Traffic signal control method and system based on deep learning

Through the deep learning-based traffic signal control method, using preprocessed data and real-time traffic flow information to optimize traffic signal control, the problem of insufficient traffic flow perception and prediction in the existing technology is solved, and more accurate and efficient traffic signal management is achieved.

CN119068663APending Publication Date: 2024-12-03SMART ROAD NETWORK (TANGSHAN) ARTIFICIAL INTELLIGENCE TECH RES INST CO LTD
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202410503444.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-04-25
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

Existing traffic signal control methods are difficult to fully perceive and predict road traffic flow, resulting in inaccurate signal control and inefficient efficiency.

Method used

Using a deep learning-based method, by establishing a preprocessed pass data set, constructing independent date characteristics and normal date characteristics, performing multi-period clustering, obtaining current date and time node data, and combining real-time traffic information collected by the surveillance camera, optimized Traffic signal control data determines the limit adjustment space and realizes accurate traffic signal control.

Benefits of technology

It improves the accuracy and efficiency of traffic signal control, can better adapt to dynamically changing traffic conditions, and reduce traffic congestion and accidents.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure HDA0004809445200000011
    Figure HDA0004809445200000011
  • Figure HDA0004809445200000021
    Figure HDA0004809445200000021
  • Figure HDA0004809445200000022
    Figure HDA0004809445200000022
Patent Text Reader

Abstract

The invention discloses a traffic signal control method and system based on deep learning, and relates to the field of intelligent traffic. The method comprises the following steps: establishing a traffic data set; performing traffic data set data segmentation through the independent date features and the normal date features, completing multi-period clustering of deep learning, and generating traffic proportion features; performing feature matching by taking the date data and the node data as basic features to generate a proportion matching result; carrying out real-time traffic flow collection through a monitoring camera, and generating compensation feedback data based on a collection result; determining basic traffic signal control data based on a proportion matching result, and determining a limit adjustment space; and optimizing the basic traffic signal control data according to the compensation feedback data, and carrying out optimization result constraint to complete traffic signal control. The technical problem that traffic flow is difficult to comprehensively perceive and predict in the prior art is solved, and the technical effects of performing accurate control based on deep learning and improving the accuracy and efficiency of traffic signal control are achieved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of intelligent transportation, and in particular to a traffic signal control method and system based on deep learning. Background Art

[0002] Traffic signal control is an important part of urban traffic management and plays a key role in improving the efficiency of urban traffic operation and reducing congestion and accidents. In the past few decades, with the continuous acceleration of urbanization, traffic congestion and frequent accidents have become important issues worldwide, which has prompted people to continuously study and improve traffic signal control technology. Existing traffic signal control methods can often only partially perceive and predict traffic flow on the road, and it is difficult to fully grasp the actual situation of road traffic, resulting in inaccurate and inefficient signal control. The traffic signal control method based on deep learning can improve the accuracy and efficiency of signal control by learning a large amount of traffic data, comprehensively perceive and predict road traffic flow, and improve the accuracy and efficiency of signal control. Summary of the invention

[0003] The embodiments of the present application provide a traffic signal control method and system based on deep learning, which solves the technical problem in the prior art that it is difficult to fully perceive and predict traffic flow.

[0004] In view of the above problems, the embodiments of the present application provide a zebra crossing detection method and system based on image recognition.

[0005] A first aspect of an embodiment of the present application provides a traffic signal control method based on deep learning, the method comprising:

[0006] Establishing a traffic data set, wherein the traffic data set is obtained by collecting historical road traffic data, wherein the traffic data set is a preprocessed data set, and the preprocessing method includes noise reduction processing and time series compensation processing;

[0007] Constructing independent date features and normal date features, segmenting the traffic data set by using the independent date features and the normal date features, and completing multi-period clustering of deep learning by using the data segmentation results to generate traffic ratio features of the intersection;

[0008] Acquire date data and time node data of the current date, use the date data and time node data as basic features, perform feature matching of the traffic ratio feature, and generate a ratio matching result;

[0009] Real-time traffic flow collection is performed through surveillance cameras, and compensation feedback data is generated based on the collection results;

[0010] Determining basic traffic signal control data based on the ratio matching result, and determining a limit adjustment space;

[0011] The basic traffic signal control data is optimized according to the compensation feedback data, and the optimization result is constrained by the limit adjustment space to complete the traffic signal control.

