Traffic signal monitoring system and method

Abstract

A traffic signal monitoring system and method belong to the field of intelligent traffic, and the system comprises a fusion processing module, an anomaly detection module, a control optimization module, a feedback optimization module and an anomaly recording module. According to the invention, the depth map neural network is adopted to carry out structured modeling on the intersection traffic map, the intersection feature information is dynamically extracted, the traffic jam index is calculated in real time, the accurate analysis and prediction of the traffic flow dynamic state are realized, and the intelligent level of the control strategy is improved. The motion trails of pedestrians, non-motor vehicles and motor vehicles in the video stream are analyzed based on a time sequence behavior recognition network, traffic violation behaviors such as red light running, retrogradation and road occupation are automatically recognized, and punishment suggestions are automatically generated in combination with a license plate recognition result. The current traffic state and the historical traffic mode database are utilized, the optimal signal timing strategy in different scenes is explored independently in combination with the deep reinforcement learning algorithm, real-time dynamic adjustment of the multi-phase signal lamp is achieved, and the road traffic efficiency is improved.

Description

Technical Field

[0001] This invention belongs to the field of intelligent transportation technology, and more specifically, relates to a traffic signal monitoring system and method. Background Technology

[0002] With the continuous growth of urban motor vehicle ownership, traffic pressure is increasing day by day. Traditional traffic signal control systems mainly rely on fixed cycles or inductive control methods, which lack the ability to perceive and respond to real-time traffic conditions. They are difficult to adapt to the complex and ever-changing urban traffic environment and are prone to problems such as traffic congestion and low traffic efficiency.

[0003] On the other hand, existing traffic violation monitoring systems mostly rely on rule-driven or manual review methods. They typically collect video from fixed cameras, then rely on backend systems for license plate recognition and violation assessment. This approach is inefficient, prone to misjudgments, and fails to meet the real-time monitoring and enforcement needs of large-scale urban road networks. Furthermore, current traffic signal control systems lack effective coordination mechanisms, often only optimizing at individual intersections and failing to achieve overall signal coordination and control across intersections or regions. They also cannot quickly adjust to emergencies such as traffic accidents or emergency lane congestion. With the development of artificial intelligence technologies such as deep learning, graph neural networks, behavior recognition, reinforcement learning, and edge computing, how to integrate multi-source traffic data to build efficient traffic perception and decision-making models, and achieve accurate identification of violations, dynamic scheduling of traffic signals, and intelligent coordination between intersections, has become a critical technical problem that urgently needs to be solved in the field of intelligent transportation. Summary of the Invention

[0004] The purpose of this section is to outline some aspects of embodiments of the present invention and to briefly describe some preferred embodiments. Simplifications or omissions may be made in this section, as well as in the abstract and title of this application, to avoid obscuring the purpose of these documents; however, such simplifications or omissions should not be construed as limiting the scope of the invention.

[0005] In view of the problems of the above-mentioned or existing traffic signal monitoring systems and methods, the present invention is proposed.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: This invention provides a traffic signal monitoring system, including: a fusion processing module, used to perform time synchronization and spatial alignment based on collected multi-source traffic information, construct a unified traffic state vector, and use a deep learning-based graph neural network model to extract features of the current intersection state to obtain a real-time traffic congestion index; The anomaly detection module is used to analyze the trajectories of pedestrians and vehicles in the video stream using a behavior recognition model to detect traffic violations. The control optimization module is used to dynamically adjust the green light time ratio of each phase based on the traffic congestion index and historical traffic pattern database. It uses reinforcement learning algorithm to autonomously learn the optimal signal control strategy under different traffic scenarios to achieve coordinated control between multiple intersections. The feedback optimization module is used to upload monitoring results and signal adjustment strategies to the cloud traffic management platform, and at the same time send early warning information to nearby driver terminals and intelligent traffic signs through edge nodes; The anomaly recording module is used to structurally store image evidence of abnormal behavior and automatically associate it with license plate recognition results to generate penalty suggestions.

[0007] As a preferred embodiment of the traffic signal monitoring system of the present invention, the system comprises: performing time synchronization and spatial alignment based on collected multi-source traffic information to construct a unified traffic state vector, including: Synchronization and alignment of data streams from various sensors are performed based on a unified timestamp system; when there is frame rate inconsistency or data delay, interpolation or nearest neighbor matching strategies are used to compensate for the time of low-frequency data; the data coordinates of various sensors are uniformly mapped to the geographic coordinate system or the local coordinate system of traffic intersections, and spatial coordinate transformation is performed using the external parameter calibration parameters between sensors to ensure the consistency of the spatial location of multi-source targets. For spatially overlapping targets, a fusion matching process is performed. By combining Kalman filtering with the Hungarian algorithm, the same target entity perceived by different sensors is identified. The ID consistency of the targets in the matching results is maintained to construct a stable cross-frame target trajectory. The state vectors of each target are mapped to the grid state map of the entire traffic intersection to form a matrix tensor representing the overall traffic state of the intersection.

