CN116580574A - Road traffic multidirectional dynamic control method based on traffic flow monitoring - Google Patents
Road traffic multidirectional dynamic control method based on traffic flow monitoring Download PDFInfo
- Publication number
- CN116580574A CN116580574A CN202310374168.8A CN202310374168A CN116580574A CN 116580574 A CN116580574 A CN 116580574A CN 202310374168 A CN202310374168 A CN 202310374168A CN 116580574 A CN116580574 A CN 116580574A
- Authority
- CN
- China
- Prior art keywords
- vehicle
- intersection
- time
- traffic flow
- vehicles
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000012544 monitoring process Methods 0.000 title claims abstract description 71
- 238000000034 method Methods 0.000 title claims abstract description 61
- 239000011159 matrix material Substances 0.000 claims abstract description 55
- 238000007726 management method Methods 0.000 claims abstract description 24
- 238000004364 calculation method Methods 0.000 claims description 31
- 238000010586 diagram Methods 0.000 claims description 18
- 238000004891 communication Methods 0.000 claims description 17
- 238000009825 accumulation Methods 0.000 claims description 14
- 238000013473 artificial intelligence Methods 0.000 claims description 14
- 238000001514 detection method Methods 0.000 claims description 14
- 230000006870 function Effects 0.000 claims description 14
- 238000000605 extraction Methods 0.000 claims description 13
- 238000007405 data analysis Methods 0.000 claims description 9
- 238000012549 training Methods 0.000 claims description 9
- 230000002159 abnormal effect Effects 0.000 claims description 7
- 238000003825 pressing Methods 0.000 claims description 5
- 238000012937 correction Methods 0.000 claims description 4
- 238000013135 deep learning Methods 0.000 claims description 4
- 238000011478 gradient descent method Methods 0.000 claims description 3
- 238000003062 neural network model Methods 0.000 claims description 3
- 238000013528 artificial neural network Methods 0.000 claims description 2
- 230000003993 interaction Effects 0.000 claims description 2
- 238000013499 data model Methods 0.000 abstract description 8
- 238000012545 processing Methods 0.000 abstract description 8
- 230000007547 defect Effects 0.000 abstract description 4
- 238000011217 control strategy Methods 0.000 abstract description 3
- 238000005265 energy consumption Methods 0.000 abstract description 3
- 230000001186 cumulative effect Effects 0.000 description 17
- 238000005516 engineering process Methods 0.000 description 13
- 238000010276 construction Methods 0.000 description 5
- 238000004458 analytical method Methods 0.000 description 3
- 101150039322 outE gene Proteins 0.000 description 3
- 238000011176 pooling Methods 0.000 description 3
- 238000013527 convolutional neural network Methods 0.000 description 2
- 238000005192 partition Methods 0.000 description 2
- 238000012800 visualization Methods 0.000 description 2
- 230000001419 dependent effect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/07—Controlling traffic signals
- G08G1/08—Controlling traffic signals according to detected number or speed of vehicles
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/01—Detecting movement of traffic to be counted or controlled
- G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
- G08G1/0125—Traffic data processing
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/01—Detecting movement of traffic to be counted or controlled
- G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
- G08G1/0137—Measuring and analyzing of parameters relative to traffic conditions for specific applications
- G08G1/0145—Measuring and analyzing of parameters relative to traffic conditions for specific applications for active traffic flow control
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Pure & Applied Mathematics (AREA)
- Mathematical Physics (AREA)
- Theoretical Computer Science (AREA)
- Chemical & Material Sciences (AREA)
- Computational Mathematics (AREA)
- Analytical Chemistry (AREA)
- Mathematical Optimization (AREA)
- Mathematical Analysis (AREA)
- Evolutionary Biology (AREA)
- Operations Research (AREA)
- Probability & Statistics with Applications (AREA)
- Bioinformatics & Computational Biology (AREA)
- Algebra (AREA)
- Life Sciences & Earth Sciences (AREA)
- Databases & Information Systems (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention relates to the technical field of traffic dynamic management and control, in particular to a road traffic multidirectional dynamic management and control method based on traffic flow monitoring, which can dynamically adjust road traffic management and control measures according to the running state of road traffic, further improve traffic efficiency and reduce traffic energy consumption, and aims at the defects and shortcomings in the prior art, the method for conducting multidirectional dynamic management and control on large traffic flow based on traffic flow monitoring, which can monitor and predict traffic flow data in real time, convert the traffic flow data into a matrix and a coordinate data model in a multidimensional manner, analyze and process the matrix and the coordinate data model by an artificial intelligent algorithm, and further output a red-green lamp management and control strategy efficiently and accurately, so that traffic management and control capability is effectively improved; compared with the prior art, the method has the advantages that the dimension reduction reconstruction is carried out on the real-time traffic data such as videos, pictures and radar signals, so that the storage capacity and the storage cost of the traffic data are greatly reduced, and the processing efficiency of the data is improved.
Description
Technical field:
the invention relates to the technical field of traffic dynamic control, in particular to a road traffic multidirectional dynamic control method based on traffic flow monitoring, which can dynamically adjust road traffic control measures according to road traffic running states, further improve traffic efficiency and reduce traffic energy consumption.
The background technology is as follows:
with the increase of the number of motor vehicles, the problems of traffic jam, low vehicle passing efficiency and the like are increased, and the problems of long time for waiting for a red light when a vehicle is parked, difficult smooth passing of large traffic flow, high energy consumption, high pollution and the like are highlighted due to the fact that traffic data processing technology is relatively backward, and traffic management work bears huge pressure.
The existing method for relieving traffic management pressure mainly comprises road traffic extension, traffic total amount control and traffic reduction (number limiting traffic and the like), bus and bicycle travel encouragement, tide lane traffic staggering traffic and the like, but the effect is not obvious. At present, a technology for improving management and control experience and strategy by analyzing historical traffic data is also available, and a green wave passing technology is typical, but through technical analysis and practice, the green wave passing technology is only applicable to specific passing directions and passing speeds, and the practicability is poor.
At present, an artificial intelligence algorithm technology is also adopted in a small amount, traffic flow data are collected and analyzed in real time, traffic light timing is optimized, and traffic control capacity is improved, however, practice proves that the methods are firstly incomplete in data feature mining, low in data utilization efficiency and relatively general in technical method, secondly, the 'black box' operation of artificial intelligence is too dependent, the content of data analysis and operation cannot be quantitatively explained, the 'AI uncertainty' problem cannot be avoided, and if the technology is widely applied to public safety fields such as traffic control, potential safety hazards cannot be avoided.
The invention comprises the following steps:
aiming at the defects and shortcomings in the prior art, the invention provides the road traffic multidirectional dynamic control method based on traffic flow monitoring, which can monitor the real-time state of road traffic and give out accurate traffic control instructions according to the real-time state, thereby effectively improving traffic efficiency and reducing congestion.