[0012] A second aspect of an embodiment of the present application provides a traffic signal control system based on deep learning, the system comprising:

[0013] A data set module, wherein the data set module is used to establish a traffic data set, wherein the traffic data set is obtained by collecting historical road traffic data, wherein the traffic data set is a preprocessed data set, and the preprocessing method includes noise reduction processing and time series compensation processing;

[0014] A feature module, wherein the feature module is used to construct independent date features and normal date features, perform data segmentation of the traffic data set through the independent date features and the normal date features, and perform multi-period clustering of deep learning through the data segmentation results to generate traffic ratio features of the intersection;

[0015] A matching module, the matching module is used to obtain date data and time node data of the current date, and use the date data and node data as basic features to perform feature matching of the traffic ratio feature to generate a ratio matching result;

[0016] A compensation feedback module, which is used to collect real-time traffic flow through a monitoring camera and generate compensation feedback data based on the collected results;

[0017] A determination module, the determination module is used to determine basic traffic signal control data based on the ratio matching result, and determine a limit adjustment space;

[0018] An optimization module is used to optimize the basic traffic signal control data according to the compensation feedback data, and to constrain the optimization result through the limit adjustment space to complete the traffic signal control.

[0019] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0020] First, the historical traffic data of the road is collected, and noise reduction and time series compensation are performed to obtain the traffic data set. Then, independent date features and normal date features are constructed, and the traffic data set is segmented using the independent date features and normal date features, and multi-period clustering of deep learning is performed through the data segmentation results to generate the traffic proportion features. Next, the date data and time node data of the current date are obtained, and the date data and node data are used as basic features to perform feature matching and generate proportion matching results. Real-time traffic flow is collected through monitoring cameras, and compensation feedback data is generated based on the collection results. Then, the basic traffic signal control data is determined based on the proportion matching results, and the limit adjustment space is determined. Finally, the basic traffic signal control data is optimized based on the compensation feedback data, and the optimization result is constrained by the limit adjustment space to complete the traffic signal control. The above method solves the technical problem of the difficulty in fully perceiving and predicting traffic flow in the prior art, and achieves the technical effect of accurate control based on deep learning and improving the accuracy and efficiency of traffic signal control. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0022] Figure 1 A flow chart of a traffic signal control method based on deep learning provided in an embodiment of the present application;

[0023] Figure 2 A schematic diagram of a process of dynamically regulating traffic signals by predicting duration in a traffic signal control method based on deep learning provided in an embodiment of the present application;

[0024] Figure 3 A schematic diagram of the structure of a traffic signal control system based on deep learning provided in an embodiment of the present application.

[0025] Explanation of the reference numerals: data set module 11 , feature module 12 , matching module 13 , compensation feedback module 14 , determination module 15 , optimization module 16 . DETAILED DESCRIPTION

[0026] The embodiments of the present application solve the technical problem that it is difficult to fully perceive and predict traffic flow in the prior art by providing a traffic signal control method and system based on deep learning.

[0027] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0028] It should be noted that the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules that are not explicitly listed or inherent to these processes, methods, products or devices.

[0029] Embodiment 1

[0030] like Figure 1 As shown, an embodiment of the present application provides a traffic signal control method based on deep learning, wherein the method includes:

[0031] Establishing a traffic data set, wherein the traffic data set is obtained by collecting historical road traffic data, wherein the traffic data set is a preprocessed data set, and the preprocessing method includes noise reduction processing and time series compensation processing;

[0032] With the acceleration of urbanization, traffic flow has increased dramatically, and traffic signal control technology has emerged, playing an important role in ensuring road traffic safety and optimizing traffic flow. In order to cope with the increasingly severe urban traffic problems, traffic signal control technology is also constantly improving and developing. For this reason, a traffic signal control method based on deep learning is proposed. By learning a large amount of traffic data, more accurate traffic flow perception and prediction can be achieved, which can better adapt to dynamically changing traffic conditions, improve the intelligence level of traffic management systems, and thus improve urban traffic conditions.

[0033] Traffic signal control is widely used in traffic scenarios such as urban roads, highways and intersections. Select any traffic scenario and collect historical road traffic data through traffic monitoring systems, GPS trajectory data, mobile phone signaling data and other methods. The historical road traffic data refers to historical data that records the operation of road traffic, including but not limited to information such as vehicle volume, vehicle speed and road conditions. Then, the collected historical traffic data is preprocessed including noise reduction and timing compensation. The timing compensation can use a trust evaluation strategy to process time series data. In this strategy, the closer the time series data point is to the current node, the more credible it is, so it can better reflect the current actual situation, rather than the uncertainty brought by historical data. Then, a traffic data set is established based on the preprocessed data. The traffic data set is a data set for the traffic field, which is used to record and analyze road traffic conditions. The establishment of a traffic data set can classify and mark different road types and traffic scenarios, providing more comprehensive and accurate data support for analysis and decision-making in the traffic field.

