A method and system for intersection signal control

Abstract

The application relates to the technical field of traffic signal control, and provides a kind of intersection signal control method and system, the method comprises: obtaining abnormal information of rail transit operation;According to the abnormal information, perception and analysis are carried out, the type of ground traffic congestion state is determined;According to the type of ground traffic congestion state, the corresponding signal scheduling strategy is executed;Signal scheduling strategy includes: the type of ground traffic congestion state is bus congestion state, the scheduling of downstream channel dredging and centralized release bus is executed;And the type of ground traffic congestion state is the interference state of temporary parking of connection vehicle, the scheduling of compressing interference lane green light and prolonging non-interference key lane green light is executed.The application has the advantages of being able to effectively deal with the problem of ground traffic congestion caused by abnormal operation of rail transit, and improving the intersection traffic efficiency.

Description

Technical Field

[0001] This invention relates to the technical field of traffic signal control, and specifically to a method and system for controlling intersection signals. Background Technology

[0002] In urban traffic management, intersections near rail transit stations often bear enormous traffic pressure. These intersections are not only key nodes on urban arterial roads, but also important hubs for rail transit passengers to enter and exit stations and transfer to surface transportation. Their design typically takes into account the connection needs of various transportation modes such as pedestrians, surface buses, taxis, and ride-hailing services. For example, wide pedestrian crossings are set up in all directions of the intersection, and multiple surface bus stops are set up in the downstream exit lanes, especially the east-west exit lanes. At the same time, dedicated temporary parking areas for ride-hailing services and taxis are planned in the surrounding area to meet the diverse connection needs of passengers.

[0003] However, when the rail transit system experiences sudden operational anomalies, such as power outages causing train stoppages or large passenger congestion, existing signaling methods become problematic. These anomalies rapidly alter passenger travel strategies, forcing those planning to continue using rail transit to switch to surface public transportation, ride-hailing services, or taxis. This unexpected surge in passengers, far exceeding the scale and speed that signal control systems can predict based on conventional data, creates a chain reaction on the surrounding surface traffic environment. Ultimately, localized congestion at intersections quickly escalates into widespread traffic inefficiency, with some directions even experiencing complete traffic standstill, causing severe and lasting impacts on the entire urban road network.

[0004] To address the aforementioned issues, existing technologies urgently need improvement. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this application provides an intersection signal control method and system, which has the advantages of effectively addressing ground traffic congestion caused by abnormal operation of rail transit and improving intersection traffic efficiency.

[0006] This application provides a method for controlling intersection signals, including:

[0007] Obtain abnormal information about rail transit operations;

[0008] Based on the abnormal information, the type of ground traffic congestion is determined;

[0009] According to the type of the ground traffic congestion state, a corresponding signal scheduling strategy is executed; the signal scheduling strategy includes: when the type of the ground traffic congestion state is a bus congestion state, a scheduling of downstream channel dredging and centralized release of buses is executed; and when the type of the ground traffic congestion state is a transfer vehicle temporary parking interference state, a scheduling of compressing the green light of the interfered lane and prolonging the green light of the key lane not interfered is executed.

[0010] Through the above scheme, the ground traffic congestion type can be intelligently identified according to the rail transit abnormal information, and targeted scheduling can be executed to effectively alleviate traffic congestion.

[0011] To further solve the problem, the present application also proposes that when the type of the ground traffic congestion state is a bus congestion state, the scheduling of downstream channel dredging and centralized release of buses includes:

[0012] When the type of the ground traffic congestion state is a bus congestion state, a downstream exit lane dredging phase is started;

[0013] In the downstream exit lane dredging phase, green light pulses and red light buffers are periodically executed;

[0014] During the red light buffer, the vehicle queue length in the U-turn lane area is monitored;

[0015] According to the vehicle queue length, the green light pulse is adjusted, and the release rhythm of the downstream vehicle is adjusted;

[0016] When the downstream exit main lane is empty, and the vehicle queue length does not reach a preset overflow warning threshold, the downstream exit lane dredging phase is ended;

[0017] After the downstream exit lane dredging phase is ended, a bus centralized release phase is started.

[0018] Through the above scheme, for the bus congestion state, through downstream channel dredging and centralized release, the congestion of the bus platform area is effectively alleviated, and the bus passing efficiency is improved.

[0019] To perfect the solution, the present application also proposes that during the red light buffer, the step of monitoring the vehicle queue length in the U-turn lane area includes:

[0020] Obtaining video image information of the U-turn lane area;

[0021] Obtaining millimeter wave radar detection information of the U-turn lane area;

[0022] According to the video image information, a vehicle target is identified and a first vehicle queue length is calculated;

[0023] According to the millimeter wave radar detection information, a vehicle target is identified and a second vehicle queue length is calculated;

[0024] According to the video image information and the millimeter wave radar detection information, the effectiveness of the video image information and the millimeter wave radar detection information is evaluated, and an effectiveness evaluation result is obtained.

[0025] According to the effectiveness evaluation result, the first vehicle queue length and the second vehicle queue length are fused, and a vehicle queue length of the U-turn lane is obtained.

[0026] Through the above scheme, the multi-source data fusion method is used to accurately monitor the vehicle queue length of the U-turn lane, and the data accuracy and reliability are improved.

[0027] To perfect the solution, the application further proposes that the step of evaluating the effectiveness of the video image information and the millimeter wave radar detection information according to the video image information and the millimeter wave radar detection information includes:

[0028] The image quality of the video image information is analyzed, and an image quality analysis result is obtained.

[0029] The signal characteristics of the millimeter wave radar detection information are analyzed, and a signal characteristic analysis result is obtained.

[0030] The consistency of the corresponding vehicle target in the video image information and the corresponding vehicle target in the millimeter wave radar detection information is compared, and a consistency result is obtained.

[0031] According to the image quality analysis result, the signal characteristic analysis result and the consistency result, the effectiveness of the video image information and the millimeter wave radar detection information is judged, and an effectiveness evaluation result is obtained.

[0032] Through the above scheme, through the multi-dimensional evaluation mechanism, the effectiveness of the video image and the millimeter wave radar detection information is ensured, and reliable data is provided for the queue length calculation.

[0033] To perfect the solution, the application further proposes that the step of fusing the first vehicle queue length and the second vehicle queue length according to the effectiveness evaluation result to obtain the vehicle queue length of the U-turn lane includes:

[0034] According to the effectiveness evaluation result, the historical effectiveness evaluation results of the video image information and the millimeter wave radar detection information within a preset time period are obtained.

[0035] Based on the effectiveness evaluation result and the historical effectiveness evaluation result, a dynamic weight of the video image information and the millimeter wave radar detection information is determined.

[0036] Based on the dynamic weight, the first vehicle queue length and the second vehicle queue length are weighted and fused to obtain a preliminary fused queue length.

[0037] The preliminary fusion queue length is time series smoothed to obtain the vehicle queue length of the U-turn lane.

[0038] Through the above scheme, the multi-source data is dynamically fused and smoothed, and the accuracy and stability of the queue length estimation are improved.

[0039] To perfect the solution, the application further proposes that the step of determining the dynamic weights of the video image information and the millimeter wave radar detection information based on the effectiveness evaluation result and the historical effectiveness evaluation result comprises:

[0040] Obtain the current effectiveness index of the video image information and the millimeter wave radar detection information;

[0041] Obtain the historical effectiveness trend index of the video image information and the millimeter wave radar detection information;

[0042] Based on the current effectiveness index and the historical effectiveness trend index, calculate the comprehensive reliability score of the video image information and the millimeter wave radar detection information;

[0043] According to the comprehensive reliability score, determine the dynamic weights of the video image information and the millimeter wave radar detection information.

