A traffic signal control system based on vehicle flow detection

Abstract

This invention discloses a traffic signal control system based on vehicle flow detection, belonging to the field of intelligent transportation technology. The system includes modules for vehicle flow acquisition, dual-dimensional vehicle flow analysis, dynamic phase stitching, and signal control execution. The vehicle flow acquisition module acquires raw data for each lane; the dual-dimensional analysis module generates dynamic vehicle flow feature values ​​through spatial correlation analysis and temporal fluctuation period extraction; the dynamic phase stitching module decomposes and reassembles fixed phases accordingly, combines a conflict matrix and priority to generate a dynamic phase sequence, and determines the duration based on the vehicle flow ratio and fluctuation period; the signal control module executes the control. This invention solves the problems of insufficient detection accuracy and rigid timing in traditional systems, improves intersection traffic efficiency, and is both practical and economical.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of intelligent transportation, in particular to a traffic signal control system based on traffic flow detection. BACKGROUND

[0002] With the continuous development of urban traffic, traffic congestion problems are increasingly serious, and the performance of the traffic signal control system directly affects the efficiency of the intersection. At present, most of the traditional traffic signal control systems are based on fixed timing schemes, or only adjust the signal duration through simple traffic flow statistics, which has many shortcomings.

[0003] In terms of traffic flow detection, the traditional system usually only independently counts the traffic flow of a single lane, ignoring the correlation between adjacent lanes. For example, in the left turn lane and the straight lane of an intersection, when the left turn traffic flow suddenly increases, it may affect the traffic flow of the straight lane, but the traditional detection method cannot capture this correlation, resulting in low accuracy of traffic flow detection.

[0004] In terms of signal timing, the traditional system adopts fixed phase division, such as east-west straight, east-west left turn, north-south straight, and north-south left turn. However, in actual traffic, the traffic flow of each lane changes dynamically, and the fixed phase cannot adapt to such changes. For example, during a certain period, the east-west straight lane has very small traffic flow, while the east-west left turn lane has very large traffic flow, but because they are in the same fixed phase, the straight lane will occupy part of the green light time, causing waste of green light resources and affecting the efficiency of the intersection.

[0005] Therefore, there is an urgent need for a traffic signal control system that can accurately detect traffic flow and flexibly adjust signal timing. SUMMARY

[0006] The purpose of the present application is to provide a traffic signal control system based on traffic flow detection to solve the problems raised in the background.

[0007] To achieve the above purpose, the present application provides the following technical solution: a traffic signal control system based on traffic flow detection, comprising a traffic flow acquisition module, a two-dimensional traffic flow analysis module, a dynamic phase splicing module and a signal control execution module, the traffic flow acquisition module acquires original traffic flow data of each lane of the intersection through traffic detection equipment, wherein the traffic detection equipment includes but is not limited to a camera and a radar;

[0008] The two-dimensional traffic flow analysis module is used to process the original traffic flow data obtained by the traffic flow acquisition module to obtain dynamic traffic flow characteristic values of each lane, and the specific working steps are as follows:

[0009] Step 1, obtaining the original traffic volume data of each lane collected by the traffic volume collection module in a continuous time period, the data being expressed by the number of vehicles in a unit time;

[0010] Step 2, performing spatial dimension lane correlation analysis: calculating the synchronism coefficient of the traffic volume change of adjacent lanes in a unit time, the calculation formula of the synchronism coefficient being , wherein represents the synchronism coefficient of lane and lane , wherein represents the traffic volume of lane in the first unit time, represents the traffic volume of lane in the first unit time, represents the number of units of time for statistics; the closer the synchronism coefficient is to 1, the higher the synchronism of the traffic volume change of the two lanes, and the greater the correlation degree;

[0011] Step 3, performing time dimension traffic fluctuation period extraction: adopting the method of sliding window combined with Fourier transform, setting the length of the sliding window as 5 minutes and the sliding step as 1 minute; performing Fourier transform on the traffic volume data in each sliding window to obtain the distribution of the traffic volume in the frequency domain, and extracting the main fluctuation period, i.e. the time interval in which the traffic flow presents periodic change in the time period;

[0012] Step 4, comprehensively combining the spatial dimension lane correlation degree and the time dimension traffic fluctuation period to generate the dynamic traffic volume characteristic value of each lane; the dynamic traffic volume characteristic value is a comprehensive parameter including the traffic volume, the lane correlation coefficient and the fluctuation period, and is used to reflect the dynamic characteristics of the lane traffic volume;

[0013] The dynamic phase splicing module is used for splitting and recombining the traditional fixed phase according to the dynamic traffic volume characteristic value output by the double-dimension traffic volume analysis module, generating a dynamic phase sequence, and determining the length of each dynamic phase, and the specific working steps are as follows:

[0014] Step S1, receiving the dynamic traffic volume characteristic value of each lane output by the double-dimension traffic volume analysis module;

[0015] Step S2, setting the lane traffic volume difference threshold and the correlation degree threshold: when the difference of the traffic volume size in the dynamic traffic volume characteristic value of two lanes exceeds the traffic volume difference threshold, and the correlation degree coefficient of the two lanes is lower than the correlation degree threshold, it is determined that the traffic volume characteristic difference of the two lanes is large;

[0016] Step S3, splitting the traditional fixed phase: traversing each fixed phase, the lanes with large traffic flow feature difference are split from the original phase;

[0017] Step S4, recombining the split lanes: the split lanes are recombined with the lanes in other fixed phases with similar traffic flow features, i.e., the lanes with a traffic flow size difference less than a traffic flow difference threshold and a correlation coefficient higher than a correlation threshold, to form new dynamic phases;

[0018] Step S5, determining the duration of each dynamic phase: according to the dynamic traffic flow feature values of the lanes in the dynamic phase, the green light duration required by the dynamic phase is calculated by weighted average, and the weight is the proportion of the traffic flow size of each lane to the total traffic flow of the dynamic phase. Meanwhile, according to the traffic fluctuation period, the green light duration of the corresponding dynamic phase is increased in the peak period of the fluctuation period and appropriately reduced in the trough period.