[0034] Constructing independent date features and normal date features, segmenting the traffic data set by using the independent date features and the normal date features, and completing multi-period clustering of deep learning by using the data segmentation results to generate traffic ratio features of the intersection;

[0035] In order to improve data processing efficiency and reduce computing costs, the constructed traffic data set can be segmented using independent date features and normal date features. To this end, it is necessary to construct independent date features and normal date features. Specifically, feature analysis is performed on the acquired traffic data set to understand the distribution and regularity of the data and extract time-related features from the data, such as year, month, day, week, season, etc. The independent date features include time-related features such as date and seasonal factors, which can reflect the impact of different dates on traffic data. The normal date features include periodic factors such as the day of the week and the days of the month, which can reflect the impact of different date cycles on traffic data. The independent date features can be constructed to calculate the corresponding independent date features for each date. For example, the similarity between different dates can be calculated, or the date can be used as an input feature to directly perform model training. The normal date features can be constructed to calculate the corresponding normal date features for the periodic changes of the date. For example, the average traffic flow per week or month can be calculated, or the date cycle can be used as an input feature to directly perform model training. Next, according to the independent date features and the normal date features, the traffic data set is segmented into multiple subsets, each subset having similar independent date features and normal date features. This segmentation helps to improve the efficiency and effectiveness of data processing, because data with similar features can be more easily analyzed and processed in the same subset. Finally, based on the data segmentation results, a deep learning algorithm is used to train and cluster each subset, complete the multi-period clustering of deep learning, and generate the traffic ratio feature. The multi-period clustering is a clustering method based on time series data, which can cluster time series data in different time periods. The purpose of multi-period clustering is to discover patterns and trends in time series data in order to better understand and analyze the evolution of data. The deep learning algorithm can automatically learn and extract features from the data, and divide the data points into different categories or clusters according to the similarity of these features. In this process, various deep learning models can be used, such as neural networks, convolutional neural networks, autoencoders, etc. Exemplarily, the convolutional neural network CNN can effectively identify various patterns and features in traffic signal images, and then use K-Means clustering to cluster the data in each time period to discover patterns and trends in the data. The deep learning algorithm can be trained and optimized in each subset to obtain more accurate traffic ratio features. These features can reflect the traffic flow ratio in different time periods and help formulate and optimize traffic signal control strategies. After clustering is completed, for each cluster, the corresponding traffic ratio feature can be calculated. The traffic ratio feature can reflect the traffic flow ratio in different time periods, such as the traffic volume and speed in different time periods. These features are very useful for application scenarios such as traffic signal control, because they can help formulate and optimize traffic signal control strategies and improve road traffic efficiency and safety.The whole process aims to cluster and analyze traffic data using data segmentation and deep learning algorithms to generate more accurate and useful traffic ratio features.

[0036] Furthermore, the multi-period clustering of deep learning is completed through the data segmentation results, and the method also includes:

[0037] Obtaining a non-peak clustering result, and generating a non-peak time interval based on the non-peak clustering result;

[0038] When the time node data of the corresponding date is within the idle time interval, a dynamic determination instruction is generated;

[0039] Execute real-time image acquisition through the dynamic determination instruction to generate real-time longitudinal and transverse vehicle flow information;

[0040] When there are vehicles in any direction of traffic and the adjacent traffic is empty, the dynamic judgment is passed;

[0041] Traffic signals are dynamically adjusted based on the determination results.

[0042] After completing the multi-period clustering of deep learning through the data segmentation results, the cluster to which each time node belongs is determined according to the timestamp of the traffic data, and the characteristics of each cluster are obtained. Then, the cluster to which each time node belongs is traversed to find the cluster representing the off-peak traffic situation, and the off-peak clustering results are obtained. For the cluster representing the off-peak period, the time interval covered by it is determined, so as to obtain the off-peak time interval corresponding to each cluster. For example, if a cluster represents a low-peak traffic period, then the time interval covered by the cluster is the off-peak time interval. The off-peak clustering result is the clustering result obtained after clustering the traffic data according to a specific clustering algorithm, such as K-Means or DBSCAN, etc., representing the off-peak traffic situation, that is, the time period with low traffic flow and slow speed. When the time node data of the corresponding date is within the off-peak time interval, the system will generate a dynamic judgment instruction. The dynamic judgment instruction is an instruction for the system to dynamically judge the current traffic situation based on the off-peak clustering results and time node data that have been obtained, such as whether it is in an off-peak traffic state, or whether the timing of the traffic lights needs to be adjusted. In this process, the system first extracts and matches the time node data, that is, compares the time node data with the obtained idle time clustering results. If the time node data has a high degree of feature matching with a certain idle time clustering result, the system will generate a corresponding dynamic judgment instruction. This instruction is dynamically generated, that is, the system will update it in real time according to real-time traffic data and traffic conditions to ensure the best control of traffic lights.