[0044] Through the above scheme, the data weights are dynamically adjusted based on the current and historical effectiveness indexes, so that the fusion result is more adaptive and robust.

[0045] To perfect the solution, the application further proposes that the step of time series smoothing the preliminary fusion queue length to obtain the vehicle queue length of the U-turn lane comprises:

[0046] An exponential smoothing method is used to process the preliminary fusion queue length to obtain the vehicle queue length of the U-turn lane.

[0047] Through the above scheme, the exponential smoothing method is used to effectively filter out data noise, so that the queue length estimation is more stable and reliable.

[0048] To perfect the solution, the application further proposes that the step of time series smoothing the preliminary fusion queue length to obtain the vehicle queue length of the U-turn lane comprises:

[0049] Obtain the historical data of the preliminary fusion queue length;

[0050] The historical data of the preliminary fusion queue length is weighted and averaged to obtain the vehicle queue length of the U-turn lane.

[0051] Through the above scheme, the historical data is processed by weighted average to further smooth the queue length estimation and improve its stability.

[0052] To perfect the solution, the application further provides that the type of the ground traffic congestion state is a temporary parking interference state of the feeder vehicle, and the step of performing the scheduling of compressing the green light of the interfered lane and extending the green light of the key lane not interfered includes:

[0053] When the type of the ground traffic congestion state is the temporary parking interference state of the feeder vehicle, real-time traffic flow information of each direction in the intersection of the rail transit station is acquired;

[0054] According to the real-time traffic flow information, the influence of the scheduling of compressing the green light of the interfered lane and extending the green light of the key lane not interfered on the traffic efficiency of the intersection is evaluated, and an influence evaluation result is obtained;

[0055] According to the influence evaluation result, the compression duration of the green light of the interfered lane and the extension duration of the green light of the key lane not interfered are adjusted;

[0056] The scheduling is performed according to the adjusted compression duration and extension duration.

[0057] Through the above solution, for the temporary parking interference of the feeder vehicle, the green light duration is dynamically adjusted through evaluation of the influence, so that the traffic is effectively guided and the interference is reduced.

[0058] An intersection signal control system for performing intersection signal control, comprising:

[0059] An abnormal information acquisition module for acquiring abnormal information of rail transit operation;

[0060] A congestion type determination module for determining the type of the ground traffic congestion state according to the abnormal information through sensing and analysis;

[0061] A signal scheduling control module for performing a corresponding signal scheduling strategy according to the type of the ground traffic congestion state; the signal scheduling strategy includes: when the type of the ground traffic congestion state is a bus congestion state, performing the scheduling of guiding and centrally releasing the buses in the downstream channel; and when the type of the ground traffic congestion state is a temporary parking interference state of the feeder vehicle, performing the scheduling of compressing the green light of the interfered lane and extending the green light of the key lane not interfered.

[0062] Through the above solution, a system implementation solution is provided, which can effectively perform the above signal control method and realize intelligent traffic management.

[0063] In summary, the intersection signal control method and system provided by the application can effectively alleviate traffic congestion by acquiring rail transit abnormal information, intelligently identifying the type of ground traffic congestion and performing targeted scheduling, thereby having the advantages of being able to effectively deal with the problem of ground traffic congestion caused by abnormal operation of rail transit and improving the traffic efficiency of the intersection. BRIEF DESCRIPTION OF DRAWINGS

[0064] Figure 1 Method flow chart of a signal control method for an intersection according to one embodiment of the present application;

[0065] Figure 2 Method flow chart of a signal control method for an intersection according to one embodiment of the present application;

[0066] Figure 3 Method flow chart of a signal control method for an intersection according to one embodiment of the present application;

[0067] Figure 4 Method flow chart of a signal control method for an intersection according to one embodiment of the present application;

[0068] Figure 5 Method flow chart of a signal control method for an intersection according to one embodiment of the present application;

[0069] Figure 6 Method flow chart of a signal control method for an intersection according to one embodiment of the present application;

[0070] Figure 7 Method flow chart of a signal control method for an intersection according to one embodiment of the present application;

[0071] Figure 8 Method flow chart of a signal control method for an intersection according to one embodiment of the present application;

[0072] Figure 9 System block diagram of a signal control system for an intersection according to one embodiment of the present application;

[0073] BRIEF DESCRIPTION OF DRAWINGS

[0074] 1. A signal control system for an intersection; 11. An abnormal information acquisition module; 12. A stagnation type determination module; 13. A signal scheduling control module. DETAILED DESCRIPTION

[0075] The technical solutions in the present application will be described clearly and completely below in conjunction with the drawings in the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. The components of the present application described and shown in the drawings herein can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the claimed present application, but only represents selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present application.

[0076] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0077] Traditional intersection signal control methods suffer from the inability to accurately determine the type of traffic congestion when sudden operational anomalies occur in the rail transit system. This limitation prevents the system from implementing effective signal scheduling strategies based on the instantaneous and unexpected superposition of changes in actual traffic flow, thereby exacerbating intersection traffic congestion and reducing the overall traffic efficiency of the intersection.

[0078] For example, suppose in a core urban area, at an intersection near a rail transit station, when a power outage causes trains to stop, forcing a large number of passengers to turn to surface buses and shuttle vehicles, the existing signal control system can only passively adjust based on local flow sensor data. The system cannot identify whether this congestion is caused by a specific type of rail transit anomaly, such as queue overflow caused by buses stopping for extended periods at platforms, or lane obstruction caused by ride-hailing vehicles and taxis temporarily stopping on the roadside. This lack of information prevents the system from distinguishing the root causes and scope of impact of different congestion patterns, thus hindering the implementation of targeted traffic management measures, resulting in the intersection facing disorderly traffic flow overlap and conflicts in a short period of time.

[0079] In response, this application proposes an intersection signal control method, combining... Figure 1 As shown, it includes:

[0080] S1, obtain abnormal information about rail transit operation;

[0081] S2, based on abnormal information, senses and analyzes to determine the type of ground traffic congestion;

[0082] S3, based on the type of ground traffic congestion, execute the corresponding signal dispatching strategy; the signal dispatching strategy includes: if the type of ground traffic congestion is bus congestion, execute the downstream channel diversion and centralized release of buses; and if the type of ground traffic congestion is temporary stop interference of shuttle vehicles, execute the dispatching to shorten the green light of the affected lane and extend the green light of the unaffected key lane.