The invention is achieved by the following measures:
the road traffic multidirectional dynamic control method based on traffic flow monitoring is characterized by comprising the following steps of:
step 1: constructing a traffic intersection area model: taking an intersection stop line and an extension line of the stop line as boundaries, setting a central intersection area of a traffic intersection as a control area, and setting road extension areas in multiple directions as monitoring areas; setting a region from which data flows out from the monitoring region to the control region as an out region and a region from which data flows into the monitoring region from the control region as an in region in each monitoring region;
step 2: and (3) planning a traffic path for the traffic intersection area model established in the step (1): each out area of the upper intersection monitoring area is used as a starting point, and is communicated with the out area of the intersection according to the left row, the straight row and the right row, so that a communication relation diagram of the out area of the intersection and the out area of the upper intersection is constructed; when a special vehicle exists at the intersection, other communication relation diagrams can be reconstructed for joint use;
Step 3: data acquisition and index extraction: the method comprises the steps that video, picture and radar signal information of vehicle conditions are acquired in real time according to K areas for an out area of a traffic intersection and one or the next N out areas, original data are acquired for storage and use, N is more than or equal to 1, K is more than or equal to 1, and the value of K depends on the total length of the out area and the section length of the area;then, detecting and tracking the characteristics, the positions and the motion states of the vehicle, extracting characteristic information indexes, and pressing the extracted characteristic information indexes into m i *n i The matrix is used for data feature storage and use, i is more than 0 and less than or equal to K; then, extracting data analysis indexes of the stored data features, wherein the data analysis indexes comprise the number and positions of vehicles in each traveling direction in K areas, the vehicle density in K areas, the inflow accumulation amount of the vehicles and the outflow accumulation amount of the vehicles, and the vehicles in each traveling direction in K areas comprise on-road vehicles and vehicles to be driven;
step 4: judging the traffic state of the intersection, wherein the judgment comprises the judgment of the multi-directional outflow of the large traffic at the upper intersection and the judgment of the existing large traffic at the intersection;
step 5: predicting the passing time of the existing large traffic flow at the intersection:
Step 5-1: predicting the traffic time of the to-be-driven vehicle at the intersection: acquiring a to-be-released vehicle flow release time curve through data set training, firstly acquiring a vehicle number and a pass time data set, setting the vehicle number as x, the vehicle pass time as y, and the acquired sample group number as Q, and marking the i-th sample as (x) i ，y i ) The method comprises the steps of carrying out a first treatment on the surface of the Next, let the vehicle passing time be g (x), and find g (x), there are two implementations:
1) Calculating the average passing time or weighted average passing time of samples under each vehicle quantity x as a predicted value of the vehicle passing time or determining the predicted value by using a gradient descent method to achieve the aim of reducing the predicted error, finally obtaining a calculation error of R, then correcting the error, setting an error threshold value of Rth, finishing the calculation when R is less than Rth, and calculating x when R is less than Rth i Error of single point, let r i ＝g(x i )-y i Setting a single-point error threshold as rth, when r i When r is larger than rth, the point is taken as an abnormal value to be removed, and the calculation process is not participated;
2) The vehicle passing time is calculated by a formula, and a vehicle passing time prediction function g (x) is obtained, wherein the g (x) can be a piecewise function or a non-piecewise function, and a polynomial fitting formula of the g (x) in each segment is as follows:
g(x；α 0 ，α 2 ，…，α n )＝α 0 x n +α 1 x n-1 +α 2 x n-2 +…+α n
X of K groups of samples i The fitting error is set asBy calculating the prediction function and true y i Error among them, so that the error is minimized to determine the parameter set alpha 0 ，α 1 ，…，α n ；
Then, performing fitting error correction, setting an error threshold as Rth, finishing fitting when R is smaller than Rth, and calculating x when R is smaller than Rth i Error of single point, let r i ＝g(x i ；α 0 ，α 1 ，…，α n )-y i Setting a single-point error threshold as rth, when r i When r is larger than rth, the point is removed as an abnormal value and does not participate in the next round of fitting process;
step 5-2: calculating the corresponding position of the vehicle flow release moment: the vehicle to be driven obtained through the step 4 is marked as N await The on-road vehicle is denoted as N onwayl Predicting the current vehicle flow release time to be released, wherein the prediction result is g (N) await ) Then at a known in-transit vehicle flow velocity v onway On the premise of (1) carrying out the calculation of the large traffic flow distance threshold value (namely, the position of the large traffic flow at the release moment) according to the release time. According to the passing time, a time identity is established:
wherein δ is the allowable time error term, t await For the time required for the traffic to pass through the monitoring zone,
t onway to the monitoring area for the in-transit trafficTime to control zone boundary;
obtaining a distance threshold L of the in-transit vehicle flow onway ：
Step 5-3: solving the release time and release time of the cart flow:
under the condition that the area reaches the priority release, determining the optimal passing time and release time according to the conditions of the on-road large traffic flow, other on-road vehicles and vehicles to be released, wherein certain errors are allowed in the time and the time, and the release time is the time when the large traffic flow reaches a parking line when no other on-road vehicles or vehicles to be released are in front of the large traffic flow according to the actual time adjustment; when an on-road vehicle or a vehicle to be released exists in front of the large traffic flow, the releasing time is the large traffic flow reaching distance threshold L onway The release time T is calculated as follows:
T＝(g(N await +N onwayl )+δ)+(h(N onway |v onway )+γ)，
wherein h (N) onway |v onway ) At a speed v onway Lower N onway The prediction result of the passing time of the vehicle, delta and gamma are time reserved items;
step 6: predicting the multi-directional outflow of the large traffic flow at the upper intersection:
step 6-1: judging whether the vehicle outflow accumulation amount forms a large vehicle flow, calculating vehicle outflow accumulation amount data of an OUT region through a multi-directional vehicle characteristic matrix of the upper intersection, calculating the vehicle density of a management and control region through a vehicle position matrix of the management and control region, and setting a density threshold value as N 6 The threshold value of the vehicle outflow integrated quantity is N 7 When the density exceeds a threshold value and the outflow accumulation amount of the vehicle in the OUT area also exceeds the threshold value, according to the length of the OUT area of the intersection and the speed of the vehicle flow, calculating the time when the large vehicle flow is expected to reach the parking line of the intersection, outputting a signal and time when the large vehicle flow possibly exists, otherwise, continuing the next step;
Step 6-2: judging whether the vehicle to be released is largeThe traffic flow obtains the running direction in the monitoring area and the number of vehicles to be driven in each direction through the vehicle feature matrix of the multi-direction OUT area of the upper intersection and the communication relation diagram of the step 2, and sets a threshold N 8 Judging whether the to-be-driven vehicles in all directions and the to-be-driven vehicles accumulated in different directions form a large traffic flow, when the number of to-be-driven vehicles is larger than a set threshold value, according to the length of an OUT area of the intersection and the speed of the traffic flow, calculating the time when a green light is put on, the large traffic flow is expected to reach the stop line of the intersection, outputting a signal and time when the large traffic flow possibly exists, otherwise, continuing the next step;
step 6-3: judging whether the total of the vehicles on the way and the vehicles to be released form a large vehicle flow, acquiring the number of the vehicles to be driven in a monitoring area and the number of the vehicles on the way in a designated area through a vehicle characteristic matrix of a multi-directional OUT area of the upper intersection, and setting a threshold N 9 Judging whether the total number forms a large traffic flow, when the total number is larger than a set threshold value, according to the length of an OUT area of the intersection and the traffic flow speed, solving the time when a green light is put, when the large traffic flow is expected to reach a stop line of the intersection, outputting a signal and time when the large traffic flow possibly exists, and if not, continuing the next step;
Step 6-4: judging whether a large traffic flow exists in transit, outputting a signal and a position of the large traffic flow possibly existing in each direction by using a vehicle feature matrix of a multi-directional OUT area of the upper intersection by using the method in the step 4-3, outputting the time when the large traffic flow is expected to reach a stop line of the intersection according to the length of the OUT area of the intersection and the speed of the traffic flow, otherwise, giving up tracking of the large traffic flow, and considering that the large traffic flow which does not flow OUT in multiple directions exists at the upper intersection;
step 7: accurate control of output to cart flow:
step 7-1: according to the prediction of the incoming large traffic flow at the intersection and the detection result of the existing number of vehicles at the intersection, accurate control is implemented, if the incoming large traffic flow at the intersection is predicted or the existing large traffic flow at the intersection is detected through the step 4, the release time of the large traffic flow, namely the time when a green light is started and the release time are determined through the method in the step 5, and the step 7-2: according to the prediction result of the multi-direction flowing OUT of the large traffic flow at the upper intersection, implementing accurate control, when the large traffic flow is not predicted at the intersection, if the multi-direction flowing OUT of the large traffic flow at the upper intersection is predicted by the step 6, predicting the time of the large traffic flow reaching the parking line of the intersection by combining the signal lamp release time of the upper intersection and the length and the speed of the OUT zone of the road section, and implementing the control of the intersection, wherein the release time of the large traffic flow after reaching the parking line is determined according to the final vehicle number of the large traffic flow detected at the intersection; when the large traffic flows out from the upper intersection in multiple directions and the large traffic flows are detected at the same time, the traffic of the large traffic flows at the intersection is preferentially considered, so that the release time of the large traffic flows, namely the time of turning on the green lamp and the release time are determined.
In the step 3 of the invention, the characteristics, the vehicle position and the vehicle motion condition of the vehicle are detected, tracked and the characteristic information indexes are extracted, and the method can be realized by utilizing an image recognition and signal recognition artificial intelligence algorithm, and the extracted characteristic information indexes are calculated according to m i *n i The matrix is used for data feature storage and use, i is more than 0 and less than or equal to K; or the data characteristic indexes are obtained for storage and use by directly obtaining interaction information with the vehicle.
