Lane queuing overflow condition detection method

Abstract

According to the lane queuing overflow condition detection method provided by the invention, the signal lamp real-time timing data of the to-be-monitored intersection and the vehicle queuing data collected by the video detector installed at the to-be-monitored intersection are fused, and the lane queuing overflow condition is detected based on the road vehicle entering data collected by the video detector in the non-overflow period. The lane queuing length of the road to be detected when overflowing queuing occurs is estimated, the queuing length identification limitation of video detection is broken through, the basic environment of each lane of the existing intersection is not changed, new hardware is not needed, and no equipment or construction cost is needed.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent traffic control, and in particular to a method for detecting lane queue overflow. Background Art

[0002] In recent years, with the increase in the number of motor vehicles, urban roads have become more and more congested, and queue overflows in the entrance lanes of urban intersections have occurred frequently. In traffic control methods, in many cases, such as automatic adjustment of intersection signal light cycles or intelligent adjustment of tidal roads, it is necessary to first identify the specific situation of vehicle queue overflow on the lane.

[0003] In the prior art, the identification of the length of vehicle queues on lanes is mostly done by collecting data through monitoring equipment. For example, the patent with application number CN202110916264.1 sets a front-end traffic flow collector and a rear-end traffic flow collector, which are installed in the middle of each entrance lane of the intersection to be detected. The front-end traffic flow collector is aimed at the front direction of the vehicle on the entrance lane, and the rear-end traffic flow collector is aimed at the rear direction of the vehicle on the entrance lane to collect the specific queue position of the vehicle.

[0004] Because a large number of video surveillance devices are installed on urban roads, traffic flow collectors usually reuse existing video surveillance devices. However, in actual use, it is found that the recognition range of video surveillance devices installed at many intersections cannot meet the required range for calculation. For example, when the queue at the intersection overflows, the vehicle queue in the lane is usually 100-200 meters, and the video surveillance equipment usually has the ability to identify the queue within 100 meters due to pixels, angles, focal lengths, etc. Therefore, in order to meet the vehicle queue data collection at the intersection, additional auxiliary detectors are required, such as: multi-target microwave radar detectors, magnetic coils, etc., which leads to an increase in overall costs. In some cities, due to cost issues, it is impossible to ensure that all intersections can install auxiliary detectors in a timely manner, which in turn affects the application and popularity of intelligent traffic control technology at urban intersections. Summary of the invention

[0005] In order to solve the problem that the monitoring range of some video surveillance equipment cannot meet the needs of lane overflow detection at intersections, the present invention provides a lane queue overflow detection method, which can utilize existing video surveillance equipment and estimate the lane queue overflow without adding new auxiliary detection equipment.

[0006] The technical solution of the present invention is as follows: a lane queue overflow detection method, characterized in that it comprises the following steps:

[0007] S1: Obtain the monitoring range of the video detector set in the to-be-monitored import direction of the to-be-monitored intersection, and record the maximum length covering the lane in the monitoring range as Lmax;

[0008] S2: Select the to-be-monitored lane, and calculate the number of vehicles merging into the to-be-monitored lane per unit time VUnit in the i-th cycle i ;

[0009] Obtain the length of the to-be-monitored lane, denoted as: Llane; Llane≥Lmax;

[0010] S3: Calculate the space length QV occupied by the average single vehicle in the queue on the to-be-monitored lane i ;

[0011] S4: Compare the queue length Q detected by the video detector at the end moment of the green light of the lane in the i-th cycle on the to-be-monitored lane i with the maximum length Lmax of the lane covered by the monitoring range;

[0012] If Q i <Lmax, then the value of the predicted queue length QR at the current moment of the to-be-monitored lane i is Q i ;

[0013] Otherwise, when Q i =Lmax, execute step S5 to estimate QR i ;

[0014] S5: The estimated value QR i of the predicted queue length QR at the current moment of the to-be-monitored lane i is calculated as follows:

[0015] QR i ' = QR i-1 +(VUnit i *C i -VO i )*QV i ;