[0083] The abnormal information of rail transit operation refers to event data that occurs outside normal operation of the rail transit system and affects passenger travel patterns and ground traffic conditions. It can be obtained in various forms, such as train delays, outages, and fault reports obtained through the rail transit operation data center interface, or data such as platform passenger flow aggregation and increased exit gate flow obtained through sensor networks. The purpose is to provide data input for subsequent ground traffic state judgment. The perception and analysis of abnormal information to determine the type of ground traffic congestion state refers to the use of data processing and pattern recognition techniques to identify the congestion or congestion patterns faced by the current intersection area ground traffic based on the obtained rail transit abnormal information. It can be achieved using machine learning algorithms, expert system rule bases, or threshold-based judgment logic, such as by analyzing passenger outflow size, bus line impact, and demand changes for connecting vehicles to determine whether it is a bus platform area congestion or a local lane disturbance caused by temporary parking of connecting vehicles. The purpose is to convert rail transit abnormalities into ground traffic problem types to facilitate scheduling measures. The signal scheduling strategy refers to pre-set or dynamically generated traffic signal timing adjustment schemes for different ground traffic congestion state types. It can be achieved using pre-set scheduling scheme libraries, real-time optimization algorithms, or rule-based decision systems. The purpose is to optimize ground traffic flow by adjusting green light duration, phase sequence, and other factors of intersection traffic lights under rail transit abnormal conditions. When the main problem of ground traffic is determined to be long-term parking or queuing congestion of buses in the platform or exit lane area, the system adopts a combined signal control strategy that starts the dedicated relief phase of the downstream exit lane, combines periodic green light pulses with red light buffers, and concentrates on releasing bus fleets under certain conditions. The purpose is to solve the bottleneck of bus traffic and ensure that buses can leave the platform area, thereby alleviating congestion at the intersection. When the main problem of ground traffic is determined to be temporary parking of connecting vehicles such as taxis or online car services on the roadside, causing a decrease in the capacity of some lanes, the system adopts an asymmetric signal control strategy that compresses the green light of the disturbed lane and extends the green light of the key lane that is not disturbed. It can be achieved by monitoring traffic flow and queue length in the disturbed lane in real time, evaluating the impact on overall traffic efficiency, and dynamically adjusting the green light duration of related lanes. The purpose is to optimize the capacity of undisturbed lanes by sacrificing the travel time of some disturbed lanes while ensuring overall traffic efficiency, thereby alleviating the impact of local disturbances on the intersection.

[0084] To further illustrate the embodiments of the present application, the following is described by an example. In some embodiments, the present application is implemented as follows. First, obtaining abnormal information of rail transit operation can be achieved by real-time data exchange with the data interface of the urban rail transit operation management center, for example, receiving train operation timetable, real-time position, fault alarm, passenger retention quantity and other data streams through the API interface. These data streams can be received by a data acquisition and preprocessing module and preliminarily processed. Subsequently, according to the abnormal information for sensing and analyzing, determining the type of ground traffic retention state can be completed by an analysis module. A set of rule-based system can be preset inside the module, or a model can be trained. For example, when receiving the abnormal information of "a train on a certain line is out of service, and passengers will transfer to buses", the analysis module will combine historical data and real-time passenger flow prediction model to determine that the type of current ground traffic retention state is "bus retention state". If the abnormal information of "passengers are retained at a certain platform, and connecting vehicles are pouring into the surrounding area" is received, it can be determined as "connecting vehicle temporary parking interference state". Once the type of ground traffic retention state is determined, the signal scheduling control module will execute the corresponding signal scheduling strategy according to the preset scheduling logic. Specifically, when it is determined as a bus retention state, the signal scheduling control module will send instructions to the intersection signal machine to start the dredging phase for the downstream exit lane, and cooperate with the green light timing scheme of bus centralized release. For example, the green light time of the bus dedicated lane can be temporarily extended, and the signals of adjacent intersections are coordinated to ensure that the bus can pass through. When it is determined as a connecting vehicle temporary parking interference state, the signal scheduling control module will calculate the real-time traffic flow data of the disturbed lane and the undisturbed key lane, and dynamically adjust the green light time according to the preset optimization target. For example, the green light time of the disturbed lane is compressed, and this part of the time is allocated to the straight or left turning lane to maintain the overall traffic. These scheduling instructions are sent to the on-site signal controller through wired or wireless communication network, and the specific light color conversion is executed by the signal controller.

[0085] Optionally, in combination with Figure 2 As shown in the figure, when the type of ground traffic retention state in step S3 is a bus retention state, the step of executing the scheduling of dredging the downstream channel and centralized releasing the bus includes:

[0086] A1, when the type of ground traffic retention state is a bus retention state, start the dredging phase of the downstream exit lane;

[0087] A2, in the dredging phase of the downstream exit lane, periodically execute green light pulse and red light buffer;

[0088] A3, during the red light buffer period, monitor the vehicle queue length in the U-turn lane area;

[0089] A4, adjust the green pulse according to the vehicle queue length, to adjust the release rhythm of the downstream vehicle;

[0090] A5, when the downstream exit main lane is empty and the vehicle queue length does not reach the preset overflow warning threshold, end the downstream exit lane dredging phase;

[0091] A6, after ending the downstream exit lane dredging phase, start the bus concentrated release phase.

[0092] The vehicle queue length refers to the queuing length or number of vehicles waiting for traffic in a specific lane or area, which can be realized by visual recognition, radar detection or geomagnetic induction technology. The overflow warning threshold refers to a critical value preset for judging whether the vehicle queue is about to reach a saturated or overflow state, which can be a vehicle number, a distance or a time length, and its purpose is to avoid the spread of downstream traffic congestion to the upstream area.

[0093] In some preferred embodiments, the application is implemented as follows. When the intersection signal control system identifies that the ground traffic congestion state is a bus congestion state through its congestion type judgment module, the signal scheduling control module will immediately send instructions to the signal lamp controller to start the downstream exit lane dredging phase. In this dredging phase, the signal lamp controller will alternately execute green pulse and red buffer according to the preset period, for example every 30 seconds. The duration of the green pulse can be set to 5 to 10 seconds to allow part of the vehicles to pass; the duration of the red buffer can be set to 20 to 25 seconds to provide time for the evacuation of downstream vehicles. During the red buffer period, the system will use the video detector and geomagnetic sensor installed in the U-turn lane area to obtain real-time vehicle information of the area and calculate the vehicle queue length. For example, the number of vehicles is identified by image processing algorithm, or the presence of vehicles is detected by geomagnetic sensor and accumulated. The signal scheduling control module will dynamically adjust the duration of the subsequent green pulse according to the real-time monitored vehicle queue length. For example, if the vehicle queue length exceeds a certain intermediate threshold, the green pulse duration can be shortened by 2 seconds; if the queue length continues to increase and approaches the overflow warning threshold, the green pulse can be further shortened or suspended. When the system confirms through the vehicle detector of the downstream exit main lane that there is no vehicle in the lane, and the vehicle queue length in the U-turn lane area is lower than the preset overflow warning threshold (for example, the queue length is less than 5 vehicles or the queue length corresponding to the queuing area does not exceed the U-turn lane range), the signal scheduling control module will immediately end the downstream exit lane dredging phase. Then, the system will start the bus concentrated release phase, at this time, the signal lamp related to the bus traffic direction will be kept green for a long time, for example, for 30 to 60 seconds, to ensure that the stranded buses can pass through the intersection quickly and concentratedly.

[0094] Optionally, in combination with Figure 3 As shown, the step of monitoring the vehicle queue length of the U-turn lane area during the red light buffer period includes:

[0095] A31, acquiring video image information of the U-turn lane area;

[0096] A32, acquiring millimeter wave radar detection information of the U-turn lane area;

[0097] A33, identifying vehicle targets and calculating a first vehicle queue length according to the video image information;

[0098] A34, identifying vehicle targets and calculating a second vehicle queue length according to the millimeter wave radar detection information;

[0099] A35, evaluating the effectiveness of the video image information and the millimeter wave radar detection information according to the video image information and the millimeter wave radar detection information, and obtaining an effectiveness evaluation result;

[0100] A36, fusing the first vehicle queue length and the second vehicle queue length according to the effectiveness evaluation result, and obtaining a vehicle queue length of the U-turn lane.