The step 4 of the invention specifically comprises the following steps:
step 4-1: judging whether the intersection has an incoming large traffic flow or not, and judging in a first mode or a second mode: in a first mode, vehicle density or vehicle number data in a historical time T of the intersection is obtained through a vehicle feature matrix of an intersection monitoring area converging into an entrance area, and a density threshold value or the vehicle number is set to be N 1 And predicting whether the large traffic flow exists or not according to the vehicle density or the vehicle number at the intersection. In the time interval T, when the fluctuation in density or number of vehicles exceeds a threshold, there may be a large traffic; establishing a two-dimensional convolution time sequence deep learning network model, judging whether a large traffic flow to be imported exists at the intersection, acquiring a vehicle position matrix or a video image in a time interval T as an input end of the neural network model, taking the acquired multidimensional information matrix as a sample, taking the large traffic flow as a label of the sample, constructing a training data set, constructing a convolution neural network structure, and sending the vehicle concentration matrix at different moments into the network to predict whether the large traffic flow exists;
Step 4-2: if the intersection is judged to be remitted into a large traffic flow, predicting the large traffic flow number through the number of vehicles, the passing time and the average outflow quantity of the vehicles at the last intersection, and if the number of the vehicles is larger than a set threshold value, outputting signals, positions and the number of the vehicles in the large traffic flow, wherein the signals, the positions and the number of the vehicles possibly exist in the large traffic flow;
step 4-3: judging whether the intersection has large traffic flow or not, calculating the vehicle density and the vehicle quantity data in K areas through the vehicle feature matrix of the OUT area of the intersection, acquiring the vehicle inflow accumulated quantity data in the history time T, and setting a vehicle inflow accumulated quantity threshold value N 2 Vehicle density threshold N 3 Number of vehicles threshold N 4 On-road traffic flow zone length threshold N 5 When the inflow accumulation amount of the vehicles is larger than a set threshold value, outputting a signal and a position where a large vehicle flow possibly exists; by means of continuously expanding K small areas, the vehicle density or the average vehicle number in the areas is calculated iteratively, the largest detection area with the vehicle density larger than a threshold value or the average vehicle number larger than the threshold value is output, if the length of the detection area is larger than the length threshold value of the large traffic flow area, signals and positions where large traffic flows possibly exist are output, the vehicle number of the large traffic flow in the middle and the other vehicle numbers in the middle are obtained, and the vehicle number of the large traffic flow in the middle is recorded as N onway The other vehicles in transit are counted as N onwayl The method comprises the steps of carrying out a first treatment on the surface of the Otherwise, giving up the tracking of the large traffic flow, and considering that the large traffic flow on the way does not exist at the intersection;
step 4-4: acquiring the running direction in the monitoring area and the number of vehicles to be driven in each direction by utilizing the vehicle characteristic matrix of the OUT area of the intersection, and recording the number of the traffic flow to be driven in the lane as N await 。
The invention also comprises a step 8 for acquiring the position of the special vehicle and controlling the passing of the multiple intersections, and specifically comprises the following steps:
step 8-1: judging whether a special vehicle exists at the intersection, dividing an intersection monitoring area into K areas, directly reading vehicle marking information or acquiring special vehicle information through a vehicle feature matrix, and acquiring the special vehicle if a certain vehicle is marked (or identified by a feature) as the special vehicleSetting the total time of the current moment and the departure moment as T, and step 8-2: calculating green light passing conditions of N intersections in front, setting management and control (determined by driving routes) that N intersections in front participate in green light passing, acquiring the positions of special vehicles and the vehicle characteristic matrix of N intersection monitoring areas in front, calculating the time for the special vehicles to reach the ith intersection parking line in front according to the speed and position information of the special vehicles if the special vehicles do not have departure time information (namely are on the road), and setting the time as ST i I is more than or equal to 0 and less than or equal to N, the 0 th intersection is the intersection, and if a special vehicle has departure time information, the vehicle speed participates in calculation by using a set fixed vehicle speed; then, calculating the total number of the vehicles in front of the special vehicles in the monitoring area of the ith intersection, calculating the total traffic time of the vehicles in front of the monitoring area of the ith intersection by using the method for predicting the traffic time in the step 5, and setting the total traffic time as AT i The method comprises the steps of carrying out a first treatment on the surface of the Finally, setting the time reservation item as gamma, if the special vehicle does not have departure time information, when ST i ≤AT i The green light release is immediately carried out at the ith intersection by +gamma, if the special vehicle has departure time information, the green light release needs to be postponed by T,
step 8-3: and (3) managing and controlling the traffic of a plurality of intersections, and when special vehicles exist in the intersections, managing and controlling the plurality of intersections by simultaneously calculating whether the release conditions of the front N intersections are met.
Aiming at the defects and shortcomings in the prior art, the invention provides a method for multi-direction dynamic management and control of large traffic flow based on traffic flow monitoring, which can monitor and predict traffic flow data in real time, then convert the traffic flow data into a matrix and coordinate data model in a multi-dimensional way, analyze and process the traffic flow data by an artificial intelligent algorithm, and further output a red-green lamp control strategy efficiently and accurately, thereby effectively improving traffic control capability; compared with the prior art, (1) through carrying out dimension reduction reconstruction on real-time traffic data such as videos, pictures and radar signals, firstly, the storage capacity and the storage cost of the traffic data are greatly reduced, the processing efficiency of the data is improved, secondly, a series of quantifiable technical indexes are obtained through dimension reduction reconstruction, the traffic data processing becomes more accurate and standard, and thirdly, a series of quantifiable technical indexes are combined with digital twinning, virtual engines and other technologies, so that the real historical traffic data can be virtually restored; (2) The technical index data is subjected to multidimensional feature analysis by using a coordinate curve and a matrix data model, so that a data rule can be acquired more accurately, abnormal data values can be found, traffic data processing visualization is realized, the problems of 'black box' operation, 'AI uncertainty' and the like of artificial intelligence can be solved, and roads are widened for the wide application of the artificial intelligence technology in public safety fields such as traffic control and the like; (3) The coordinate curve, the matrix data model and the algorithm innovatively introduced by the invention can be fused with artificial intelligence technology and product depth, and can provide reliable technical support for intelligent traffic control, vehicle-road coordination and automatic driving after being continuously accumulated and optimized.
Description of the drawings:
fig. 1 is a flow chart of the present invention.
FIG. 2 is a flow chart of the present invention for judging traffic flow at the intersection.
FIG. 3 is a flow chart of the present invention for judging the traffic time of the large traffic flow at the intersection.
FIG. 4 is a flow chart of a particular vehicle management and control in accordance with the present invention.
Fig. 5 is a schematic diagram of traffic intersection region model construction in embodiment 1 of the present invention.
FIG. 6 is a schematic diagram of a two-dimensional convolution time series deep learning network model in embodiment 3 of the present invention.
Fig. 7 is a diagram showing the communication relationship between a plurality of intersections in embodiment 4 of the present invention.
The specific embodiment is as follows:
the invention will be further described with reference to the drawings and examples.
Example 1:
in the method, the traffic road conditions are partitioned, two road junction traffic communication diagrams are constructed, traffic information such as road conditions is acquired, the large traffic flow of the last road junction is predicted in real time, and the traffic situation of the road junction is adjusted. The specific process is as follows:
step 1: region construction: constructing a traffic intersection area model, taking a stop line and an extension line of the stop line as boundaries, setting a central intersection area of the traffic intersection as a management and control area, and setting road extension areas in multiple directions as a monitoring area; setting a region from which data flows out from the monitoring region to the control region as an out region and a region from which data flows into the monitoring region from the control region as an in region in each monitoring region;
Step 2: planning a traffic path of the area model: each out area of the upper intersection monitoring area is used as a starting point, and is communicated with the out area of the intersection according to the left row, the straight row and the right row, so that a communication relation diagram of the out area of the intersection and the out area of the upper intersection is constructed as shown in an attached figure 7;
step 3: data acquisition and index extraction: setting the length of out areas in multiple directions of a previous intersection to be 1250 meters, dividing an area by 50 meters, collecting video, pictures and radar signal information of vehicle conditions in real time according to 25 areas in the out areas in all directions of a previous traffic road, and acquiring original data for storage and use; then, using image recognition and signal recognition artificial intelligence algorithm to detect and track the vehicle characteristics, vehicle position and vehicle movement condition and extract characteristic information index, then pressing m i *n i The matrix is used for data feature storage and use, i is more than 0 and less than or equal to 25; finally, the stored data features are further subjected to data analysis index extraction calculation, and specifically the method comprises the number and positions (including vehicles on the way and vehicles on the way) of vehicles in each running direction in 25 areas, the vehicle density in 25 areas, the vehicle inflow accumulation amount and the vehicle outflow accumulation amount, and the method is shown in the attached drawing.