[0008] As a preferred embodiment of the traffic signal monitoring system of the present invention, the system comprises: performing time synchronization and spatial alignment based on collected multi-source traffic information to construct a unified traffic state vector, including: Synchronization and alignment of data streams from various sensors are performed based on a unified timestamp system; when there is frame rate inconsistency or data delay, interpolation or nearest neighbor matching strategies are used to compensate for the time of low-frequency data; the data coordinates of various sensors are uniformly mapped to the geographic coordinate system or the local coordinate system of traffic intersections, and spatial coordinate transformation is performed using the external parameter calibration parameters between sensors to ensure the consistency of the spatial location of multi-source targets. For spatially overlapping targets, a fusion matching process is performed. By combining Kalman filtering with the Hungarian algorithm, the same target entity perceived by different sensors is identified. The ID consistency of the targets in the matching results is maintained to construct a stable cross-frame target trajectory. The state vectors of each target are mapped to the grid state map of the entire traffic intersection to form a matrix tensor representing the overall traffic state of the intersection.

[0009] As a preferred embodiment of the traffic signal monitoring system of the present invention, the system utilizes a behavior recognition model to analyze the trajectories of pedestrians and vehicles in the video stream to detect traffic violations, including: Based on actual intersection maps or image segmentation models, semantic regions are divided; each semantic region is represented by a polygon mask, and the target trajectory points are judged to determine the current target's location. The traffic light is currently red. During the red light duration t∈[t1,t2], the vehicle trajectory crosses the stop line area: ∃t∈[t1,t2],(x t ,y t )∈R before the stop line, and (x t+1 ,y t+1 )∈R 路口核心区 This was determined to be a red light violation. Define the legal passage direction vector Calculate the target trajectory direction vector If the included angle between the two satisfies: This can be interpreted as the target exhibiting reverse behavior in the current area; When the red light is on, if the vehicle remains stationary for more than the set time and the center of gravity of the detection frame crosses the center line of the stop line area, it can be considered as parking beyond the stop line. When a pedestrian is detected moving from pedestrian zone R 人行道 Entering the motor vehicle lane area R 车道 And the trajectory did not pass through the zebra crossing area R. 斑马线 If so, it will be marked as an illegal crossing.

[0010] As a preferred embodiment of the traffic signal monitoring system of the present invention, the system dynamically adjusts the green light time ratio of each phase based on a traffic congestion index and a historical traffic pattern database, and autonomously learns the optimal signal control strategy under different traffic scenarios using a reinforcement learning algorithm to achieve coordinated control among multiple intersections, including: Define action A t The proposed adjustment scheme for the green light duration allocation for each phase within the current cycle is as follows: Let the number of phases be k, and the current action be: A t =[g1,g2,...,g k ] Among them, g i The green light duration for the i-th phase; Design a composite reward function R for traffic optimization. t Taking into account traffic efficiency, congestion level, and delay time: Among them, Q i (t) represents the number of vehicles queuing in the i-th direction, W i (t) represents the average waiting time of vehicles passing through per unit time, CI avg (t) represents the average congestion index of the current intersection, where α, β, and γ are weighting coefficients; The optimal timing policy is trained using a reinforcement learning algorithm, and the process is as follows: Initialize the Q-network or policy network π. θ Experience replay pool D, in the simulation environment, input state S t Select action A t Get reward R t With the next state S t+1 , will (S t A t ,R t ,S t+1 Store the samples in the experience pool, sample a small batch of samples from the experience pool, and update the policy parameters θ or the Q function.

[0011] As a preferred embodiment of the traffic signal monitoring system of the present invention, the system uploads monitoring results and signal adjustment strategies to a cloud-based traffic management platform, and simultaneously issues early warning information to nearby driver terminals and intelligent traffic signs via edge nodes, including: After receiving the local traffic monitoring results, the edge computing nodes deployed at the intersection determine whether there are any events or abnormal situations that require warning. The edge nodes send warning information to the on-board terminals of nearby vehicles via V2X. The information includes the current congestion situation at the intersection, the remaining time of the traffic light, warnings of dangerous behaviors, and suggested detours.

[0012] As a preferred embodiment of the traffic signal monitoring system of the present invention, the system includes: structured storage of image evidence of abnormal behavior, automatic association with license plate recognition results, and generation of penalty suggestions, including: The system performs license plate recognition on vehicles detected in illegal images, uses a deep learning model to locate the license plate position, and identifies character sequences through an OCR network with convolution and attention mechanisms. Results with low confidence (<0.8) are discarded, and multiple candidate results are supported for comparison.