[0101] The video image information of the U-turn lane area is obtained by installing a video monitoring device above the U-turn lane area to collect visual data streams in real time, which aims to provide visual feature information of the vehicle, such as the shape, color, position, etc. of the vehicle. The millimeter wave radar detection information of the U-turn lane area is obtained by deploying a millimeter wave radar sensor in the U-turn lane area to obtain radar detection data in real time, which aims to provide physical attribute information of the vehicle, such as the distance, speed, existence, etc. of the vehicle. According to the video image information, the vehicle target is identified and the first vehicle queue length is calculated, which means that image processing and computer vision algorithms are used to analyze the video image information, automatically detect and locate the vehicle individuals in the image, and based on these identified vehicle targets, the vehicle queue length under the video perspective is determined through counting or spatial arrangement analysis, which aims to obtain a preliminary estimate of the vehicle queue from the visual dimension. According to the millimeter wave radar detection information, the vehicle target is identified and the second vehicle queue length is calculated, which means that radar signal processing algorithms are used to analyze the millimeter wave radar detection information to identify the vehicle targets within the radar beam coverage, and according to the distance information and number of these targets, the vehicle queue length detected by the radar is calculated, which aims to obtain a preliminary estimate of the vehicle queue from the radar detection dimension. The effectiveness of the video image information and the millimeter wave radar detection information is evaluated to obtain the effectiveness evaluation result, which means that the quality and reliability of the sensor data from two different sources are analyzed to determine their availability and accuracy under the current environmental conditions. Specifically, it can be a comprehensive judgment of image clarity, lighting conditions, radar signal strength, noise level and consistency between the two information sources. Its purpose is to provide a reliable basis for subsequent information fusion to ensure the accuracy of the fusion result. The first vehicle queue length and the second vehicle queue length are fused to obtain the vehicle queue length of the U-turn lane, which means that according to the effectiveness evaluation result, appropriate algorithms or strategies are used to comprehensively process the vehicle queue lengths calculated from video images and millimeter wave radars respectively to obtain a more accurate and robust final vehicle queue length. Its purpose is to overcome the limitations of single sensor through multi-source information complementation to improve measurement accuracy and reliability.

[0102] In some preferred embodiments, the application is implemented as follows. To accurately monitor the vehicle queue length in the U-turn lane during red light buffering, a high-resolution network camera can be deployed to obtain video image information of the U-turn lane area, while a 77GHz vehicle-mounted millimeter wave radar sensor is installed to obtain millimeter wave radar detection information of the area. Specifically, after obtaining the video image information, a real-time target detection model based on deep learning, such as YOLOv5, can be used to identify vehicle targets in the video frames and calculate the first vehicle queue length. This can include perspective transformation correction of the identified vehicle bounding boxes, mapping them to the actual ground coordinate system, and then determining the queue length through counting and arrangement analysis. At the same time, for millimeter wave radar detection information, a clustering algorithm such as DBSCAN can be used to process radar point cloud data, identify vehicle targets, and calculate the second vehicle queue length based on the distance and number of these targets. Further, to evaluate the effectiveness of the video image information and the millimeter wave radar detection information, image quality analysis can be performed on the video image, such as calculating the sharpness, brightness and contrast of the image, and setting corresponding quality thresholds. At the same time, signal feature analysis can be performed on the millimeter wave radar detection information, such as evaluating the signal-to-noise ratio of the radar echo and the density of the target point cloud. In addition, consistency between the vehicle targets identified in the video image and the vehicle targets detected by the millimeter wave radar can also be compared, such as by matching the position and number of vehicles. Based on these analysis results, a comprehensive effectiveness evaluation result can be obtained, such as through a multi-factor weighted model or a fuzzy logic system to judge the reliability of the current two information sources. Finally, based on the effectiveness evaluation result, the first vehicle queue length and the second vehicle queue length can be dynamically fused. This can include determining the dynamic weights of the video image information and the millimeter wave radar detection information based on the current evaluation result and the historical effectiveness evaluation results in a predetermined time period. For example, when the video image quality decreases, its weight can be reduced, while the weight of the millimeter wave radar can be increased. Then, based on these dynamic weights, the first vehicle queue length and the second vehicle queue length are weighted and fused to obtain a preliminary fused queue length. To further improve the stability of the queue length, the preliminary fused queue length can be subjected to time series smoothing processing, such as exponential smoothing, to obtain the final vehicle queue length of the U-turn lane. This processing can effectively filter out transient noise and measurement fluctuations, providing more stable and reliable queue length data.

[0103] Optionally, in combination with Figure 4 As shown, the step of A35 evaluating the effectiveness of the video image information and the millimeter wave radar detection information based on the video image information and the millimeter wave radar detection information to obtain an effectiveness evaluation result includes:

[0104] A351, performing image quality analysis on the video image information to obtain an image quality analysis result;

[0105] A352, performing signal feature analysis on the millimeter wave radar detection information to obtain a signal feature analysis result;

[0106] A353, comparing the consistency of the corresponding vehicle targets in the video image information and the millimeter wave radar detection information to obtain a consistency result;

[0107] A354, judging the effectiveness of the video image information and the millimeter wave radar detection information according to the image quality analysis result, the signal feature analysis result and the consistency result to obtain an effectiveness evaluation result.

[0108] Among them, image quality analysis refers to evaluating the visual attributes of video image data, which can be specifically realized by analyzing the clarity, brightness, contrast, noise level and whether there is occlusion or blur of the image, etc. The purpose is to judge whether the video image information is affected by environmental factors such as light, rain, fog or sensor failure, so as to evaluate its reliability for vehicle target identification.

[0109] Among them, signal feature analysis refers to evaluating the original signal or processed signal features detected by the millimeter wave radar, which can be specifically realized by analyzing the strength, signal-to-noise ratio, Doppler shift, target scattering cross section (RCS) and whether there are false targets or clutter interference, etc. The purpose is to judge whether the millimeter wave radar detection information is affected by environmental interference, multipath effect or sensor performance degradation, so as to evaluate its reliability for vehicle target identification.

[0110] Among them, consistency refers to the matching degree of the vehicle targets identified in the video image information and the millimeter wave radar detection information in spatial position, motion state or quantity, which can be specifically realized by comparing the position coordinates, velocity vectors, size estimates and target quantities of the vehicle targets identified by the two kinds of sensor data after time and space alignment. The purpose is to verify whether the information from two different sources points to the same physical entity, so as to judge whether there is obvious deviation or error between them.

[0111] The effectiveness evaluation result refers to a comprehensive judgment on whether the video image information and millimeter-wave radar detection information are reliable and suitable for subsequent vehicle queue length estimation. Specifically, it can be achieved by setting a threshold or adopting logical rules, combining image quality analysis results, signal feature analysis results, and consistency results, and outputting a quantitative score or classification label representing the effectiveness of the information. Its purpose is to provide a reliable basis for subsequent fusion of data from different sensors and avoid introducing low-quality or inconsistent data into the final queue length calculation.

[0112] In some preferred embodiments, this application is implemented as follows: When performing image quality analysis on video image information, image processing algorithms can be used, such as calculating the image's sharpness score (e.g., gradient magnitude based on the Laplacian operator), average brightness, contrast (e.g., grayscale range based on histogram equalization), and noise level (e.g., residual analysis after Gaussian filtering). These indicators can be quantified and compared with preset thresholds. For example, if the sharpness score is below a certain threshold, or the brightness is too high or too low, the image quality analysis result can indicate poor video image information quality.