[0019] Step S6, generating a dynamic phase sequence: determining the order of the dynamic phases according to the conflict relationship between the dynamic phases;

[0020] The signal control execution module is configured to receive the dynamic phase sequence and the duration of each dynamic phase generated by the dynamic phase splicing module, control the on-off state of the traffic signal light, and realize signal control of the intersection traffic.

[0021] Preferably, the specific implementation logic of generating the dynamic traffic flow feature values of each lane in step 4 is as follows:

[0022] Step 4.1, establishing a three-dimensional feature matrix: taking the lane as the row index, and taking the traffic flow size, the lane correlation coefficient, and the fluctuation period as the column index, an initial feature matrix is constructed , wherein represents the traffic flow size of lane , i.e., the number of vehicles per unit time , which is the average value of , represents the average synchronization coefficient of lane and the adjacent lane, i.e. , which is the average value of , is the adjacent lane of lane , and represents the main fluctuation period of lane ;

[0023] Step 4.2, feature standardization: the three-dimensional feature matrix is standardized to eliminate the dimension influence, and the standardization formula is: , wherein , represents all elements in the column of matrix , and the standardized feature value;

[0024] Step 4.3, dynamic weight distribution: the dynamic weight of each feature dimension is calculated by using the improved entropy method, first, the information entropy of the i-th feature is calculated , , wherein is the total number of lanes, is the minimum value to avoid zero value in logarithmic operation; then the weight is calculated: The smaller the information entropy, the higher the discrete degree of the feature, and the greater the weight.

[0025] Step 4.4, weighted fusion to generate feature value: the standardized feature value is weighted and summed with the corresponding weight to obtain the dynamic traffic flow feature value of each lane . .

[0026] Preferably, the specific implementation logic of determining the length of each dynamic phase in step S5 is as follows:

[0027] Step S5.1, calculate the basic green time: let the dynamic phase contain a lane set , the traffic flow of lane is , wherein , the total traffic flow of the dynamic phase is , the weight of lane is , and the calculation formula of the basic green time is: , wherein is the dynamic traffic flow feature value of lane , and is the basic conversion coefficient;

[0028] Step S5.2, determine the fluctuation period adjustment coefficient: let the average fluctuation period of each lane in the dynamic phase be , that is take the average value of the fluctuation periods of each lane , and the relative position of the current time in the fluctuation period is , wherein ; the adjustment coefficient is calculated by using the improved sine function, wherein is the fluctuation influence factor, wherein is set according to the fluctuation intensity of the intersection traffic flow; when is in the first half of the period, the value is positive,​ in the second half of the cycle, is negative,

[0029] Step S5.3, calculate the final green time: and limit it to the minimum green time and the maximum green time , i.e. .

[0030] Preferably, the generating of the dynamic phase sequence in step S6 is implemented as follows:

[0031] Step S6.1, construct a conflict matrix: judge all dynamic phases in pairs, if there is a cross conflict between the driving directions of the two dynamic phases (such as one phase contains east-west straight lane and the other phase contains south-north left turn lane, the driving paths of the two traffic flows cross), mark it as "conflict" in the conflict matrix, represented by the number 1; if there is no cross conflict between the driving directions of the two dynamic phases (such as one phase contains east-west straight lane and the other phase contains west-east straight lane, the driving paths of the two traffic flows are parallel and do not cross), mark it as "compatible", represented by the number 0; the conflict matrix is a square matrix, where n is the total number of dynamic phases, and the matrix element represents the conflict relationship between the dynamic phase and the dynamic phase , represents conflict, represents compatibility;

[0032] Step S6.2, calculate the priority of the dynamic phase: calculate the priority according to the final green time of the dynamic phase and the characteristic value of the dynamic traffic flow, and the priority calculation formula is , where is the final green time of the dynamic phase , is the lane set contained in the dynamic phase , is the number of lanes in the set, is a weight coefficient, and , used to balance the influence of the green time and the characteristic value of the traffic flow on the priority; the larger the priority value, the more urgent the traffic demand of the dynamic phase;

[0033] Step S6.3, generate an initial phase sequence: sort the dynamic phases according to the priority from high to low to form an initial dynamic phase sequence;

[0034] ​Step S6.4, adjusting the phase sequence to eliminate conflicts: traversing the initial phase sequence, for two adjacent dynamic phases, if the conflict matrix is marked as "conflict", the order is kept unchanged; if marked as "compatible", it is judged whether it can be combined into a combined phase, if the total signal period length can be reduced after combination, i.e. the combined phase length takes the maximum of the lengths of the two dynamic phases, then combination is performed, otherwise the order is kept unchanged; after the above adjustment, the final dynamic phase sequence without conflict and high efficiency is formed.