Vehicle quantity feature extraction matrix at a certain moment:
vehicle density feature extraction matrix at a certain moment:
vehicle outflow cumulative amount: (the cumulative amount was calculated in 4 small time periods and then the outflow amount was calculated from the moment forward for a while by addition)
| Out | T1 | T2 | T3 | T4 | ALL |
| OutE | 0 | 0 | 0 | 0 | 0 |
| OutS | 0 | 0 | 0 | 0 | 0 |
| OutN | 3 | 2 | 2 | 1 | 8 |
Vehicle inflow cumulative amount: (the cumulative amount is calculated in 4 small time periods and then the addition calculation is performed, and the inflow amount is calculated for a period of time before this moment)
| Out | T1 | T2 | T3 | T4 | ALL |
| OutE | 16 | 6 | 6 | 2 | 40 |
| OutS | 1 | 2 | 5 | 2 | 10 |
| OutN | 1 | 3 | 0 | 2 | 6 |
Step 4: predicting the multi-directional outflow of the large traffic flow at the upper intersection:
step 4-1: it is determined whether the outflow cumulative amount of the vehicle constitutes a large traffic flow. The vehicle outflow accumulated quantity data of each OUT area calculated in the step 3 is set to be 30, and the vehicle outflow accumulated quantity threshold value is set to be more than the threshold value, so that whether large vehicle flows exist cannot be judged in the step, and the next step is continued;
step 4-2: and judging whether the vehicles to be released form a large vehicle flow or not. And (3) acquiring the number of vehicles to be driven in each direction in the monitoring area through the vehicle feature matrix of the multi-direction OUT area of the upper intersection and the communication relation diagram of the step (2), setting a threshold value as 60, and judging whether the vehicles to be driven in each direction and the accumulated vehicles to be driven in different directions form a large vehicle flow. The number of to-be-driven vehicles in the OUTE zone is 20, the number of to-be-driven vehicles in the OUTS zone is 10, the total number of to-be-driven vehicles in the OUTN zone is 30, and when the number of to-be-driven vehicles is smaller than a set threshold value, no large traffic flow exists at present, and the judgment of the next step is continued; because the three OUT areas do not form large traffic flow, a threshold value 55 of the total number of vehicles is set, the threshold value of the density is 0.5, the total number of vehicles in the OUTS area is calculated to be 60, the density is 0.038, the OUTS area meets the condition that the total number of vehicles is larger than the threshold value and the density is smaller than the threshold value of the density, therefore, the traffic flow of the OUTS area is compressed by utilizing a red light control mode to obtain large traffic flow, and the OUTE area and the OUTN area are calculated in the same way;
Step 4-3: and judging whether the in-transit vehicle flow exists. Through step 3, vehicle density and vehicle number data in 25 areas (each area is 50 m long) and vehicle inflow cumulative amount data in a certain history time are calculated, a vehicle inflow cumulative amount threshold 50, a vehicle density threshold 0.6, an average vehicle number threshold 20 and an on-road traffic flow area length threshold of 140 m are set. The calculated vehicle outflow cumulative amounts of the OUTE, OUTS, OUTN area are all larger than the threshold value, so that the vehicle density and the average vehicle number in the iterative calculation area are continuously expanded to the left and right sides, the vehicle density calculation of the OUTE area is taken as an example, each small area of the first round is calculated as an independent area, the calculation result is shown in the step 3, 5 areas with the vehicle density larger than 0.6 are calculated, the densities are 0.66,0.61,0.88,0.72,0.6 respectively, 0.66 is the first area close to a parking line and is a waiting area, the remaining four areas are eliminated, the left and right areas are expanded by one area respectively in the second round of iteration, the vehicle density value of the calculation area is 0.54,0.74,0.73,0.48, 2 areas are larger than the threshold value, the third round of iteration is carried out, the calculated vehicle density value is 0.59,0.58, and the calculated vehicle density values are smaller than the threshold value, and the specific calculation table is shown in the drawing. Outputting a detection area with the density of 0.74 in the second round of iteration meeting the condition, wherein the position of the head of the traffic flow in the area is 900 meters away from the parking line of the intersection, and the length is 150 meters, and because the length of the detection area is larger than the length threshold value of the large traffic flow area in the way, the time when the large traffic flow is expected to reach the parking line of the intersection is output according to the length of the OUT area of the intersection, the position of the parking line of the intersection on the large traffic flow distance of the intersection and the traffic flow speed;
Iterative calculation result table:
step 5: accurate control of output to cart flow:
and according to the prediction result of the multi-directional outflow large traffic flow at the upper intersection, accurate control is implemented. 4, predicting that an on-road large traffic flow exists in an OUTE zone of the upper intersection, and determining the time for the large traffic flow to reach a parking line according to the final number of vehicles in the large traffic flow detected by the intersection by combining the time for the large traffic flow to pass through an upper intersection signal lamp, the length of the OUT zone of the road section and the speed of the traffic flow, and predicting the time for the large traffic flow to reach the parking line of the intersection;
example 2:
in the method, traffic information such as road conditions is acquired by setting partitions on the traffic road conditions and constructing a traffic path communication graph, large traffic flow information (including the number and the positions) in the intersection is acquired in real time, and traffic conditions of the intersection are adjusted by combining traffic flow information to be released of the intersection. The specific process is as follows:
step 1: region construction: constructing a traffic intersection area model, taking a stop line and an extension line of the stop line as boundaries, setting a central intersection area of the traffic intersection as a management and control area, and setting road extension areas in multiple directions as a monitoring area; setting a region from which data flows out from the monitoring region to the control region as an out region and a region from which data flows into the monitoring region from the control region as an in region in each monitoring region;
Step 2: planning a traffic path of the area model: each out area of the upper intersection monitoring area is used as a starting point, and is communicated with the out area of the intersection according to the left row, the straight row and the right row, so that a communication relation diagram of the out area of the intersection and the out area of the upper intersection is constructed; the drawing is the same as in example 1.
Step 3: data acquisition and index extraction: setting the length of an outE area of the intersection to be 1250 meters, dividing the total three lanes into straight lanes by 50 meters, collecting video, pictures and radar signal information of vehicle conditions in real time according to 25 areas in the outE area, and acquiring original data for storage and use; then, using image recognition and signal recognition artificial intelligence algorithm to detect and track the vehicle characteristics, vehicle position and vehicle movement condition and extract characteristic information index, then pressing m i *n i The matrix is used for data feature storage and use, i is more than 0 and less than or equal to 25; finally, further carrying out data analysis index extraction calculation on the stored data characteristics, designating the numerical value of each vehicle according to the length of the vehicle by adopting a thermodynamic diagram mode, such as 5 for a common car and 15 for a bus, assigning the vehicle at the designated position in 25 areas, setting the area covered by the numerical value to be 3 times of the size of the vehicle, and then calculating the total number of the thermal values of the vehicle in 25 areas, as shown in the attached drawing Showing; the cumulative amount of vehicle inflow and the cumulative amount of vehicle outflow in the outE area are counted as shown in the drawing.
Vehicle quantity feature extraction matrix at a certain moment:
vehicle thermodynamic numerical matrix at a certain moment:
vehicle outflow cumulative amount: (the cumulative amount was calculated in 4 small time periods and then the outflow amount was calculated from the moment forward for a while by addition)
| Out | T1 | T2 | T3 | T4 | ALL |
| OutE | 0 | 0 | 0 | 0 | 0 |
Vehicle inflow cumulative amount: (the cumulative amount is calculated in 4 small time periods and then the addition calculation is performed, and the inflow amount is calculated for a period of time before this moment)
| Out | T1 | T2 | T3 | T4 | ALL |
| OutE | 16 | 6 | 6 | 2 | 40 |
Step 4: the existing large traffic flow detection at the intersection:
step 4-1: referring to the implementation method of step 4-3 in embodiment 1, through the characteristic matrix of the number of vehicles in the OUT area of the intersection, the thermal value data of the vehicles in 25 areas are calculated, the inflow cumulative amount data of the vehicles in the history time T is calculated, the inflow cumulative amount threshold 60, the average thermal value threshold 220 and the length threshold 150 of the large traffic flow area are set. The calculated vehicle inflow cumulative amount is 40, which is smaller than the set threshold value, and the signal that the large vehicle flow possibly exists and the position cannot be output at the stepA location; then, the next calculation is carried out, the average thermal value of vehicles in the 25 small areas is continuously expanded leftwards and rightwards, the average thermal value of vehicles in the areas is calculated in an iterative manner, the largest detection area with the average thermal value of vehicles being larger than a threshold value is output, if the length of the detection area is larger than the length threshold value of the area of the large traffic flow in the middle, signals and the positions of the large traffic flow possibly exist are output, the calculation result is shown in the attached drawing, the detection area with the average thermal value of 254 in the third iteration meeting the condition is output, the position of the head of the large traffic flow in the area is located at the position, which is away from an intersection stop line by 450 m, the length is 250 m, and the number of vehicles is 82, and the length of the detection area is larger than the length threshold value of the area of the large traffic flow in the middle, so that the large traffic flow in the middle can be considered as the large traffic flow N in the middle onway =82, other vehicles in transit are noted N onwayl ＝8；
Iterative calculation result table:
step 4-2: the number of vehicles to be driven in the monitoring area is obtained by utilizing the characteristic matrix of the number of vehicles in the OUT area of the intersection, and the number of the vehicles to be driven in the lane is recorded as N await ＝20；
Step 5: predicting the traffic time of the traffic flow at the intersection:
step 5-1: and acquiring a to-be-released vehicle flow release time curve through data set training. First, a data set of the number of vehicles and the passing time is acquired, the number of vehicles is x, the passing time of vehicles is y, the number of acquired sample groups is K, and the i-th sample group is marked as (x) i ，y i ) The method comprises the steps of carrying out a first treatment on the surface of the Next, let the vehicle passing time be g (x), and determine g (x). The vehicle passing time is calculated by a formula, and a vehicle passing time prediction function g (x) is obtained, wherein the g (x) can be a piecewise function or a non-piecewise function, and a polynomial fitting formula of the g (x) in each segment is as follows:
g(x；α 0 ，α 2 ，…，α n )＝α 0 x n +α 1 x n-1 +α 2 x n-2 +…+α n
sample K groupX of the book i The fitting error is set asBy calculating the prediction function and true y i Error among them, so that the error is minimized to determine the parameter set alpha 0 ，α 1 ，…，α n 。
Then, performing fitting error correction, setting an error threshold as Rth, finishing fitting when R is smaller than Rth, and calculating x when R is smaller than Rth i Error of single point, let r i ＝g(x i ；α o ，α 1 ，…，α n )-y i Setting a single-point error threshold as rth, when r i And when r is larger than rth, the point is removed as an abnormal value and does not participate in the next round of fitting process.
