Traffic intersection road condition regulation and control method and system based on machine visual perception

Abstract

The invention discloses a traffic intersection road condition regulation and control method and system based on machine visual perception, and relates to the technical field of intelligent traffic. Historical data of traffic flow and pedestrian flow leading to different directions of a traffic intersection are obtained, and the number of green lights in each time interval in each direction is obtained; calculating the traffic flow level and the pedestrian flow level of each green light by combining the entropy theory, and obtaining the total flow level of the green light, and extracting the first-level, second-level and third-level time intervals of each day according to the morning, noon and evening; a first-level determined time interval, a first-level hidden time interval, a second-level determined time interval, a second-level hidden time interval, a third-level determined time interval and a third-level hidden time interval in the normal date set and the special date set are obtained, and green light duration adjustment is carried out according to the levels of the same time interval of the intersection in different directions. The traffic efficiency of the traffic intersection is effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent transportation, and specifically to a traffic intersection road condition regulation method and system based on machine vision perception. Background Art

[0002] With the rapid development of urbanization, traffic congestion has become an issue that cannot be ignored in urban development. Especially for the control of traffic intersection signal lights, traditional signal light control methods, such as fixed-time control, switch signal lights at set time intervals, completely ignoring the dynamic changes of actual traffic flow. During peak hours or when traffic flow suddenly changes, obvious deficiencies will be exposed. And some slightly intelligent induction control methods, although they can adjust the signal light duration according to the real-time traffic flow situation to a certain extent, they only focus on the single factor of traffic flow. This control method ignores the differences in vehicle types. Different types of vehicles require different times to pass through the intersection. Large vehicles, due to their large body size and slow start, take a longer time to pass through the intersection. If this factor is not considered, it may cause waste of green light time and affect the overall traffic efficiency.

[0003] At the same time, the existing signal light adjustment methods do not combine traffic flow and pedestrian flow. In urban traffic, pedestrians are also important traffic participants. At intersections near some schools and commercial areas, the pedestrian flow is very large during specific periods. However, the current signal light control fails to fully consider this situation, resulting in short green light time and pedestrians being unable to safely pass through the intersection, affecting traffic order and safety.

[0004] Therefore, there is an urgent need for a new method and system that can achieve intelligent regulation of traffic intersection signal lights, thereby improving the traffic efficiency of traffic intersections. Summary of the Invention

[0005] Aiming at the deficiencies of the prior art, the present invention provides a traffic intersection road condition regulation method and system based on machine vision perception, which solves the problems that traditional signal light regulation does not comprehensively consider vehicle and pedestrian flow and is difficult to adapt to complex traffic conditions.

[0006] To achieve the above objectives, the present invention is realized through the following technical solutions: A traffic intersection road condition regulation method and system based on machine vision perception, including:

[0007] Obtain historical data of traffic flow and pedestrian flow in different directions of a traffic intersection. For each direction, divide a day into twenty-four intervals per hour, obtain the number of times the green light appears in each time interval, and combine the entropy theory to obtain the traffic flow and pedestrian flow levels each time the green light appears. Combine the two to obtain the total flow level of this green light. Determine the level of this time interval according to the proportion of different levels of total flow in the time interval, and obtain a level sequence of twenty-four time intervals;

[0008] Divide the dates into a normal date set and a special date set according to the proportion of the number of first-level in the time interval, and extract the first-level, second-level, and third-level time intervals of each day in the early, middle, and late parts respectively, and put them into the corresponding early first-level set, early second-level set, early third-level set, middle first-level set, middle second-level set, middle third-level set, late first-level set, late second-level set, and late third-level set, to obtain the first-level determined time interval, first-level hidden time interval, second-level determined time interval, second-level hidden time interval, third-level determined time interval, and third-level hidden time interval in the normal date set and the special date set.