[0043] After the system automatically generates dynamic instructions, it obtains real-time traffic images or videos through image acquisition devices (such as cameras or laser radars, etc.). These images or videos contain the driving conditions of vehicles on the road, such as the location, speed, and driving direction of the vehicles. Then, the system uses image processing and computer vision technology to analyze these images or videos and extract the characteristics and behavior patterns of the vehicles. For example, these data are analyzed and processed by image processing and computer vision technology to obtain information such as the number, speed, and driving direction of the vehicles. Finally, the system generates real-time longitudinal and transverse traffic flow information based on these characteristics and behavior patterns. The real-time longitudinal and transverse traffic flow information refers to the vehicle flow information obtained in real time in the traffic scene, including the number, speed, and driving direction of the vehicles, as well as the status of traffic lights and traffic conditions. These real-time longitudinal and transverse traffic flow information can be transmitted to the traffic signal control system through the network to realize real-time control and adjustment of traffic lights, thereby improving the traffic efficiency and safety of the road. When there are vehicles in any direction of traffic and the adjacent traffic is empty, it means that the traffic demand in that direction is high, while the traffic demand in the other direction is low. In this case, the system will dynamically determine that the vehicle can pass through the intersection safely without causing interference or danger to vehicles in the other direction. Next, the system will dynamically adjust the traffic signal based on the judgment result. For example, if there are a large number of vehicles in this direction and the speed is slow, the system can appropriately extend the green light time in this direction to ease traffic pressure. If there are a small number of vehicles in the other direction and the speed is fast, the system can appropriately shorten the green light time in this direction to avoid wasting traffic resources. This way of dynamically adjusting traffic signals can effectively improve the efficiency and safety of road traffic and reduce traffic congestion and traffic accidents.

[0044] Acquire date data and time node data of the current date, use the date data and time node data as basic features, perform feature matching of the traffic ratio feature, and generate a ratio matching result;

[0045] The system will obtain the current date and time node information, including date (such as year, month, day), time (such as hours, minutes, seconds) and possible other related data, such as holidays, weather conditions, etc. After obtaining the current date data and time node data, these data will be used as the basis for subsequent feature matching. The system will use the trained model or algorithm to match and predict based on the acquired date and time node data, as well as possible other features such as traffic flow. For example, the system may compare the current time and date with historical data to predict the possible traffic flow ratio. After completing the feature matching, the system will generate a ratio matching result. The ratio matching result refers to the prediction result related to the traffic flow ratio or traffic conditions generated by matching the acquired date data and time node data of the current date with the pre-trained model or algorithm. This ratio matching result can be a relatively accurate prediction of the traffic flow ratio, or a measure of the traffic conditions. This series of processes can effectively predict traffic conditions and optimize the control of traffic lights, thereby improving the efficiency and safety of road traffic.

[0046] Real-time traffic flow collection is performed through surveillance cameras, and compensation feedback data is generated based on the collection results;

[0047] Real-time traffic information is collected through surveillance cameras on the road section, and then the collected data is processed and analyzed using video processing technology and computer vision technology. According to the results of processing and analysis, compensation feedback data can be generated, including information related to traffic flow and road safety performance, such as the number of vehicles, vehicle speed, road occupancy, and traffic events. For example, if the traffic flow on a certain road section is large, the system can automatically extend the green light time of the road section to relieve traffic pressure; if the frequency of traffic events on a certain road section is high, the system can automatically adjust the flashing frequency of traffic lights or increase the number of traffic surveillance cameras to improve road safety performance. These data can be used to adjust traffic signal control strategies in real time, such as adjusting the flashing frequency of traffic lights or extending the green light time, etc., to improve traffic flow and road safety performance. Among them, the compensation feedback data refers to the data generated after processing and analysis of real-time traffic data collected by sensors such as surveillance cameras, which is used to adjust and optimize traffic signal control strategies.

[0048] Determining basic traffic signal control data based on the ratio matching result, and determining a limit adjustment space;