Step 5-2: position determination at the release moment of the cart flow: predicting the current passing time of the to-be-passed vehicle, and predicting the passing time of the to-be-passed vehicle 20 to be 15 seconds, namely g (N) await ) Because it was detected in the previous step that there were a small number of vehicles in transit in front of the large traffic near the area to be released, the number of vehicles in transit was incorporated into the number to be released, the release time was calculated, and the transit time at 28 vehicles to be released was predicted to be 18 seconds, i.e., g (N) await +N onway1 ) =g (28) =18, then at a known in-transit vehicle flow velocity v onway On the premise of 30km/h, the distance threshold value of the large traffic flow (namely, the position of the large traffic flow at the release moment) is obtained according to the release time. According to the passing time, a time identity is established:
where δ is an allowable time error term, in this embodiment set to 0,
t await for the time required for the traffic to pass through the monitoring zone,
t onway to the in-transit vehicleAnd (5) reaching the boundary of the monitoring area and the control area.
Obtaining a distance threshold L of the in-transit vehicle flow onway ：
Calculated, L onway = (g (28) +0) 40=18×30=540 meters.
Step 5-3: obtaining the release time and release time of the cart flow:
under the condition that the area reaches the priority release, 82 on-road vehicles are calculated, and other on-road vehicles 8 and vehicles to be released 20 are calculated. Because other vehicles in front of the large traffic flow and vehicles to be released are arranged, the releasing moment is the arrival distance threshold L of the large traffic flow onway The time of 540 m, that is, the time of releasing the large traffic when the large traffic is 540 m from the stop line of the intersection, is immediately released because the detected position of the large traffic is less than 540 m from the stop line of the intersection.
The release time T is calculated as follows:
T＝(g(N await +N onwayl )+δ)+(h(N onway |v onway )+γ)，
wherein h (N) onway |v onway ) At a speed v onway Lower N onway As a result of predicting the passage time of the vehicle, δ and γ are time reservation items, and all are set to 0 in this embodiment.
The transit time of the cart flow is calculated as t= (g (28) +0) + (h (82|30) +0) =18+50=68 seconds.
Step 6: accurate control of output to cart flow:
according to the detection result of the existing number of vehicles at the intersection, accurate control is implemented. The existing large traffic flow at the intersection is detected through the step 4, the number of vehicles in the large traffic flow is 82, 28 vehicles to be released and 28 vehicles in the road are shared in front, and the release time (namely, the time of turning on a green light) and the release time of the large traffic flow are determined through the method in the step 5, wherein the release time is the instant release at the moment, and the release time is 68 seconds.
Example 3:
in the method, traffic information such as road conditions is obtained by setting partitions on traffic road conditions and constructing a traffic path communication graph, large traffic flow information (comprising the number and the positions) which is collected by the intersection is predicted in real time, and traffic conditions of the intersection are adjusted by combining traffic flow information to be discharged of the intersection. The specific process is as follows:
step 1: region construction: constructing a traffic intersection area model, taking a stop line and an extension line of the stop line as boundaries, setting a central intersection area of the traffic intersection as a management and control area, and setting road extension areas in multiple directions as a monitoring area; setting a region from which data flows out from the monitoring region to the control region as an out region and a region from which data flows into the monitoring region from the control region as an in region in each monitoring region;
step 2: planning a traffic path of the area model: each out area of the upper intersection monitoring area is used as a starting point, and is communicated with the out area of the intersection according to the left row, the straight row and the right row, so that a communication relation diagram of the out area of the intersection and the out area of the upper intersection is constructed; the drawing is the same as in example 1.
Step 3: data acquisition and index extraction: setting the length of an out area of the intersection to be 1250 meters, dividing the out area into three lanes by 50 meters, collecting video, pictures and radar signal information of the vehicle condition in real time according to 25 areas in the out area, and acquiring original data for storage and use; then, using image recognition and signal recognition artificial intelligence algorithm to detect and track the vehicle characteristics, vehicle position and vehicle movement condition and extract characteristic information index, then pressing m i *n i The matrix is used for data feature storage and use, i is more than 0 and less than or equal to 25; and finally, further carrying out data analysis index extraction calculation on the stored data features, wherein the data analysis index extraction calculation specifically comprises the number and the positions of vehicles in 25 areas and the density of the vehicles in 25 areas.
Step 4: the prediction of the traffic flow entering at the intersection is as follows:
step 4-1: and establishing a two-dimensional convolution time sequence deep learning network model, and predicting whether a large vehicle flow exists or not as shown in figure 6.
First, obtainAnd preprocessing the acquired image or the vehicle position matrix by the vehicle position matrix or the video image in the time interval T. Carrying out real-time snapshot on a lane in a certain distance (the area closest to the control area) from the upper road opening through a camera, and carrying out image stitching; continuously acquiring a plurality of lane images at the same position in T time from the beginning of the driving of the first vehicle, and taking the acquired multidimensional information matrix as an input end of a neural network model; then, a prediction model is built and parameter training optimization is carried out. And taking the acquired multidimensional information matrix as a sample (model input) and whether the acquired multidimensional information matrix is a label (model output) taking the on-road traffic flow as the sample, and constructing a training data set to build a CNN convolutional neural network structure. The method comprises the steps of enabling a plurality of time data acquired in a time interval T to enter a two-dimensional convolution layer respectively, enabling the number of filters to be 32, enabling the step length to be 2, enabling the convolution kernel size to be 7*7, enabling a layer of maximum pooling layer to be connected later, enabling the step length to be 2, enabling the convolution kernel size to be 3*3, then entering two layers of the same convolution layer, enabling the filters to be 64 and 128 respectively, enabling other parameters to be the same as those of a first layer, finally connecting a layer of average pooling layer, outputting linear characteristic layer results through a full connection layer, splicing linear characteristics at different time according to time sequence, sending the linear characteristics into a one-dimensional convolution network, enabling the number of the filters to be 32, 64 and 128 in sequence, enabling the step length to be 2, enabling the convolution kernel size to be 7, enabling the step length to be 2, enabling the convolution kernel size to be 2 to be the largest pooling layer to be connected later, and finally building a Dense Dense layer network, and predicting whether on-road traffic flow exists. The network model block diagram is shown in the attached drawing. Calculating loss of the network model by using the cross entropy loss function, and setting a prediction result as The true label is y, and the cross entropy loss function formula is as follows: />Until the model loss is smaller than a set threshold L, finishing model training, and storing calculation parameters in the model for real-time prediction; and finally, predicting the traffic flow. Inputting the multidimensional information matrix into a model to obtain a prediction result as to whether the in-transit vehicle flow exists,and the position information of the on-road traffic flow can be obtained together through the arrangement positions of the cameras.
Step 4-2: through prediction, the number threshold value is set to be 50, the number of vehicles to be released in a period of time of the last intersection is 80, and the accumulated outflow quantity of the vehicles is 40, wherein the number of vehicles to be released in the period of time of the last intersection is greater than the set threshold value, so that signals possibly with large traffic are output, the predicted number of the large traffic is 80, the position of the large traffic is 1200 meters away from the stop line of the intersection, and the large traffic is positioned at the position of the out area of the intersection just before the large traffic is released;
step 5: predicting the traffic time of the traffic flow at the intersection:
step 5-1: and acquiring a to-be-released vehicle flow release time curve through data set training. First, a data set of the number of vehicles and the passing time is acquired, the number of vehicles is x, the passing time of vehicles is y, the number of acquired sample groups is K, and the i-th sample group is marked as (x) i ，y i ) The method comprises the steps of carrying out a first treatment on the surface of the Next, let the vehicle passing time be g (x), and determine g (x). And calculating the average passing time or the weighted average passing time of the samples under each vehicle quantity x, and taking the average passing time or the weighted average passing time as a predicted value of the vehicle passing time, or determining the predicted value by using a gradient descent method, so as to achieve the aim of reducing the predicted error, and finally obtaining the calculated error as R. Then, correction of the error is performed, an error threshold value is set as Rth, when R < Rth, calculation is finished, and when R < Rth, x is calculated i Error of single point, let r i ＝g(x i )-y i Setting a single-point error threshold as rth, when r i When r is larger than rth, the point is taken as an abnormal value to be removed, and the calculation process is not participated.