[0013] A traffic signal monitoring method includes: performing time synchronization and spatial alignment based on collected multi-source traffic information, constructing a unified traffic state vector, and using a deep learning-based graph neural network model to extract features of the current intersection state to obtain a real-time traffic congestion index. By using behavior recognition models to analyze the trajectories of pedestrians and vehicles in video streams, traffic violations can be detected. Based on the traffic congestion index and historical traffic pattern database, the green light time ratio of each phase is dynamically adjusted, and the reinforcement learning algorithm is used to autonomously learn the optimal signal control strategy under different traffic scenarios to achieve coordinated control between multiple intersections. The monitoring results and signal adjustment strategies are uploaded to the cloud-based traffic management platform, and warning information is sent to nearby driver terminals and intelligent traffic signs through edge nodes. The system stores image evidence of abnormal behavior in a structured manner and automatically associates it with license plate recognition results to generate penalty suggestions.

[0014] A computing device, the computing device comprising: At least one processor, memory, and input / output unit; The memory is used to store computer programs, and the processor is used to call the computer programs stored in the memory to execute the steps of the traffic signal monitoring system.

[0015] A computer-readable storage medium includes instructions that, when executed on a computer, cause the computer to perform steps of a traffic signal monitoring system.

[0016] The beneficial effects of this invention are as follows: This invention employs a deep graph neural network to perform structured modeling of intersection traffic maps, dynamically extracts intersection feature information, and calculates the traffic congestion index in real time, achieving accurate analysis and prediction of traffic flow dynamics and improving the intelligence level of control strategies. Based on temporal behavior recognition network analysis of the movement trajectories of pedestrians, non-motorized vehicles, and motorized vehicles in video streams, it automatically identifies traffic violations such as running red lights, driving against traffic, and occupying lanes, and automatically generates penalty suggestions based on license plate recognition results, achieving contactless and intelligent traffic enforcement. Utilizing a database of current traffic conditions and historical traffic patterns, combined with deep reinforcement learning algorithms, it autonomously explores the optimal signal timing strategies under different scenarios, achieving real-time dynamic adjustment of multi-phase traffic lights, improving road traffic efficiency, and alleviating traffic congestion. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of a traffic signal monitoring system provided in an embodiment of the present invention.

[0019] Figure 2 This is a flowchart of a traffic signal monitoring method provided in an embodiment of the present invention.

[0020] Figure 3 A schematic diagram of the structure of a medium according to an embodiment of the present invention is shown.

[0021] Figure 4 A schematic diagram of the structure of a computing device according to an embodiment of the present invention is shown.

[0022] In the accompanying drawings, the same or corresponding reference numerals indicate the same or corresponding parts. Detailed Implementation

[0023] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0024] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0025] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0026] Example The following is for reference. Figures 1-4 , Figure 2 This is a flowchart illustrating a traffic signal monitoring method according to an embodiment of the present invention. It should be noted that the embodiments of the present invention can be applied to any applicable scenario.

[0027] Figure 2 The flowchart of a traffic signal monitoring method provided in an embodiment of the present invention, shown below, includes: S1: Based on the collected multi-source traffic information, time synchronization and spatial alignment are performed to construct a unified traffic state vector. A graph neural network model based on deep learning is used to extract features of the current intersection state to obtain the real-time traffic congestion index.

[0028] Preferably, a unified timestamp system is used to synchronize and align data streams from various sensors; when there is inconsistent frame rate or data delay, interpolation or nearest neighbor matching strategies are used to compensate for the time of low-frequency data; the data coordinates of various sensors are uniformly mapped to the geographic coordinate system or the local coordinate system of traffic intersections, and spatial coordinate transformation is performed using the external parameter calibration parameters between sensors to ensure the consistency of the spatial location of multi-source targets. For spatially overlapping targets, a fusion matching process is performed. By combining Kalman filtering with the Hungarian algorithm, the same target entity perceived by different sensors is identified. The ID consistency of the targets in the matching results is maintained to construct a stable cross-frame target trajectory. The state vectors of each target are mapped to the grid state map of the entire traffic intersection to form a matrix tensor representing the overall traffic state of the intersection.

[0029] Preferably, the target traffic intersection is divided into several grid areas or traffic units, with each area serving as a node in the graph; the connection relationship between nodes is set according to the actual road topology and traffic flow direction to form a directed graph, and an initial feature vector is assigned to each node; The intersection embedding vector is input into a trained regression neural network or decision function, and the output is a real-valued traffic congestion index or congestion level.