[0016] Among them, QR i-1 is the predicted queue length in the (i - 1)-th cycle of the to-be-monitored lane;

[0017] VUnit i is the number of vehicles merging into the to-be-monitored lane per unit time in the i-th cycle;

[0018] C i is the signal cycle duration of the i-th signal cycle of the to-be-monitored intersection;

[0019] QV iCalculate the average length of space occupied by a single vehicle queue in the lane for the i-th cycle;

[0020] VO i The number of vehicles passing through the lane at the end of the green light in the i-th cycle is the number of vehicles leaving the detection area;

[0021] S6: Compare QR i ' and the value of Lmax;

[0022] If QR i '>Lmax, then the queue length QR of the lane to be monitored is predicted at the current time point i The value of is QR i ';

[0023] Otherwise, the queue length QR of the lane to be monitored is predicted at the current time point i The value of is Lmax;

[0024] S7: Compare QR i ' and lane to be monitored Llane;

[0025] If QR i '>Llan, it is considered that there is an overflow risk in the monitored lane and a warning message is given;

[0026] Otherwise, steps S2 to S7 are executed in a loop.

[0027] It is further characterized by:

[0028] In step S2, the number of vehicles merging into the lane to be monitored per unit time in the i-th cycle is VUnit i The calculation of includes the following steps:

[0029] a1: Starting from the 0th cycle to the i-1th cycle, all cycles with queue length less than Lmax are counted and recorded as: calculation cycle;

[0030] a2: Counting the total number of vehicles imported in all the calculation cycles detected by the video detector, recorded as: the total number of vehicles imported for calculation;

[0031] a3: VUnit is calculated by the total number of vehicles imported into the calculation and the total duration of the calculation cycle. i ;

[0032] In step S2, the number of vehicles merging into the lane to be monitored per unit time in the i-th cycle is VUnit i The calculation formula is:

[0033]

[0034] Among them, Q n At the end of the green light time of the nth cycle lane, the video detector detects the length of the lane queue;

[0035] VI n The number of vehicles entering the detection area at the end of the green light of the nth cycle lane;

[0036] C n is the duration of the nth signal cycle at the intersection;

[0037] sgn(x) is the sign function, and its specific value is:

[0038] When x>0, sgn(x)=1; when x=0, sgn(x)=0; when x<0, sgn(x)=-1;

[0039] In step S3, the average space length QV occupied by a single vehicle queue in the lane to be monitored i The calculation method is:

[0040] QV i =Q i / VS i ;

[0041] Among them, QV i Q is the average length of space occupied by a single vehicle queue in the i-th signal cycle monitoring lane, i is the lane queue length detected by the video detector at the end time of the green light of the i-th cycle lane, Q i >0;VS i is the number of vehicles stranded in the lane of the detection area at the end of the green light in the i-th cycle, VS i >0;

[0042] In step S6, the queue length QR of the lane to be monitored is predicted and calculated at the current time point i The calculation formula is:

[0043] QR i =max(Lamx,QR i-1 +(VUnit i *C i -VO i )*QV i );

[0044] In step a1, the method for determining the 0th cycle is:

[0045] b1: The video detector counts the queue length of each lane of each entrance road corresponding to the intersection to be monitored in real time;

[0046] b2: Once it is detected that the total road queue length on the lane to be monitored is equal to Lmax, the corresponding cycle is denoted as: the starting cycle;

[0047] b3: Starting from the starting cycle, push forward m cycles according to the preset minimum number of calculation cycles m, and set it as the 0th cycle;

[0048] Based on the 0th cycle, estimate the queue lengths of all subsequent cycles where queue overflow is likely to occur;

[0049] In step b2, the value of m is obtained as follows: among all complete signal cycles included in the 10 minutes before the start time of the starting cycle, find the number of signal cycles where the total road queue length < Lmax, and set it as m.