[0113] When performing signal feature analysis on millimeter-wave radar detection information, the signal strength, signal-to-noise ratio (SNR), and stability of Doppler frequency shift for each detected target can be extracted. For example, if the signal strength of a target is lower than the background noise level, or its SNR is lower than a preset threshold, or its Doppler frequency shift changes in a short period of time, the signal feature analysis results can indicate that the millimeter-wave radar detection information is subject to interference or instability.

[0114] When comparing the consistency between vehicle targets identified in video images and those detected by millimeter-wave radar, spatiotemporal registration can be performed first. For example, the pixel coordinates of the vehicle targets in the video image can be converted to actual geographic coordinates and matched with the geographic coordinates of the targets detected by the millimeter-wave radar. After successful matching, the consistency of the matched targets' size, velocity vector, and target ID can be compared. For instance, if a large vehicle identified in the video corresponds to a small target in the radar data, or if their velocities are opposite, or if multiple vehicles are identified in the video but only one is detected by the radar in a certain area, the consistency result can indicate a discrepancy between the two.

[0115] Finally, in judging the effectiveness of the video image information and the millimeter wave radar detection information according to the image quality analysis result, the signal feature analysis result and the consistency result, a multi-factor weighted decision model or a rule-based expert system can be used. For example, different weights can be given to the image quality, the signal feature and the consistency, and then a comprehensive effectiveness score is calculated. If the comprehensive score is higher than a certain effectiveness threshold, it is judged that the information has effectiveness; otherwise, it is judged as invalid or low effectiveness. For example, when the image quality analysis result shows that the image is clear, the signal feature analysis result shows that the signal is stable, and the consistency result shows that the matching degree is high, it can be judged that both kinds of information have high effectiveness; on the contrary, if any or more indicators are poor, the effectiveness evaluation result is correspondingly reduced.

[0116] Optionally, in combination with Figure 5 As shown in FIG. 6, the step of A36 of fusing the first vehicle queue length and the second vehicle queue length to obtain the vehicle queue length of the U-turn lane according to the effectiveness evaluation result includes:

[0117] A361, according to the effectiveness evaluation result, obtaining the historical effectiveness evaluation results of the video image information and the millimeter wave radar detection information in a preset time period;

[0118] A362, determining the dynamic weights of the video image information and the millimeter wave radar detection information based on the effectiveness evaluation result and the historical effectiveness evaluation result;

[0119] A363, based on the dynamic weights, weighting and fusing the first vehicle queue length and the second vehicle queue length to obtain a preliminary fused queue length;

[0120] A364, performing time series smoothing processing on the preliminary fused queue length to obtain the vehicle queue length of the U-turn lane.

[0121] The time series smoothing processing refers to processing a series of data arranged in time sequence to eliminate short-term fluctuations or noise and reveal potential trends or more stable values. It can be realized by using various algorithms such as moving average method, exponential smoothing method, Kalman filter, etc. The purpose is to improve the stability and reliability of the data sequence and reduce the influence of instantaneous errors on subsequent decisions.

[0122] In some preferred embodiments, the application is implemented as follows: after obtaining the validity evaluation result, the system maintains a historical data buffer for storing the validity evaluation results of video image information and millimeter wave radar detection information every second within the past five minutes. When a new validity evaluation result is generated, the system extracts the historical validity evaluation results within a preset time period from the buffer. Then, in order to determine the dynamic weights of the video image information and the millimeter wave radar detection information, the system can calculate the weighted average of the current validity evaluation result and the historical validity evaluation result, for example, giving the current validity evaluation result a weight of 0.6 and the historical validity evaluation result a weight of 0.4, to obtain the respective comprehensive reliability scores. Then, according to these comprehensive reliability scores, the dynamic weights of the video image information and the millimeter wave radar detection information are calculated through normalization processing or a Softmax function, ensuring that the sum of the weights of the two is 1. For example, if the comprehensive reliability score of the video image information is higher, its dynamic weight can be set to 0.7, and the dynamic weight of the millimeter wave radar detection information is 0.3. Subsequently, based on these dynamic weights, the first vehicle queue length identified through the video image information and the second vehicle queue length identified through the millimeter wave radar detection information are weighted and fused, for example, the preliminary fused queue length = dynamic weight_video * first vehicle queue length + dynamic weight_radar * second vehicle queue length. Finally, in order to perform time series smoothing on the preliminary fused queue length, the system can use the exponential smoothing method. Specifically, the new U-turn lane vehicle queue length can be obtained by weighted averaging of the current preliminary fused queue length and the smoothed queue length at the previous time, for example, new queue length = a * current preliminary fused queue length + (1-a) * smoothed queue length at the previous time, where a is a smoothing coefficient, which can be set to 0.3, for example, so as to effectively eliminate transient fluctuations and obtain a more stable U-turn lane vehicle queue length.

[0123] Optionally, in combination with the above-mentioned Figure 6 The step of determining the dynamic weights of the video image information and the millimeter wave radar detection information based on the validity evaluation result and the historical validity evaluation result in A362 comprises:

[0124] A3621, obtaining the current validity indicators of the video image information and the millimeter wave radar detection information;

[0125] A3622, obtaining the historical validity trend indicators of the video image information and the millimeter wave radar detection information;

[0126] A3623, calculating the comprehensive reliability scores of the video image information and the millimeter wave radar detection information based on the current validity indicators and the historical validity trend indicators;

[0127] A3624, determines the dynamic weight of video image information and millimeter wave radar detection information according to the comprehensive reliability score.

[0128] Wherein, the current effectiveness index refers to the quantitative value at the current time for measuring the quality and reliability degree of video image information and millimeter wave radar detection information, which can be represented by image definition, illumination intensity, signal-to-noise ratio, target detection confidence and other parameters, and its purpose is to reflect the availability of sensor data in the immediate environment. The historical effectiveness trend index refers to the statistical quantity reflecting the change law of the effectiveness of video image information and millimeter wave radar detection information in the past period of time, which can be calculated by using the average value, variance, change rate or time series model prediction value of the historical effectiveness index, and its purpose is to evaluate the stability and reliability of sensor data in long-term operation. The comprehensive reliability score refers to the quantitative result of the comprehensive evaluation of the reliability of video image information and millimeter wave radar detection information based on the current effectiveness index and the historical effectiveness trend index, which can be calculated by using weighted average, fuzzy logic reasoning, machine learning model and other methods, and its purpose is to provide a single evaluation value that comprehensively reflects the quality of sensor data. The dynamic weight refers to the contribution proportion of real-time adjustment according to the comprehensive reliability score when fusing video image information and millimeter wave radar detection information, which can be determined by using normalization processing, Sigmoid function mapping or rule-based lookup table, and its purpose is to ensure that in different environmental conditions, the sensor data with higher reliability occupies a larger proportion in the fusion result.