[0030] Furthermore, all sensors are connected to a unified synchronous clock system during data acquisition, assigning a unified timestamp T to each frame of sensed data. i If there is frame rate inconsistency or data arrival delay, linear interpolation or nearest neighbor interpolation is used to compensate for low-frequency data streams so that time-consistent data windows can be obtained during target fusion. A local two-dimensional ground coordinate system is established with the traffic intersection as the reference coordinate origin; the target points in the local detection coordinate system are mapped to a unified coordinate system by using the external parameter matrix of each type of sensor when it is deployed on site. For the set of targets sensed by all sensors under a unified time window, a target cost matrix is ​​constructed, and the cost of each item is estimated based on spatial distance (such as Euclidean distance) or Bayesian probability. The Hungarian algorithm is used to find the minimum matching path in the cost matrix to obtain the one-to-one correspondence between targets. The Kalman filter is used to perform trajectory fusion and prediction update for the matched targets to handle the noise difference problem of multi-source data and form a single target state estimation vector. Each target is assigned a globally unique ID, and a trajectory maintenance list is established. If a target fails to match continuously for more than a set number of frames, it is marked as "disappeared." If a new target continues to fail to match, it is marked as "added." The temporal trajectory of the target is maintained across frames, recording its six-dimensional state vector (x, y, vx, vy, ax, ay). The traffic intersection is divided into several grid cells, and each cell records the statistical value of the targets falling into it. A traffic state tensor with dimensions C×H×W is constructed, where C represents the state channel. The tensor is used as input to the subsequent graph neural network to reflect the traffic scene structure and state distribution in real time.

[0031] Furthermore, standard graph convolutional neural networks or graph attention networks are used for graph-level representation learning; After propagation through several GCN layers, the high-dimensional embedding vector h of each node is obtained. iAggregate all nodes in the entire map or target area to form an overall intersection state representation H; input the aggregated intersection embedding vector H into the trained regression neural network model f(⋅) to output the corresponding congestion index or congestion level.

[0032] S2: Utilize behavior recognition models to analyze the trajectories of pedestrians and vehicles in video streams to detect traffic violations.

[0033] Preferably, semantic regions are divided based on actual intersection maps or image segmentation models; each semantic region is represented by a polygon mask, and the target trajectory points are judged to determine the current target's location. The traffic light is currently red. During the red light duration t∈[t1,t2], the vehicle trajectory crosses the stop line area: ∃t∈[t1,t2],(x t ,y t )∈R before the stop line, and (x t+1 ,y t+1 )∈R 路口核心区 This was determined to be a red light violation. Define the legal passage direction vector Calculate the target trajectory direction vector If the included angle between the two satisfies: This can be interpreted as the target exhibiting reverse behavior in the current area; When the red light is on, if the vehicle remains stationary for more than the set time and the center of gravity of the detection frame crosses the center line of the stop line area, it can be considered as parking beyond the stop line. When a pedestrian is detected moving from pedestrian zone R 人行道 Entering the motor vehicle lane area R 车道 And the trajectory did not pass through the zebra crossing area R. 斑马线 If so, it will be marked as an illegal crossing.

[0034] Furthermore, after training with the deep learning segmentation model DeepLabv3+, the following features were identified in the image: Area R before the stop line: The coordinates of the polygon mask boundary points are [(580, 850), (800, 850), (800, 880), (580, 880)]; Core area R of the intersection: The mask coordinates are [(500, 600), (1400, 600), (1400, 800), (500, 800)]; The system uses YOLOv5 + DeepSORT to identify and track the target vehicle, obtaining the target trajectory point sequence for vehicle number Vehicle_027: { "VehicleID": "Vehicle_027", "Trajectory Point": [ {"t": "17:48:14.120", "x": 640, "y": 870}, {"t": "17:48:14.160", "x": 642, "y": 865}, {"t": "17:48:14.200", "x": 646, "y": 860}, {"t": "17:48:14.240", "x": 655, "y": 845}, {"t": "17:48:14.280", "x": 670, "y": 830} ] } Analysis determines that point 1 (640, 870) ∈ R (stop line), and point 5 (670, 830) ∈ R (core area). And the time "17:48:14.120" ∈ [17:48:12.000, 17:48:32.000] (red light status). The logical conditions for running a red light are met.

[0035] S3: Based on the traffic congestion index and historical traffic pattern database, the green light time ratio of each phase is dynamically adjusted. The reinforcement learning algorithm is used to autonomously learn the optimal signal control strategy under different traffic scenarios to achieve coordinated control between multiple intersections.