[0035] Preferably, the specific implementation steps of the step S6.4 of adjusting the phase sequence to eliminate conflicts are as follows:

[0036] Step S6.4.1, initializing the adjustment sequence: taking the initial phase sequence as the sequence to be adjusted, denoted as , wherein , and is the initial sorted dynamic phase;

[0037] Step S6.4.2, setting the traversal pointer: let the pointer start traversing from the first phase of the sequence;

[0038] Step S6.4.3, judging the relationship between adjacent phases: when , the current phase and the next phase are obtained, and the value of in the conflict matrix is queried; if , the pointer is incremented by 1, and the next set of adjacent phases is traversed; if , step S6.4.4 is executed;

[0039] Step S6.4.4, evaluating the feasibility of combination: calculating the total length of the first two phases before combination , and the length of the combined phase after combination ; if , it is determined that the combination is feasible; otherwise, it is determined that the combination is not feasible;

[0040] Step S6.4.5, performing combination or keeping the order: if the combination is feasible, combining and into a combined phase , whose length is , and replacing and with in the sequence to be adjusted, the sequence length is reduced by 1, the pointer remains unchanged, and the relationship between the new combined phase and the next phase is checked; if the combination is not feasible, the pointer is incremented by 1, and the order of the two phases is kept unchanged;

[0041] Step S6.4.6, Repeated Traversal Adjustment: When When the time is up, complete one round of traversal; if there is a phase merging operation in this round of traversal, then set the pointer back to 1 and execute steps S6.4.3 to S6.4.5 again on the updated sequence; if there is no merging operation in this round of traversal, then stop adjusting.

[0042] Step S6.4.7: Output the final sequence: After the above iterative adjustment, the sequence to be adjusted is the dynamic phase sequence with no conflict and the optimal total period duration.

[0043] Compared with existing technologies, the advantages of this invention are: improved traffic flow detection accuracy: the dual-dimensional traffic flow analysis module breaks through the traditional single-lane independent statistical mode, calculating the synchronicity coefficient of traffic flow changes in adjacent lanes through spatial dimension, capturing the traffic flow correlation between lanes; simultaneously, in the time dimension, a sliding window combined with Fourier transform is used to extract the traffic flow fluctuation cycle, comprehensively reflecting the dynamic change law of traffic flow. Through the construction of a three-dimensional feature matrix, standardization processing, and improved entropy method dynamic weight allocation, the generated dynamic traffic flow feature values ​​can accurately characterize the magnitude, correlation, and periodicity of traffic flow in each lane, providing reliable data support for subsequent signal timing.

[0044] Dynamic signal timing optimization is achieved through a dynamic phase splicing module that splits and reassembles traditional fixed phases based on dynamic traffic flow characteristics. By setting traffic flow difference thresholds and correlation thresholds, lanes with significant characteristic differences are split and recombined into dynamic phases, avoiding the waste of green light resources under fixed phases. When determining the phase duration, a weighted average of lane traffic flow proportions is used to calculate the base duration, and a fluctuation cycle adjustment coefficient is used to dynamically adapt the signal timing by increasing green light duration during peak hours and decreasing it during off-peak hours, making the signal timing more closely match actual traffic flow needs.

[0045] Improving intersection traffic efficiency: By constructing a conflict matrix to identify phase conflict relationships, and combining priority sorting and iterative merging adjustments to generate conflict-free dynamic phase sequences, the orderly flow of traffic is ensured. The merging optimization of compatible phases further shortens the total signal cycle time, reduces vehicle waiting time, effectively alleviates intersection traffic congestion, and improves overall traffic efficiency.

[0046] Balancing practicality and economy: The system's traffic flow acquisition module and signal control execution module adopt mature detection equipment such as cameras and radar, as well as conventional control logic, reducing development difficulty and cost; the core innovative module focuses on traffic flow analysis and dynamic phase adjustment, ensuring technological advancement while facilitating practical application and making it suitable for traffic signal control scenarios at various urban intersections. Attached Figure Description

[0047] Fig. 1The system structure schematic diagram of the present application;

[0048] Fig. 2 The two-dimensional traffic volume analysis module working process schematic diagram of the present application;

[0049] Fig. 3 The dynamic phase splicing module working process schematic diagram of the present application. DETAILED DESCRIPTION

[0050] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the scope of protection of the present application.

[0051] Please refer to Figs. 1-3 The present application provides a technical solution: a traffic signal control system based on traffic volume detection, comprising a traffic volume collection module, a two-dimensional traffic volume analysis module, a dynamic phase splicing module and a signal control execution module. The traffic volume collection module collects original traffic volume data of each lane at an intersection through a traffic detection device, wherein the traffic detection device includes but is not limited to a camera and a radar.

[0052] The specific traffic volume collection module is the data input core of the traffic signal control system. Through the traffic detection device such as a camera and a radar, the original traffic volume data of each lane at the intersection in a continuous time period is collected in real time, and the data is expressed by the number of vehicles in a unit time (such as 10 seconds). Its role is to provide basic data support for the subsequent two-dimensional traffic volume analysis module, so as to ensure that the system can accurately analyze and adjust the signal timing based on the real traffic flow. This module uses mature traffic detection technology to realize real-time and accurate perception of the traffic flow of each lane, which is the premise of intelligent control of the whole system;

[0053] The two-dimensional traffic volume analysis module is used for processing the original traffic volume data obtained by the traffic volume collection module to obtain the dynamic traffic volume characteristic value of each lane. The specific working steps are as follows:

[0054] Step 1: Obtain the original traffic volume data of each lane in a continuous time period collected by the traffic volume collection module, and the data is expressed by the number of vehicles in a unit time.