[0129] In some preferred embodiments, the step of determining the dynamic weights of the video image information and the millimeter wave radar detection information based on the effectiveness evaluation result and the historical effectiveness evaluation result can be implemented as follows: First, the current effectiveness indicators of the video image information and the millimeter wave radar detection information are obtained. For example, for the video image information, its image definition, brightness, contrast, and the vehicle target confidence output by the target detection algorithm can be analyzed as the current effectiveness indicators. For the millimeter wave radar detection information, its signal-to-noise ratio, target distance accuracy, target speed accuracy, and target tracking stability can be obtained as the current effectiveness indicators. These indicators can be calculated in real time or directly read from the sensor interface. Second, the historical effectiveness trend indicators of the video image information and the millimeter wave radar detection information are obtained. This can involve maintaining a historical effectiveness indicator database within a sliding time window. For example, the average and standard deviation of each current effectiveness indicator of the video image information and the millimeter wave radar detection information in the past five minutes can be calculated as their historical effectiveness trend indicators. In addition, the historical performance of each sensor effectiveness indicator under different weather or lighting conditions can also be recorded as a trend reference. Then, based on the current effectiveness indicators and the historical effectiveness trend indicators, the comprehensive reliability scores of the video image information and the millimeter wave radar detection information are calculated. Specifically, a comprehensive reliability function can be defined for each sensor, which weights and sums the current effectiveness indicators (e.g., with a weight of 0.6) and the historical effectiveness trend indicators (e.g., with a weight of 0.4). For example, a threshold can be set, and if the current definition is lower than the threshold, the current effectiveness score will be reduced; if the historical average signal-to-noise ratio is low for a long time, the historical trend score will also be low. Finally, these weighted scores are normalized to the range of 0 to 1 to obtain the respective comprehensive reliability scores. Finally, the dynamic weights of the video image information and the millimeter wave radar detection information are determined according to the comprehensive reliability scores. For example, a Softmax function or a simple proportional allocation method can be used. If the comprehensive reliability score of the video image information is 0.8 and the comprehensive reliability score of the millimeter wave radar detection information is 0.6, the total score is 1.4, and then the weight of the video image information is set to 0.8 / 1.4 and the weight of the millimeter wave radar detection information is set to 0.6 / 1.4. In this way, the sensor with a higher comprehensive reliability score is given a larger weight in the subsequent fusion process, thereby ensuring the accuracy of the fusion result.

[0130] Optionally, the step of A364 performing time series smoothing on the preliminary fusion queue length to obtain the vehicle queue length of the U-turn lane includes:

[0131] An exponential smoothing method is used to process the preliminary fusion queue length to obtain the vehicle queue length of the U-turn lane.

[0132] The exponential smoothing method is a time series prediction method that assigns different weights to historical data, with more recent data being given greater weight, to predict future values or smooth current data. It can be implemented in various forms such as single exponential smoothing, double exponential smoothing, or triple exponential smoothing, with the goal of effectively filtering out random fluctuations and noise in the data while maintaining sensitivity to changes in data trends.

[0133] Optionally, in combination with Figure 7 As shown in A364, the step of performing time series smoothing on the preliminary merged queue length to obtain the vehicle queue length of the U-turn lane includes:

[0134] A3641, obtaining historical data of the preliminary merged queue length;

[0135] A3642, performing weighted average processing on the historical data of the preliminary merged queue length to obtain the vehicle queue length of the U-turn lane.

[0136] The weighted average processing refers to assigning a weight to each data point in a set of data, then multiplying each data point by its corresponding weight, and finally adding all the products and dividing by the sum of all weights to obtain a weighted average value. It can be implemented in various ways such as linear weighting, exponential decay weighting, or sliding window weighting, with the goal of more accurately reflecting the influence of different historical data points on the current state, thereby improving the accuracy of the estimate.

[0137] In some preferred embodiments, the historical data of the preliminary merged queue length is processed by weighted average, which can be implemented in the form of exponential decay weighting. For example, the system can maintain a historical sequence of preliminary merged queue lengths, such as the queue length data of the last N time steps. When calculating the current vehicle queue length of the U-turn lane, these historical data points are assigned a weight that decays over time, i.e., the closer the data point is to the current time, the greater its weight, and the farther away the data point is, the smaller its weight. For example, a decay factor a (0 < a < 1) can be set, and the current queue length can be calculated by the weighted sum of historical data points, where each historical data point has a weight of a power of a. In this way, the system can ensure that the latest data has a greater impact on the result, while still using early historical data to smooth short-term fluctuations, thereby obtaining a U-turn lane vehicle queue length that is responsive to current changes and has good stability. This processing method can effectively reduce the queue length estimation bias caused by sensor transient errors or temporary fluctuations in traffic flow, allowing the signal scheduling system to make decisions based on more reliable data.

[0138] Optionally, in combination with Figure 8As shown, when the type of the ground traffic congestion state in step S3 is the feeder vehicle temporary stop interference state, the step of performing the scheduling of compressing the green light of the interfered lane and extending the green light of the key lane not interfered includes:

[0139] B1, when the type of the ground traffic congestion state is the feeder vehicle temporary stop interference state, obtaining real-time traffic flow information of each direction within the intersection of the rail transit station;

[0140] B2, according to the real-time traffic flow information, evaluating the influence of the scheduling of compressing the green light of the interfered lane and extending the green light of the key lane not interfered on the traffic efficiency of the intersection, to obtain an influence evaluation result;

[0141] B3, according to the influence evaluation result, adjusting the compression duration of the green light of the interfered lane and the extension duration of the green light of the key lane not interfered;

[0142] B4, performing the scheduling according to the adjusted compression duration and extension duration.

[0143] Wherein, the real-time traffic flow information refers to the data set about the running state of vehicles in each direction within the intersection at a specific time point, which is obtained through various traffic sensors or data sources, and can specifically include traffic volume, vehicle speed, queue length, lane occupancy rate, etc., and its purpose is to provide accurate and dynamic data basis for subsequent scheduling decisions. Further, evaluating the influence of the scheduling of compressing the green light of the interfered lane and extending the green light of the key lane not interfered on the traffic efficiency of the intersection to obtain an influence evaluation result means that based on the obtained real-time traffic flow information, using traffic simulation models, prediction algorithms or expert systems, the overall traffic capacity, average delay, queue length and other indicators of the intersection that may be caused by different green light duration adjustment schemes are quantitatively analyzed, so as to obtain the potential influence degree of each scheme on the traffic efficiency, and its purpose is to quantify the pros and cons of different scheduling schemes to provide basis for optimization decisions. On this basis, adjusting the compression duration of the green light of the interfered lane and the extension duration of the green light of the key lane not interfered means that according to the influence evaluation result, using optimization algorithms, rule engines or machine learning models, dynamically calculating and determining the specific amount of time that the green light of the interfered lane should be shortened and the specific amount of time that the green light of the key lane not interfered should be increased, and its purpose is to relieve local interference while maximizing the balance of the overall traffic flow of the intersection to avoid the generation of new congestion points.