[0036] Preferably, define action A t The proposed adjustment scheme for the green light duration allocation for each phase within the current cycle is as follows: Let the number of phases be k, and the current action be: A t =[g1,g2,...,g k ] Among them, g i The green light duration for the i-th phase; Design a composite reward function R for traffic optimization. t Taking into account traffic efficiency, congestion level, and delay time: Among them, Qi (t) represents the number of vehicles queuing in the i-th direction, W i (t) represents the average waiting time of vehicles passing through per unit time, CI avg (t) represents the average congestion index of the current intersection, where α, β, and γ are weighting coefficients; The optimal timing policy is trained using a reinforcement learning algorithm, and the process is as follows: Initialize the Q-network or policy network π. θ Experience replay pool D, in the simulation environment, input state S t Select action A t Get reward R t With the next state S t+1 , will (S t A t ,R t ,S t+1 Store the samples in the experience pool, sample a small batch of samples from the experience pool, and update the policy parameters θ or the Q function.

[0037] Furthermore, video detectors are deployed at each entrance lane to track vehicle traffic volume per unit time. Virtual induction coils are set up to monitor the number of vehicles queuing and the average waiting time in real time. The time range is weekday morning peak hours (7:00–9:00), with a data collection frequency of once per minute. The collected data is the queue length Q in each direction. i (t), waiting time W i (t), and the congestion index CI calculated by the traffic simulator. avg (t), using SUMO to build a simulated road network, and using real collected data to construct traffic flow input; for episode in range(N): Reset the environment and obtain the initial state S_t. while the period has not been terminated: Select action A_t using strategy πθ The environment performs action A_t, returns reward R_t, and the next state S_{t+1}. Store (S_t, A_t, R_t, S_{t+1}) into the experience pool D. If the update conditions are met: Sample from D Update the policy network θ using the PPO loss function. Update state: S_t ← S_{t+1} Training effect: The green light timing optimization mechanism based on continuous action space avoids the limitations of discrete action constraints.

[0038] S4: Upload the monitoring results and signal adjustment strategies to the cloud-based traffic management platform, and simultaneously send early warning information to nearby driver terminals and intelligent traffic signs through edge nodes.

[0039] Preferably, after receiving the local traffic monitoring results, the edge computing node deployed at the intersection determines whether there are any events or abnormal situations that require warning. The edge node sends the warning information to the vehicle terminals of nearby vehicles via V2X. The information includes the current congestion situation at the intersection, the remaining time of the traffic light, warnings of dangerous behaviors, and suggested detours.

[0040] Furthermore, by deploying image semantic segmentation models locally on edge computing nodes, accurate division of zebra crossings, stop lines, and core areas can be achieved; sliding window judgment is adopted for vehicle / pedestrian trajectories to improve the accuracy and stability of abnormal event judgment; the congestion index is calculated by comprehensively considering three factors: queue length change rate, average vehicle speed, and throughput rate. Once the congestion index exceeds the threshold of >0.75, V2X broadcast suggestions for detours are triggered.

[0041] Edge nodes read traffic light change times in real time and combine them with trajectory prediction models to predict whether a vehicle will run a red light. The system integrates "light countdown + current traffic behavior + suggested speed" to create precise in-vehicle alerts. Local event inference and zero uplink latency enable microsecond-level response scenarios. All anomaly detection and judgment are completed locally on the edge nodes, eliminating the need for cloud access. A multi-threaded asynchronous broadcast mechanism keeps information latency within 40ms, meeting the requirements for autonomous driving response.

[0042] S5: Structure and store image evidence of abnormal behavior, automatically associate it with license plate recognition results, and generate penalty suggestions.

[0043] Preferably, license plate recognition is performed on vehicles detected in illegal images. A deep learning model is used to locate the license plate position, and a convolutional + attention mechanism OCR network is used to recognize character sequences. Results with low confidence (<0.8) are discarded, and comparison of multiple candidate results is supported.

[0044] Furthermore, illegal image frames are acquired from the illegal event detection module, and all vehicle location boxes are marked using a YOLOv8 or PP-YOLOv2 vehicle detection model. For each vehicle box, an image sub-region is extracted and sent to the next processing step. For the images within the vehicle boxes, a specially trained license plate detection model is used to locate the license plate. The model outputs the license plate region coordinates (x, y, w, h), such as: { "plate_box": [512, 388, 128, 36], "confidence": 0.94 } The input image is fed into an OCR network with the following structure: Input image → CNN encoder (ResNet) → Feature sequence ↓ Self-attention layer (Transformer Encoder) ↓ Seq2Seq decoder (LSTM or Transformer Decoder) ↓ Output character sequence If the confidence score of the output character sequence is less than 0.80, the recognition result is discarded. The OCR network supports outputting Top-k candidate results, which are then matched and compared with the vehicle passage records for the violation time period. If the Top-1 match is successful, the record is valid; if the Top-1 match is invalid, try matching Top-2 / Top-3. If all matches fail, the violation record will be marked as "pending manual review".