[0009] As a further solution of the present invention, when performing specific green light time scheduling, it is necessary to determine the level of the intersection in the same time interval in different directions, preset the basic green light duration of each direction, calculate the level difference between the lowest-level time interval and the highest-level time interval in different directions, and for the direction higher than the lowest level, the green light time increases by 5 seconds for each higher level, and the green light time of the lowest-level direction decreases by 5 seconds accordingly.

[0010] As a further solution of the present invention, the specific method for obtaining the traffic flow and pedestrian flow levels each time a green light is on according to the entropy theory is as follows:

[0011] For the type of vehicle, define large vehicle = 3, medium vehicle = 2, small vehicle = 1, and for the type of person, define slow pedestrian = 3, medium-speed pedestrian = 2, fast pedestrian = 1, to obtain the type vectors C = {c1, c2,..., cm1}, P = {p1, p2,..., pm2}, where m1 and m2 are the number of vehicles passing through during this green light cycle;

[0012] Calculate the net passing time of each vehicle and each person. The net passing time is the total green light time minus the intermediate stop time, and obtain the time vectors Tnet1 = {cnet1, cnet2,..., cnetm} and Tnet2 = {pnet1, pnet2,..., pnetm} respectively;

[0013] Perform min-max normalization processing on Tnet1 and Tnet2;

[0014] According to the formula Calculate the type entropy, where the type distribution probability pi = the number of vehicles or the number of people in the i-th category / m;

[0015] According to the formula Calculate the time entropy, where calculate the probability of each interval

[0016] Calculate the stop entropy according to the formula H(S) = -r * log2r - (1 - r) * log2(1 - r), where the stop mark

[0017] According to the formula calculate the weights of the entropies in each dimension;

[0018] According to the formula Score = w T *S T +w Tnorm *S Tnorm +w S *S S calculate the comprehensive score;

[0019] If Score >= Score2, set the vehicle or pedestrian flow level of this green light to high; if Score1 <= Score < Score2, set the vehicle or pedestrian flow level of this green light to medium; if Score < Score1, set the vehicle or pedestrian flow level of this green light to low.

[0020] As a further solution of the present invention, after the vehicle flow and pedestrian flow sequences A and B in each time interval are classified, the classification sequences A' and B' are obtained, and then according to the flow level addition rule, the total flow level sequence C of this time interval is obtained. The flow level addition rule is: if one of the corresponding moments in A' and B' is low, set the flow level of this moment to low; if there is no low but there is medium in the corresponding moment, set the flow level of this moment to medium; otherwise, set the flow level of this moment to high.

[0021] As a further solution of the present invention, calculate the proportions pers1, pers2, and pers3 of low, medium, and high in the total flow level sequence C. If max(pers1, pers2, pers3) = per1, set this time interval as the first-level time interval; if max(pers1, pers2, pers3) = per2, set this time interval as the second-level time interval; if max(pers1, pers2, pers3) = per3, set this time interval as the third-level time interval.

[0022] As a further solution of the present invention, find out the proportion first of the number of first-level intervals in the total number of intervals in a day. If first >= nim, classify this day into the normal date set; if first < nim, classify this day into the special date set, and record the specific dates of these special dates, where nim is the proportion threshold.

[0023] As a further solution of the present invention, the time intervals in the early, middle, and late parts need to be divided according to 24 hours a day. Divide the time interval (0, 12] into early, divide the time interval (12, 18] into middle, and divide the time interval (18, 24] into late.

[0024] As a further solution of the present invention, multiple groups of historical data in the same direction of the intersection are obtained, the intersection of time intervals of the same level is taken as the determined time interval, the difference between the union and the intersection is obtained to obtain the hidden time interval, and finally the time intervals sorted by level are: first-level determined time interval, first-level hidden time interval, second-level determined time interval, second-level hidden time interval, third-level determined time interval, and third-level hidden time interval.