[0144] The scheme of the present application realizes fine management of signal scheduling strategy by introducing the acquisition and analysis of real-time traffic flow information. When the system determines that the ground traffic congestion state is the temporary parking interference state of the feeder vehicle, it is no longer simply to execute the preset green light compression and extension, but first to acquire the real-time traffic flow information of each direction in the intersection of the rail transit station. These information, such as traffic volume, queue length and vehicle speed, can fully reflect the current traffic running condition of the intersection. Based on these real-time data, the system can evaluate the influence of different green light compression and extension time combinations on the overall traffic efficiency of the intersection, so as to obtain the quantitative evaluation result. This evaluation process considers the traffic load of each direction of the intersection, avoiding the situation of only focusing on local problems and ignoring the overall efficiency. It is precisely because of this comprehensive evaluation that the system can dynamically adjust the compression time of the disturbed lane and the extension time of the key lane according to the evaluation result. This adjustment is no longer a fixed value, but is optimized according to the real-time traffic condition and the influence prediction of the overall efficiency, ensuring that the temporary parking interference of the feeder vehicle is relieved while the negative impact on other direction traffic flow is minimized. Finally, the system executes scheduling according to the adjusted compression time and extension time, so that the signal timing can more flexibly and intelligently respond to sudden traffic events. Compared with the basic scheme of executing the preset scheduling strategy according to the type of congestion state, the scheme of the present application further introduces the perception of real-time traffic flow information, the evaluation of scheduling influence and the dynamic adjustment of green light time after identifying the temporary parking interference state of the feeder vehicle. This in-depth analysis and optimization mechanism enables the original strategy of "compressing the green light of the disturbed lane and extending the green light of the key lane" to change from extensive execution to fine management. In this way, not only can the local congestion caused by the temporary parking of the feeder vehicle be effectively relieved, but also the secondary congestion caused by excessive adjustment can be avoided, so as to solve the specific traffic problem while significantly improving the overall traffic efficiency and traffic flow balance of the intersection.

[0145] In some preferred embodiments, when the system identifies that the ground traffic congestion state is a temporary stop interference state of the feeder vehicle, a set of refined scheduling processes can be started. Specifically, obtaining real-time traffic flow information of each direction in the intersection of the rail transit station can be through various sensor devices such as loop detectors, video detectors, and millimeter wave radars deployed at each entrance and exit of the intersection, real-time collection of vehicle flow, vehicle average speed, vehicle queue length, and lane occupancy rate and other data, and aggregation of these data to the central control unit for processing. According to these real-time traffic flow information, the influence of scheduling of compressing the green light of the disturbed lane and extending the green light of the key lane not affected on the intersection traffic efficiency is evaluated, and the influence evaluation result can be obtained by establishing a microscopic traffic simulation model, such as VISSIM or SUMO, taking real-time traffic flow data as input, simulating the running conditions of the intersection in the future period of time under different green light compression and extension schemes, including average delay, saturation, queue overflow risk and other indicators of each direction, so as to quantify the influence of each scheme on the overall traffic efficiency. For example, the green light of the disturbed lane can be compressed by 5 seconds, 10 seconds, and 15 seconds, and the green light of the key lane not affected can be extended by a corresponding time length, and then the influence of these schemes on the total delay time of the intersection is compared. On this basis, according to the influence evaluation result, adjusting the compression time length of the green light of the disturbed lane and the extension time length of the green light of the key lane not affected can be through an optimization algorithm, such as a genetic algorithm or a particle swarm optimization algorithm, taking minimizing the total delay of the intersection or maximizing the traffic capacity as the objective function, combining the evaluation result, and dynamically calculating the optimal green light compression time length and extension time length. For example, if the evaluation result shows that compressing the green light of the disturbed lane by 10 seconds and extending the green light of the key lane by 10 seconds can obtain the best overall traffic efficiency, the system will determine to adopt this scheme. Finally, performing scheduling according to the adjusted compression time length and extension time length can be through a signal control interface, downloading the calculated new green light time length parameters to the intersection signal, and the signal immediately switches the signal light according to the new timing scheme, thereby realizing dynamic management of traffic flow.

[0146] An intersection signal control system for performing intersection signal control, comprising Figure 9 As shown in the figure, the intersection signal control system 1 comprises:

[0147] An abnormal information acquisition module 11 for acquiring abnormal information of rail transit operation;

[0148] A congestion type determination module 12 for sensing and analyzing the abnormal information to determine the type of ground traffic congestion state;

[0149] The signal scheduling control module 13 is configured to execute a corresponding signal scheduling strategy according to the type of the ground traffic congestion state, wherein the signal scheduling strategy comprises: when the type of the ground traffic congestion state is a bus congestion state, executing scheduling of guiding and releasing buses in a downstream lane; and when the type of the ground traffic congestion state is a temporary interference state of a feeder vehicle, executing scheduling of compressing a green light of an interfered lane and prolonging a green light of a key lane which is not interfered.

[0150] The abnormal information acquisition module is a unit for collecting and processing rail transit operation state data, and can be a data interface or a sensor network, and is used to provide real-time and accurate original data for subsequent traffic state analysis; the congestion type determination module is a unit for in-depth analysis and pattern recognition of the acquired abnormal information to determine the specific nature of the current ground traffic congestion or congestion, and can be a processor with a built-in machine learning algorithm or an expert rule base, and is used to convert the original abnormal information into a specific traffic state type for signal scheduling decision; and the signal scheduling control module is a unit for generating and executing a corresponding traffic signal timing scheme according to the determined ground traffic congestion state type, and can be a programmable logic controller or a traffic signal machine, and is used to optimize the passing efficiency of the intersection and relieve the traffic pressure by dynamically adjusting the green light duration and phase sequence of the signal light.

[0151] The scheme of the present application realizes intelligent control of intersection signals through modular design, thereby coping with traffic problems that may occur during rail transit abnormalities. Specifically, the abnormal information acquisition module serves as the input end of the system and is responsible for real-time acquisition of abnormal information of rail transit operation, such as train delay or suspension, etc., to provide basic data for subsequent decision-making. These abnormal information is then passed to the stranded type determination module, which conducts in-depth perception and analysis on the received abnormal information to identify the specific stranded state type of the current ground traffic, such as bus stranded state or interference state of temporary stop of feeder vehicles. This determination result is the key to the accurate scheduling of the system. Finally, the signal scheduling control module selects and executes the corresponding signal scheduling strategy according to the output result of the stranded type determination module. For example, when the bus stranded state is determined, the system will execute the scheduling of downstream channel dredging and centralized release of buses; when the interference state of temporary stop of feeder vehicles is determined, the system will execute the scheduling of compressing the green light of the disturbed lane and prolonging the green light of the key lane that is not disturbed. This cooperative working mechanism enables the system to dynamically adjust the signal timing according to the actual situation and optimize the passing efficiency of the intersection. The system as a carrier for implementing the above-mentioned intersection signal control method enables the method to be efficiently and automatically executed. By specificizing the method steps into operable system modules, the system of the present application can respond to rail transit abnormalities in real time, convert abstract scheduling logic into actual signal control instructions, thereby overcoming the limitations of the method in execution efficiency and real-time performance, ensuring smooth operation of intersection traffic flow under sudden traffic incidents, and avoiding the decline or even paralysis of the overall passing efficiency of the intersection due to the abnormal surge or stagnation of single-mode traffic flow.

[0152] In some preferred embodiments, the intersection signal control system can be implemented as follows: the abnormal information acquisition module can be a server connected to the rail transit operation data center through an API interface, which is configured with a data collection program and can receive and parse real-time train operation status, fault reports and other abnormal information from the data center. The stagnation type determination module can be a high-performance edge computing unit, which is pre-installed with a traffic flow pattern recognition model based on deep learning. The model can accurately identify various ground traffic stagnation states such as bus stagnation, temporary parking of feeder vehicles and other disturbances by analyzing abnormal information and real-time traffic data from roadside sensors such as video detectors and ground coils. The signal scheduling control module can be an intelligent traffic signal, which is built-in with a variety of preset signal scheduling strategy libraries and connected to the stagnation type determination module through Ethernet. When the stagnation type determination module outputs a specific stagnation state type, the signal scheduling control module can immediately retrieve and execute the corresponding signal scheduling strategy from the strategy library, for example, in the bus stagnation state, the signal machine can automatically adjust the green light phase of the downstream exit lane and cooperate with the dedicated phase for centralized release of buses; in the temporary parking of feeder vehicles and other disturbances, the signal machine can dynamically compress the green light duration of the disturbed lane while extending the green light duration of the key lane that is not disturbed, to ensure the optimization of the overall traffic efficiency of the intersection. The entire system is managed and monitored through an integrated software platform, which can achieve automatic and intelligent control of the intersection signal.