[0045] After introducing the method of exemplary embodiments of the present invention, the following references are made. Figure 1 An exemplary embodiment of the present invention provides a traffic signal monitoring system, which includes: The fusion processing module is used to perform time synchronization and spatial alignment based on the collected multi-source traffic information, construct a unified traffic state vector, and use a deep learning-based graph neural network model to extract features of the current intersection state to obtain the real-time traffic congestion index. The anomaly detection module is used to analyze the trajectories of pedestrians and vehicles in the video stream using a behavior recognition model to detect traffic violations. The control optimization module is used to dynamically adjust the green light time ratio of each phase based on the traffic congestion index and historical traffic pattern database. It uses reinforcement learning algorithm to autonomously learn the optimal signal control strategy under different traffic scenarios to achieve coordinated control between multiple intersections. The feedback optimization module is used to upload monitoring results and signal adjustment strategies to the cloud traffic management platform, and at the same time send early warning information to nearby driver terminals and intelligent traffic signs through edge nodes; The anomaly recording module is used to structurally store image evidence of abnormal behavior and automatically associate it with license plate recognition results to generate penalty suggestions.

[0046] After introducing the methods and systems of exemplary embodiments of the present invention, the following references are made. Figure 3 A computer-readable storage medium according to exemplary embodiments of the present invention will be described, please refer to... Figure 3The computer-readable storage medium shown is an optical disc 30, on which a computer program (i.e., a program product) is stored. When the computer program is run by a processor, it implements the steps described in the above-described method implementation, such as: performing time synchronization and spatial alignment based on collected multi-source traffic information to construct a unified traffic state vector; using a deep learning-based graph neural network model to extract features of the current intersection state to obtain a real-time traffic congestion index; using a behavior recognition model to analyze the trajectories of pedestrians and vehicles in the video stream to detect traffic violations; dynamically adjusting the green light time ratio of each phase based on the traffic congestion index and a historical traffic pattern database; using a reinforcement learning algorithm to autonomously learn the optimal signal control strategy under different traffic scenarios to achieve coordinated control between multiple intersections; uploading the monitoring results and signal adjustment strategies to a cloud-based traffic management platform, while simultaneously issuing warning information to nearby driver terminals and intelligent traffic signs through edge nodes; and structurally storing image evidence of abnormal behavior and automatically associating it with license plate recognition results to generate penalty suggestions. The specific implementation methods of each step will not be repeated here.

[0047] It should be noted that examples of the computer-readable storage medium may also include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other optical and magnetic storage media, which will not be elaborated here.

[0048] After introducing the methods, apparatus, and media of exemplary embodiments of the present invention, the following references are made. Figure 4 A computing device for traffic signal monitoring according to an exemplary embodiment of the present invention.

[0049] Figure 4 A block diagram is shown of an exemplary computing device 40 suitable for implementing embodiments of the present invention. The computing device 40 may be a computer system or a server. Figure 4 The computing device 40 shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of the present invention.

[0050] like Figure 4 As shown, the components of computing device 40 may include, but are not limited to: one or more processors or processing units 401, system memory 402, and bus 403 connecting different system components (including system memory 402 and processing unit 401).

[0051] The computing device 40 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by the computing device 40, including volatile and non-volatile media, and removable and non-removable media.

[0052] System memory 402 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 4021 and / or cache memory 4022. Computing device 40 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, ROM 4023 may be used to read and write non-removable, non-volatile magnetic media (…). Figure 4 (Not shown in the image, usually referred to as "hard drive"). Although not shown in... Figure 4 The diagram illustrates that disk drives for reading and writing to removable non-volatile disks (e.g., "floppy disks") and optical disc drives for reading and writing to removable non-volatile optical discs (e.g., CD-ROMs, DVD-ROMs, or other optical media) can be provided. In these cases, each drive can be connected to bus 403 via one or more data media interfaces. System memory 402 may include at least one program product having a set (e.g., at least one) of program modules configured to perform the functions of the embodiments of the present invention.

[0053] A program / utility 4025 having a set (at least one) of program modules 4024 may be stored, for example, in system memory 402, and such program modules 4024 include, but are not limited to, an operating system, one or more application programs, other program modules, and program data, each or some combination of these examples may include an implementation of a network environment. Program modules 4024 typically perform the functions and / or methods described in the embodiments of the present invention.

[0054] The computing device 40 can also communicate with one or more external devices 404 (such as a keyboard, pointing device, display, etc.). This communication can be performed via the input / output (I / O) interface 405. Furthermore, the computing device 40 can also communicate with one or more networks (e.g., local area networks (LANs), wide area networks (WANs), and / or public networks, such as the Internet) via a network adapter 406. Figure 4 As shown, network adapter 406 communicates with other modules of computing device 40 (such as processing unit 401) via bus 403. It should be understood that, although... Figure 4 As not shown, it can be used in conjunction with computing device 40 with other hardware and / or software modules.

