater than 1, outputting traffic congestion information data of the in-zone road section, and controlling the vehicle to flow in by using the red light.

Step 6 of the invention also comprises accident monitoring and control, and when x exists in a certain lane through calculation in step 6-1 _ad If =1, the traffic accident information data is output.

The method also comprises the step 6 of dredging congestion control, and a control area is obtained through calculation in the step 6-1When x is present _cross And if the traffic congestion information data is not less than 1, outputting the traffic congestion information data of the road sections of the control area, and controlling the vehicles to flow in the corresponding out area by using red lights.

Step 6 of the method also comprises low-efficiency forbidden control, the index of an out area is obtained through calculation in step 6-1, and under the state of green light, when the release index of a certain lane is 0, red light is switched, and other lanes are reselected.

Compared with the prior art, the method has the advantages that (1) the real-time traffic data such as videos, pictures and radar signals are subjected to dimensionality reduction reconstruction, so that the storage capacity and the storage cost of the traffic data are greatly reduced, the data processing efficiency is improved, secondly, a series of quantifiable technical indexes are obtained through dimensionality reduction reconstruction, the traffic data processing becomes more accurate and standard, thirdly, the series of quantifiable technical indexes are combined with technologies such as digital twin and virtual engines, and the real historical traffic data can be virtually restored; (2) By using a coordinate curve and a matrix data model to perform multidimensional characteristic analysis on technical index data, the data rule can be more accurately acquired, the abnormal data value can be found, and the traffic data processing visualization can be realized, so that the problems of artificial intelligent black box operation, uncertain AI and the like can be solved, and roads are widened for the wide application of the artificial intelligent technology to the public safety fields of traffic control and the like; (3) The coordinate curve, the matrix data model and the algorithm innovated by the invention can be deeply fused with artificial intelligence technology and products, and can provide reliable technical support for intelligent traffic control, vehicle-road cooperation and automatic driving after continuous accumulation and optimization.

Description of the drawings:

FIG. 1 is a system flow diagram of the present invention.

Fig. 2 is a regional view of an intersection in embodiment 1 of the present invention.

Fig. 3 is a road junction passage route diagram in embodiment 1 of the present invention.

Fig. 4 is a schematic view of the road condition of the monitoring area in embodiment 1 of the present invention.

Fig. 5 is a diagram of an all-road-condition information matrix in embodiment 1 of the present invention.

Fig. 6 is a schematic view of the vehicle condition of the monitoring area in embodiment 1 of the present invention.

Fig. 7 is a schematic view of a traffic condition of a management area in embodiment 1 of the present invention.

Fig. 8 is a diagram of a monitoring area vehicle position information matrix in embodiment 1 of the present invention.

Fig. 9 is a diagram of a management area vehicle position information matrix in embodiment 1 of the present invention.

Fig. 10 is a graph illustrating the traffic inflow data in the control area according to embodiment 1 of the present invention.

Fig. 11 is a sequence diagram of the management area vehicle inflow data in embodiment 1 of the present invention.

Fig. 12 is a graph illustrating traffic flow data of a regulation area according to embodiment 1 of the present invention.

Fig. 13 is a schematic view of vehicle speed tracking data in a management area according to embodiment 1 of the present invention.

Fig. 14 is a diagram of a management area vehicle speed tracking data matrix in embodiment 1 of the present invention.

Fig. 15 is a graph showing a vehicle density distribution data curve of a management area in embodiment 1 of the present invention.

Fig. 16 is a road condition matrix diagram of the lane line in embodiment 2 of the present invention.

Fig. 17 is a road width road condition matrix diagram in embodiment 2 of the present invention.

FIG. 18 is a graph showing the IN zone vehicle inflow data IN the embodiment 2 of the present invention.

FIG. 19 is a graph showing a vehicle speed trace data IN the IN area according to the embodiment 2 of the present invention.

FIG. 20 is a graph showing the average vehicle speed trace data IN the IN zone IN accordance with example 2 of the present invention.

FIG. 21 is a diagram of a matrix of IN area vehicle speed tracking data according to embodiment 2 of the present invention.

FIG. 22 is a graph showing a data curve of the IN zone vehicle density distribution IN accordance with example 2 of the present invention.

Fig. 23 is a graph showing monitored data of vehicles to be released in the OUT region in accordance with embodiment 3 of the present invention.

Fig. 24 is a graph showing traffic flow monitoring data in the OUT area in the embodiment 3 of the present invention.

Fig. 25 is a graph showing a time-fitted data curve of the passing time of the vehicle to be released in the OUT region in embodiment 3 of the present invention.