[0055] Step 2: Perform spatial dimension lane correlation analysis: calculate the synchronism coefficient of the traffic volume change of adjacent lanes in a unit time, and the calculation formula of the synchronism coefficient is , wherein represents the traffic volume of lane and lane synchronization coefficient, representing the lane traffic volume in the first unit time, representing the lane traffic volume in the first unit time, representing the statistical unit time quantity; the closer the synchronization coefficient is to 1, the higher the synchronization of the traffic volume changes of the two lanes, and the greater the correlation degree;

[0056] Step 3, time dimension traffic fluctuation cycle extraction: a sliding window combined with Fourier transform is used, the sliding window length is set to 5 minutes, and the sliding step is 1 minute; Fourier transform is performed on the traffic volume data in each sliding window to obtain the distribution of the traffic volume in the frequency domain, and the main fluctuation cycle is extracted, that is, the time interval at which the traffic presents periodic changes in the time period;

[0057] Step 4, comprehensive lane correlation degree in the spatial dimension and traffic fluctuation cycle in the time dimension to generate dynamic traffic characteristic values of each lane; the dynamic traffic characteristic value is a comprehensive parameter including traffic volume, lane correlation degree coefficient and fluctuation cycle, which is used to reflect the dynamic characteristics of the lane traffic; the specific implementation logic is as follows:

[0058] Step 4.1, establish a three-dimensional feature matrix: take the lane as the row index, and take the traffic volume, the lane correlation degree coefficient and the fluctuation cycle as the column index to construct the initial feature matrix , wherein representing the traffic volume of the lane , that is, the number of vehicles in the unit time is the average value of the traffic volume of the lane , the average synchronization coefficient of the lane and the adjacent lane, that is is the average value of the synchronization coefficient of the lane and the adjacent lane, is the adjacent lane of the lane , and represents the main fluctuation cycle of the lane

[0059] Step 4.2, feature standardization: the three-dimensional feature matrix is standardized to eliminate the dimension influence, and the standardization formula is: , wherein , represents all elements in the first column of the matrix , and is the standardized feature value;

[0060] Step 4.3, dynamic weight distribution: the dynamic weight of each feature dimension is calculated by using the improved entropy method, first, the information entropy of the first feature is calculated is the total number of lanes, is the minimum value to avoid zero value in logarithmic operation; then the weight is calculated: The smaller the information entropy, the higher the degree of dispersion of the feature, and the greater the weight;

[0061] Step 4.4, weighted fusion to generate feature value: the weighted sum of the standardized feature value and the corresponding weight is obtained to get the dynamic traffic flow feature value of each lane

[0062] The dynamic phase splicing module is used to split and recombine the traditional fixed phase according to the dynamic traffic flow feature value output by the double-dimensional traffic flow analysis module, generate a dynamic phase sequence, and determine the length of each dynamic phase. The specific working steps are as follows:

[0063] Step S1, receiving the dynamic traffic flow feature value of each lane output by the double-dimensional traffic flow analysis module;

[0064] Step S2, setting lane traffic flow difference threshold and correlation threshold: when the difference of dynamic traffic flow feature value between two lanes exceeds the traffic flow difference threshold, and the correlation coefficient of the two lanes is lower than the correlation threshold, it is determined that the traffic flow feature difference of the two lanes is large;

[0065] Step S3, splitting the traditional fixed phase: traversing the lanes in each fixed phase, the lanes with large traffic flow feature difference are split from the original phase;

[0066] Step S4, recombining the split lanes: the split lanes are recombined with the lanes in other fixed phases with similar traffic flow features, i.e. the difference of traffic flow size is less than the traffic flow difference threshold and the correlation coefficient is higher than the correlation threshold, to form a new dynamic phase;

[0067] Step S5, determining the length of each dynamic phase: according to the dynamic traffic flow feature value of each lane in the dynamic phase, the weighted average method is used to calculate the green time length required by the dynamic phase, and the weight is the proportion of the traffic flow size of each lane to the total traffic flow of the dynamic phase. At the same time, according to the traffic fluctuation period, the green time length of the corresponding dynamic phase is increased at the peak period of the fluctuation period and appropriately reduced at the trough period; the specific implementation logic is as follows:

[0068] Step S5.1, calculating the basic green time length: assuming that the dynamic phase contains a lane set​​​​ , the traffic volume of the lane , wherein , the total traffic volume of the dynamic phase , the weight of the lane , the basic green time , wherein is the dynamic traffic volume characteristic value of the lane is the basic conversion coefficient;

[0069] Step S5.2, determining the fluctuation period adjustment coefficient: let the average fluctuation period of each lane in the dynamic phase , that is , take the average value of the fluctuation periods of each lane , and the relative position of the current time in the fluctuation period is , wherein ; the adjustment coefficient is calculated using the improved sine function, wherein is the fluctuation influence factor, wherein is set according to the fluctuation intensity of the intersection traffic flow; when is in the first half of the period, the value is positive, ; when in the second half, the value is negative, ;

[0070] Step S5.3, calculating the final green time: , and limiting it to the minimum green time and the maximum green time , that is .

[0071] Step S6, generating a dynamic phase sequence: according to the conflict relationship between each dynamic phase, determine the order of the dynamic phase; the specific implementation steps are as follows:

[0072] ​​​​​​Step S6.1, constructing a conflict matrix: judge all dynamic phases in pairs, if there is intersection conflict between the lane driving directions corresponding to two dynamic phases (for example, one phase contains east-west straight lane, and the other phase contains south-north left turn lane, the driving paths of the two vehicle flows intersect), mark it as “conflict” in the conflict matrix, and use the number 1 to represent; if there is no intersection conflict between the lane driving directions corresponding to two dynamic phases (for example, one phase contains east-west straight lane, and the other phase contains west-east straight lane, the driving paths of the two vehicle flows are parallel and do not intersect), mark it as “compatible”, and use the number 0 to represent; the conflict matrix is a square matrix of , , where n is the total number of dynamic phases, and the matrix element represents the conflict relationship between the dynamic phase and the dynamic phase , represents conflict, represents compatibility;

[0073] Step S6.2, calculating the priority of the dynamic phase: the priority is calculated comprehensively according to the final green time of the dynamic phase and the characteristic value of the dynamic vehicle flow, and the priority calculation formula is , where is the final green time of the dynamic phase , is the lane set contained by the dynamic phase , is the number of lanes in the set, is a weight coefficient, and , used to balance the influence of the green time and the characteristic value of the vehicle flow on the priority; the larger the priority value is, the more urgent the traffic demand of the dynamic phase is;

[0074] Step S6.3, generating an initial phase sequence: sort the dynamic phases in order of priority from high to low to form an initial dynamic phase sequence;

[0075] Step S6.4, adjusting the phase sequence to eliminate conflict: traverse the initial phase sequence, and for two adjacent dynamic phases, if the conflict matrix is marked as “conflict”, keep the order unchanged; if it is marked as “compatible”, judge whether it can be combined into a combined phase, if the total signal cycle time can be reduced after combination, that is, the combined phase time is the maximum of the time of the two dynamic phases, then combine, otherwise keep the order unchanged; after the above adjustment, a conflict-free and efficient dynamic phase sequence is finally formed; the specific implementation steps are as follows:

[0076] Step S6.4.1, initializing the adjustment sequence: taking the initial phase sequence as the sequence to be adjusted, denoted as , where to The dynamic phase after initial sorting;

[0077] Step S6.4.2, Set the traversal pointer: Set the pointer Iterate from the first phase of the sequence;

[0078] Step S6.4.3: Determine the relationship between adjacent phases: When At that time, obtain the current phase. and the next phase Query the conflict matrix The value; if Then the pointer Increment by 1 and continue iterating through the next group of adjacent phases; if Then proceed to step S6.4.4;

[0079] Step S6.4.4: Assess the feasibility of merging: Calculate the total duration of the two phases before merging. Duration of the combined phase after merging ;like If the conditions are met, the merger is deemed feasible; otherwise, the merger is deemed infeasible.

[0080] Step S6.4.5: Perform the merge or maintain the order: If the merge is feasible, and Merge into combined phases Its duration is and in the sequence to be adjusted replace and Decrease the sequence length by 1, pointer Keep the current position unchanged and continue checking the relationship between the new combined phase and the next phase; if merging is not feasible, the pointer... Increment by 1, maintaining the phase order unchanged;

[0081] Step S6.4.6, Repeated Traversal Adjustment: When When the time is up, complete one round of traversal; if there is a phase merging operation in this round of traversal, then set the pointer back to 1 and execute steps S6.4.3 to S6.4.5 again on the updated sequence; if there is no merging operation in this round of traversal, then stop adjusting.

[0082] Step S6.4.7: Output the final sequence: After the above iterative adjustment, the sequence to be adjusted is the dynamic phase sequence with no conflict and the optimal total period duration.

[0083] The signal control execution module is used to receive the dynamic phase sequence and duration of each dynamic phase generated by the dynamic phase splicing module, control the on / off state of the traffic lights, and realize signal control of traffic at the intersection.

[0084] The following will be described in conjunction with specific embodiments:

[0085] Implementation scenario: Take a typical crossroad as an example, which contains east, west, south and north directions, each of which is provided with a straight lane, a left-turn lane and a right-turn lane, a total of 12 lanes. The vehicle flow collection module adopts the combination of high-definition camera and radar to collect the number of vehicles in each lane within a unit time (10 seconds).

[0086] Implementation process of each module: Vehicle flow collection module: In a continuous time period, such as the morning peak of 7:30-8:30, the original vehicle flow data of 12 lanes is collected. For example, the vehicle flow data of the east straight lane in multiple 10-second unit times is 15, 18, 20, 17, etc.; the corresponding data of the east left-turn lane is 8, 10, 12, 9, etc.

[0087] Two-dimensional vehicle flow analysis module

[0088] Step 1: Receive the original vehicle flow data of 12 lanes collected by the vehicle flow collection module in the above continuous time period.

[0089] Step 2: Perform spatial dimension lane correlation analysis. Take the east straight lane (lane ) and the east left-turn lane (lane ) as an example, calculate the synchronization coefficient. Assuming that the number of units of time is , the formula is used to calculate, and the synchronization coefficient of the two lanes is , which indicates that the synchronization of the vehicle flow change of the two lanes is low, and the correlation is small. Similarly, the synchronization coefficients between other adjacent lanes are calculated;

[0090] Step 3: Perform time dimension vehicle flow fluctuation period extraction. Set the sliding window length to 5 minutes and the sliding step to 1 minute. Fourier transform is performed on the vehicle flow data in each sliding window, for example, in the window of 7:30-7:35, the vehicle flow of the east straight lane presents obvious periodic distribution in the frequency domain, and the main fluctuation period is extracted as 3 minutes, that is, the vehicle flow of the lane presents periodic change every 3 minutes.

[0091] Step 4: Generate dynamic vehicle flow characteristic values of each lane;

[0092] Step 4.1: Establish a three-dimensional feature matrix M. For the east straight lane (lane ), is the average value of the vehicle flow in a unit time, which is calculated as 18; For the average synchronization coefficient with adjacent lanes (eastbound left-turn lane and eastbound right-turn lane), assuming that the synchronization coefficient of the eastbound straight lane and the eastbound right-turn lane is 0.6, then ; The main fluctuation period is 3 minutes. Similarly, the three-dimensional feature matrix of the other 11 lanes is constructed;

[0093] Step 4.2: Feature standardization. Taking the column of traffic volume size as an example, assuming that the minimum value of the first column in the matrix is 5 and the maximum value is 25, then for the eastbound straight lane , the standardized value is . Similarly, the standardization processing of the lane correlation coefficient and the fluctuation period is completed;

[0094] Step 4.3: Dynamic weight distribution. Assuming that the total number of lanes , the information entropy of the first feature (traffic volume size) is calculated , and the weight is calculated by the formula , and the weight is ; similarly, the weight , is calculated. Then the weight ; ; ;

[0095] Step 4.4: Weighted fusion to generate feature values. The dynamic traffic volume feature value of the eastbound straight lane , assuming , , then ; similarly, the dynamic traffic volume feature values of the other 11 lanes are calculated;

[0096] Dynamic phase splicing module:

[0097] Step S1: Receive the dynamic traffic volume feature values of 12 lanes output by the two-dimensional traffic volume analysis module.

[0098] Step S2: Set the lane traffic volume difference threshold to 0.2 and the correlation threshold to 0.5.

[0099] Step S3: Split the traditional fixed phase. The traditional fixed phase such as the eastbound straight and left-turn phase, since the difference between the traffic volume size of the eastbound straight lane and the left-turn lane is calculated to be more than 0.2, and the correlation coefficient 0.3 is lower than 0.5, it is determined that the traffic volume feature difference of the two lanes is large, and the eastbound left-turn lane is split from the fixed phase. Similarly, other fixed phases are split.

[0100] ​Step S4: recombining the split lanes. Comparing the east left-turn lane with the south left-turn lane, the difference in traffic volume is less than 0.2, and the correlation coefficient is higher than 0.5, so they are recombined to form a new dynamic phase . Similarly, the recombination of other split lanes is completed to form dynamic phases 、 , etc.

[0101] Step S5: determining the duration of each dynamic phase:

[0102] Step S5.1: calculating the basic green time. Taking the dynamic phase (including the east left-turn lane and the south left-turn lane) as an example, the traffic volume of lane (the east left-turn lane) is , the traffic volume of lane (the south left-turn lane) is , and the total traffic volume of the dynamic phase is . The weight of lane is , and the weight of lane is . Given that , , and the basic conversion coefficient is seconds, the basic green time is seconds;

[0103] Step S5.2: determining the fluctuation period adjustment coefficient. The average fluctuation period of each lane in the dynamic phase is , the fluctuation period of the east left-turn lane is 4 minutes, and the fluctuation period of the south left-turn lane is 5 minutes, so the fluctuation period adjustment coefficient is minutes. The relative position of the current time within the fluctuation period is minutes (in the first half of the period), the fluctuation influence factor is . The adjustment coefficient is .

[0104] Step S5.3: calculating the final green time. seconds, the minimum green time is set to seconds, the maximum green time is seconds, and the final seconds. Similarly, the final green times of other dynamic phases are calculated.

[0105] Step S6: generating a dynamic phase sequence:

[0106] Step S6.1: constructing a conflict matrix. For the dynamic phases 、 , Wait for pairwise combination judgments, such as (Turn left when heading east or south) and There is an intersection and conflict between (eastbound and westbound straight traffic) in the conflict matrix. ; and (Right turns from north and west) No intersection or conflict. .

[0107] Step S6.2: Calculate the dynamic phase priority. For example, its final green light duration Seconds, assuming the total green light duration for all dynamic phases is 100 seconds. The sum of the dynamic traffic flow feature values ​​in the included lane set is The number of lanes is 2. Weighting coefficient. Priority Similarly, the priority of other dynamic phases can be calculated.

[0108] Step S6.3: Generate the initial phase sequence. Sort the phase sequences from highest to lowest priority to obtain the initial phase sequence. ;

[0109] Step S6.4: Adjust the phase sequence to eliminate collisions. Traverse the initial sequence, and The conflict matrix is ​​marked as 1, and the order remains unchanged; and The conflict matrix is ​​marked as 0 to assess the feasibility of the merger. Total duration before merger. Combined phase duration ,like Then merge and If a combination of phases is used, the sequence is maintained; otherwise, it is left as is. After multiple rounds of iterative adjustments, a conflict-free and efficient dynamic phase sequence is finally formed.

[0110] The signal control execution module receives the dynamic phase sequence and duration of each phase generated by the dynamic phase splicing module, and controls the traffic lights to turn on and off according to this sequence and duration, achieving precise signal control of traffic at the intersection. In the aforementioned morning rush hour scenario, through the control of this system, traffic flow in each lane at the intersection can proceed more orderly, effectively alleviating traffic congestion.

[0111] The application discloses a traffic signal control system based on vehicle flow detection and relates to the technical field of intelligent traffic. The system comprises a vehicle flow acquisition module, a double-dimension vehicle flow analysis module, a dynamic phase splicing module and a signal control execution module. The vehicle flow acquisition module collects original vehicle flow data of each lane at an intersection through cameras, radars and other equipment; the double-dimension vehicle flow analysis module generates dynamic vehicle flow characteristic values containing vehicle flow size, correlation coefficient and fluctuation period through space-dimension lane correlation degree analysis (synchronization coefficient calculation) and time-dimension vehicle flow fluctuation period extraction (sliding window combined with Fourier transform); the dynamic phase splicing module splits and reorganizes the traditional fixed phase according to the characteristic values, generates a dynamic phase sequence through conflict matrix identification, priority order sorting and iterative merging adjustment, and dynamically determines the phase length in combination with the vehicle flow proportion and fluctuation period; and the signal control execution module controls the on-off of the signal lamp according to the dynamic phase sequence and length.

[0112] While the embodiments of the application have been illustrated and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made therein without departing from the spirit and scope of the application, which is defined by the appended claims and their equivalents.

Claims

1. A traffic signal control system based on vehicle flow detection, characterized by, The traffic flow acquisition module, the double-dimension traffic flow analysis module, the dynamic phase splicing module and the signal control execution module are included, the traffic flow acquisition module acquires the original traffic flow data of each lane at the intersection through traffic detection equipment, wherein the traffic detection equipment includes but is not limited to a camera and a radar; The double-dimension traffic flow analysis module is used for processing the original traffic flow data acquired by the traffic flow acquisition module to obtain the dynamic traffic flow characteristic value of each lane, and the specific working steps are as follows: Step 1, acquire the original traffic flow data of each lane in the continuous time period collected by the traffic flow acquisition module, and the data is expressed by the number of vehicles in a unit time; Step 2, lane correlation degree analysis in spatial dimension: calculate the synchronism coefficient of the traffic flow change of adjacent lanes in unit time, and the calculation formula of the synchronism coefficient is Wherein represents the synchronism coefficient of lane and lane , represents the traffic flow of lane in the first unit time, represents the traffic flow of lane in the first unit time, represents the number of unit time for statistics; The closer the synchronism coefficient is to 1, the higher the synchronism of the traffic flow changes of the two lanes is, and the greater the correlation degree is; Step 3, time-dimension traffic flow fluctuation period extraction: a sliding window combined with Fourier transform is adopted, the length of the sliding window is set to 5 minutes, and the sliding step is 1 minute; Fourier transform is performed on the traffic flow data in each sliding window to obtain the distribution of the traffic flow in the frequency domain, and the main fluctuation period is extracted, that is, the time interval in which the traffic flow presents periodic change in the time period; Step 4, the lane correlation degree in the space dimension and the traffic flow fluctuation period in the time dimension are integrated to generate the dynamic traffic flow characteristic value of each lane; the dynamic traffic flow characteristic value is a comprehensive parameter including the traffic flow size, the lane correlation degree coefficient and the fluctuation period, and is used to reflect the dynamic characteristics of the lane traffic flow; The dynamic phase splicing module is used for splitting and recombining the traditional fixed phase according to the dynamic traffic flow characteristic value output by the double-dimension traffic flow analysis module to generate a dynamic phase sequence and determine the length of each dynamic phase, and the specific working steps are as follows: Step S1, receive the dynamic traffic flow characteristic value of each lane output by the double-dimension traffic flow analysis module; Step S2, set the lane traffic flow difference threshold and the correlation degree threshold: when the difference between the traffic flow sizes in the dynamic traffic flow characteristic values of two lanes exceeds the traffic flow difference threshold, and the correlation degree coefficient of the two lanes is lower than the correlation degree threshold, it is determined that the traffic flow characteristic difference of the two lanes is large; Step S3, split the traditional fixed phase: traverse the lanes in each fixed phase, and split the lanes with large traffic flow characteristic difference from the original phase; Step S4, recombine the split lanes: the split lanes are recombined with the lanes in other fixed phases with similar traffic flow characteristics, that is, the lanes with a traffic flow size difference less than the traffic flow difference threshold and a correlation degree coefficient higher than the correlation degree threshold, to form a new dynamic phase; Step S5, determine the length of each dynamic phase: according to the dynamic traffic flow characteristic value of each lane in the dynamic phase, the green light duration required by the dynamic phase is calculated by weighted average, and the weight is the proportion of the traffic flow size of each lane to the total traffic flow of the dynamic phase; at the same time, according to the traffic flow fluctuation period, the green light duration of the corresponding dynamic phase is increased at the peak period of the fluctuation period, and is appropriately reduced at the trough period; Step S6, generate a dynamic phase sequence: determine the sequence of the dynamic phases according to the conflict relationship between the dynamic phases; The signal control execution module is configured to receive the dynamic phase sequence and the time length of each dynamic phase generated by the dynamic phase splicing module, control the on-off state of the traffic signal lamp, and realize signal control of the intersection traffic.

2. The traffic signal control system based on vehicle flow detection as claimed in claim 1 wherein: The specific implementation logic of generating the dynamic traffic feature value of each lane in step 4 is as follows: Step 4.1, establish a three-dimensional feature matrix: take the lane as the row index, take the traffic volume, the lane correlation coefficient, and the fluctuation period as the column index, and construct an initial feature matrix wherein represents the traffic volume of the lane , that is, the number of vehicles per unit time represents the average value of the traffic volume of the lane , represents the average synchronization coefficient of the lane with the adjacent lane, that is represents the average value of the synchronization coefficient of the lane with the adjacent lane , represents the adjacent lane of the lane , represents the main fluctuation period of the lane ; Step 4.2, feature standardization: the three-dimensional feature matrix is standardized to eliminate the dimension effect, and the standardization formula is: wherein , represents all elements in the i-th column of the matrix , , and is the standardized feature value; Step 4.3, dynamic weight distribution: the dynamic weight of each feature dimension is calculated by using the improved entropy method, first calculate the information entropy of the first feature , , wherein is the total number of lanes, is the minimum value to avoid zero value in logarithmic operation; then calculate the weight : , the smaller the information entropy, the higher the discrete degree of the feature, and the greater the weight;​ Step 4.4, weighted fusion to generate feature value: the normalized feature value is weighted and summed with the corresponding weight to obtain the dynamic traffic flow feature value of each lane : .

3. The traffic signal control system based on vehicle flow detection as claimed in claim 1 wherein: The specific implementation logic of determining the time length of each dynamic phase in step S5 is as follows: Step S5.1, calculating the basic green light duration: assuming that the dynamic phase contains a lane set , the traffic volume of lane is , wherein , the total traffic volume of the dynamic phase is , the weight of lane is , and the calculation formula of the basic green light duration is: , wherein is the dynamic traffic volume characteristic value of lane , and is a basic conversion coefficient; Step S5.2: Determine the fluctuation period adjustment coefficient: Assume the dynamic phase The average fluctuation period of each lane is ,Right now Take the fluctuation period of each lane The average value, and the relative position within the fluctuation cycle at the current moment is ,in The adjustment coefficient is calculated using an improved sine function. ,in The factors affecting volatility include, among which The setting is based on the intensity of traffic flow fluctuations at the intersection; when When in the first half of the cycle, The value is positive. In the latter half, The value is negative. ; Step S5.3, calculate final green time: and limit to minimum green time and maximum green time i.e. .

4. The traffic signal control system based on vehicle flow detection as claimed in claim 1 wherein: The specific implementation steps of generating the dynamic phase sequence in step S6 are as follows: Step S6.1, constructing a conflict matrix: all dynamic phases are combined in pairs to determine, if there is a crossing conflict between the lane driving directions corresponding to two dynamic phases, mark "conflict" in the conflict matrix, represented by the number 1; if there is no crossing conflict between the lane driving directions corresponding to two dynamic phases, mark "compatible", represented by the number 0; the conflict matrix is a square matrix of , , where n is the total number of dynamic phases, and the matrix element represents the conflict relationship between the dynamic phase and the dynamic phase , represents conflict, represents compatibility; Step S6.2, calculating the dynamic phase priority: the priority is calculated according to the final green time of the dynamic phase and the dynamic traffic volume characteristic value, and the priority calculation formula is wherein is the final green time of the dynamic phase , is the lane set contained in the dynamic phase , is the number of lanes in the set, is the weight coefficient, and , used to balance the influence of the green time and the traffic volume characteristic value on the priority; the greater the priority value, the more urgent the traffic demand of the dynamic phase; is the dynamic traffic volume characteristic value of each lane; Step S6.3, generating an initial phase sequence: sorting the dynamic phases in order from high to low priority to form an initial dynamic phase sequence; Step S6.4, adjusting the phase sequence to eliminate conflicts: traversing the initial phase sequence, for two adjacent dynamic phases, if the conflict matrix is marked as "conflict", the order is kept unchanged; if it is marked as "compatible", it is judged whether it can be combined into a combined phase, if the total signal cycle time length can be reduced after combination, that is, the combined phase time length takes the maximum value of the time lengths of the two dynamic phases, then the combination is performed, otherwise the order is kept unchanged; after the above adjustment, the final dynamic phase sequence without conflict and high efficiency is formed.

5. The traffic signal control system based on vehicle flow detection as claimed in claim 4 wherein: The specific implementation steps of step S6.4 of adjusting the phase sequence to eliminate conflicts are as follows: Step S6.4.1, initialization of adjustment sequence: take the initial phase sequence as the sequence to be adjusted, denoted as wherein to is the initial sorted dynamic phase; Step S6.4.2, Set traversal pointer: let pointer traverse from the first phase of the sequence; Step S6.4.3, judging adjacent phase relationship: when , the current phase and the next phase are obtained, the value in the conflict matrix at is inquired; if , the pointer is incremented by 1, and the next set of adjacent phases is continued to be traversed; if , step S6.4.4 is executed; Step S6.4.4, evaluate merging feasibility: calculate total duration of the two phases before merging , duration of the combined phase after merging ; if , then decide that merging is feasible; Otherwise, it is determined that the combination is not feasible; Step S6.4.5: Perform the merge or maintain the order: If the merge is feasible, and Merge into combined phases Its duration is and in the sequence to be adjusted replace and Decrease the sequence length by 1, pointer Keep the current position unchanged and continue checking the relationship between the new combined phase and the next phase; if merging is not feasible, the pointer... Increment by 1, maintaining the phase order unchanged; Step S6.4.6, repeat the adjustment of the traversal: when a round of traversal is completed; if there is a phase-merging operation in the current round of traversal, the pointer is reset to 1, and steps S6.4.3 to S6.4.5 are executed again on the updated sequence; if there is no merging operation in the current round of traversal, the adjustment is stopped; Step S6.4.7, output the final sequence: after the above iterative adjustment, the adjusted sequence is the dynamic phase sequence without conflict and with optimal total cycle time length.

Images