[0050] A method for detecting lane queue overflow provided by this application fuses the real-time signal timing data of the signal lights at the intersection to be monitored with the vehicle queue data collected by the video detector installed at the intersection to be monitored. Based on the road vehicle entry data collected by the video detector during non-overflow cycles, it estimates the lane queue length when the road to be detected has an overflow queue, breaking through the queue length recognition limit of video detection, without changing the basic environment of each lane at the existing intersection, without the need to invest in new hardware, and without any equipment and construction costs. Description of the Drawings

[0051] Figure 1 It is a structural schematic diagram of the road to be monitored;

[0052] Figure 2 It is a flow chart of the method for detecting lane queue overflow in this application. Detailed Embodiments

[0053] As Figure 2 shown, the present invention includes a method for detecting lane queue overflow, which includes the following steps.

[0054] In an actual intersection, usually each video detector will monitor multiple lanes in the same direction simultaneously. In this application, each lane is respectively set as the lane to be monitored, and the method for detecting lane queue overflow in this application performs detection for each lane to be monitored respectively.

[0055] S1: As Figure 1 shown, confirm the visible area of the video detector set at the to-be-monitored import direction of the to-be-monitored intersection, that is, obtain the monitoring range of the video detector, and record the maximum length of the lane covered by the monitoring range as Lmax. The signal light cycle under the fixed cycle release scheme at the intersection is denoted as C, and the green light passing time of the lane within a single cycle is denoted as G. The lane queue length detected by the video detector at the end moment of the lane green light is denoted as Q.

[0056] S2: Select the lane to be monitored, and calculate the number of vehicles merging into the lane to be monitored per unit time, VUnit, in the i-th cycle. i Specifically, it includes the following steps:

[0057] a1: Starting from the 0-th cycle and ending at the (i - 1)-th cycle, count all the cycles with a queue length less than Lmax, and record them as: calculation cycles.

[0058] In step a1, the determination method of the 0-th cycle is as follows:

[0059] b1: The video detector continuously counts the queue lengths of each lane of each approach corresponding to the intersection to be monitored.

[0060] b2: Once it is detected that the total road queue length on the lane to be monitored is equal to Lmax, record the corresponding cycle as: start cycle; when the video detector detects that the total road queue length on the lane to be monitored is equal to Lmax, it is considered that there is a possibility of queue overflow in the direction of the approach to be monitored at the intersection.

[0061] b3: Starting from the start cycle, push forward m cycles according to the preset minimum number of calculation cycles m, and set it as the 0-th cycle; based on the 0-th cycle, estimate the queue lengths of all subsequent cycles with the possibility of queue overflow.

[0062] Among them, the value of m is obtained as follows: among all the complete signal cycles included in the 10 minutes before the start time of the start cycle, find the number of signal cycles with a queue length of the total road queue length < Lmax, which is m.

[0063] In this embodiment, assume that the total road queue length collected by the video detector at 7:20 is Lmax, then set the corresponding cycle as the start cycle; look forward 10 minutes from the start cycle, which includes 4 complete signal cycles, and the total queue lengths are all less than Lmax. Then record these four signal cycles in chronological order as: the 0-th to 3-th cycles; the lane queue lengths of these 4 signal cycles are all less than Lmax, and the specific values can be accurately collected based on the video collector; for the start cycle, i = 4, and estimate the queue length of the 4-th cycle according to the subsequent steps.

[0064] a2: Count the total number of vehicles merging in all the calculation cycles detected by the video detector, and record it as: total number of merging vehicles for calculation.

[0065] a3: Calculate VUnit by dividing the total number of merging vehicles for calculation by the total time length of the calculation cycles. i .

[0066] That is, the number of vehicles merging into the lane to be monitored per unit time in the i-th cycle is VUnit i =Total number of vehicles entering the calculation / total length of the calculation cycle. The corresponding calculation formula is:

[0067]

[0068] Among them, Q n At the end of the green light time of the nth cycle lane, the video detector detects the length of the lane queue;

[0069] VI n The number of vehicles entering the detection area at the end of the green light of the nth cycle lane;

[0070] C n is the duration of the nth signal cycle at the intersection;

[0071] sgn(x) is the sign function, and its specific value is:

[0072] When x>0, sgn(x)=1; when x=0, sgn(x)=0; when x<0, sgn(x)=-1.