[0049] The basic traffic signal control data is predicted or calculated based on the obtained ratio matching results, including the initial setting time of various traffic lights, such as the length of time of red and green lights, the flashing frequency of yellow lights, etc. In addition, the basic traffic signal control data may also include preset traffic signal control schemes for special situations, such as holidays, bad weather, etc. For example, if the predicted traffic flow ratio is high, the system may automatically extend the green light time of the corresponding road section. The basic traffic signal control data refers to a set of data calculated based on the ratio matching results and other relevant data, which is used to guide the real-time control of traffic lights. When generating basic traffic signal control data, the system will consider multiple factors, such as real-time traffic flow, road conditions, vehicle speed, weather, etc., in order to generate the optimal traffic signal control data. In addition, the system may also predict future traffic conditions based on historical data and models, so as to adjust the basic traffic signal control data in advance. Finally, based on this basic traffic signal control data, the limit adjustment space can be determined. The limit adjustment space usually refers to the minimum and maximum time lengths of traffic lights that can be adjusted under the premise of ensuring traffic safety. For example, for a specific traffic intersection, its limit adjustment space may be the minimum and maximum value of the green light time under the premise of ensuring the safety of vehicles and pedestrians. The method of determining the basic traffic signal control data and determining the limit adjustment space based on the proportional matching results can dynamically adjust the control data of the traffic light according to the current traffic conditions to optimize the traffic flow and improve the traffic efficiency of the road.

[0050] The basic traffic signal control data is optimized according to the compensation feedback data, and the optimization result is constrained by the limit adjustment space to complete the traffic signal control.

[0051] In order to make traffic signal control more reasonable and efficient, the basic traffic signal control data can be optimized according to the compensation feedback data. For example, if the compensation feedback data shows that the traffic flow of a certain section of road is higher than expected, the system may automatically shorten the green light time of the corresponding section of road. Then, the system will further adjust the optimized traffic signal control data according to the limit adjustment space to ensure that the generated traffic signal control data is within an acceptable range. Finally, the system completes the traffic signal control according to the adjusted traffic signal control data. This dynamic adjustment of the traffic signal control strategy can adapt to the ever-changing traffic conditions and improve the traffic efficiency and safety of the road.

[0052] Further, the method for optimizing traffic signal control also includes:

[0053] Set continuous breakthrough threshold;

[0054] When the optimization result continuously triggers the limit adjustment space, and the number of triggers meets the continuous breakthrough threshold, a breakthrough update result is generated;

[0055] Traffic signal control is performed based on the breakthrough update result.

[0056] In order to further optimize the traffic signal control, the system will set a continuous breakthrough threshold to determine whether the optimization result continuously triggers the limit adjustment space. The continuous breakthrough threshold refers to a conditional judgment triggered when the number of times the optimization result continuously triggers the limit adjustment space reaches or exceeds a specific threshold during the traffic signal control optimization process. The setting of this threshold is to balance the sensitivity and stability of the optimization process. If the threshold is set too high, the optimization result may not respond to the change of traffic flow in time, thereby affecting the road traffic efficiency; if the threshold is set too low, the optimization result may be too sensitive and frequently trigger the breakthrough update result, thereby affecting the stability of traffic. The system monitors the changes in the optimization results. If the optimization result continuously triggers the limit adjustment space and the number of triggers reaches the set continuous breakthrough threshold, the system will consider this as a breakthrough situation. Among them, the setting of the continuous breakthrough threshold needs to consider multiple factors, such as the size of traffic flow, the speed of change of traffic flow, the sensitivity of the traffic signal control strategy, etc. Specifically, the continuous breakthrough threshold can be determined by methods such as analysis of historical data and model prediction. When a breach occurs, the system will generate a breach update result, including adjustments to the time settings of traffic lights such as green light time, red light time, and yellow light time, and may also include adjustments to the timing of traffic lights. The breach update result refers to a set of new basic traffic signal control data generated by the system according to the specific situation after the continuous breach threshold is triggered. This new set of basic traffic signal control data is the result of adjustment and optimization on the original basis, and is used to better adapt to current and future traffic conditions. Finally, the system will use this breach update result to control traffic signals, including real-time adjustment of the light timing of traffic lights to better adapt to current traffic conditions. When a breach occurs, the traffic signal control strategy can be adjusted and optimized in a timely manner to ensure the efficiency and safety of road traffic. At the same time, by setting the continuous breach threshold, it is possible to avoid adjusting the traffic signal control strategy too frequently, thereby maintaining the stability and continuity of traffic.

[0057] Further, such as Figure 2 As shown, the traffic signal is dynamically regulated by predicting the duration, and the method further includes:

[0058] Establish signal communication between adjacent signal lights;

[0059] When any traffic light performs dynamic regulation, the triggering direction of the corresponding traffic light is obtained;

[0060] Locating adjacent traffic lights based on the triggered travel direction, and calculating predicted duration based on adjacent distances;

[0061] The traffic signal is dynamically regulated by the predicted duration.