[0055] The processing unit 401 executes various functional applications and data processing by running programs stored in the system memory 402. For example, it performs time synchronization and spatial alignment based on collected multi-source traffic information to construct a unified traffic state vector; it uses a deep learning-based graph neural network model to extract features from the current intersection state to obtain a real-time traffic congestion index; it analyzes the trajectories of pedestrians and vehicles in the video stream using a behavior recognition model to detect traffic violations; it dynamically adjusts the green light time ratio of each phase based on the traffic congestion index and a historical traffic pattern database; it uses reinforcement learning algorithms to autonomously learn the optimal signal control strategy under different traffic scenarios to achieve coordinated control between multiple intersections; it uploads the monitoring results and signal adjustment strategies to the cloud-based traffic management platform, and simultaneously sends early warning information to nearby driver terminals and intelligent traffic signs through edge nodes; it structures and stores image evidence of abnormal behavior, automatically associates it with license plate recognition results, and generates penalty suggestions. The specific implementation methods of each step will not be repeated here. It should be noted that although several units / modules or sub-units / sub-modules of the traffic signal monitoring system are mentioned in the detailed description above, this division is merely exemplary and not mandatory. In fact, according to embodiments of the present invention, the features and functions of two or more units / modules described above can be embodied in one unit / module. Conversely, the features and functions of one unit / module described above can be further divided and embodied by multiple units / modules.

[0056] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0057] In the several embodiments provided by this invention, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the coupling or direct coupling or communication connection shown or discussed may be through some communication interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0058] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0059] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0060] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0061] Finally, it should be noted that the above-described embodiments are merely specific implementations of the present invention, used to illustrate the technical solutions of the present invention, and not to limit it. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments within the technical scope disclosed in the present invention, or make equivalent substitutions for some of the technical features; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

[0062] Furthermore, although the operations of the method of the present invention are described in a specific order in the accompanying drawings, this does not require or imply that these operations must be performed in that specific order, or that all the operations shown must be performed to achieve the desired result. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into one step, and / or one step may be broken down into multiple steps.

Claims

1. A traffic signal monitoring system, characterized in that, include: The fusion processing module is used to perform time synchronization and spatial alignment based on the collected multi-source traffic information, construct a unified traffic state vector, and use a deep learning-based graph neural network model to extract features of the current intersection state to obtain the real-time traffic congestion index. The anomaly detection module is used to analyze the trajectories of pedestrians and vehicles in the video stream using a behavior recognition model to detect traffic violations. The control optimization module is used to dynamically adjust the green light time ratio of each phase based on the traffic congestion index and historical traffic pattern database. It uses reinforcement learning algorithm to autonomously learn the optimal signal control strategy under different traffic scenarios to achieve coordinated control between multiple intersections. The feedback optimization module is used to upload monitoring results and signal adjustment strategies to the cloud traffic management platform, and at the same time send early warning information to nearby driver terminals and intelligent traffic signs through edge nodes; The anomaly recording module is used to structurally store image evidence of abnormal behavior and automatically associate it with license plate recognition results to generate penalty suggestions.

2. The traffic signal monitoring system as described in claim 1, characterized in that, The process of time synchronization and spatial alignment based on collected multi-source traffic information to construct a unified traffic state vector includes: Synchronization and alignment of data streams from various sensors are performed based on a unified timestamp system; when there is frame rate inconsistency or data delay, interpolation or nearest neighbor matching strategies are used to compensate for the time of low-frequency data; the data coordinates of various sensors are uniformly mapped to the geographic coordinate system or the local coordinate system of traffic intersections, and spatial coordinate transformation is performed using the external parameter calibration parameters between sensors to ensure the consistency of the spatial location of multi-source targets. For spatially overlapping targets, a fusion matching process is performed. By combining Kalman filtering with the Hungarian algorithm, the same target entity perceived by different sensors is identified. The ID consistency of the targets in the matching results is maintained to construct a stable cross-frame target trajectory. The state vectors of each target are mapped to the grid state map of the entire traffic intersection to form a matrix tensor representing the overall traffic state of the intersection.

3. The traffic signal monitoring system as described in claim 1, characterized in that, The method employs a deep learning-based graph neural network model to extract features from the current intersection state, obtaining a real-time traffic congestion index, including: The target traffic intersection is divided into several grid areas or traffic units, with each area serving as a node in the graph. The connection relationships between nodes are set according to the actual road topology and traffic flow direction to form a directed graph, and an initial feature vector is assigned to each node. The intersection embedding vector is input into a trained regression neural network or decision function, and the output is a real-valued traffic congestion index or congestion level.