The specific implementation mode is as follows:

the invention is further described below with reference to the drawings and examples.

Example 1:

the method comprises the steps of setting partitions for traffic road conditions, constructing a traffic path communication graph, performing dimensionality reduction reconstruction and multidimensional characteristic analysis on traffic information such as road conditions, and utilizing an algorithm to control the congestion state of a control area in real time and initialize and control the congestion of the control area in real time. The specific process is as follows:

step 1: a regional model of a certain cross traffic intersection is constructed (as shown in fig. 2), a central intersection region of the traffic intersection is set as a control region by taking a stop line and an extension line of the stop line as boundaries, and road extension regions in multiple directions are set as monitoring regions; then setting the area of data flowing from the monitoring area to the control area as an OUT area and setting the area of data flowing from the control area to the monitoring area as an IN area IN each monitoring area; in the present embodiment, the east, south, west and north are respectively represented by letters E, S, W and N, for example: OUTE represents an OUT region where data flows OUT from the west-east to the management region, and INE represents an IN region where data is injected from the management region to the monitoring region from the west-east.

Step 2: planning a traffic path model of the intersection (as shown in fig. 3): and (3) taking each OUT area of the monitoring area as a starting point, communicating with the corresponding IN area according to the three directions of left row, straight row and right row, and constructing a communication relation graph of the OUT area and the IN area.

And step 3: and (3) carrying out dimensionality reduction reconstruction on the original data: the method comprises four steps of data acquisition, road condition dimension reduction reconstruction, vehicle condition dimension reduction reconstruction and dimension reduction information storage:

step 3-1: data acquisition: in a monitoring area and a control area, a map, a road engineering drawing and a camera are utilized to collect road condition information in a shunting section, wherein the road condition information comprises the length and width of a lane of a road, the position of a lane line, the position of a sidewalk, a road side access and exit, a turnout, an inter-road barrier, the degree of road curvature and gradient information; the method comprises the following steps of collecting video, pictures and radar signal information of vehicle characteristics, traffic flow and vehicle motion states at a branch section;

step 3-2: road condition dimension reduction reconstruction (as shown in fig. 4 and 5): (1) Firstly, establishing a rectangular coordinate system and setting a scale to form an exact corresponding relation between a real road condition and the coordinate system, in this example, as shown in fig. 4, it is set that the real road condition length L of a certain road section out area is 1400 meters, the maximum road condition width W is 20 meters, it is set that each scale of an x-axis of the coordinate system maps the real road condition length of 50 meters, and each scale of a y-axis maps the real road condition width of 2 meters, as shown in fig. 4, 28 scales are taken on the x-axis, and 10 scales are taken on the y-axis to completely map the real road condition, fig. 5 is a schematic diagram of matrix storage for the boundary size and the stop line position of the road section, in fig. 5, 0 represents traffic lane information, 1 represents roadside information, and 2 represents stop line information; (2) Setting a line segment in a coordinate system to correspond to a stop line of a traffic intersection, marking and reconstructing road condition length and width size, lane line position, sidewalk position, road side entrance and exit, fork, road barrier, road bending degree and slope information of an out area in the coordinate system by taking the line segment as a starting point, setting a line segment in positions of coordinates x =0 and y =10 to correspond to the stop line of the traffic intersection, and marking and reconstructing the road condition information of the length and width size, lane line position, up-down slope and the like of the out area in the coordinate system by taking the line segment as the starting point; (3) Dividing the road marked and reconstructed in the coordinate system into K areas for m _i *n _i The grid is divided, and then the information in the divided grid is stored by using a matrix, in this example, we set K =28, that is, 28 parts are divided in equal proportion according to the scale of the horizontal axis of the coordinate, at this time, if we want to analyze the road condition of the 16 th grid area (i = 16) (as shown in fig. 4 and 5, corresponding to the real road length of 50 meters and the width of 12 meters), we can further divide the area by using 2 meters by 2 meters of grid, so as to form a 25 x 6 matrix (m is 2 meters by 2 meters of grid), and then we can obtain the matrix ₁₆ ＝25，n ₁₆ = 6) is configured to store traffic information of the 16 th mesh area;