[0062] Optionally, traffic flow can be optimized by dynamically adjusting traffic signals, which can be achieved by establishing signal communication between adjacent traffic lights. First, the system can use wireless communication technologies such as Bluetooth, WiFi, etc., or use other traffic signal control systems, such as traffic signal control centers, to achieve signal communication between adjacent traffic lights. When any traffic light needs to be dynamically adjusted, the system will calculate the triggering direction of the corresponding traffic light based on the collected data and the preset algorithm model. The triggering direction is calculated based on factors such as real-time traffic flow, traffic speed, road conditions, etc., and is used to guide the regulation of traffic lights. For example, when the traffic flow in a certain direction is large, the green light time in that direction can be automatically extended to relieve traffic pressure; on the contrary, when the traffic flow in a certain direction is small, the green light time in that direction can be appropriately shortened to reduce the vacant time in the direction and effectively utilize road resources.

[0063] Based on the triggered travel direction, the system can locate adjacent traffic lights, which are traffic lights adjacent to the current traffic light, and may be adjacent horizontally, vertically, or diagonally. Then, the system calculates the travel time (i.e., predicted duration) between the current traffic light and the adjacent traffic light based on the adjacent distance. Finally, through the predicted duration, the system can dynamically adjust traffic signals, including adjustments to traffic light time settings such as green light time, red light time, and yellow light time, and may also include adjustments to traffic light timing, etc. This method can help the system better understand the traffic flow conditions of adjacent traffic lights, thereby better adjusting traffic signals to improve road traffic efficiency and safety.

[0064] Furthermore, the method further comprises:

[0065] Reading the vehicle data of the adjacent direction of the triggering passing direction from the adjacent signal light, and if the reading result is empty, the vehicle is determined to have passed;

[0066] Reading pedestrian data in the adjacent direction of the triggering passing direction from the adjacent signal light, if the reading result is empty, the pedestrian is determined to have passed;

[0067] When both the vehicle determination and the pedestrian determination are passed, the dynamic regulation of the traffic signal is performed according to the predicted duration.

[0068] Preferably, according to the calculated triggering direction of passage, the system reads the traffic data of vehicles and pedestrians of adjacent traffic lights and makes a judgment. If the read vehicle data is an empty result, it means that the vehicle in the adjacent direction has passed, and the vehicle can be judged to have passed. If the read pedestrian data is an empty result, it means that the pedestrian in the adjacent direction has passed, and the pedestrian can be judged to have passed. When both the vehicle judgment and the pedestrian judgment are passed, it means that the traffic in this direction has passed safely. At this time, the dynamic regulation of traffic signals can be performed according to the predicted duration. The scheme of dynamic regulation of traffic signals based on factors such as real-time traffic flow and traffic speed can effectively improve the traffic efficiency and safety of roads.

[0069] Further, the method for updating the traffic data set includes:

[0070] Generate result control evaluation based on traffic signal control results;

[0071] Generate feature update data through the result control evaluation;

[0072] The traffic data set is continuously updated according to the feature update data.

[0073] Traffic conditions are complex and changeable. In order to help the system better understand and grasp the current traffic conditions, the traffic data set also needs to be updated. Specifically, the system uses a preset algorithm model to calculate the traffic signal control results based on multiple factors such as real-time traffic flow, traffic speed, road conditions, etc. The traffic signal control results include adjustments to traffic signal time settings such as green light time, red light time, and yellow light time, and may also include adjustments to traffic signal timing. Then, the system will calculate some evaluation indicators based on the traffic signal control results, including green light utilization rate, average vehicle waiting time, average vehicle speed, etc., to evaluate the effect of traffic signal control results. The result control evaluation is used to evaluate the rationality and effectiveness of traffic signal control results. The system will compare the calculated evaluation indicators with the preset standards. Among them, the preset standards are established based on traffic engineering theory, empirical data or expert opinions, etc., to determine whether the traffic signal control results are reasonable and effective. According to the comparison results, the system will generate corresponding result control evaluation. If the evaluation index meets the preset standard and the traffic signal control result effectively improves the road traffic efficiency, then the evaluation result may be positive; if the evaluation index does not meet the preset standard, or the traffic signal control result does not significantly improve the road traffic efficiency, then the evaluation result may be negative. For example, it can be evaluated whether the adjusted traffic signal light timing conforms to the changes in traffic flow, whether it improves the road traffic efficiency, etc. According to the generated result control evaluation, the system can perform self-feedback and adjustment. If the evaluation result is negative, the system can automatically adjust the traffic signal control strategy, such as adjusting the green light time, red light time, etc., so as to better adapt to the actual traffic conditions in the signal control of the next time period. Through the feedback result control evaluation, it can help the system better understand and grasp the current traffic conditions, so as to carry out more accurate and efficient traffic signal control and improve the road traffic efficiency and safety.