4. The traffic signal monitoring system as described in claim 1, characterized in that, The method of using behavior recognition models to analyze the trajectories of pedestrians and vehicles in video streams to detect traffic violations includes: Based on actual intersection maps or image segmentation models, semantic regions are divided; each semantic region is represented by a polygon mask, and the target trajectory points are judged to determine the current target's location. The traffic light is currently red. During the red light duration t∈[t1,t2], the vehicle trajectory crosses the stop line area: ∃t∈[t1,t2],(x t ,y t )∈R before the stop line, and (x t+1 ,y t+1 )∈R 路口核心区 This was determined to be a red light violation. Define the legal passage direction vector Calculate the target trajectory direction vector If the included angle between the two satisfies: This can be interpreted as the target exhibiting reverse behavior in the current area; When the red light is on, if the vehicle remains stationary for more than the set time and the center of gravity of the detection frame crosses the center line of the stop line area, it can be considered as parking beyond the stop line. When a pedestrian is detected moving from pedestrian zone R 人行道 Entering the motor vehicle lane area R 车道 And the trajectory did not pass through the zebra crossing area R. 斑马线 If so, it will be marked as an illegal crossing.

5. The traffic signal monitoring system as described in claim 1, characterized in that, Based on a traffic congestion index and historical traffic pattern database, the system dynamically adjusts the green light time ratio for each phase and uses reinforcement learning algorithms to autonomously learn the optimal signal control strategy for different traffic scenarios, achieving coordinated control among multiple intersections, including: Define action A t The proposed adjustment scheme for the green light duration allocation for each phase within the current cycle is as follows: Let the number of phases be k, and the current action be: A t =[g1,g2,...,g k ] Among them, g i The green light duration for the i-th phase; Design a composite reward function R for traffic optimization. t Taking into account traffic efficiency, congestion level, and delay time: Among them, Q i (t) represents the number of vehicles queuing in the i-th direction, W i (t) represents the average waiting time of vehicles passing through per unit time, CI avg (t) represents the average congestion index of the current intersection, where α, β, and γ are weighting coefficients; The optimal timing policy is trained using a reinforcement learning algorithm. The process is as follows: Initialize the Q-network or policy network π. θ Experience replay pool D, in the simulation environment, input state S t Select action A t Get reward R t With the next state S t+1 , will (S t A t ,R t ,S t+1 Store the samples in the experience pool, sample a small batch of samples from the experience pool, and update the policy parameters θ or the Q function.

6. The traffic signal monitoring system as described in claim 1, characterized in that, The process of uploading monitoring results and signal adjustment strategies to the cloud-based traffic management platform, and simultaneously issuing early warning information to nearby driver terminals and intelligent traffic signs via edge nodes, includes: After receiving the local traffic monitoring results, the edge computing nodes deployed at the intersection determine whether there are any events or abnormal situations that require warning. The edge nodes send warning information to the on-board terminals of nearby vehicles via V2X. The information includes the current congestion situation at the intersection, the remaining time of the traffic light, warnings of dangerous behaviors, and suggested detours.

7. The traffic signal monitoring system as described in claim 1, characterized in that, The process of structuring and storing image evidence of abnormal behavior, automatically associating it with license plate recognition results, and generating penalty recommendations includes: The system performs license plate recognition on vehicles detected in illegal images, uses a deep learning model to locate the license plate position, and identifies character sequences through an OCR network with convolution and attention mechanisms. Results with low confidence (<0.8) are discarded, and multiple candidate results are supported for comparison.

8. A traffic signal monitoring method, characterized in that, include: Based on the collected multi-source traffic information, time synchronization and spatial alignment are performed to construct a unified traffic state vector. A graph neural network model based on deep learning is used to extract features of the current intersection state to obtain the real-time traffic congestion index. By using behavior recognition models to analyze the trajectories of pedestrians and vehicles in video streams, traffic violations can be detected. Based on the traffic congestion index and historical traffic pattern database, the green light time ratio of each phase is dynamically adjusted, and the reinforcement learning algorithm is used to autonomously learn the optimal signal control strategy under different traffic scenarios to achieve coordinated control between multiple intersections. The monitoring results and signal adjustment strategies are uploaded to the cloud-based traffic management platform, and warning information is sent to nearby driver terminals and intelligent traffic signs through edge nodes. The system stores image evidence of abnormal behavior in a structured manner and automatically associates it with license plate recognition results to generate penalty suggestions.

9. A computing device, the computing device comprising: At least one processor, memory, and input / output unit; The memory is used to store computer programs, and the processor is used to call the computer programs stored in the memory to execute the steps of the traffic signal monitoring system as described in any one of claims 1 to 7.

10. A computer-readable storage medium comprising instructions that, when executed on a computer, cause the computer to perform the steps of the traffic signal monitoring system as claimed in any one of claims 1 to 7.

Images