Step 5-2: position determination at the release moment of the cart flow: in this embodiment, a non-passing vehicle in front of a large traffic flow detected at the current intersection is set, the number of vehicles to be released is 20, the current time for releasing the large traffic flow is predicted, and the prediction result is g (N await ) =15, then at a known in-transit vehicle flow velocity v onway On the premise of=30, the large traffic flow distance threshold (i.e., the position at the large traffic flow release time) is obtained according to the release time. Obtaining the in-transitDistance threshold L of traffic flow onway ：
L onway ＝(g(N await +0)+0)*v onway = (15×30) =450 meters.
Step 5-3: obtaining the release time and release time of the cart flow:
under the condition that the area reaches the priority release, the conditions of 80 on-road large traffic flows, 0 on-road vehicles and 20 vehicles to be released are calculated. Since the vehicles to be released are in front of the large traffic flow, the release time is the large traffic flow reaching distance threshold L onway At the time of =450 meters, that is, 450 meters from the stop line of the intersection, the large traffic is released. The release time T is calculated as follows:
T＝(g(N await )+δ)+(h(N onway |v onway )+γ)，
wherein h (N) onway |v onway ) At a speed v onway Lower N onway As a result of the prediction of the passage time of the vehicle,
delta and gamma are time reserved items, and in the embodiment, the preset value is 0.
The transit time of the cart flow is calculated as t= (g (20) +0) + (h (80|30) +0) =15+49=64 seconds.
Step 6: accurate control of output to cart flow:
and according to the prediction result of the incoming traffic flow at the intersection, implementing accurate control. And 4, predicting that the intersection is converged into a large traffic flow, wherein the number of vehicles in the large traffic flow is 80, and 20 vehicles are to be released in front, and determining release time (namely, the time of starting a green light) and release time of the large traffic flow by the method in the step 5, wherein the release time is the time when the large traffic flow reaches 450 meters from the stop line of the intersection, and the release time is 64 seconds.
Example 4:
in the method, traffic road conditions of a plurality of intersections are partitioned, a traffic path communication graph is constructed, traffic information and feature matrixes of the intersections are obtained, and therefore all-road traffic green light control is carried out on special vehicles marked at the intersections. The specific process is as follows:
step 1: region construction: constructing a traffic intersection area model, taking a stop line and an extension line of the stop line as boundaries, setting a central intersection area of the traffic intersection as a management and control area, and setting road extension areas in multiple directions as a monitoring area; setting a region from which data flows out from the monitoring region to the control region as an out region and a region from which data flows into the monitoring region from the control region as an in region in each monitoring region;
step 2: planning a passing path of the multi-intersection regional model: and (3) taking an out area of the intersection monitoring area as a starting point, communicating with the out area of the next intersection in the traveling direction according to the straight line, constructing a communication relation diagram of the out area of the intersection and the out area of the next intersection, and constructing a communication relation diagram of a plurality of intersections by the method, as shown in figure 7.
Step 3: data acquisition and index extraction: setting the length of out areas of four intersections from the intersection to the lower part to be 750 meters, 500 meters, 1000 meters and 2000 meters respectively, dividing an area by 50 meters, collecting video, pictures and radar signal information of vehicle conditions in real time according to 15, 10, 20 and 40 areas respectively, and acquiring original data for storage and use; then, for each intersection area, using image recognition and signal recognition artificial intelligence algorithm to detect and track vehicle characteristics, vehicle position and vehicle movement condition and extract characteristic information index, then according to m i *n i The matrix is used for storing and using data characteristics, wherein i is more than 0 and less than or equal to the number of areas; and finally, respectively calculating the number of vehicles in the four intersection monitoring areas.
Step 4: position acquisition and multi-intersection traffic control of special vehicles:
step 4-1: judging whether a special vehicle exists at the intersection. The special vehicle is set to be received at the current moment of the intersection system, the position of the special vehicle is 600 meters away from the stop line of the intersection, the departure moment is 7 points and 30 minutes, the fixed vehicle speed is set to be 40km/h, and the special vehicle driving route is that two continuous intersections directly move and then the next intersection rotates left, and the special vehicle directly moves to the destination after the left rotation. The current time is known to be 7 points and 20 minutes.
Step 4-2: and calculating green light release conditions of the intersection and the front 3 intersections. Setting the management and control that 3 intersections participate in green light release in front of the intersection, namely the next intersection OAnd the UTE area, the lower intersection OUTE area and the lower intersection OUTN area acquire the vehicle characteristic matrixes of the intersection and the front 3 intersection monitoring areas. The known fixed speed of the special vehicle is 40km/h, the distance from the stop line of the intersection is 600 meters, the time for reaching the stop line of the intersection and the three intersection in front is calculated, and the calculation results are ST in sequence 0 =15 seconds, ST 1 =27.5 seconds, ST 2 =52.5 seconds, ST 3 =102.5 seconds; then, the total number of vehicles in front of the special vehicles in the monitoring areas of the intersection and the front three intersections is calculated, the calculation results are 20, 100, 120 and 80 in sequence, the total traffic time of the vehicles in front of the monitoring areas of the intersection and the front three intersections is calculated by using the method for predicting the traffic time in the step 5 of the embodiment 2, and the calculation results are AT in sequence 0 =15 seconds, AT 1 =31 seconds, AT 2 =43 seconds, AT 3 =37 seconds; finally, setting the time reservation item to be gamma=10 seconds, and according to conditions, meeting ST at the current time point intersection, the next intersection and the next intersection i ≤AT i And (3) carrying out green light release, wherein the condition is not met by the lower crossing, the green light release is not carried out, and the green light release time of all the crossings is postponed for 10 minutes by referring to the departure time of the special vehicle.
Step 4-3: and carrying out multi-intersection traffic control. In order to ensure the smoothness of the road traffic of the special vehicles, the traffic capacity of the vehicles passing through a plurality of intersections at the current moment is calculated, and when the special vehicles start, the intersections, the lower intersections and the lower intersections all need to pass green lights. Wherein the vehicle traffic capacity is updated once at intervals.
Aiming at the defects and shortcomings in the prior art, the invention provides a method for multi-direction dynamic management and control of large traffic flow based on traffic flow monitoring, which can monitor and predict traffic flow data in real time, then convert the traffic flow data into a matrix and coordinate data model in a multi-dimensional way, analyze and process the traffic flow data by an artificial intelligent algorithm, and further output a red-green lamp control strategy efficiently and accurately, thereby effectively improving traffic control capability; compared with the prior art, (1) through carrying out dimension reduction reconstruction on real-time traffic data such as videos, pictures and radar signals, firstly, the storage capacity and the storage cost of the traffic data are greatly reduced, the processing efficiency of the data is improved, secondly, a series of quantifiable technical indexes are obtained through dimension reduction reconstruction, the traffic data processing becomes more accurate and standard, and thirdly, a series of quantifiable technical indexes are combined with digital twinning, virtual engines and other technologies, so that the real historical traffic data can be virtually restored; (2) The technical index data is subjected to multidimensional feature analysis by using a coordinate curve and a matrix data model, so that a data rule can be acquired more accurately, abnormal data values can be found, traffic data processing visualization is realized, the problems of 'black box' operation, 'AI uncertainty' and the like of artificial intelligence can be solved, and roads are widened for the wide application of the artificial intelligence technology in public safety fields such as traffic control and the like; (3) The coordinate curve, the matrix data model and the algorithm innovatively introduced by the invention can be fused with artificial intelligence technology and product depth, and can provide reliable technical support for intelligent traffic control, vehicle-road coordination and automatic driving after being continuously accumulated and optimized.