[0074] After the result control evaluation is completed, the system will generate feature update data according to the evaluation results. The feature update data refers to the update or adjustment of the original feature vector according to the result control evaluation, including the modification, addition or deletion of the feature vector. Finally, the system will continue to update the traffic data set based on the feature update data, including the update of traffic flow data and the update of vehicle speed data, so as to better reflect the current traffic conditions. In this way, the system can continuously adapt to and optimize changes in traffic conditions, improve the accuracy and efficiency of traffic signal control, and thus improve the traffic efficiency and safety of the road. At the same time, these updated data will also be used for the calculation and evaluation of the next traffic signal control results, thus forming a closed-loop process. In other words, the system can continuously optimize and improve its own algorithms and models through the feedback mechanism to better adapt to changes in traffic conditions.

[0075] In summary, the embodiments of the present application have at least the following technical effects:

[0076] The present application first collects historical road traffic data, and performs noise reduction and time series compensation processing to obtain a traffic data set. Then, independent date features and normal date features are constructed, and the traffic data set is segmented using the independent date features and normal date features, and multi-period clustering of deep learning is performed through the data segmentation results to generate the traffic ratio features. Next, the date data and time node data of the current date are obtained, and the date data and node data are used as basic features to perform feature matching and generate a ratio matching result. Real-time traffic collection is then collected through surveillance cameras, and compensation feedback data is generated based on the collection results. Then, the basic traffic signal control data is determined based on the ratio matching results, and the limit adjustment space is determined. Finally, the basic traffic signal control data is optimized based on the compensation feedback data, and the optimization result constraints are performed through the limit adjustment space to complete the traffic signal control. The above method solves the technical problem that it is difficult to fully perceive and predict traffic flow in the prior art, and achieves the technical effect of accurate control based on deep learning and improving the accuracy and efficiency of traffic signal control.

[0077] Embodiment 2

[0078] Based on the same inventive concept as the traffic signal control method based on deep learning in the aforementioned embodiment, Figure 3 As shown, the present application provides a traffic signal control system based on deep learning, and the system and method embodiments in the present application are based on the same inventive concept. The system includes:

[0079] Data set module 11, feature module 12, matching module 13, compensation feedback module 14, determination module 15, optimization module 16.

[0080] A data set module 11, wherein the data set module 11 is used to establish a traffic data set, wherein the traffic data set is obtained by collecting historical road traffic data, wherein the traffic data set is a preprocessed data set, and the preprocessing method includes noise reduction processing and time series compensation processing;

[0081] A feature module 12, wherein the feature module 12 is used to construct independent date features and normal date features, perform data segmentation of the traffic data set through the independent date features and the normal date features, and perform multi-period clustering of deep learning through the data segmentation results to generate traffic ratio features of the intersection;

[0082] A matching module 13, the matching module 13 is used to obtain date data and time node data of the current date, and use the date data and node data as basic features to perform feature matching of the traffic ratio feature to generate a ratio matching result;

[0083] A compensation feedback module 14, which is used to collect real-time traffic flow through a monitoring camera and generate compensation feedback data based on the collected results;

[0084] A determination module 15, the determination module 15 is used to determine basic traffic signal control data based on the ratio matching result, and determine a limit adjustment space;

[0085] The optimization module 16 is used to optimize the basic traffic signal control data according to the compensation feedback data, and constrain the optimization result through the limit adjustment space to complete the traffic signal control.

[0086] Furthermore, the feature module 12 is used to execute the following method:

[0087] Obtaining a non-peak clustering result, and generating a non-peak time interval based on the non-peak clustering result;

[0088] When the time node data of the corresponding date is within the idle time interval, a dynamic determination instruction is generated;

[0089] Execute real-time image acquisition through the dynamic determination instruction to generate real-time longitudinal and transverse vehicle flow information;

[0090] When there are vehicles in any direction of traffic and the adjacent traffic is empty, the dynamic judgment is passed;

[0091] Traffic signals are dynamically adjusted based on the determination results.

[0092] Furthermore, the compensation feedback module 14 is used to perform the following method:

[0093] Establish signal communication between adjacent signal lights;

[0094] When any traffic light performs dynamic regulation, the triggering direction of the corresponding traffic light is obtained;

[0095] Locating adjacent traffic lights based on the triggered travel direction, and calculating predicted duration based on adjacent distances;

[0096] The traffic signal is dynamically regulated by the predicted duration.

[0097] Furthermore, the compensation feedback module 14 is used to perform the following method:

[0098] Reading the vehicle data of the adjacent direction of the triggering passing direction from the adjacent signal light, and if the reading result is empty, the vehicle is determined to have passed;

[0099] Reading pedestrian data in the adjacent direction of the triggering passing direction from the adjacent signal light, if the reading result is empty, the pedestrian is determined to have passed;

[0100] When both the vehicle determination and the pedestrian determination are passed, the dynamic regulation of the traffic signal is performed according to the predicted duration.