[0025] As a further solution of the present invention, if it is found that the levels of different directions of the intersection are the same within a certain time interval, the sizes of the two determined time intervals are compared first, and the green light time of the intersection direction with the larger determined time interval is increased preferentially. If the two determined time intervals are the same, the hidden time intervals are continued to be compared, and the comparison is carried out downwards step by step.

[0026] A traffic intersection traffic control system based on machine vision perception, comprising: a data acquisition module, a flow level module, a time interval level module, and a green light time control module;

[0027] Data acquisition module, which obtains video data from all directions of the traffic intersection, analyzes the video data using machine vision algorithms, identifies vehicles and pedestrians, and counts their number, type, passing time, and residence time, and transmits the information to the traffic level module;

[0028] Traffic level module, which calculates the type entropy, time entropy, and stay entropy of vehicle and pedestrian flow at each green light in the time interval, obtains its comprehensive score, divides its level, and transmits it to the time interval level module;

[0029] The time interval level module obtains the level of the time interval by combining the levels of vehicle flow and pedestrian flow, and divides the date into a normal date set and a special date set according to the proportion of the first level, and extracts the first level, second level, and third level time intervals of each day according to the three parts of morning, middle, and evening, and puts them into the corresponding early first level set, early second level set, early third level set, middle first level set, middle second level set, middle third level set, late first level set, late second level set, and late third level set, and obtains the first level determined time interval, the first level hidden time interval, the second level determined time interval, the second level hidden time interval, the third level determined time interval, and the third level hidden time interval in the normal date set and the special date set, and transmits them to the green light time control module;

[0030] Green light time control module, which determines the level of the same time interval at the intersection in different directions, pre-sets the basic green light duration for each direction, calculates the level difference between the lowest level time interval and the highest level time interval in different directions, and for directions higher than the lowest level, the green light time increases by 5 seconds for each higher level, and the green light time for the corresponding lowest level direction decreases by 5 seconds.

[0031] The present invention provides a traffic intersection road condition regulation method and system based on machine vision perception. Compared with the prior art, it has the following beneficial effects:

[0032] (1) The present invention uses machine vision perception technology to process historical traffic data and combines entropy theory to classify the traffic flow and pedestrian flow levels, which can more accurately reflect the actual traffic conditions at the traffic intersection and make the evaluation of traffic conditions more comprehensive and scientific;

[0033] (2) By dividing the time interval levels and classifying the dates, the present invention can distinguish the time periods and dates with different traffic flow characteristics, provide a more targeted basis for traffic regulation, dynamically adjust the green light duration according to different levels of time intervals, and effectively improve the traffic efficiency at the traffic intersection. Description of the Drawings

[0034] Figure 1 is the flowchart of the steps of the present invention;

[0035] Figure 2 is the system principle block of the present invention. Detailed Embodiments

[0036] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0037] Such as Figure 1 , the present application provides a traffic intersection road condition regulation method and system based on machine vision perception, including:

[0038] Using the high-definition cameras and sensors installed at the traffic intersection, based on machine vision perception technology, obtain the traffic flow and pedestrian flow data in different directions at the intersection under the green light. Taking one day [0, 24) as a historical data cycle, divide it into one interval per hour, that is, [0, 1), [1, 2), [2, 3),..., [23, 24). For each interval, record the number of times n that the green light appears, and the number of vehicles and people ai, bi passing through at the end of each green light, i ∈ [1, n], so as to obtain the sequences A = {a1, a2,..., an} and B = {b1, b2,..., bn} of the traffic flow and pedestrian flow passing through each time under the green light in each interval;

[0039] Calculate the traffic flow level and pedestrian flow level each time according to the traffic flow and pedestrian flow sequences of each interval. The specific steps for calculating the flow level are as follows:

[0040] (1) Construct feature vectors

[0041] For the type of vehicle, define large vehicle (truck / bus) = 3, medium vehicle (SUV / MPV) = 2, small vehicle (sedan) = 1. For the type of person, define fast pedestrian (young adults, rarely staying) = 1, medium-speed pedestrian (middle-aged and elderly, carrying items) = 2, slow pedestrian (old people, children, mobile phone users) = 3, to obtain the type vectors C = {c1, c2,..., cm1}, P = {p1, p2,..., pm2}, where m1 is the number of vehicles passing through during this green light cycle, and m2 is the number of people passing through during this green light cycle;

[0042] Calculate the net passing time of each vehicle and each person = total green light passing time - intermediate staying time, respectively obtaining the time vectors Tnet1 = {cnet1, cnet2,..., cnetm}, Tnet2 = {pnet1, pnet2,..., pnetm}, and perform min-max normalization on Tnet1 and Tnet2;

[0043] (2) Calculation of multi-dimensional entropy values

[0044] First is the type entropy. The higher the entropy value, the more diverse the vehicle types and person types, and the lower the passing efficiency may be. By calculating the type distribution probability pi = the number of vehicles or people of the i-th type / m. As described above, the vehicle types are divided into three categories: large, medium, and small, and the person types are divided into three categories: fast, medium-speed, and slow. According to the formula Calculate the type entropy;

[0045] Next is the time entropy, which reflects the stability of the passing efficiency of vehicles or people. The higher the entropy value, the greater the difference in passing times of vehicles or people. For example, some vehicles or people pass quickly, while some pass slowly due to staying, and the flow stability is worse. Divide the standardized net passing time into k = 5 equally spaced intervals, and calculate the probability of each interval According to the formula Calculate the time entropy;

[0046] Finally is the staying entropy. When the entropy value is close to 1, the staying and non-staying vehicles or people each account for half, indicating that abnormal events have a large interference on the traffic flow, such as pedestrians crossing the road or vehicle breakdowns. Define the staying mark Calculate the staying probability Calculate the staying entropy according to the formula H(S) = -r * log2r - (1 - r) * log2(1 - r);

[0047] (3) Comprehensive score calculation

[0048] According to the formula Calculate the weights of the entropies in each dimension. The smaller the entropy value, the higher the weight, because a smaller entropy indicates that the dimension is more ordered. Convert each entropy value to a score between 0 and 100. The smaller the entropy, the higher the score, representing that the traffic flow is more ordered. The specific conversion formula is, S T =(1 - H(T)) * 100, S Tnorm =(1 - H(Tnorm)) * 100, S S =(1 - H(S)) * 100; According to the formula Score = w T *S T +w Tnorm *S Tnorm +w S *S S Calculate the comprehensive score;

[0049] (IV) Classification Rules

[0050] If Score >= 80, with a single vehicle or person dominant, concentrated passing time, and no or few stops, set the traffic flow level of this green light cycle for vehicles or pedestrians as high; if 50 <= Score < 80, with a mixture of vehicle or person types, different passing times, and occasional stops, set the traffic flow level of this green light cycle for vehicles or pedestrians as medium; if Score < 50, with complex vehicle or person types, large differences in passing times, and frequent stops, set the traffic flow level of this green light cycle for vehicles or pedestrians as low.

[0051] After the above traffic flow level classification for the sequences A and B of vehicle flow and pedestrian flow at each green light within each interval, obtain the level sequences A' and B'. At this time, it is necessary to obtain the combined total traffic flow level sequence C according to the traffic flow level addition rule. The specific traffic flow level addition rule is as follows: low > medium > high, that is, if there is a low in the corresponding moments of A' and B', set the traffic flow level at this moment as low; if there is no low but there is a medium in the corresponding moments, set the traffic flow level at this moment as medium; otherwise, set the traffic flow level at this moment as high;

[0052] For example, during the evening rush hour in the interval [17, 18), the vehicle flow level sequence A' = {medium, high, medium} and the pedestrian flow level sequence B' = {high, low, medium} at a certain intersection. According to the rule, at the first moment, A' is medium and B' is high, there is no low but there is a medium, so the corresponding total traffic flow level sequence C at this moment is medium; at the second moment, A' is high and B' is low, there is a low, so the corresponding total traffic flow level sequence C at this moment is low; at the third moment, A' is medium and B' is medium, there is no low, there is a medium, so the corresponding total traffic flow level sequence C at this moment is medium. Finally, obtain the total traffic flow level sequence C = {medium, low, medium};

[0053] Collect the proportions pers1, pers2, and pers3 of low, medium, and high in the total traffic flow level sequence C. If max(pers1, pers2, pers3) = per1, then set this time interval as the first-level time interval, indicating that the vehicle types are chaotic and the quantity is large, or the people are miscellaneous and numerous in this time interval; if max(pers1, pers2, pers3) = per2, then set this time interval as the second-level time interval, indicating that the vehicle types are average or the pedestrian flow is average in this time interval; if max(pers1, pers2, pers3) = per3, then set this time interval as the third-level time interval, indicating that the vehicle types are single and the quantity is small or the number of people is small in this time interval.

[0054] For example, if the proportions of low, medium, and high in the total traffic flow level sequence C in the above time interval [17, 18) are 0.4, 0.3, and 0.3 respectively, and max(0.4, 0.3, 0.3) = 0.4, that is, pers1 is the largest, then this time interval is the first-level time interval, indicating that the traffic condition at this intersection is relatively complex and the traffic pressure is relatively large during this time period.

[0055] Perform the above level division for all 24 time intervals in a day to obtain the time level sequence for 24 hours of a single day. Taking the historical data of a month as an example, by analyzing the time level sequence of each day, divide the dates again, and calculate the proportion first of the number of first-level intervals in the total number of intervals in each day. If first >= nim, where nim is a preset threshold, then classify this day into the normal date set; if first < nim, then classify this day into the special date set and record the specific dates of these special dates.

[0056] The dates are divided into the normal date set and the special date set because there are different traffic flow patterns on different dates. The traffic flow on normal dates usually follows a certain daily pattern, while abnormal traffic flow changes will occur on special dates, such as holidays and days when large-scale events are held. Through this division, it is possible to analyze and process traffic data of different types of dates more targeted and formulate more reasonable traffic control strategies.

[0057] For the normal date set, long-term data analysis can be carried out based on its stable traffic rules to establish relatively fixed traffic control strategies; for the special date set, it can be analyzed separately to predict possible traffic congestion points in advance and formulate special response plans, thereby improving the pertinence and effectiveness of traffic control. For example, during holidays, the distribution of traffic flow is very different from that in daily life, and the traffic pressure around commercial areas and tourist attractions will increase significantly. By identifying holidays as special dates, it is possible to increase the police deployment and adjust the signal timing in these areas in advance to avoid traffic congestion.

[0058] For the normal date set and the special date set, extract the first-level, second-level, and third-level time intervals of each day in the morning, afternoon, and evening respectively, and put them into the morning first-level set, morning second-level set, morning third-level set, afternoon first-level set, afternoon second-level set, afternoon third-level set, evening first-level set, evening second-level set, and evening third-level set correspondingly;

[0059] The time intervals of the morning, afternoon, and evening need to be divided according to 24 hours of a day. The time interval (0, 12] is divided into the morning, the time interval (12, 18] is divided into the afternoon, and the time interval (18, 24] is divided into the evening. There are first-level, second-level, and third-level time intervals in different time intervals;

[0060] Obtain multiple groups of historical data in the same direction at the intersection, perform intersection operation and union operation on the time intervals in these 9 sets, take the obtained intersection as the determined time interval, and then obtain the hidden time interval by taking the difference between the union and the intersection. Finally, the time intervals sorted by level are the first-level determined time interval, first-level hidden time interval, second-level determined time interval, second-level hidden time interval, third-level determined time interval, and third-level hidden time interval;

[0061] For example, in the set of normal dates, for Date 1, the early first-level set is \((6, 8]\), the early second-level set is \((8, 9]\), the early third-level set is \([0, 6], (9, 11]\), the middle first-level set is \((11, 12], (13, 14]\), the middle second-level set is \((12, 13]\), the middle third-level set is \((14, 17]\), the late first-level set is \((17, 19]\), the late second-level set is \((19, 21]\), the late third-level set is \((21, 24]\); for Date 2, the early first-level set is \((7, 8]\), the early second-level set is \((8, 10]\), the early third-level set is \([0, 7], (10, 11]\), the middle first-level set is \((11, 12], (13, 15]\), the middle second-level set is \((12, 13]\), the middle third-level set is \((15, 17]\), the late first-level set is \((17, 20]\), the late second-level set is \((20, 21]\), the late third-level set is \((21, 24]\); Gathering the early first-level set, early second-level set, early third-level set, middle first-level set, middle second-level set, middle third-level set, late first-level set, late second-level set, and late third-level set of Date 1 and Date 2 together to find the intersection, the first-level determined time interval is \((7, 8], (11, 12], (13, 14], (17, 19]\), the first-level hidden time interval is \((6, 7], (14, 15], (19, 20]\), the second-level determined time interval is \((8, 9], (12, 13], (20, 21]\), the second-level hidden time interval is \((9, 10], (19, 20]\), the third-level determined time interval is \([0, 6], (10, 11], (15, 17]\), the third-level hidden time interval is \((6, 7], (9, 10], (14, 15], (21, 24]\). The specific rank order of the above time intervals is: first-level determined time interval > first-level hidden time interval > second-level determined time interval > second-level hidden time interval > third-level determined time interval > third-level hidden time interval. In the above example, \((19, 20]\) in the first-level hidden time interval repeats with the second-level hidden time interval. At this time, it needs to be assigned to the time interval with a higher level. At the same time, \((19, 20]\) in the second-level hidden time interval repeats with the first-level hidden time interval, and \((6, 7], (9, 10], (14, 15]\) in the third-level hidden time interval also repeat with those of a higher level than it, and they need to be deleted from it. Finally, the first-level determined time interval is \((7, 8], (11, 12], (13, 14], (17, 19]\), the first-level hidden time interval is \((6, 7], (14, 15], (19, 20]\), the second-level determined time interval is \((8, 9], (12, 13], (20, 21]\), the second-level hidden time interval is \((9, 10]\), the third-level determined time interval is \([0, 6], (10, 11], (15, 17]\), the third-level hidden time interval is \((21, 24]\).

[0062] When conducting specific green light time scheduling, adjust the green light time according to the time interval level division in each direction at each intersection. For example, in the time interval (6, 7], the time interval level corresponding to the front of the intersection is the first-level determined time interval, the left side corresponds to the third-level determined time interval, and the right side corresponds to the second-level hidden time interval. Since the traffic pressure in the first-level time interval is the greatest, at this time, it is necessary to increase the green light time in front of the intersection and reduce the green light time on the left and right sides;

[0063] The basic green light duration for each direction is TT. Calculate the level difference grade between the lowest-level time interval and the highest-level time interval. For directions higher than the lowest level, the green light time increases by 5 seconds for each higher level, and the green light time for the lowest-level direction decreases by 5 seconds accordingly;

[0064] If it is found that the levels of different directions at an intersection are the same within a certain time interval, first compare the sizes of their determined time intervals, and preferentially increase the green light time for the direction of the intersection with the larger determined time interval to prevent more serious traffic jams as the duration increases. If the determined time intervals of the two are also the same, then continue to compare the hidden time intervals and compare level by level downward;

[0065] Assume that the basic green light duration TT for three directions at an intersection is 30 seconds. At a certain moment, the front of the intersection is the first-level determined time interval, the left side is the second-level hidden time interval, and the right side is the third-level determined time interval. The level difference grade between the first level and the third level is 4. Then the green light time in front of the intersection is extended by 4×5 = 20 seconds and becomes 50 seconds; the green light time on the left side is extended by 5 seconds and becomes 35 seconds; the green light time on the right side is reduced by 20 seconds and becomes 10 seconds. By this way of dynamically adjusting the green light time, it is possible to reasonably allocate road resources according to the real-time traffic conditions at the intersection and effectively relieve traffic congestion.

[0066] Some of the data in the above formula are numerically calculated after removing their dimensions, and the content not described in detail in this specification belongs to the prior art well-known to those skilled in the art.

[0067] The above embodiments are only used to illustrate the technical method of the present invention and not to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical method of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A traffic intersection road condition regulation method based on machine vision perception, characterized in that include: Obtain historical data on vehicle and pedestrian flows in different directions at traffic intersections. For each direction, divide a day into 24 intervals per hour, calculate the number of green lights in each time interval, and combine the entropy theory to calculate the level of vehicle and pedestrian flows at each green light. Combine the two to obtain the total flow level of the green light. Determine the level of the time interval according to the proportion of different levels of the total flow in the time interval, and obtain the level sequence of the 24 time intervals. According to the proportion of the number of first-level sequence in the time interval, the dates are divided into a normal date set and a special date set, and the first-level, second-level, and third-level time intervals of each day are extracted according to the three parts of morning, middle and evening, and put them into the corresponding early first-level set, early second-level set, early third-level set, middle first-level set, middle second-level set, middle third-level set, late first-level set, late second-level set, and late third-level set to obtain the first-level determined time interval, first-level hidden time interval, second-level determined time interval, second-level hidden time interval, third-level determined time interval, and third-level hidden time interval in the normal date set and the special date set.

2. The traffic intersection road condition regulation method based on machine vision perception according to claim 1, characterized in that, When performing specific green light time scheduling, it is necessary to determine the level of the same time interval in different directions at the intersection, pre-set the basic green light duration for each direction, calculate the level difference between the lowest level time interval and the highest level time interval in different directions, and for directions higher than the lowest level, the green light time increases by 5 seconds for each higher level, and the corresponding green light time for the lowest level direction decreases by 5 seconds.

3. The traffic intersection road condition regulation method based on machine vision perception according to claim 1, wherein According to the entropy theory, the specific method for obtaining the vehicle flow and pedestrian flow levels at each green light is: For the type of vehicle, define large vehicle = 3, medium vehicle = 2, small vehicle = 1; for the type of person, define slow pedestrian = 3, medium pedestrian = 2, fast pedestrian = 1, and get the type vector C = {c1, c2, ..., cm1}, P = {p1, p2, ..., pm2}, m1, m2 are the number of vehicles passing through the green light cycle; Calculate the net passing time of each vehicle and each person. The net passing time is the total green light time minus the stopover time, and obtain the time vectors Tnet1 = {cnet1, cnet2, ..., cnetm} and Tnet2 = {pnet1, pnet2, ..., pnetm} respectively; Perform min-max normalization on Tnet1 and Tnet2; According to the formula calculate the type entropy, where the type distribution probability pi = the number of vehicles or the number of people in the i-th category / m; According to the formula calculate the time entropy, where the probability of each interval is calculated The residence entropy is calculated according to the formula H(S) = -r * log2r - (1 - r) * log2(1 - r), where the residence mark residence probability According to the formula Calculate the weights of the entropy of each dimension; According to the formula Score = w T *S T +w Tnorm *S Tnorm +w S *S S calculate the comprehensive score; If Score>=Score2, the vehicle or pedestrian flow level of the green light is set to high; if Score1<=Score <Score2,将该次绿灯的车或人流量等级设为中;若Score<Score1,将该次绿灯的车或人流量等级设为低。 4. The traffic intersection road condition regulation method based on machine vision perception according to claim 1, wherein After the vehicle flow and pedestrian flow sequences A and B in each time interval are divided into levels, the level sequences A' and B' are obtained, and then the total flow level sequence C of the time interval is obtained according to the flow level addition rule. The flow level addition rule is: if one of the corresponding moments of A' and B' is low, the flow level at that moment is set to low; if there is no low but medium at the corresponding moment, the flow level at that moment is set to medium; otherwise, the flow level at that moment is set to high.

5. The traffic intersection road condition regulation method based on machine vision perception according to claim 1, characterized in that Calculate the proportion of low, medium and high in the total traffic level sequence C, pers1, pers2, pers3. If max(pers1, pers2, pers3) = per1, set the time interval as the first-level time interval; if max(pers1, pers2, pers3) = per2, set the time interval as the second-level time interval; if max(pers1, pers2, pers3) = per3, set the time interval as the third-level time interval.

6. The traffic intersection road condition regulation method based on machine vision perception according to claim 1, characterized in that, Find the ratio of the number of first-level intervals to the total number of intervals in a day first. If first>=nim, then the day is included in the normal date set. If first <nim,则将该天归入特殊日期集合,并记录这些特殊日期的具体日期,其中,nim为比例阈值。 7. The traffic intersection road condition regulation method based on machine vision perception according to claim 1, characterized in that The time intervals of morning, noon and evening need to be divided according to 24 hours a day, and the time interval (0,12] is divided into morning, the time interval (12,18] is divided into noon, and the time interval (18,24] is divided into evening.

8. The traffic intersection road condition regulation method based on machine vision perception according to claim 1, characterized in that Multiple groups of historical data in the same direction of the intersection are obtained, and the intersection of time intervals of the same level is taken as the determined time interval. The hidden time interval is obtained by subtracting the union and the intersection. Finally, the time intervals sorted by level are: first-level determined time interval, first-level hidden time interval, second-level determined time interval, second-level hidden time interval, third-level determined time interval, and third-level hidden time interval.

9. The traffic intersection road condition regulation method based on machine vision perception according to claim 1, wherein, If it is found that the levels of different directions at an intersection are the same within a certain time interval, the sizes of the two determined time intervals are compared first, and the green light time of the intersection direction with the larger determined time interval is increased first. If the two determined time intervals are the same, the hidden time intervals are compared again, step by step downward.

10. A traffic intersection road condition regulation system based on machine vision perception, which is used to execute the traffic intersection road condition regulation method based on machine vision perception according to any one of claims 1-9, and is characterized in that, The module includes: data acquisition module, flow level module, time interval level module, green light time control module; Data acquisition module, which obtains video data from all directions of the traffic intersection, analyzes the video data using machine vision algorithms, identifies vehicles and pedestrians, and counts their number, type, passing time, and residence time, and transmits the information to the traffic level module; Traffic level module, which calculates the type entropy, time entropy, and stay entropy of vehicle and pedestrian flow at each green light in the time interval, obtains its comprehensive score, divides its level, and transmits it to the time interval level module; Time interval level module. This module obtains the level of the time interval by combining the traffic flow level and the pedestrian flow level, and divides the dates into the normal date set and the special date set according to the proportion of the number of the first level. It extracts the first-level, second-level, and third-level time intervals of each day in the morning, afternoon, and evening respectively, and puts them into the corresponding early first-level set, early second-level set, early third-level set, middle first-level set, middle second-level set, middle third-level set, late first-level set, late second-level set, and late third-level set, obtaining the first-level determined time intervals, first-level hidden time intervals, second-level determined time intervals, second-level hidden time intervals, third-level determined time intervals, and third-level hidden time intervals in the normal date set and the special date set, and transmits them to the green light time regulation module; Green light time regulation module. This module determines the level of the intersection at the same time interval in different directions, preset the basic green light duration for each direction, calculates the level difference between the lowest-level time interval and the highest-level time interval in different directions. For the direction higher than the lowest level, the green light time increases by 5 seconds for each higher level, and the green light time of the lowest-level direction decreases by 5 seconds accordingly.

Images