[0153] Through the above technical solutions, the present application provides an intersection signal control system, which can effectively obtain abnormal information of rail transit operation and accurately determine the type of ground traffic stagnation state based on these information. By executing customized signal scheduling strategies according to different stagnation types, such as scheduling for bus stagnation state to dredge downstream channels and centrally release buses, and scheduling for temporary parking of feeder vehicles and other disturbances to compress the green light of disturbed lanes and extend the green light of key lanes that are not disturbed, the system can achieve intelligent and dynamic control of the intersection signal. This enables the system to timely and accurately respond to changes in ground traffic flow caused by rail transit abnormalities, effectively alleviating traffic stagnation and congestion, and avoiding the decline in overall traffic efficiency of the intersection caused by the sudden increase or stagnation of a single traffic mode, thereby improving the traffic capacity and traffic management efficiency of the intersection.

[0154] The above only describes the embodiments of the present application and is not used to limit the protection scope of the present application. For those skilled in the art, the present application can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method of intersection signal control, characterized by, The method comprises the following steps: acquiring abnormal information of rail transit operation; sensing and analyzing the abnormal information to determine the type of ground traffic congestion state; executing corresponding signal scheduling strategy according to the type of ground traffic congestion state; the signal scheduling strategy comprises: when the type of ground traffic congestion state is bus congestion state, executing scheduling of downstream channel dredging and centralized release of buses; and when the type of ground traffic congestion state is temporary parking interference state of feeder vehicle, executing scheduling of compressing green light of interfered lane and prolonging green light of key lane not interfered; when the type of ground traffic congestion state is bus congestion state, the step of executing scheduling of downstream channel dredging and centralized release of buses comprises: starting downstream exit lane dredging phase when the type of ground traffic congestion state is bus congestion state; periodically executing green light pulse and red light buffer in the downstream exit lane dredging phase; monitoring vehicle queue length of U-turn lane area during the red light buffer; adjusting the green light pulse according to the vehicle queue length to adjust the release rhythm of downstream vehicles; ending the downstream exit lane dredging phase when the downstream exit main lane is empty and the vehicle queue length does not reach a preset overflow warning threshold; starting bus centralized release phase after ending the downstream exit lane dredging phase; when the type of ground traffic congestion state is temporary parking interference state of feeder vehicle, the step of executing scheduling of compressing green light of interfered lane and prolonging green light of key lane not interfered comprises: acquiring real-time traffic flow information of each direction in the intersection of rail transit station when the type of ground traffic congestion state is temporary parking interference state of feeder vehicle; evaluating the influence of the scheduling of compressing green light of interfered lane and prolonging green light of key lane not interfered on intersection passing efficiency according to the real-time traffic flow information to obtain influence evaluation result; adjusting the compression duration of the green light of the interfered lane and the prolongation duration of the green light of the key lane not interfered according to the influence evaluation result; executing the scheduling according to the adjusted compression duration and prolongation duration.

2. The method of claim 1, wherein The step of monitoring vehicle queue length of U-turn lane during the red light buffer comprises: acquiring video image information of the U-turn lane area; acquiring millimeter wave radar detection information of the U-turn lane area; identifying vehicle targets and calculating first vehicle queue length according to the video image information; identifying vehicle targets and calculating second vehicle queue length according to the millimeter wave radar detection information; evaluating the effectiveness of the video image information and the millimeter wave radar detection information according to the video image information and the millimeter wave radar detection information to obtain effectiveness evaluation result; fusing the first vehicle queue length and the second vehicle queue length according to the effectiveness evaluation result to obtain vehicle queue length of the U-turn lane.

3. The method of claim 2, wherein The step of evaluating the effectiveness of the video image information and the millimeter wave radar detection information according to the video image information and the millimeter wave radar detection information to obtain effectiveness evaluation result comprises: performing image quality analysis on the video image information to obtain image quality analysis result; performing signal feature analysis on the millimeter wave radar detection information to obtain a signal feature analysis result; comparing the consistency of the corresponding vehicle target in the video image information and the corresponding vehicle target in the millimeter wave radar detection information to obtain a consistency result; judging the effectiveness of the video image information and the millimeter wave radar detection information according to the image quality analysis result, the signal feature analysis result and the consistency result to obtain an effectiveness evaluation result.

4. The method of claim 2, wherein The step of fusing the first vehicle queue length and the second vehicle queue length according to the effectiveness evaluation result comprises: According to the effectiveness evaluation result, the historical effectiveness evaluation results of the video image information and the millimeter wave radar detection information within a preset time period are obtained; Based on the effectiveness evaluation result and the historical effectiveness evaluation result, the dynamic weight of the video image information and the millimeter wave radar detection information is determined; Based on the dynamic weight, the first vehicle queue length and the second vehicle queue length are weighted and fused to obtain a preliminary fused queue length; The preliminary fused queue length is subjected to time series smoothing processing to obtain the vehicle queue length of the U-turn lane.

5. The method of claim 4, wherein The step of determining the dynamic weight of the video image information and the millimeter wave radar detection information based on the effectiveness evaluation result and the historical effectiveness evaluation result comprises: The current effectiveness index of the video image information and the millimeter wave radar detection information is obtained; The historical effectiveness trend index of the video image information and the millimeter wave radar detection information is obtained; Based on the current effectiveness index and the historical effectiveness trend index, the comprehensive reliability score of the video image information and the millimeter wave radar detection information is calculated; According to the comprehensive reliability score, the dynamic weight of the video image information and the millimeter wave radar detection information is determined.

6. The method of claim 4, wherein The step of performing time series smoothing processing on the preliminary fused queue length to obtain the vehicle queue length of the U-turn lane comprises: The exponential smoothing method is used to process the preliminary fused queue length to obtain the vehicle queue length of the U-turn lane.

7. The method of claim 4, wherein The step of performing time series smoothing processing on the preliminary fused queue length to obtain the vehicle queue length of the U-turn lane comprises: The historical data of the preliminary fused queue length is obtained; The historical data of the preliminary fused queue length is subjected to weighted average processing to obtain the vehicle queue length of the U-turn lane.

8. An intersection signal control system for carrying out the intersection signal control method according to claim 7, characterized by It comprises: an abnormal information acquisition module for acquiring abnormal information of rail transit operation; a stagnation type determination module for sensing and analyzing the abnormal information to determine the type of ground traffic stagnation state; a signal scheduling control module for executing corresponding signal scheduling strategies according to the type of ground traffic stagnation state; The signal scheduling strategy comprises: when the type of ground traffic stagnation state is a bus stagnation state, executing a scheduling of downstream channel dredging and centralized release of buses; and when the type of ground traffic stagnation state is a temporary parking of feeder vehicle interference state, executing a scheduling of compressing the green light of the interfered lane and prolonging the green light of the key lane not interfered.

Images