[0101] Furthermore, the determining module 15 is used to execute the following method:

[0102] Set continuous breakthrough threshold;

[0103] When the optimization result continuously triggers the limit adjustment space, and the number of triggers meets the continuous breakthrough threshold, a breakthrough update result is generated;

[0104] Traffic signal control is performed based on the breakthrough update result.

[0105] Furthermore, the optimization module 16 is used to perform the following method:

[0106] Generate result control evaluation based on traffic signal control results;

[0107] Generate feature update data through the result control evaluation;

[0108] The traffic data set is continuously updated according to the feature update data.

[0109] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. Other embodiments are within the scope of the attached claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0110] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

[0111] This specification and drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. A traffic signal control method based on deep learning, characterized in that: The method comprises: Establishing a traffic data set, wherein the traffic data set is obtained by collecting historical road traffic data, wherein the traffic data set is a preprocessed data set, and the preprocessing method includes noise reduction processing and time series compensation processing; Constructing independent date features and normal date features, segmenting the traffic data set by using the independent date features and the normal date features, and completing multi-period clustering of deep learning by using the data segmentation results to generate traffic ratio features of the intersection; Acquire date data and time node data of the current date, use the date data and time node data as basic features, perform feature matching of the traffic ratio feature, and generate a ratio matching result; Real-time traffic flow collection is performed through surveillance cameras, and compensation feedback data is generated based on the collection results; Determining basic traffic signal control data based on the ratio matching result, and determining a limit adjustment space; The basic traffic signal control data is optimized according to the compensation feedback data, and the optimization result is constrained by the limit adjustment space to complete the traffic signal control.

2. The method according to claim 1, characterized in that The method further comprises: Obtaining a non-peak clustering result, and generating a non-peak time interval based on the non-peak clustering result; When the time node data of the corresponding date is within the idle time interval, a dynamic determination instruction is generated; Execute real-time image acquisition through the dynamic determination instruction to generate real-time longitudinal and transverse vehicle flow information; When there are vehicles in any direction of traffic and the adjacent traffic is empty, the dynamic judgment is passed; Traffic signals are dynamically adjusted based on the determination results.

3. The method according to claim 1, characterized in that The method further comprises: Establish signal communication between adjacent signal lights; When any traffic light performs dynamic regulation, the triggering direction of the corresponding traffic light is obtained; Locating adjacent traffic lights based on the triggered travel direction, and calculating predicted duration based on adjacent distances; The traffic signal is dynamically regulated by the predicted duration.

4. The method according to claim 3, characterized in that The method further comprises: Reading the vehicle data of the adjacent direction of the triggering passing direction from the adjacent signal light, and if the reading result is empty, the vehicle is determined to have passed; Reading pedestrian data in the adjacent direction of the triggering passing direction from the adjacent signal light, if the reading result is empty, the pedestrian is determined to have passed; When both the vehicle determination and the pedestrian determination are passed, the dynamic regulation of the traffic signal is performed according to the predicted duration.

5. The method according to claim 1, characterized in that The method further comprises: Set continuous breakthrough threshold; When the optimization result continuously triggers the limit adjustment space, and the number of triggers meets the continuous breakthrough threshold, a breakthrough update result is generated; Traffic signal control is performed based on the breakthrough update result.

6. The method according to claim 1, characterized in that The method further comprises: Generate result control evaluation based on traffic signal control results; Generate feature update data through the result control evaluation; The traffic data set is continuously updated according to the feature update data.

7. A traffic signal control system based on deep learning, characterized in that: The system comprises: A data set module, wherein the data set module is used to establish a traffic data set, wherein the traffic data set is obtained by collecting historical road traffic data, wherein the traffic data set is a preprocessed data set, and the preprocessing method includes noise reduction processing and time series compensation processing; A feature module, wherein the feature module is used to construct independent date features and normal date features, perform data segmentation of the traffic data set through the independent date features and the normal date features, and perform multi-period clustering of deep learning through the data segmentation results to generate traffic ratio features of the intersection; A matching module, the matching module is used to obtain date data and time node data of the current date, and use the date data and node data as basic features to perform feature matching of the traffic ratio feature to generate a ratio matching result; A compensation feedback module, which is used to collect real-time traffic flow through a monitoring camera and generate compensation feedback data based on the collected results; A determination module, the determination module is used to determine basic traffic signal control data based on the ratio matching result, and determine a limit adjustment space; An optimization module is used to optimize the basic traffic signal control data according to the compensation feedback data, and to constrain the optimization result through the limit adjustment space to complete the traffic signal control.

Citation Information

Cited By

  • Safety facility system suitable for highway grade crossing with large traffic volume

    CN119992829A

  • A safety system for high-volume highway flat intersections

    CN119992829B