Claims (4)
1. The road traffic multidirectional dynamic control method based on traffic flow monitoring is characterized by comprising the following steps of:
step 1: constructing a traffic intersection area model: taking an intersection stop line and an extension line of the stop line as boundaries, setting a central intersection area of a traffic intersection as a control area, and setting road extension areas in multiple directions as monitoring areas; setting a region from which data flows out from the monitoring region to the control region as an out region and a region from which data flows into the monitoring region from the control region as an in region in each monitoring region;
step 2: and (3) planning a traffic path for the traffic intersection area model established in the step (1): each out area of the upper intersection monitoring area is used as a starting point, and is communicated with the out area of the intersection according to the left row, the straight row and the right row, so that a communication relation diagram of the out area of the intersection and the out area of the upper intersection is constructed; when a special vehicle exists at the intersection, other communication relation diagrams can be reconstructed for joint use;
step 3: data acquisition and index extraction: traffic route control for out area and last or next N number of traffic intersectionsThe out areas in multiple directions collect video, picture and radar signal information of the vehicle condition in real time according to K areas, original data are obtained for storage and use, N is more than or equal to 1, K is more than or equal to 1, and the value of K depends on the total length of the out areas and the slicing length of the areas; then, detecting and tracking the characteristics, the positions and the motion states of the vehicle, extracting characteristic information indexes, and pressing the extracted characteristic information indexes into m i *n i The matrix is used for data feature storage and use, i is more than 0 and less than or equal to K; then, extracting data analysis indexes of the stored data features, wherein the data analysis indexes comprise the number and positions of vehicles in each traveling direction in K areas, the vehicle density in K areas, the inflow accumulation amount of the vehicles and the outflow accumulation amount of the vehicles, and the vehicles in each traveling direction in K areas comprise on-road vehicles and vehicles to be driven;
step 4: judging the traffic state of the intersection, wherein the judgment comprises the judgment of the multi-directional outflow of the large traffic at the upper intersection and the judgment of the existing large traffic at the intersection;
step 5: predicting the passing time of the existing large traffic flow at the intersection:
step 5-1: predicting the traffic time of the to-be-driven vehicle at the intersection: acquiring a to-be-released vehicle flow release time curve through data set training, firstly acquiring a vehicle number and a pass time data set, setting the vehicle number as x, the vehicle pass time as y, and the acquired sample group number as Q, and marking the i-th sample as (x) i ,y i ) The method comprises the steps of carrying out a first treatment on the surface of the Next, let the vehicle passing time be g (x), and find g (x), there are two implementations:
1) Calculating the average passing time or weighted average passing time of samples under each vehicle quantity x as a predicted value of the vehicle passing time or determining the predicted value by using a gradient descent method to achieve the aim of reducing the predicted error, finally obtaining a calculation error of R, then correcting the error, setting an error threshold value of Rth, finishing the calculation when R is less than Rth, and calculating x when R is less than Rth i Error of single point, let r i ＝g(x i )-y i Setting a single point error threshold as rth, when r i When < rth, this point is taken as valid data, when r > rth, this pointAs outlier rejection, the method does not participate in the calculation process;
2) The vehicle passing time is calculated by a formula, and a vehicle passing time prediction function g (x) is obtained, wherein the g (x) can be a piecewise function or a non-piecewise function, and a polynomial fitting formula of the g (x) in each segment is as follows:
g(x；α 0 ,α 2 ,...,α n )＝α 0 x n +α 1 x n-1 +α 2 x n-2 +…+α n
x of K groups of samples i The fitting error is set asBy calculating the prediction function and true y i Error among them, so that the error is minimized to determine the parameter set alpha 0 ,α 1 ,…,α n ；
Then, performing fitting error correction, setting an error threshold as Rth, finishing fitting when R is smaller than Rth, and calculating x when R is smaller than Rth i Error of single point, let r i ＝h(x i ；α 0 ,α 1 ,…,α n )-y i Setting a single-point error threshold as Rth, when r i When r is larger than rth, the point is removed as an abnormal value and does not participate in the next round of fitting process;
step 5-2: calculating the corresponding position of the vehicle flow release moment: the vehicle to be driven obtained through the step 4 is marked as N await The on-road vehicle is denoted as N onwayl Predicting the current vehicle flow release time to be released, wherein the prediction result is g (N) await ) Then at a known in-transit vehicle flow velocity v onway On the premise of (1) carrying out the calculation of the large traffic flow distance threshold value (namely, the position of the large traffic flow at the release moment) according to the release time. According to the passing time, a time identity is established:
wherein δ is the allowable time error term, t await For the time required for the traffic to pass through the monitoring zone,
t onway the time when the on-road traffic flow reaches the boundary of the monitoring area and the control area;
obtaining a distance threshold L of the in-transit vehicle flow onway ：
Step 5-3: solving the release time and release time of the cart flow:
under the condition that the area reaches the priority release, determining the optimal passing time and release time according to the conditions of the on-road large traffic flow, other on-road vehicles and vehicles to be released, wherein certain errors are allowed in the time and the time, and the release time is the time when the large traffic flow reaches a parking line when no other on-road vehicles or vehicles to be released are in front of the large traffic flow according to the actual time adjustment; when an on-road vehicle or a vehicle to be released exists in front of the large traffic flow, the releasing time is the large traffic flow reaching distance threshold L onway The release time T is calculated as follows:
T＝(g(N await +N onwayl )+δ)+(h(N onway |v onway )+γ),
wherein h (N) onway |v onway ) At a speed v onway Lower N onway The prediction result of the passing time of the vehicle, delta and gamma are time reserved items;
Step 6: predicting the multi-directional outflow of the large traffic flow at the upper intersection:
step 6-1: judging whether the vehicle outflow accumulation amount forms a large vehicle flow, calculating vehicle outflow accumulation amount data of an OUT region through a multi-directional vehicle characteristic matrix of the upper intersection, calculating the vehicle density of a management and control region through a vehicle position matrix of the management and control region, and setting a density threshold value as N 6 The threshold value of the vehicle outflow integrated quantity is N 7 When the density exceeds the threshold value and the vehicle outflow accumulation amount of the OUT region exceeds the threshold value, the large value is obtained according to the length of the OUT region of the intersection and the vehicle flow speedOutputting a signal and time when the traffic flow is expected to reach the stop line of the intersection, if not, continuing the next step;
step 6-2: judging whether vehicles to be released form a large traffic flow, acquiring the running direction in a monitoring area and the number of vehicles to be released in each direction through the vehicle characteristic matrix of the multi-directional OUT area of the upper intersection and the communication relation diagram of the step 2, and setting a threshold N 8 Judging whether the to-be-driven vehicles in all directions and the to-be-driven vehicles accumulated in different directions form a large traffic flow, when the number of to-be-driven vehicles is larger than a set threshold value, according to the length of an OUT area of the intersection and the speed of the traffic flow, calculating the time when a green light is put on, the large traffic flow is expected to reach the stop line of the intersection, outputting a signal and time when the large traffic flow possibly exists, otherwise, continuing the next step;
Step 6-3: judging whether the total of the vehicles on the way and the vehicles to be released form a large vehicle flow, acquiring the number of the vehicles to be driven in a monitoring area and the number of the vehicles on the way in a designated area through a vehicle characteristic matrix of a multi-directional OUT area of the upper intersection, and setting a threshold N 9 Judging whether the total number forms a large traffic flow, when the total number is larger than a set threshold value, according to the length of an OUT area of the intersection and the traffic flow speed, solving the time when a green light is put, when the large traffic flow is expected to reach a stop line of the intersection, outputting a signal and time when the large traffic flow possibly exists, and if not, continuing the next step;
step 6-4: judging whether a large traffic flow exists in transit, outputting a signal and a position of the large traffic flow possibly existing in each direction by using a vehicle feature matrix of a multi-directional OUT area of the upper intersection by using the method in the step 4-3, outputting the time when the large traffic flow is expected to reach a stop line of the intersection according to the length of the OUT area of the intersection and the speed of the traffic flow, otherwise, giving up tracking of the large traffic flow, and considering that the large traffic flow which does not flow OUT in multiple directions exists at the upper intersection;
step 7: accurate control of output to cart flow:
step 7-1: according to the prediction of the incoming large traffic flow at the intersection and the detection result of the existing number of vehicles at the intersection, accurate control is implemented, if the incoming large traffic flow at the intersection is predicted or the existing large traffic flow at the intersection is detected through the step 4, the release time of the large traffic flow, namely the time when a green light is started and the release time are determined through the method in the step 5, and the step 7-2: according to the prediction result of the multi-direction flowing OUT of the large traffic flow at the upper intersection, implementing accurate control, when the large traffic flow is not predicted at the intersection, if the multi-direction flowing OUT of the large traffic flow at the upper intersection is predicted by the step 6, predicting the time of the large traffic flow reaching the parking line of the intersection by combining the signal lamp release time of the upper intersection and the length and the speed of the OUT zone of the road section, and implementing the control of the intersection, wherein the release time of the large traffic flow after reaching the parking line is determined according to the final vehicle number of the large traffic flow detected at the intersection; when the large traffic flows out from the upper intersection in multiple directions and the large traffic flows are detected at the same time, the traffic of the large traffic flows at the intersection is preferentially considered, so that the release time of the large traffic flows, namely the time of turning on the green lamp and the release time are determined.