[0073] Because when a queue overflow occurs in the monitored lane, the video detector can only detect vehicles with a maximum length of Lmax. It is impossible to determine whether subsequent vehicles overflow through the data collected by the video detector. Similarly, the number of vehicles that actually enter the monitored lane when overflow occurs cannot be accurately collected by the video detector. Therefore, when calculating the vehicle input data of the monitored road, the data of the signal period with queue overflow is not included in the calculation. In this application, the symbol function sgn(x) is used to control whether the data of each period is included in the calculation. When the lane queue length Q collected by the video detector corresponding to the nth period is n When it is equal to Lmax, it means that there is a possibility of lane queue overflow in the nth cycle, and the corresponding number of merging vehicles is not included in the calculation.

[0074] S3: Calculate the average space length QV occupied by a single vehicle queue in the lane to be monitored i .

[0075] Usually, vehicles queue up continuously, and the queue value changes in direct proportion to the number of stranded vehicles, so:

[0076] QV i =Q i / VS i ;

[0077] Among them, QV i Q is the average length of space occupied by a single vehicle queue in the i-th signal cycle monitoring lane, iis the lane queue length detected by the video detector at the end of the green light for the i-th cycle lane, Q i > 0; VS i is the number of vehicles remaining in the lane within the detection area at the end of the green light for the i-th cycle, VS i > 0.

[0078] S4: Compare the lane queue length Q i detected by the video detector at the end of the green light for the i-th cycle of the lane to be monitored with the maximum length Lmax of the lane covered by the monitoring range;

[0079] If Q i < Lmax, then the predicted queue length QR i at the current time point of the lane to be monitored takes the value of Q i , indicating that there is no overflow risk for the lane to be monitored within this signal cycle;

[0080] Otherwise, when Q i = Lmax, execute step S5 to estimate QR i .

[0081] When there is an overflow at the intersection, the queue at the intersection exceeds the limit value recognized by the detector in the detection area, that is, the lane queue length Q i = Lmax. At this time, the predicted queue length QR i ≥ Lmax, and it is necessary to calculate QRi by combining the plan cycle, green light time, number of passing vehicles, and historical data.

[0082] When the vehicles in the lane to be monitored queue up at the intersection beyond the detection area, if the number of vehicles entering the lane within the cycle can be replaced by VUnit i * C i , then at this time, the queue increase value can be converted by the net increase in the number of vehicles in the lane (the difference between the number of entering vehicles and the number of leaving vehicles) and the space length occupied by a single vehicle in the queue, and finally QR i is calculated.

[0083] S5: The estimated value QR i of the predicted queue length QR i at the current time point of the lane to be monitored is calculated as follows:

[0084] QR i ' = QR i-1 + (VUnit i * C i - VO i ) * QV i ;

[0085] Among them, QR i-1is the predicted queue length of the lane to be monitored in the i-1th cycle;

[0086] VUnit i is the number of vehicles merging into the lane to be monitored per unit time in the i-th cycle;

[0087] C i is the duration of the i-th signal cycle at the intersection to be monitored;

[0088] QV i Calculate the average length of space occupied by a single vehicle queue in the lane for the i-th cycle;

[0089] VO i The number of vehicles passing through the lane at the end of the green light in the i-th cycle is the number of vehicles leaving the detection area.

[0090] S6: Compare QR i ' and the value of Lmax;

[0091] Because the queue of vehicles is changing dynamically, the number of vehicles entering the lane to be monitored in the i-th cycle is estimated based on the vehicle data entering the lane to be monitored in the previous i-1 cycles. There is a certain probability that QR i ' is less than Lmax, so if the estimated QR i ' is less than Lmax, then the data Lmax collected by the video detector in the i-th cycle shall prevail, and QR i The value of ' is set to Lmax to ensure that the method of the present application is more practical; in this case, there is no risk of queue overflow in the lane to be monitored.

[0092] If QR i '>Lmax, then the queue length QR of the lane to be monitored is predicted at the current time point i The value of is QR i ';

[0093] Otherwise, the queue length QR of the lane to be monitored is predicted at the current time point i The value of is Lmax.

[0094] That is: the queue length QR of the lane to be monitored is predicted at the current time point i The calculation formula is:

[0095] QR i =max(Lamx,QR i-1 +(VUnit i *C i -VO i )*QV i ).

[0096] S7: Compare QR i ' and lane to be monitored Llane;

[0097] If QR i '>Llan, it is considered that there is an overflow risk in the monitored import direction of the monitored intersection, and a warning message is given;

[0098] Otherwise, steps S2 to S7 are executed in a loop.

[0099] Assume that the maximum queue value Lmax of the video detector area in this embodiment is 80 meters. Specific data of each cycle on the lane to be monitored is shown in Table 1.

[0100] Table 1: Data example of lanes to be monitored

[0101]

[0102] Based on the data in Table 1, according to the value of the green light end video queue Q collected by the video detector, in the two signal cycles, the Q value detected by the video detector is Lmax=80, that is, there is a possibility of queue overflow. In this embodiment, m is 4, that is, 4 cycles are pushed forward from the first cycle where the Lmax value appears, and it is set to the 0th cycle, then Q4=Q5=80. From the 0th cycle to the 3rd cycle, when the green light ends, the length of the vehicle queue on the road to be monitored does not exceed the detection area.

[0103] First, calculate whether there is a queue overflow when i=4, and the queue length if there is a queue overflow.

[0104] According to QV i =Q i / VS i , calculate QV4 = 80 / 17 = 4.706;

[0105] According to the formula:

[0106]

[0107] Calculate the number of vehicles merging into the lane per unit time when i=4:

[0108] Vunit4=(10+10+12+13) / (100+100+100+120)=0.107;

[0109] Since Q4=Lmax=80, we can calculate:

[0110] QR4'=QR3+(VUnit3*C3-VO3)*QV4=78+(0.107*120-10)*4.706=91.365;

[0111] QR4'>Lmax, then the value of QR4 is 91.365.

[0112] Based on the same process, QR5 is calculated as:

[0113] QR5=91.365+(0.107*120-11)*5)=100.565;

[0114] If QR n 'If it exceeds the length of the lane to be monitored, it indicates that there is an overflow risk on the lane to be monitored in the entrance direction of the intersection.

[0115] After using the technical solution of the present invention, there is no need to add additional hardware equipment. The signal timing data can be used in the queue detection and recognition data. The existing video can be used to detect detailed indicators such as the number of vehicles in a fixed rectangular area, the queue length, the number of vehicles passing the stop line, etc., so the length of the vehicle queue on the monitored road can be estimated, thereby improving the queue length recognition ability in intelligent transportation.

Claims

1. A lane queue overflow detection method, characterized in that: It includes the following steps: S1: Obtain the monitoring range of the video detector set in the to-be-monitored import direction of the to-be-monitored intersection, and record the maximum length covering the lane in the monitoring range as Lmax; S2: Select the lane to be monitored and calculate the number of vehicles merging into the lane to be monitored per unit time VUnit in the i-th cycle i ; Obtain the length of the to-be-monitored lane, denoted as: Llane; Llane≥Lmax; S3: Calculate the average space length QV occupied by a single vehicle queue in the lane to be monitored i ; S4: Compare the lane queue length Q detected by the video detector at the end time of the green light of the lane in the i-th cycle i The maximum length of the lane covered by the monitoring range is recorded as Lmax; If Q i < Lmax, then the predicted queuing length QR of the lane to be monitored at the current moment is calculated i and the value of QR is Q i ; Otherwise , Q i =Lmax, execute step S5 to estimate QR i ; S5: Predict and calculate the queue length QR of the lane to be monitored at the current time point i The estimated value of QR i The calculation method of ' is: QR i ' = QR i-1 +(VUnit i *C i -VO i )*QV i ; Among them, QR i-1 is the predicted queue length of the lane to be monitored in the i-1th cycle; VUnit i is the number of vehicles merging into the lane to be monitored per unit time in the i-th cycle; C i is the duration of the i-th signal cycle at the intersection to be monitored; QV i Calculate the average length of space occupied by a single vehicle queue in the lane for the i-th cycle; VO i The number of vehicles passing through the lane at the end of the green light in the i-th cycle is the number of vehicles leaving the detection area; S6: Compare QR i ' and the value of Lmax; If QR i ' > Lmax, then the queue length QR of the lane to be monitored is predicted at the current time point i The value of is QR i '; Otherwise, the queue length QR of the lane to be monitored is predicted at the current time point i The value of is Lmax; S7: Compare QR i ' and lane to be monitored Llane; If QR i '>Llan, it is considered that there is an overflow risk in the monitored lane and a warning message is given; Otherwise, loop through steps S2 to S7.

2. The lane queue overflow detection method according to claim 1, characterized in that: In step S2, the number of vehicles merging into the lane to be monitored per unit time in the i-th cycle is VUnit i The calculation of includes the following steps: a1: Starting from the 0th cycle and up to the (i - 1)th cycle, count all the cycles with a queue length less than Lmax, and denote them respectively as: calculation cycles; a2: Count the total number of vehicles that have merged in all the calculation cycles detected by the video detector, denoted as: total number of merged vehicles for calculation; a3: VUnit is calculated by the total number of vehicles imported into the calculation and the total duration of the calculation cycle. i .

3. The lane queue overflow detection method according to claim 2, characterized in that: In step S2, the number of vehicles merging into the lane to be monitored per unit time in the i-th cycle is VUnit i The calculation formula is: ； Among them, Q n At the end of the green light time of the nth cycle lane, the video detector detects the length of the lane queue; VI n The number of vehicles entering the detection area at the end of the green light of the nth cycle lane; C n is the duration of the nth signal cycle at the intersection; sgn(x) is the sign function, and its specific values are: When x > 0, sgn(x) = 1; when x = 0, sgn(x) = 0; when x < 0, sgn(x) = -1.

4. The lane queue overflow detection method according to claim 1, characterized in that: In step S3, the average space length QV occupied by a single vehicle queue in the lane to be monitored i The calculation method is: QV i =Q i / VS i ; Among them, QV i Q is the average length of space occupied by a single vehicle queue in the i-th signal cycle monitoring lane, i is the lane queue length detected by the video detector at the end time of the green light of the i-th cycle lane, Q i >0;VS i is the number of vehicles stranded in the lane of the detection area at the end of the green light in the i-th cycle, VS i >0.

5. The lane queue overflow detection method according to claim 1, characterized in that: In step S6, the queue length QR of the lane to be monitored is predicted and calculated at the current time point i The calculation formula is: QR i = max(Lamx,QR i-1 +(VUnit i *C i -VO i )*QV i )。 6. The lane queue overflow detection method according to claim 2, characterized in that: In step a1, the method for determining the 0th cycle is: b1: The video detector continuously counts the queue lengths of each lane of each approach corresponding to the to-be-monitored intersection in real time; b2: Once it detects that the total road queue length on the to-be-monitored lane is equal to Lmax, then denote the corresponding cycle as: start cycle; b3: Starting from the start cycle, push forward m cycles according to the preset minimum number of calculation cycles m, and set it as the 0th cycle; Based on the 0th cycle, estimate the queue lengths of all subsequent cycles with the possibility of queue overflow.

7. A lane queue overflow detection method according to claim 6, characterized in that: In step b2, the value-taking method of m is: among all the complete signal cycles included in the 10 minutes before the start time of the start cycle, find the number of signal cycles with a road queue total length < Lmax, and set it as m.

Images