step 3-3: vehicle condition dimension reduction reconstruction (as shown in fig. 6 and 8) (1)Collecting video, picture and radar signal information of vehicle conditions in K regions in real time, detecting and tracking vehicle characteristics, vehicle positions and vehicle motion conditions by using image identification and signal identification artificial intelligence algorithm, extracting characteristic information indexes, and calculating the characteristic information indexes according to m _i *n _i The matrix stores data, so that a vehicle condition coordinate graph can be generated in real time in the road condition graph after data superposition, thereby realizing visual display of the road condition and the vehicle condition information, in this example, the step 3-2 is continued to take the 16 th network region of the road section as an example, as shown in fig. 8, the road section is in a matrix (m is a 25 x 6 matrix) of the road section ₁₆ ＝25，n ₁₆ = 6) recording road condition information, and using the same 25 × 6 matrix (m) ₁₆ ＝25，n ₁₆ = 6), in fig. 8, 0 represents a position not occupied by the vehicle, and 1 represents a position occupied by the vehicle;

step 3-4: and D, storing dimension reduction information: (as shown in fig. 5, 8, 9, 11, 14, 16, 17 and 21) storing the acquired multidimensional matrix data of road conditions and vehicle conditions, comparing the matrix data with video data and image data can find that the storage space required by the matrix data is greatly reduced;

and 4, step 4: extracting dimension reduction reconstruction data of the control area to form a new coordinate curve or matrix data model, and performing multi-dimensional characteristic analysis, wherein the method specifically comprises the following steps:

(1) Establishing a vehicle inflow data model of a control area: (as shown in fig. 10), setting a statistical time period t in a vehicle condition matrix in a monitoring area, calculating the inflow quantity of the vehicle in a time-sharing manner by taking t as a unit, (for example, t =1 second in the present example), summing the inflow quantity at each moment to form a new sequence, then establishing a coordinate system by the quantity of time to form a vehicle inflow data curve for characteristic analysis (the sequence is shown in fig. 11);

(2) Establishing a vehicle outflow data model of a control area: (as shown in fig. 12), setting a statistical time duration t (for example, t =1 second in this example) for a vehicle condition matrix in a monitoring area, calculating or accumulating the vehicle outflow number in a time-sharing manner by taking t as a unit to form a new matrix, and then establishing a coordinate system by using the time x number to form a vehicle outflow data curve for feature analysis;

(3) Establishing a vehicle speed tracking data model of a control area: (as shown in fig. 13) the vehicle condition matrix in each zone is monitored, the speed data of each vehicle is extracted, the continuous speed of multiple vehicles is calculated to form a new matrix, then a coordinate system is established by time-speed, and a vehicle speed data curve is generated in real time for feature analysis. Wherein the new matrix formed is defined as u × v (e.g., u =11, v =16 in this example), and is composed of a plurality of time-varying vehicle speed sequences, u represents the length of the vehicle speed time sequence, and v represents the number of vehicles (see fig. 14, where a value of-1 in the matrix represents that a vehicle is not present at that time);

(4) Establishing a vehicle density distribution data model of a control area: (as shown in fig. 15), monitoring a vehicle condition matrix in the area, calculating the vehicle coverage rate in the matrix in real time to form a new matrix, then establishing a coordinate system according to the ratio of time to the coverage rate, and generating a vehicle density distribution curve in real time to perform characteristic analysis;

and 5: multi-dimensional feature data hybrid calculation: and monitoring the congestion state of the control area, wherein the calculation and prediction results of the part are used as the precondition for dynamic traffic control.

Step 5-1: and (5) acquiring vehicle speed characteristics (specifically, see fig. 13, the characteristics include duration of time that the speed is steeply reduced to 0, the number of vehicles at the speed is 0 and the like) and vehicle density information (specifically, see fig. 15) which are changed along with time in the control area by using the data models (3) and (4) in the step 4, setting a density threshold value of D =0.8 and a duration threshold value of P =4, monitoring a plurality of vehicles of which the speed is steeply reduced to 0 and the duration of which the vehicle density is greater than D exceeds P, and possibly generating an accident in the control area, wherein the accident state AD =1.

Step 5-2: acquiring characteristic information of inflow and outflow of vehicles, vehicle speed and vehicle density changing along with time in the control area by using the data models (1), (2), (3) and (4) in the step 4, and calculating the congestion state of the control area by using a machine learning algorithm, wherein the steps are as follows: 1. obtaining traffic flow characteristics of the control area for a period of time from the data models in steps 4 (1) and 2), including traffic flow information flowing into the control area and traffic flow information flowing out of the control area (see fig. 10 and 12 in detail, and the sequence is see fig. 11); 2. from the step4 (3) acquiring a vehicle speed curve (or matrix) in the control area from the data model (specifically, see fig. 13, and a schematic matrix see fig. 14), and extracting a speed change characteristic of 10 seconds from the curve; 3. acquiring the vehicle density (see fig. 15) and the vehicle position matrix (see fig. 7 and 9) in the control area at a certain time from the data model in the step 4 (4), and setting the congestion threshold PC of the control area by combining the three parts of information _th And =0.8, training and predicting by using a multi-input deep learning network, and judging that the control area is in a congestion state.

Step 6: congestion clearing (initializing red light, clearing data):

let x _cross In order to manage the congestion status of the area,

determining lane presence x _cross =1, therefore, traffic congestion information data of the road segment in the management area is output, and the vehicle inflow is controlled by using the red light (red light is initialized);

example 2:

the present example performs the following operations for IN zone congestion, overflow prevention:

by partitioning road conditions, constructing a traffic path link map, performing dimensionality reduction reconstruction and multidimensional characteristic analysis on traffic information such as the road conditions, and the like, and utilizing an algorithm to control the congestion state of an IN area IN real time and perform dynamic traffic control.

The method comprises the following specific steps:

step 1: same as step 1 in example 1.

Step 2: same as step 2 in example 1.

And step 3: this embodiment is basically the same as step 3 in embodiment 1, but differs therefrom in that, in recording the road condition information characteristics, the matrix shown in fig. 5 is not used (the roadside information, the lane line, and the lane information are collected in the same matrix), but the lane line and the lane information are recorded in different matrices as shown in fig. 16 and 17, 2 represents the information of the lane line, 0 represents the information other than the lane line in the matrix of fig. 16, and 1 represents the roadside information and 0 represents the information of the lane in the matrix of fig. 17.

And 4, step 4: extracting dimension reduction reconstruction data of the IN area to form a new coordinate curve or matrix data model, and performing multi-dimensional characteristic analysis, wherein the method specifically comprises the following steps:

(1) Establishing an IN area vehicle inflow data model: (as shown in fig. 18), setting a statistical time t (in this example, t =1 second) for a vehicle condition matrix in a monitoring area, calculating the inflow quantity of the vehicle in a time-sharing manner by taking t as a unit, summing the inflow quantity at each moment to form a new sequence, and then establishing a coordinate system by the quantity of time to form a vehicle inflow data curve for feature analysis;

(2) Establishing an IN area vehicle speed tracking data model: (as shown in fig. 19 and 20) monitoring the vehicle condition matrix in each region, extracting the speed data of each vehicle, calculating the continuous speed of multiple vehicles to form a new matrix, then establishing a coordinate system by time-speed, and generating a vehicle speed data curve in real time for feature analysis. Wherein the new matrix formed is defined as u × v (e.g., u =11, v =32 in this example), and is composed of a plurality of time-varying vehicle speed sequences, u represents the length of the vehicle speed time sequence, and v represents the number of vehicles (see fig. 21, only the first 16 columns of value matrix are shown in fig. 21, and-1 value in the matrix represents that a vehicle does not exist at that time);

(3) Establishing an IN area vehicle density distribution data model: (as shown in fig. 22) monitoring a vehicle condition matrix in the area, calculating the vehicle coverage rate in the matrix in real time to form a new matrix, then establishing a coordinate system according to the ratio of time to the coverage rate, and generating a vehicle density distribution curve in real time to perform characteristic analysis;

and 5: and (4) acquiring characteristic information of the inflow and outflow quantity of the vehicles, the vehicle speed and the vehicle density IN the IN area along with the change of time by using the data models IN the steps (1), (2) and (3), and calculating the congestion state of the IN area by using a machine learning algorithm, wherein the steps are as follows: 1. obtaining the vehicle inflow quantity characteristics of the IN area for a period of time from the data model IN the step 4 (1) (particularly, see FIG. 18); 2. acquiring a vehicle speed curve (or matrix) IN the IN area from the data model IN the step 4 (2) (specifically, see FIG. 19 and the matrix see FIG. 21), and extracting a speed change characteristic of 10 seconds from the curve; 3. acquiring a certain time region of the control region from the data model in the step 4 (3)The IN area congestion threshold PI is set by combining the internal vehicle density (see figure 22) and the vehicle position matrix (see figure 8) and combining the three parts of information _th And =0.8, training and predicting by using a multi-input deep learning network, and judging the congestion state.

And 6: spill prevention (control of in zone vehicle inflow):

let x _in In order to be in the in-zone congestion state,

determining lane presence x _in And =1, outputting traffic jam information data of the in-zone road section, and controlling the vehicles flowing into the lane by using a red light.

Example 3:

the following operations are executed for the road conditions with no vehicle and low-efficiency forbidden in the OUT area:

by partitioning road conditions, constructing a traffic path link map, performing dimensionality reduction reconstruction and multidimensional characteristic analysis on traffic information such as the road conditions, and the like, capturing the idle state of the lane in the OUT area in real time by using an algorithm and performing dynamic traffic control. The method comprises the following specific steps:

step 1: same as step 1 in example 1.

And 2, step: same as step 2 in example 1.

And step 3: same as step 3 in example 1.

And 4, step 4: extracting OUT region dimensionality reduction reconstruction data to form a new coordinate curve or matrix data model, and performing multi-dimensional characteristic analysis, wherein the method specifically comprises the following steps:

(1) Establishing an OUT area to-be-released vehicle monitoring data model: (as shown in fig. 23), a vehicle condition matrix in the monitoring area is calculated in real time, the number of vehicles to be released at a certain time in the matrix is extracted to form a new matrix, then a coordinate system is established according to the number of time, and a monitoring curve of the number of vehicles to be released is generated to perform characteristic analysis;

(2) Establishing an OUT area on-road traffic flow monitoring data model: (as shown in fig. 24), monitoring a vehicle condition matrix in the area, calculating the number and the position of the in-transit traffic flow in real time to form a new matrix, then establishing a coordinate system by using the number of time, and generating an in-transit traffic flow monitoring curve for characteristic analysis;

(3) Establishing a passing time fitting data model of the vehicles to be released in the OUT area: (as shown in fig. 25), monitoring a vehicle condition matrix in the area, calculating the number and the passing time of a certain batch of vehicles to be released in the matrix, then establishing a coordinate system according to the number and the time, recording and fitting multiple batches of data, and obtaining a fitted data curve of the passing time of the vehicles to be released for characteristic analysis;

and 5: and (5) acquiring or predicting the number of vehicles to be released and the number of large traffic streams in the OUT area in real time by using the data models in the step (1) and the step (2) in the step (4). The method comprises the following specific steps: obtaining the number N of vehicles to be released of lanes waiting for left-turning, right-turning and straight going in the east-west-south-four directions from the data model (1) _await Information and waiting vehicle status z _await Information (see fig. 23 in particular); obtaining the number N of traffic in the east-west-south-north direction from the data model (2) _onway Information and in-transit traffic status z _onway Information (see fig. 24 in particular); obtaining the traffic flow N to be released from the data model (3) _await The predicted release time value (see fig. 25) is combined with the above three information to determine release and release time.

Step 6: low efficiency forbidden (to solve green light empty):

setting an initial state: setting all red lights (red light = 1) in the initial state of the signal lamp, and setting the threshold value of the vehicle for initial passing as N _start If the total number of vehicles on a certain lane in the monitoring area obtained in the step 5-1 exceeds the threshold value, the vehicles can be preferentially released, wherein x is set _n The total number of traffic flows required to be calculated for a certain lane in the monitoring area is calculated as follows:

x _n ＝z _await *N _await +z _onway *N _onway

determining lane x in green light state _n =0, so the lane switches red, reselecting to pass other lanes.

Example 4:

by partitioning road conditions, constructing a traffic path link map, performing dimensionality reduction reconstruction and multidimensional characteristic analysis on traffic information such as the road conditions, and the like, and utilizing an algorithm to control the congestion state of a control area in real time and perform dynamic traffic control. The present embodiment adopts the same steps as embodiment 1, wherein the machine learning algorithm part in step 5-3 is specifically as follows:

the steps of using a machine learning algorithm to predict congestion are as follows:

step 1: constructing an original input data set: respectively acquiring three parts of characteristics for describing the congestion state of a control area from a data processing model, wherein the three parts of characteristics comprise a vehicle position characteristic matrix, a vehicle inflow and outflow characteristic curve and a vehicle speed characteristic matrix or a characteristic diagram matrix, and the congestion state corresponding to the definition characteristics is used as a label; the inflow and outflow characteristics of the vehicle are obtained through an inflow and outflow curve, and the obtained length is 10 seconds; the vehicle speed characteristic is a vehicle speed diagram or a characteristic matrix after original data dimension reduction reconstruction, wherein the characteristic matrix comprises speed records of a plurality of vehicles in the area, and the acquisition length is 10 seconds; the label is defined by setting and threshold values, the scattered coverage rate of the vehicle position matrix is more than 70%, the difference value of inflow and outflow of the vehicles in the historical records is more than 4 and more than 5 seconds, the average vehicle speed in a vehicle speed graph is reduced to be less than 10 kilometers per hour, the congestion is defined, and otherwise, the traffic is smooth;

and 2, step: characteristic pretreatment: firstly, preprocessing a vehicle position characteristic matrix, forming a group of serialized matrix blocks according to a time sequence, then flattening the matrix to obtain a sequence, so that each moment is represented by a linear sequence, secondly, normalizing a vehicle inflow and outflow characteristic curve, and finally, using the preprocessed characteristic as an input characteristic of a model;

and 3, step 3: building a deep learning model: 1) Firstly, building a full attention mechanism network, inputting a preprocessed vehicle position characteristic matrix into the full connection attention mechanism network, wherein 2 encoder layers are passed through in the process, a first sublayer connection structure comprises a multi-head self-attention sublayer, a normalization layer and a residual connection, a second sublayer connection structure comprises a feed-forward full connection sublayer, a normalization layer and a residual connection, and then enabling the matrix output from an encoder to pass through a linear layer to output a linear characteristic layer result; 2) Secondly, building a convolution neural network with a residual error structure, inputting a vehicle speed characteristic matrix or a vehicle speed characteristic diagram into the convolution neural network with the residual error structure, firstly enabling data to enter the same two-dimensional convolution layer, enabling the number of filters to be 32, enabling the step length to be 2, enabling the convolution kernel size to be 7 x 7, then connecting a maximum pooling layer, enabling the step length to be 2 and enabling the convolution kernel size to be 3 x 3, then enabling the data to enter two residual error modules, enabling each residual error module to comprise two layers of convolution, enabling the parameters to be the same, enabling the number of the filters to be 32, enabling the convolution kernel size to be 3 x 3, enabling information before processing to be connected after processing of each residual error module, enabling the information to enter the next residual error module together, finally connecting an average pooling layer, and outputting a linear characteristic layer result after full connection layer; 3) Finally, splicing and merging the characteristic curve of the inflow and outflow quantity of the vehicle after normalization processing and the characteristics of the two linear layers, inputting the merged characteristic curve into a fully-connected network, and finally accessing a Softmax layer to predict the congestion state;

and 4, step 4: model training: sending the data set into the network in the step 3 for model training, calculating the loss of the network model by using a cross entropy loss function until the model loss is less than a set threshold value L, finishing the model training and storing the calculation parameters in the model, wherein the loss of the network model is calculated by using the cross entropy loss function, and the prediction result is set as

The real label is y, and the cross entropy loss function formula is as follows:

in the training process, if the loss reduction difference value of the first 10 epochs and the loss reduction difference value of the last 10 epochs is less than 0.1, the model training is finished, and calculation parameters in the model are stored for real-time prediction;

and 5: and (3) predicting the congestion state: respectively acquiring three parts of characteristics for describing the congestion state of the control area from the data processing model in real time, wherein the characteristics comprise a vehicle position characteristic matrix, a vehicle inflow and outflow characteristic curve and a vehicle speed characteristic matrix, and inputting the characteristics into a network after the preprocessing of the step 2Obtaining a prediction result PC, the result being the congestion probability, and then setting a threshold value PC _th And =0.8, if the value is larger than the threshold, the congestion state is determined, and if the value is smaller than the threshold, the congestion state is predicted.

Compared with the prior art, the method has the advantages that (1) the real-time traffic data such as videos, pictures and radar signals are subjected to dimensionality reduction reconstruction, so that on one hand, the storage capacity and the storage cost of the traffic data are greatly reduced, and the data processing efficiency is improved, on the other hand, a series of quantifiable technical indexes are obtained through dimensionality reduction reconstruction, and the traffic data processing becomes more accurate and standard; (2) By carrying out multidimensional characteristic analysis on technical index data by using a coordinate curve and a matrix data model, the method can more accurately acquire data rules, find data abnormal values and realize traffic data processing visualization, thereby solving the problems of artificial intelligent 'black box' operation, 'AI uncertainty' and the like, and widening the way for the artificial intelligent technology to be widely applied to public safety fields such as traffic control and the like; (3) The coordinate curve, the matrix data model and the algorithm innovated by the invention can be deeply fused with artificial intelligence technology and products, and can provide reliable technical support for intelligent traffic control, vehicle-road cooperation and automatic driving after continuous accumulation and optimization.

Claims

1. A traffic dynamic control method based on traffic data dimension reduction reconstruction and multidimensional analysis is characterized by comprising the following steps:

step 1: constructing a traffic intersection region model, setting a central crossing region of a traffic intersection as a control region and setting road extension regions in multiple directions as monitoring regions by taking a stop line and an extension line of the stop line as boundaries; setting the area of data flowing from the monitoring area to the control area as an out area and setting the area of data flowing from the control area to the monitoring area as an in area in each monitoring area;

step 2: planning a passing path of the area model: taking each out area of the monitoring area as a starting point, communicating with the corresponding in area according to three directions of left-going, straight-going and right-going, and constructing a communication relation graph of the out areas and the in areas;

and 3, step 3: performing information layered extraction on the original data of each region and performing dimensionality reduction reconstruction by using a matrix and coordinates, wherein the step specifically comprises data acquisition, road condition dimensionality reduction reconstruction, vehicle condition dimensionality reduction reconstruction and dimensionality reduction information storage;

and 4, step 4: extracting layered dimensionality reduction reconstruction data of each area to form a new coordinate curve or matrix data model, and carrying out multi-dimensional characteristic analysis, wherein the method specifically comprises the following steps:

(1) Establishing a monitoring data model of the vehicles to be released in the OUT area: monitoring a vehicle condition matrix in each OUT area, calculating and extracting the number of vehicles to be released at a certain time in the matrix in real time to form a new matrix, then establishing a coordinate system according to the time quantity, and generating a monitoring curve of the number of vehicles to be released in the OUT area for characteristic analysis;

(2) Establishing an on-road traffic flow monitoring data model in an OUT area: monitoring a vehicle condition matrix in each OUT area, calculating the number and the position of the on-road traffic flow in real time and forming a new matrix, then establishing a coordinate system according to the number of time, and generating an on-road traffic flow monitoring curve of the OUT area for characteristic analysis;

(4) Establishing an OUT area vehicle outflow data model (the model principle can be used for analyzing the vehicle inflow data in the control area): monitoring the vehicle condition matrix in each OUT area, setting a statistical time length t, calculating the vehicle outflow quantity in a time-sharing mode or in an accumulated mode by taking t as a unit, forming a new matrix, then establishing a coordinate system by the quantity of time, and forming a vehicle outflow data curve in the OUT area for characteristic analysis;

(5) Establishing an IN area vehicle inflow data model (the model principle can be used for analyzing vehicle outflow data IN a control area): monitoring the vehicle condition matrix of each IN area, setting statistical duration t, calculating the vehicle inflow quantity IN a time-sharing or accumulated mode by taking t as a unit, forming a new matrix, establishing a coordinate system by time-quantity, and forming a vehicle inflow data curve of the IN area for characteristic analysis;

(7) Establishing a vehicle density distribution data model of each area: monitoring the vehicle condition matrix of each region, calculating the vehicle coverage rate in the matrix in real time and forming a new matrix, then establishing a coordinate system by using the time coverage rate, and generating a vehicle density distribution curve in real time for characteristic analysis;

step 5-1: and (3) acquiring or predicting the number of vehicles to be released and the number of large traffic streams in the OUT area in real time by using the data models (1) and (2) in the step 4:

N _ath For vehicles to be released

N _oth As threshold value for on-road vehicle

Step 5-2: and (5) acquiring vehicle speed characteristics and vehicle density information which change along with time in the area by using the data models in the steps (6) and (7), setting a density threshold value as D and a duration threshold value as P, and monitoring accidents in a monitoring area and a control area by monitoring whether the vehicle speed drops to 0 steeply and the duration of the vehicle density greater than D exceeds P or not. When the density in the area exceeds D and the duration exceeds P, or the number of vehicles with the vehicle speed dropping to 0 in the area is more than or equal to 1, the traffic accident is determined to occur, otherwise, the traffic accident does not occur. Traffic accident status AD is as follows:

and 6: traffic dynamic management and control, including signal lamp management and control:

step 6-1-2: determining a passable path: determining a combination of straight lanes and left-turn lanes which can be released simultaneously (red light = 0) according to the step 2, and then determining whether the right-turn lanes can pass according to the traffic data of the combination; step 6-1-3: calculating a single lane release index: calculating a single-lane release index according to the numerical value or state output in the steps 5-1, 5-2 and 5-3, and determining whether to release and the priority of release according to the numerical value of the single-lane release index, wherein the single-lane release index is calculated according to the following formula:

x _n ＝z _await *N _await +z _onway *N _onway

let x _spec For a special vehicle (special duty vehicle) state,

let x _in In order to be in the in-zone congestion state,

let x _cross In order to manage the congestion status of the area,

let x _ad The current abnormal state of the lane is detected,

y＝f(x _spec ，x _in ，x _cross ，x _ad ，x _n )

＝C _spec x _spec +C _in x _in +C _cross x _cross +C _ad x _ad +(1-x _spec )(1-x _in )(1-x _cross )(1-x _ad )*x _n

2. The traffic dynamic control method based on traffic data dimension reduction reconstruction and multidimensional analysis according to claim 1, wherein step 3 is specifically realized by the following steps:

step 3-1: data acquisition: in a monitoring area and a control area, a map, a road engineering drawing and a camera are utilized to acquire road condition information in a shunting section, wherein the road condition information comprises lane length and width, lane line position, sidewalk position, road side access and exit, a turnout, an inter-road barrier, road bending degree and gradient information; the method comprises the steps that a branch section collects video, pictures and radar signal information of vehicle characteristics, vehicle flow and vehicle motion states;

step 3-2: road condition dimension reduction reconstruction: (1) Firstly, establishing a rectangular coordinate system and setting a scale to ensure that the actual road condition and the coordinate system form an exact corresponding relation; (2) Setting a line segment in a coordinate system to correspond to a stop line of a traffic intersection, and marking and reconstructing road condition length and width dimensions, lane line position, sidewalk position, road side entrance and exit, fork road, road barrier, road bending degree and gradient information of an out area in the coordinate system by taking the line segment as a starting point; (3) Dividing the road marked and reconstructed in the coordinate system into K areas for m _i *n _i Gridding and cutting to obtain value range m _i >0、n _i >0、K≥i>0, then storing the information in the divided grids by using a matrix;

step 3-3: vehicle condition dimensionality reduction reconstruction: (1) Collecting video, picture and radar signal information of vehicle conditions in real time according to K areas, detecting and tracking vehicle characteristics, vehicle positions and vehicle motion conditions by using image recognition and signal recognition artificial intelligence algorithms, extracting characteristic information indexes, and calculating the distance between the vehicle characteristics, the vehicle positions and the vehicle motion conditions according to m _i *n _i The matrix is used for data storage, so that a vehicle condition coordinate graph can be generated in a road condition graph in real time after data superposition, and visual display of road condition and vehicle condition information is realized;

3. The traffic dynamic control method based on traffic data dimension reduction reconstruction and multidimensional analysis according to claim 1, wherein the calculation steps of the congestion states of the control area and the IN area IN step 5-3 are as follows: step 5-3-1: constructing an original input data set: acquiring a vehicle position characteristic matrix, a vehicle inflow and outflow quantity characteristic curve and a vehicle speed characteristic matrix by using the data models of (4), (5), (6) and (7) in the step 4, and then defining a congestion state corresponding to the characteristics as a label;

step 5-3-2: characteristic pretreatment: arranging the vehicle position characteristic matrixes according to a time sequence to form a group of serialized matrix blocks, and then performing serialized tiling on each matrix block to obtain linear characteristics; carrying out normalization processing on the characteristic curve of inflow and outflow of the vehicle; and taking the preprocessed three-part features as input features of the model.

step 5-3-4: model training: sending the data set into the deep learning model in the step 5-3-3 for training, calculating the loss of the network model by using a cross entropy loss function, wherein in the training process, if the loss reduction difference value of the first n epochs and the last n epochs is smaller than a fixed value, n is larger than or equal to 1, and the fixed value is larger than 0, the model training is finished, and the calculation parameters in the model are stored;

4. The method as claimed in claim 1, wherein step 6 further includes overflow prevention control, i.e. controlling in-zone vehicle inflow,

wherein

Obtaining the index of the in area through the calculation in the step 6-1, and if x exists _in When =1, then the flow is interruptedAnd (4) traffic jam information data of the road sections out of the in area, and controlling the vehicles to flow in by using red lights.

5. The traffic dynamic control method based on traffic data dimension reduction reconstruction and multidimensional analysis according to claim 1, characterized in that step 6 further comprises accident monitoring control, and when x exists in a lane through calculation in step 6-1 _ad And if the traffic accident information data is not less than 1, outputting the traffic accident information data.

6. The method as claimed in claim 1, wherein step 6 further includes congestion dredging control, the index of the control area is obtained through calculation in step 6-1, and when x exists, the index is obtained _cross And if the traffic congestion information data is not less than 1, outputting the traffic congestion information data of the road sections of the control area, and controlling the vehicles to flow in the corresponding out area by using red lights.

7. The method as claimed in claim 1, wherein step 6 further includes low-efficiency forbidden control, the index of out zone is obtained through calculation in step 6-1, and in a green light state, when the release index of a lane is 0, a red light is switched and another lane is reselected for release.