2. The method for multidirectional dynamic control of road traffic based on traffic flow monitoring according to claim 1, wherein in step 3, the characteristics of the vehicle, the position of the vehicle, and the motion condition of the vehicle are detected, tracked, and characteristic information indexes are extracted, and the method is implemented by using image recognition and signal recognition artificial intelligence algorithms, and the extracted characteristic information indexes are calculated according to m i *n i The matrix is used for data feature storage and use, i is more than 0 and less than or equal to K; or the data characteristic indexes are obtained for storage and use by directly obtaining interaction information with the vehicle.
3. The method for multi-directional dynamic management and control of road traffic based on traffic flow monitoring according to claim 1, wherein the step 4 specifically comprises the following steps:
step 4-1: judging whether the intersection has an incoming large traffic flow or not, and judging in a first mode or a second mode:
in a first mode, vehicle density or vehicle number data in a historical time T of the intersection is obtained through a vehicle feature matrix of an intersection monitoring area converging into an entrance area, and a density threshold value or the vehicle number is set to be N 1 And predicting whether the large traffic flow exists or not according to the vehicle density or the vehicle number at the intersection. In the time interval T, when the density or When the fluctuation in the number of vehicles exceeds a threshold, a large vehicle flow may exist; establishing a two-dimensional convolution time sequence deep learning network model, judging whether a large traffic flow to be imported exists at the intersection, acquiring a vehicle position matrix or a video image in a time interval T as an input end of the neural network model, taking the acquired multidimensional information matrix as a sample, taking the large traffic flow as a label of the sample, constructing a training data set, constructing a convolution neural network structure, and sending the vehicle concentration matrix at different moments into the network to predict whether the large traffic flow exists;
step 4-2: if the intersection is judged to be remitted into a large traffic flow, predicting the large traffic flow number through the number of vehicles, the passing time and the average outflow quantity of the vehicles at the last intersection, and if the number of the vehicles is larger than a set threshold value, outputting signals, positions and the number of the vehicles in the large traffic flow, wherein the signals, the positions and the number of the vehicles possibly exist in the large traffic flow;
step 4-3: judging whether the intersection has large traffic flow or not, calculating the vehicle density and the vehicle quantity data in K areas through the vehicle feature matrix of the OUT area of the intersection, acquiring the vehicle inflow accumulated quantity data in the history time T, and setting a vehicle inflow accumulated quantity threshold value N 2 Vehicle density threshold N 3 Number of vehicles threshold N 4 On-road traffic flow zone length threshold N 5 When the inflow accumulation amount of the vehicles is larger than a set threshold value, outputting a signal and a position where a large vehicle flow possibly exists; by means of continuously expanding K small areas, the vehicle density or the average vehicle number in the areas is calculated iteratively, the largest detection area with the vehicle density larger than a threshold value or the average vehicle number larger than the threshold value is output, if the length of the detection area is larger than the length threshold value of the large traffic flow area, signals and positions where large traffic flows possibly exist are output, the vehicle number of the large traffic flow in the middle and the other vehicle numbers in the middle are obtained, and the vehicle number of the large traffic flow in the middle is recorded as N onway The other vehicles in transit are counted as N onwayl The method comprises the steps of carrying out a first treatment on the surface of the Otherwise, giving up the tracking of the large traffic flow, and considering that the large traffic flow on the way does not exist at the intersection;
step 4-4: acquiring the running direction in the monitoring area and the directions by using the vehicle characteristic matrix of the OUT area of the intersectionThe number of vehicles to be driven in the lane is recorded as N await 。
4. The method for multi-directional dynamic control of road traffic based on traffic flow monitoring according to claim 1, further comprising step 8, wherein the method comprises the following steps of:
Step 8-1: judging whether a special vehicle exists at the intersection, dividing an intersection monitoring area into K areas, directly reading vehicle marking information or acquiring special vehicle information through a vehicle feature matrix, acquiring position information, vehicle speed information, driving route and departure time information of the special vehicle if a certain vehicle is marked (or identified by a feature) as the special vehicle, setting the total time of the current time and the departure time as T,
step 8-2: calculating green light passing conditions of N intersections in front, setting management and control (determined by driving routes) that N intersections in front participate in green light passing, acquiring the positions of special vehicles and the vehicle characteristic matrix of N intersection monitoring areas in front, calculating the time for the special vehicles to reach the ith intersection parking line in front according to the speed and position information of the special vehicles if the special vehicles do not have departure time information (namely are on the road), and setting the time as ST i I is more than or equal to 0 and less than or equal to N, the 0 th intersection is the intersection, and if a special vehicle has departure time information, the vehicle speed participates in calculation by using a set fixed vehicle speed; then, calculating the total number of the vehicles in front of the special vehicles in the monitoring area of the ith intersection, calculating the total traffic time of the vehicles in front of the monitoring area of the ith intersection by using the method for predicting the traffic time in the step 5, and setting the total traffic time as AT i The method comprises the steps of carrying out a first treatment on the surface of the Finally, setting the time reservation term as θ, if the special vehicle does not have departure time information, when ST i ≤AT i The green light release is immediately carried out at the ith intersection by +gamma, if the special vehicle has departure time information, the green light release needs to be postponed by T,
step 8-3: and (3) managing and controlling the traffic of a plurality of intersections, and when special vehicles exist in the intersections, managing and controlling the plurality of intersections by simultaneously calculating whether the release conditions of the front N intersections are met.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310374168.8A CN116580574A (en) | 2023-04-10 | 2023-04-10 | Road traffic multidirectional dynamic control method based on traffic flow monitoring |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310374168.8A CN116580574A (en) | 2023-04-10 | 2023-04-10 | Road traffic multidirectional dynamic control method based on traffic flow monitoring |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116580574A true CN116580574A (en) | 2023-08-11 |
Family
ID=87536695
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310374168.8A Pending CN116580574A (en) | 2023-04-10 | 2023-04-10 | Road traffic multidirectional dynamic control method based on traffic flow monitoring |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116580574A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117496740A (en) * | 2023-12-29 | 2024-02-02 | 山东高速股份有限公司 | Method, device, equipment and storage medium for managing and controlling traffic of vehicles on expressway |
-
2023
- 2023-04-10 CN CN202310374168.8A patent/CN116580574A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117496740A (en) * | 2023-12-29 | 2024-02-02 | 山东高速股份有限公司 | Method, device, equipment and storage medium for managing and controlling traffic of vehicles on expressway |
| CN117496740B (en) * | 2023-12-29 | 2024-03-19 | 山东高速股份有限公司 | Method, device, equipment and storage medium for managing and controlling traffic of vehicles on expressway |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110728842B (en) | Abnormal driving early warning method based on reasonable driving range of vehicles at intersection | |
| CN109754597B (en) | Urban road regional congestion regulation and control strategy recommendation system and method | |
| CN103235933B (en) | A kind of vehicle abnormality behavioral value method based on HMM | |
| CN108510762B (en) | Optimal control method for intelligent signal lamp in multi-line intersection area of expressway | |
| CN111597905B (en) | Highway tunnel parking detection method based on video technology | |
| CN101364347A (en) | Detection method for vehicle delay control on crossing based on video | |
| CN116580574A (en) | Road traffic multidirectional dynamic control method based on traffic flow monitoring | |
| CN113516854A (en) | Multi-interface coordination self-adaptive control method based on card police and video detector | |
| CN113380027A (en) | Intersection traffic state parameter estimation method and system based on multi-source data | |
| CN112818935B (en) | Multi-lane congestion detection and duration prediction method and system based on deep learning | |
| CN101064061A (en) | Different species traffic information real time integrating method | |
| CN115687709A (en) | Traffic dynamic control method based on traffic data dimension reduction reconstruction and multidimensional analysis | |
| CN111161545B (en) | Intersection region traffic parameter statistical method based on video | |
| CN110738852B (en) | Intersection steering overflow detection method based on vehicle track and long and short memory neural network | |
| CN111986483A (en) | Method and device for studying and judging road congestion state based on electric alarm data collision and storage medium | |
| CN112215073A (en) | Traffic marking line rapid identification and tracking method under high-speed motion scene | |
| CN117454318B (en) | Bridge group space-time load distribution identification method based on multi-source data fusion | |
| CN113140108B (en) | Cloud traffic situation prediction method in internet-connected intelligent traffic system | |
| CN117421700B (en) | Sensor data filtering and fusing method and device for automatic driving | |
| CN114694078A (en) | Traffic behavior judgment method based on multi-target tracking | |
| CN109859475A (en) | A kind of intersection signal control method based on DBSCAN Density Clustering, apparatus and system | |
| Wen-juan et al. | Application of vision sensing technology in urban intelligent traffic control system | |
| WO2021031173A1 (en) | Traffic state recognition method based on binocular camera | |
| CN113538902B (en) | Intersection vehicle track data restoration method based on traffic state | |
| Ren et al. | LaneDraw: Cascaded lane and its bifurcation detection with nested fusion |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |