Cross-induction control method, electronic device, and computer-readable storage medium

Abstract

This application proposes a cross-traffic induction control method, electronic device, and computer-readable storage medium, including: acquiring traffic flow data of a target intersection at the current moment and the corresponding traffic light operation data; determining the comprehensive traffic demand of each light group in the target phase based on the traffic flow data; determining the comprehensive traffic demand of the target light group by combining the sensing data collected by the sensing devices deployed in the same direction as the target light group and the sensing data collected by the sensing devices deployed in the opposite direction; determining the gradient vehicle number threshold corresponding to each light group in the target phase based on the running time of the target phase and the preset correspondence between each light group in the target phase and the vehicle number threshold for different running time intervals; and generating control instructions for controlling the traffic lights at the target intersection based on the operation data, the comprehensive traffic demand of each light group in the target phase, and the gradient vehicle number threshold.

Description

Technical Field

[0001] This application relates to the field of traffic control technology, specifically to a cross-traffic sensing control method, electronic equipment, and computer-readable storage medium. Background Technology

[0002] Traffic signal control is a core tool in urban traffic management, and the quality of its control strategies directly determines the efficiency and safety level of intersections. In recent years, with the development of computer vision technology, video-based perception methods have been applied to signal control. By deploying cameras such as electronic police cameras and checkpoints at intersections, image recognition algorithms are used to detect vehicles and pedestrians, making more refined sensing control possible. However, these methods heavily rely on good optical and weather conditions. In situations with insufficient lighting at night, or low visibility conditions such as rain, snow, fog, or haze, video image quality drops sharply, leading to a significant decrease in target recognition rate and an increase in missed detection rate. This prevents the sensing control system from accurately sensing real traffic demand, resulting in problems such as "vehicles not being detected, leading to insufficient clearance time" and reduced vehicle throughput. Furthermore, in existing solutions, when sensing devices (such as cameras) go offline due to malfunctions or transmission problems, the entire sensing control system is often forced to degrade or fail, indicating insufficient system robustness. Summary of the Invention

[0003] To address the aforementioned technical problems, this application provides a cross-sensing control method, electronic device, and computer-readable storage medium that can operate stably and reliably under low visibility conditions, and can synergistically optimize vehicle traffic efficiency and pedestrian crossing safety, thereby improving the robustness of the system.

[0004] This application provides a cross-sensing control method, including: Acquire traffic flow data and corresponding traffic light operation data of the target intersection at the current moment; the traffic flow data is collected and generated by sensing devices deployed at multiple different locations and directions at the target intersection, and the operation data includes the operation plan, the current target stage, the light status of each light group in the target stage, and the running time of the target stage; Based on traffic flow data, the comprehensive traffic demand of each light group in the target phase is determined; specifically, for the target light group, the comprehensive traffic demand of the target light group is determined by combining the perception data collected by the perception devices deployed in the target light group in its own direction and the perception data collected by the perception devices deployed in the opposite direction; the target light group is any light group whose light status is green in the target phase. Based on the running time of the target phase and the pre-defined correspondence between each light group in the target phase and the vehicle number threshold for different running time intervals, determine the current gradient vehicle number threshold for each light group in the target phase. Based on operational data, the overall traffic demand of each traffic light group in the target phase, and the threshold for the number of vehicles at different gradients, control instructions are generated to control the traffic lights at the target intersection.

[0005] In one embodiment, before determining the gradient vehicle number threshold corresponding to each light group in the target stage based on the elapsed runtime of the target stage and the preset correspondence between each light group in the target stage and the vehicle number threshold for different runtime intervals, the following steps are included: Clustering algorithms are used to identify historical traffic flow data of the target intersection at different times within a historical period, and to determine several typical traffic flow patterns, including morning peak mode, evening peak mode, off-peak mode, and nighttime mode. For each traffic flow mode, the correspondence between the elapsed running time of each stage and the gradient vehicle number threshold is analyzed to form a gradient threshold strategy for each traffic flow mode. The gradient threshold strategy includes the correspondence between each light group in each stage and the vehicle number threshold for different running time intervals. Based on the current time period and traffic flow data, dynamically match the current traffic flow pattern and activate the corresponding gradient threshold strategy.

[0006] This application also provides an electronic device, including: a memory and a processor, wherein the memory stores computer program instructions for execution on the processor, and when the processor executes the computer program instructions, it implements the cross-sensing control method as described above.

[0007] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the cross-sensing control method described above.

[0008] As described above, the cross-sensing control method provided in this application solves several technical problems such as unreliable perception under low visibility, low efficiency of green light time utilization, and poor system adaptability by organically combining the cross-sensing architecture, gradient threshold decision mechanism, and closed-loop control system. Ultimately, it achieves a comprehensive technical effect of improving the efficiency and reliability of intersection traffic in harsh environments and significantly reduces the application cost of sensing control. Attached Figure Description

[0009] Figure 1 Schematic diagram of the testing area Figure 1 ; Figure 2 Schematic diagram of the testing area Figure 2 ; Figure 3 This is a flowchart illustrating a cross-sensing control method provided in an embodiment of this application. Detailed Implementation

[0010] First, the relevant terms or nouns involved in the embodiments of this application will be explained.

[0011] Traffic light clusters: These indicate the direction and turning information of at least one traffic flow that simultaneously has the right-of-way. Each traffic light cluster corresponds to a unique identifier. Using numbers to represent traffic light clusters, for example, 1 represents the east left-turn light cluster, 2 represents the east straight-ahead light cluster, 3 represents the east right-turn light cluster, 4 represents the west left-turn light cluster, 5 represents the west straight-ahead light cluster, 6 represents the west right-turn light cluster, 7 represents the south left-turn light cluster, 8 represents the south straight-ahead light cluster, 9 represents the south right-turn light cluster, 10 represents the north left-turn light cluster, 11 represents the north straight-ahead light cluster, and 12 represents the north right-turn light cluster. In practical applications, right-turn light clusters may be disregarded. It should be noted that the names of traffic light clusters can follow the rule of "entry direction + travel direction," for example, the "south straight-ahead light cluster" controls the traffic flow entering from the south entrance and traveling straight through the intersection from the north.

[0012] Sensing devices: These refer to hardware devices deployed at intersections to collect traffic flow data, including but not limited to electronic police systems, checkpoints, traffic flow cameras, radar equipment (such as 4D millimeter-wave radar), or integrated radar-visual systems. These devices are typically installed on light poles, gantries, or other supporting structures at intersections.

[0013] Phases: These include east-west straight traffic, east-west left turns, north-south straight traffic, north-south left turns, eastbound traffic permission, westbound traffic permission, southbound traffic permission, and northbound traffic permission. Each phase typically includes two traffic light groups that allow traffic to proceed. For example, the north-south straight traffic phase includes both southbound and northbound straight traffic light groups.

[0014] Traffic flow plan: This includes a combination of phase sequence and phase duration. For a traffic flow plan, the phases may be in the following order: north-south straight, north-south left turn, east-west straight, and east-west left turn, and the duration of each phase may be a preset value such as 30 seconds.

[0015] Detection Zone: This refers to a virtual area set up at an intersection to detect traffic participants. It is associated with a specific traffic light group and includes vehicle detection zones and pedestrian detection zones. The vehicle detection zone can include an exit vehicle detection zone and an entrance vehicle detection zone. The exit vehicle detection zone is located after the exit lane and is used to detect vehicles that have crossed the stop line at the intersection. Its data is mainly used to determine whether there is a risk of overflow congestion at the intersection. The entrance vehicle detection zone is located at the entrance lane of the intersection and is used to detect vehicles that are about to cross the intersection or are waiting to cross. In addition, in order to more accurately sense or count the number of entrance vehicles in each lane, the entrance vehicle detection zone can be divided into entrance vehicle sub-detection zones corresponding to each lane. In this embodiment, the entrance vehicle detection zone includes a first detection zone, a second detection zone, and a third detection zone, all covered by sensing devices at different locations and targeting the same entrance lane, as an example. For any given traffic light group, the number of traffic objects in the first detection zone bound to that light group can be determined based on the sensing data collected by the opposing sensing devices (i.e., sensing devices installed on the opposite side of the entrance direction controlled by the light group), and the number of traffic objects in the second and third detection zones bound to that light group can be determined based on the sensing data collected by the local sensing devices (i.e., sensing devices installed on the same side of the entrance direction controlled by the target light group). For example, for a north-facing straight-ahead light group, the local sensing devices refer to cameras or radars installed at the north entrance, while the opposing sensing devices refer to cameras or radars installed at the south entrance. Similarly, for a south-facing straight-ahead light group, the local sensing devices refer to cameras or radars installed at the south entrance, while the opposing sensing devices refer to cameras or radars installed at the north entrance. See also... Figure 1 For the northbound traffic lights, the sensing devices deployed in this direction correspond to detection areas a1, b1, and c1, respectively. (See reference...) Figure 2 For the south-facing straight-ahead traffic light group, the detection zones corresponding to the sensing devices deployed in this direction are a2, b2, and c2, respectively. The north-facing straight-ahead traffic light group's corresponding entrance vehicle detection zone includes both the first detection zone (a2) sensed by the south-facing camera or radar, and the second and third detection zones (b1 and c1) sensed by the north-facing camera or radar. Similarly, the south-facing straight-ahead traffic light group's corresponding entrance vehicle detection zone includes both the first detection zone (a1) sensed by the north-facing camera or radar, and the second and third detection zones (b2 and c2) sensed by the south-facing camera or radar. Pedestrian detection zones are set up in the pedestrian crossing (zebra crossing) area to detect pedestrians waiting to cross or crossing the street. It should be noted that in standard intersection phase design, for a straight-ahead traffic light group, the pedestrian detection zone associated with it during the green light phase is usually located on the left or right side of the adjacent zebra crossing area in the direction of travel indicated by that light group. For example, see [link to relevant documentation]. Figure 1 and Figure 2For a crossroads, the zebra crossing area includes the south zebra crossing area, east zebra crossing area, north zebra crossing area, and west zebra crossing area. During the north-south straight-ahead phase, the light groups that gain the right-of-way are the south and north straight-ahead light groups. At this time, the east-west pedestrian flow also gains the right-of-way. In this case, the pedestrian detection zone corresponding to the south straight-ahead light group can be the east zebra crossing area, and the pedestrian detection zone corresponding to the north straight-ahead light group can be the west zebra crossing area. It should be noted that the vehicle detection zones corresponding to the straight-ahead light group and the left-turn light group may be the same.

[0016] This embodiment provides a cross-sensing control method, such as Figure 3 As shown, this cross-traffic induction control method can be applied to traffic control intelligent devices, which are used to directly or indirectly control traffic lights at intersections. Specifically, these devices can be electronic devices such as edge computing devices that communicate with the intersection's sensing devices and traffic signal controllers. The method includes: Step S101: Obtain the traffic flow data of the target intersection at the current time and the corresponding traffic light operation data; the traffic flow data is generated by sensing devices deployed at multiple different locations and directions at the target intersection, and the operation data includes the operation plan, the current target stage, the light status of each light group in the target stage, and the running time of the target stage.

[0017] Step S102: Based on traffic flow data, determine the comprehensive traffic demand of each light group in the target phase; wherein, for the target light group, the comprehensive traffic demand of the target light group is determined by combining the perception data collected by the perception devices deployed in the target light group in its own direction and the perception data collected by the perception devices deployed in the opposite direction; the target light group is any light group whose light status is green in the target phase.

[0018] Step S103: Based on the running time of the target stage and the preset correspondence between each light group in the target stage and the vehicle number threshold for different running time intervals, determine the gradient vehicle number threshold corresponding to each light group in the target stage.

[0019] Step S104: Based on the operational data and the comprehensive traffic demand of each light group in the target stage, as well as the gradient vehicle quantity threshold, generate control instructions for controlling the traffic lights at the target intersection.

[0020] The target intersection is the intersection requiring traffic control, specifically a crossroads, T-junction, or other types of intersection, without particular limitation. Traffic flow data refers to real-time data collected by sensing devices (also known as traffic monitoring devices) deployed at multiple locations and directions at the target intersection, used to describe the traffic conditions at the target intersection. In this embodiment, the sensing devices may include video surveillance equipment and / or radar equipment. Correspondingly, the traffic flow data may include video stream data collected by video surveillance equipment at the target intersection, and may also include radar detection data collected by radar equipment such as 4D radar. Of course, traffic flow data can also be data collected by a radar-visual integrated machine, including video stream data and radar detection data. Video surveillance equipment includes, but is not limited to, devices capable of collecting video and / or images such as electronic police, checkpoints, and traffic flow cameras to provide stable video stream data. In one embodiment, video stream data of vehicle detection areas and pedestrian detection areas near the intersection can be collected by electronic police equipment, while video stream data of vehicle detection areas approaching the intersection can be collected by checkpoint equipment. It should be noted that the detection ranges of different sensing devices may partially overlap or not overlap at all.

[0021] This involves acquiring the operational data of the traffic lights at the target intersection at the current moment. This can be done in real-time or periodically, such as at 1-second intervals. Operational data refers to data acquired in real-time from the traffic signal controllers at the target intersection, describing the current operating status of the traffic lights. Operational data may include the operational plan, the target phase being executed, and the light status of each light group within the target phase. The operational plan indicates the traffic flow control rules for each direction, i.e., the timing plan number and parameters currently being executed by the traffic signal controller. This may include the traffic plan, the operational control logic, and the pedestrian number threshold, minimum green light duration, maximum green light duration, phase downgrade duration, phase advance duration, phase duration, and phase delay duration for each light group within the target phase. The target phase refers to the signal phase being executed by the traffic signal controller at the current moment, such as the "north-south straight phase." The light status of each light group refers to the current light color status (e.g., red, green, yellow) of each traffic signal group (e.g., "east straight," "west left turn") within the target phase.

[0022] The operation plan for traffic lights at the target intersection may differ at different times. For example, the operation plan from 8:00 AM to 9:00 AM may be the morning rush hour plan, while the operation plan from 9:00 AM to 5:00 PM may be the off-peak plan. Similarly, the target phase may vary at different times. For instance, at 7:11:10 AM, the target phase may be north-south straight traffic, while at 7:12:00 AM, the target phase may be north-south left turns. The light groups within a target phase can be those specific to that phase. For example, when the target phase is north-south straight traffic, the light groups in that phase include the north-south straight traffic light group and the south-south straight traffic light group, both with green lights. At the current moment, the light states of the light groups within the target phase may be the same or different; for example, some light groups may be green while others are red.

[0023] In cases where the traffic signal controllers used to control the traffic lights at the target intersection are not integrated into the traffic signal control intelligent device, the device can periodically acquire the current operating data of the traffic lights at the target intersection from the signal controllers. Furthermore, if the traffic signal control intelligent device has stored different operating schemes for the traffic lights at the target intersection, the operating scheme in the operating data can be identified as an operating scheme number, and the device can determine the corresponding operating scheme locally based on the operating scheme identifier and the current time. Of course, the traffic signal controllers can also be integrated into the traffic signal control intelligent device.

[0024] Among them, comprehensive traffic demand refers to a quantitative value representing the traffic demand in a particular direction for a given light group, obtained by fusing multi-source sensing data. Examples include the total number of vehicles in the vehicle inspection zone or the total number of pedestrians in the pedestrian inspection zone. For a target light group, the comprehensive traffic demand is determined by combining sensing data collected by sensing devices deployed in the target light group's own direction and those deployed in the opposite direction. This method can also determine the comprehensive traffic demand of the target light group when the sensing devices deployed in the target light group's own direction are temporarily unavailable or inaccurate due to obstruction, lens damage, offline status, or low visibility.

[0025] The target phase's runtime refers to the time elapsed from the start of the target phase to the current time. For each light group within the target phase, different runtime intervals can be mapped to vehicle quantity thresholds. This mapping allows for the determination of the corresponding gradient vehicle quantity threshold for that light group. The gradient vehicle quantity threshold is a dynamically changing vehicle quantity threshold based on the phase's runtime. This threshold switches according to preset runtime intervals, and its overall trend within the phase's operating cycle is non-decreasing, typically showing a gradual increase. Specifically, for a given target light group, at least two runtime intervals (e.g., first interval, second interval, third interval) are predefined, and a vehicle quantity threshold (e.g., first threshold, second threshold, third threshold) is configured for each interval. If the first interval corresponds to the first threshold, the second interval to the second threshold, and the third interval to the third threshold, then the following conditions must be met: first threshold ≤ second threshold ≤ third threshold, and third threshold > first threshold. At the current time, based on which interval the target phase's runtime falls into, the appropriate gradient vehicle quantity threshold for that light group can be dynamically determined.

[0026] Based on operational data and the comprehensive traffic demand of each traffic light group in the target phase, as well as the threshold for the number of vehicles at different gradients, control commands can be generated to control the traffic lights at the target intersection. These commands allow the traffic signal controller to adjust the state of the traffic lights and / or the phase of traffic flow, thereby achieving a synergistic optimization of vehicle traffic efficiency and pedestrian safety, while significantly reducing the application cost of sensor-based control. The control commands are generated by the intelligent traffic control equipment and sent to the traffic signal controller to directly control changes in the traffic lights. Examples include "stepping commands" (ending the current phase and entering the next phase) or "early termination commands" (ending the green light of a specified traffic light group ahead of schedule).

[0027] For example, taking the north-south straight-ahead phase of a certain intersection as the target phase, the traffic signal control intelligent device first obtains the current operating data from the traffic signal controller in real time (e.g., once per second) and the current traffic flow data from video surveillance equipment and 4D radar through a communication interface. Next, the traffic signal control intelligent device performs fusion analysis on the acquired data to determine the comprehensive traffic demand of each light group (i.e., the south-south and north-south straight-ahead light groups) in the target phase. For example, taking the north-south straight-ahead light group as an example, the video stream is first analyzed using visual algorithms such as YOLO. Then, based on the perception data collected by the electronic police at the north entrance, the number of vehicles in the north-south straight-ahead vehicle detection zone is counted. The vehicle detection zone includes a short-range detection zone (close to the stop line, i.e.,...). Figure 1 (c1) and the mid-range detection area (middle of the entrance, i.e.) Figure 1 (b1 in the text), and based on the sensing data collected by the electronic police at the south entrance, the long-distance detection area at the south entrance (i.e. Figure 2 The number of vehicles in section a2) is calculated, and then the vehicle numbers in these two sections are added together to obtain the comprehensive traffic demand for the northbound straight-ahead traffic light group. Next, based on the operational data, the comprehensive traffic demand of each traffic light group in the target phase, and the gradient vehicle number threshold, it is determined whether the current phase should be terminated, and specific control instructions are generated.

[0028] In one embodiment, traffic flow data includes video stream data, and comprehensive traffic demand includes the number of traffic objects within the detection area; based on the traffic flow data, the comprehensive traffic demand for each light group in the target phase is determined, including: Object annotation is performed on video stream data using a trained YOLO model; Based on the video stream data after object labeling, the number of traffic objects in the detection area bound to each light group in the target stage is counted. The detection area includes the import vehicle detection area, which includes the first detection area, the second detection area, and the third detection area, which are covered by sensing devices at different locations and target the same import lane. The number of vehicles in the import vehicle detection area is the sum of the number of vehicles identified in the first detection area, the second detection area, and the third detection area.

[0029] In this context, video stream data can be viewed as a sequence of dynamic images continuously collected and transmitted by video sensing devices (such as electronic police cameras, checkpoints, and traffic cameras) deployed at the target intersection. The YOLO model refers to an advanced object detection algorithm developed by Alibaba DAMO Academy. This model is based on the YOLO architecture and incorporates techniques such as a neural architecture search backbone network and reparameterized generalized FPN. It has been pre-trained on large public datasets such as COCO and Objects365, and has been specifically optimized for targets such as vehicles and pedestrians in traffic scenarios, achieving a good balance between accuracy and speed. In this embodiment, the collected low-visibility video stream can be segmented at 1 frame per second and saved as an image dataset. Initial annotation using YOLO is then performed, followed by programmatic removal of labeled objects irrelevant to the target classification, and manual removal of incorrectly labeled objects. Unlabeled target objects are added to form a training dataset, which is then used to train the YOLO model. Object annotation refers to the process of using the YOLO model to analyze each frame of the video stream, identify the traffic participants (such as vehicles and pedestrians) contained within it, mark their locations with bounding boxes, and classify their types. The first, second, and third detection zones refer to the detection areas covered by sensing devices deployed at different locations for the same entrance lane. These three detection zones are spatially distributed along the vehicle's direction of travel, corresponding to long-range, medium-range, and short-range detection functions, respectively. Those skilled in the art should understand that the terms "first," "second," and "third" are used only to distinguish different detection zones and do not indicate any priority or difference in importance.

[0030] For example, taking the northbound straight-ahead traffic light cluster as an example, the first step is to configure the intersection's sensing equipment, including: configuring a southbound electronic police system to cover the first detection zone of the northbound entrance; configuring a northbound checkpoint to cover the second detection zone of the northbound entrance; and configuring a northbound electronic police system to cover the third detection zone of the northbound entrance. Next, the three vehicle detection zones are linked to the northbound straight-ahead traffic light cluster, including: using the southbound electronic police system imagery, delineating a polygonal area, such as an octagonal area, upstream of the northbound exit lane as the first detection zone; using the northbound checkpoint imagery, delineating an octagonal area in the middle of the northbound entrance lane as the second detection zone; and using the northbound electronic police system imagery, delineating an octagonal area upstream of the northbound entrance lane as the third detection zone. Next, the traffic control intelligent equipment performs the following operations in each analysis cycle (e.g., per second): simultaneously acquires real-time video streams from the south entrance electronic police, the north entrance checkpoint, and the north entrance electronic police; uses the deployed YOLO model to perform real-time analysis on the three video streams, identifies vehicle targets in each frame of the image, and generates annotation results with location and category information; for each detection area, counts the number of vehicles labeled within that area; and adds up the number of vehicles in the three detection areas to obtain the total number of vehicles in the north straight-ahead traffic light group's designated entrance vehicle detection area.

[0031] Understandably, by summing the data, even if one inspection zone temporarily malfunctions or becomes inaccurate due to obstruction, lens damage, or visibility issues, data from other inspection zones can still provide support, ensuring the reliability and continuity of vehicle count acquisition. It should be noted that if a vehicle is detected in two inspection zones, it will only be counted once in the imported vehicle inspection zone. This improves reliability in low visibility conditions, while the collaborative enhancement of perception across multiple inspection zones ensures that even if one inspection zone malfunctions due to obstruction or equipment failure, other inspection zones can still provide valid data.

[0032] In one embodiment, based on the elapsed runtime of the target phase and the preset correspondence between each light group in the target phase and the vehicle number threshold for different runtime intervals, the gradient vehicle number threshold corresponding to each light group in the target phase is determined, including: If the running time of the target phase is within the first running time interval corresponding to the target light group and the first running time interval corresponds to the first threshold, then the gradient vehicle number threshold corresponding to the target light group is determined as the first threshold; the target light group is any light group in the target phase whose light state is green, and the start time of the first running time interval is greater than or equal to the minimum green light duration of the target phase. If the running time of the target stage is within the second running time interval corresponding to the target light group and the second running time interval corresponds to the second threshold, then the gradient vehicle number threshold corresponding to the target light group is determined as the second threshold; the second threshold is greater than or equal to the first threshold. If the running time of the target phase is within the third running time interval corresponding to the target light group and the third running time interval corresponds to the third threshold, then the gradient vehicle number threshold corresponding to the target light group is determined as the third threshold; the start time of the third running time interval is less than or equal to the maximum green light duration of the target phase, and the third threshold is greater than or equal to the second threshold.

[0033] In the sensing control process, for each light group whose light status is green in the target phase, the target running time interval corresponding to the target phase running time can be determined first based on the running time of the target phase. Then, based on the target running time interval, the correspondence between the light group and the vehicle number threshold for different running time intervals is queried to determine the vehicle number threshold corresponding to the target running time interval, and this threshold is used as the gradient vehicle number threshold for the light group. The first threshold, second threshold, and third threshold can be set according to actual needs. For example, for a certain light group, the corresponding first threshold can be set to 1, the second threshold to 2, and the third threshold to 3, etc.

[0034] For example, taking the target phase as the north-south straight-ahead phase, the runtime intervals include the first duration interval (20, 30], the second duration interval (30, 40], and the third duration interval (40, 50]. For the south-south straight-ahead light group in the north-south straight-ahead phase, the vehicle number threshold corresponding to the first duration interval (20, 30] can be 1, the vehicle number threshold corresponding to the second duration interval (30, 40] can be 1, and the vehicle number threshold corresponding to the third duration interval (40, 50) can be 2; while for the north-south straight-ahead light group in the north-south straight-ahead phase, the vehicle number threshold corresponding to the first duration interval (20, 30] can be 1, the vehicle number threshold corresponding to the second duration interval (30, 40] can be 2, and the vehicle number threshold corresponding to the third duration interval (40, 50) can be 3, etc.

[0035] In this way, through gradient design, the problem of "green light being idle and inefficient" caused by the arrival of sporadic vehicles in the later stages of the green light in traditional inductive control can be effectively avoided. This prompts the inductive control system to terminate the current stage in a timely manner when the utilization rate of the green light time decreases, and switch the right of way to other directions of traffic with demand, thereby improving the overall traffic efficiency of the intersection.

[0036] In one embodiment, before determining the gradient vehicle number threshold corresponding to each light group in the target stage based on the elapsed runtime of the target stage and the preset correspondence between each light group in the target stage and the vehicle number threshold for different runtime intervals, the following steps are included: Clustering algorithms are used to identify historical traffic flow data of the target intersection at different times within a historical period, and to determine several typical traffic flow patterns, including morning peak mode, evening peak mode, off-peak mode, and nighttime mode. For each traffic flow mode, the correspondence between the elapsed running time of each stage and the gradient vehicle number threshold is analyzed to form a gradient threshold strategy for each traffic flow mode. The gradient threshold strategy includes the correspondence between each light group in each stage and the vehicle number threshold for different running time intervals. The current traffic flow pattern is dynamically matched according to the time period to which the current time belongs, and the corresponding gradient threshold strategy is activated.

[0037] By learning from and analyzing historical traffic flow data at target intersections, the system can be endowed with predictive and self-optimizing capabilities. Specifically, cluster analysis can be used to identify typical traffic flow patterns at different times (such as morning rush hour and off-peak hours), and optimal gradient threshold strategies can be established for each pattern. Furthermore, reinforcement learning models can be employed to continuously optimize these thresholds with the goal of reducing the average delay at intersections. In addition, periodic patterns such as weekdays and holidays can be identified and applied to adjust control parameters in advance when facing foreseeable traffic changes, thereby achieving a leap from passive sensing to proactive prediction.

[0038] For example, the traffic control intelligent device can store historical data from the past month. Every day at midnight, an analysis task can be initiated: First, the K-Means clustering algorithm is used to divide past weekday data into four modes: "morning peak (7:00-9:00)," "evening peak (17:00-19:00)," "off-peak (other times)," and "nighttime (22:00-5:00 the next day)." Then, for the "north-south straight-ahead" phase, statistics are compiled: in "off-peak mode," when the phase has run for 20-30 seconds, only one vehicle entering the south-south straight-ahead light group is needed to continue traffic flow, achieving the highest efficiency; while in "morning peak mode," due to denser traffic, at least two vehicles are needed to justify continuing traffic flow. Thus, different gradient threshold strategies are generated for the "off-peak" and "morning peak" modes. At 7:00 AM that day, the system automatically switches to "morning peak" mode and activates the corresponding gradient threshold strategy for sensor control.

[0039] In one embodiment, based on operational data and the overall traffic demand of each traffic light group in the target phase, as well as the gradient vehicle quantity threshold, control instructions for controlling the traffic lights at the target intersection are generated, including: Based on the operational data and the number of traffic objects in the detection area bound to each light group in the target phase, as well as the gradient vehicle number threshold, determine the phase vehicle state value and phase pedestrian state value of the target phase at the current moment. Based on the vehicle state value and pedestrian state value of the target stage at the current moment, determine the stage signal value of the target stage at the current moment; Based on the target stage's stage signal value at the current moment, determine the target stage's stage decision value at the current moment; Based on the target stage's stage decision value at the current moment, generate control instructions for controlling the traffic lights at the target intersection.

[0040] The phase vehicle state value quantifies whether there is still traffic demand in all directions served by the target phase at the current moment; the phase pedestrian state value quantifies whether there is still traffic demand in all directions associated with the target phase at the current moment. The phase decision value characterizes the control strategy for the target phase at the current moment, indicating whether the current target phase should continue or terminate immediately. The phase signal value is a comprehensive signal derived from the current phase vehicle and pedestrian states, used to characterize the overall traffic condition and health of the phase.

[0041] In one embodiment, the detection area further includes a pedestrian detection area. The operation plan includes computational control logic and pedestrian number thresholds for the pedestrian detection areas corresponding to each light group in the target stage. Based on the operation data and the number of traffic objects in the detection areas bound to each light group in the target stage, as well as the gradient vehicle number thresholds, the stage vehicle state value and stage pedestrian state value of the target stage at the current moment are determined, including: At the current moment, if the number of vehicles in the imported vehicle detection area corresponding to the target light group is empty or invalid, the vehicle detection status of the target light group at the current moment is marked as the third preset value; if the number of vehicles in the imported vehicle detection area corresponding to the target light group is greater than or equal to the gradient vehicle number threshold, the vehicle detection status of the target light group at the current moment is marked as the first preset value; otherwise, the vehicle detection status of the target light group at the current moment is marked as the second preset value. If, in the m consecutive moments preceding the current moment, the number of times the vehicle detection status value of the target light group is greater than or equal to the number of times the vehicle detection status value of the target light group is the third preset value, then the vehicle status value of the target light group at the current moment is marked as the third preset value; if the number of times the vehicle detection status value of the target light group is the second preset value is greater than or equal to the number of times the vehicle detection status value of the target light group is the second preset value, then the vehicle status value of the target light group at the current moment is marked as the second preset value; otherwise, the vehicle status value of the target light group at the current moment is marked as the first preset value. If the number of pedestrians in the pedestrian detection area corresponding to the target light group is greater than or equal to the pedestrian number threshold, then the pedestrian status value of the target light group at the current time is marked as the first preset value; otherwise, the pedestrian status value of the target light group at the current time is marked as the second preset value. When the operation control logic is the first operation logic, mark the vehicle state value of the target stage at the current time as the minimum value of the vehicle state value of all light groups in the target stage at the current time, and mark the pedestrian state value of the target stage at the current time as the minimum value of the pedestrian state value of all light groups in the target stage at the current time. When the operation control logic is the second operation logic, if at least one light group in the target stage has a vehicle state value of the first preset value at the current time, then mark the target stage vehicle state value of the current time as the first preset value; otherwise, mark the target stage vehicle state value of the current time as the minimum value of the vehicle state values ​​of all light groups in the target stage at the current time. If at least one light group in the target stage has a pedestrian state value of the first preset value, then mark the target stage pedestrian state value of the current time as the first preset value; otherwise, mark the target stage pedestrian state value of the current time as the second preset value.

[0042] Here, vehicle status value and pedestrian status value refer to a status identifier representing the current traffic demand of a single light group, derived by comparing its corresponding detection result with the corresponding quantity threshold. Based on the vehicle status value and pedestrian status value of each light group at the current moment and the calculation control logic, the stage vehicle status value and stage pedestrian status value of the target stage at the current moment can be determined. It should be noted that, in this embodiment, obtaining the vehicle status value and pedestrian status value of each light group in the target stage at the current moment can be understood as obtaining the vehicle status value and pedestrian status value of each light group in the target stage whose light state is green at the current moment.

[0043] The gradient vehicle number threshold and pedestrian number threshold refer to the minimum number of traffic participants required to determine whether a traffic light group has a need for passage. When the number of target objects in the corresponding detection area reaches or exceeds the threshold, it is considered that there is a valid need for passage in that direction. The vehicle detection status value refers to the instantaneous status identifier obtained after judging the situation in the vehicle detection area corresponding to the target traffic light group based on a single detection (i.e., a frame of data at the current moment). If the number of vehicles in the vehicle detection area corresponding to the target traffic light group is empty or invalid, it may be due to a disconnection or transmission interruption of the sensing device associated with the target traffic light group. The first, second, and third preset values ​​can be set according to actual needs. For example, the first preset value can be set to 1, the second preset value can be set to 0, and the third preset value can be set to 2, etc. m can be set to 8 or 10, etc.; the first preset number of times is less than m, and the first preset number of times can be set to 4 or 5, etc.; the pedestrian number threshold can be set to 1 or 2, etc.

[0044] Specifically, if the number of vehicles in the import vehicle detection area corresponding to the target light group is greater than or equal to the gradient vehicle number threshold, it indicates that there are vehicles or a large number of vehicles in the import vehicle detection area corresponding to that light group at the current moment, and the vehicle detection status value of the target light group at the current moment is marked as the first preset value. If the number of vehicles in the import vehicle detection area corresponding to the target light group is less than the vehicle number threshold, it indicates that there may be no vehicles or a small number of vehicles in the vehicle detection area corresponding to that light group at the current moment, and the vehicle detection status value of the target light group at the current moment is marked as the second preset value. If, over a consecutive m time interval from the current time, the number of times the vehicle detection status value of the target light group is greater than or equal to the number of times the value of the third preset value is greater than or equal to the number of times the value of the first preset value is greater than or equal to the number of times the value of the first preset value is greater than or equal to the number of times the value of the first preset value is greater than or equal to the number of times the value of the second ...

[0045] It should be noted that if the sensing device associated with the target light group is offline, resulting in an empty or invalid vehicle count in the vehicle detection zone corresponding to the target light group, the number of pedestrians in the pedestrian detection zone corresponding to the target light group can be determined by analyzing the video stream data collected by sensing devices associated with other light groups in the traffic flow data. For example, taking a crossroads as the target intersection and a north-south straight-ahead phase as an example, if the electronic traffic enforcement camera associated with the north-south straight-ahead light group is offline, the traffic control intelligent equipment cannot obtain the number of pedestrians in the pedestrian detection zone (i.e., the western zebra crossing area) corresponding to the north-south straight-ahead light group captured by the electronic traffic enforcement camera associated with the north-south straight-ahead light group. However, since the electronic traffic enforcement camera associated with the western straight-ahead light group can capture the western zebra crossing area, the number of pedestrians in the western zebra crossing area can be determined through the video stream data collected by the electronic traffic enforcement camera associated with the western straight-ahead light group, thus determining the number of pedestrians in the pedestrian detection zone corresponding to the north-south straight-ahead light group.

[0046] The operational control logic, also known as the compact logic, is used to instruct the operational logic during the generation of vehicle and pedestrian state values ​​for the target stage. It includes a first operational logic and a second operational logic. The first operational logic, also called compact logic, takes a cautious approach to traffic demand assessment throughout the stage. If traffic demand in any direction within the stage does not reach a threshold (i.e., the state value is not "demanding"), the entire stage is considered demand-free. This effectively prevents one direction from being "over-served" while other directions are "starved" in scenarios with high traffic volume, the need for efficient queue clearing, or extremely high safety requirements, ensuring that all directions have no significant traffic demand by the end of the stage. The second operational logic, also known as loose logic, is more proactive in assessing traffic demand. If there is traffic demand in any direction within the stage, the entire stage is considered demanding. This allows for rapid response to sudden traffic surges in scenarios with low traffic volume and the need to minimize vehicle and pedestrian waiting times. It avoids missing service opportunities due to minor misjudgments by a single detector, improving system response speed and user experience.

[0047] Specifically, when the operation control logic is the second operation logic, for the stage vehicle state value, it first checks whether there is at least one vehicle light group in the target stage with a vehicle state value of the first preset value, such as 1. If so, the corresponding stage vehicle state value is directly marked as the first preset value; otherwise, it is marked as the minimum value of the vehicle state values ​​of all vehicle light groups. For the stage pedestrian state value, it first checks whether there is at least one pedestrian light group in the target stage with a pedestrian state value of the first preset value, such as 1. If so, the corresponding stage pedestrian state value is directly marked as the first preset value; otherwise, it is marked as the second preset value, such as 0. In this way, the operation logic can be configured or dynamically switched according to the time period, intersection characteristics, or real-time traffic conditions to achieve a balance between control accuracy and sensitivity, thereby improving control flexibility.

[0048] In one embodiment, determining the stage signal value of the target stage at the current moment based on the stage vehicle state value and stage pedestrian state value of the target stage at the current moment includes: If the vehicle status value of the target stage at the current moment is the first preset value and all light groups in the target stage meet the vehicle idle release confirmation rate, and the pedestrian status value of the target stage at the current moment is the second preset value and all light groups in the target stage meet the pedestrian idle release confirmation rate, then mark the stage signal value of the target stage at the current moment as the fourth preset value. If the vehicle status value of the target stage at the current moment is the third preset value and all light groups in the target stage meet the offline confirmation rate, then mark the stage signal value of the target stage at the current moment as the fifth preset value. Otherwise, the stage signal value of the target stage at the current moment is the sixth preset value.

[0049] Specifically, the vehicle vacancy confirmation rate of a light group is such that, in a consecutive n-timeframe from the current time, the number of times the number of vehicles in the import vehicle detection area corresponding to that light group is less than a preset number of vehicles is greater than or equal to a first preset occurrence number; the pedestrian vacancy confirmation rate of a light group is such that, in a consecutive n-timeframe from the current time, the number of times the number of pedestrians in the pedestrian detection area corresponding to that light group is less than a preset number of pedestrians is greater than or equal to a second preset occurrence number; and the offline confirmation rate of a light group is such that, in a consecutive n-timeframe from the current time, the number of times the number of vehicle objects in the import vehicle detection area corresponding to that light group is empty or invalid is greater than or equal to a third preset occurrence number.

[0050] Among them, the stage signal value refers to the comprehensive signal derived from the current stage vehicle and pedestrian status, used to characterize the overall traffic condition and health of the stage. It is usually represented by a multi-state discrete value. For example, the sixth preset value such as 1 can be used to indicate that there is a clear vehicle or pedestrian passage demand in at least one direction in the current stage, the fourth preset value such as 0 can be used to indicate that the vehicle and pedestrian passage demand in all directions in the current stage has been met, and the fifth preset value such as 2 can be used to indicate that the sensing device has malfunctioned and cannot obtain reliable traffic flow data.

[0051] The parameters n, the preset number of vehicles, and the preset number of pedestrians can be set according to actual needs. For example, n can be set to 6 or 8, the preset number of vehicles can be set to 2 or 4, and the preset number of pedestrians can be set to 2 or 3. The first preset occurrence count, the second preset occurrence count, and the third preset occurrence count can be set according to the size of n, and all of them are less than n. For example, the first preset occurrence count, the second preset occurrence count, and the third preset occurrence count can all be set to 2, 3, or 4. Specifically, if the vehicle status value at the current moment in the target stage is not the first preset value and / or all light groups in the target stage do not meet the vehicle idle confirmation rate, and / or the pedestrian status value at the current moment in the target stage is not the second preset value and all second-type light groups in the target stage do not meet the pedestrian idle confirmation rate, and / or the vehicle status value at the current moment in the target stage is not the third preset value and / or at least one light group in the target stage does not meet the offline confirmation rate, the target stage signal value at the current moment is marked as the sixth preset value.

[0052] In one embodiment, determining the stage decision value of the target stage at the current moment based on the stage signal value of the target stage at the current moment includes: When the first objective condition is met, the stage decision value of the target stage at the current moment is determined to be the seventh preset value; the first objective condition includes any one of the following conditions: the current moment has not reached the minimum decision time of the target stage, the minimum decision time of the stage is the difference between the stage duration of the target stage and the first preset duration threshold, and the first preset duration threshold is the sum of the stage advance duration of the target stage and the first preset duration; the current moment has reached the minimum decision time of the target stage but has not reached the maximum decision time of the target stage, and the second objective condition is met, the maximum decision time of the stage is the difference between the stage duration of the target stage and the second preset duration threshold, and the second preset duration threshold is the difference between the stage delay duration and the second preset duration; The second target condition includes any one of the following conditions: the target stage's phase signal value at the current moment is the sixth preset value and there exists a non-target stage's target vehicle signal priority request value that is not the ninth preset value; the target stage's phase signal value at the current moment is the sixth preset value and the target stage's phase pedestrian state value at the current moment is the first preset value; the target stage's phase signal value at the current moment is the fifth preset value, the target stage's elapsed duration is less than the difference between the target stage's phase downgrade duration and the third preset duration threshold, and there exists a non-target stage's target vehicle signal priority request value that is not the ninth preset value; the target stage's phase signal value at the current moment is the fifth preset value, the target stage's elapsed duration is greater than or equal to the difference between the target stage's phase downgrade duration and the fourth preset duration threshold, and the target stage's phase pedestrian state value at the current moment is the first preset value. When the third objective condition is met, the stage decision value of the objective stage at the current moment is determined to be the eighth preset value; The third objective condition includes any one of the following: the current time has reached the maximum decision time of the objective stage; the current time has reached the minimum decision time of the objective stage but has not reached the maximum decision time of the objective stage, and the fourth objective condition is satisfied. The fourth target condition includes any one of the following conditions: the target stage's stage signal value at the current moment is the fourth preset value; the target stage's stage decision value at the previous moment is the eighth preset value; the target stage's stage signal value at the current moment is the fifth preset value, the target stage's runtime is greater than or equal to the difference between the target stage's stage downgrade duration and the fifth preset duration threshold, and the target stage's pedestrian state value at the current moment is the second preset value; there exists a non-target stage's target vehicle signal priority request value of the ninth preset value and the target stage's pedestrian state value at the current moment is the second preset value.

[0053] Specifically, if, up to the current moment, the duration of the target vehicle signal priority status value of a light group in the non-target phase being the tenth preset value is greater than or equal to the third preset duration, then the target vehicle signal priority request value in the non-target phase at the current moment is marked as the ninth preset value; if, at the current moment, the number of target vehicles in the import vehicle detection area corresponding to that light group in the non-target phase is greater than the preset number of target vehicles, then the target vehicle signal priority status value of the target light group at the current moment is marked as the tenth preset value.

[0054] Since the stage vehicle state value and stage pedestrian state value reflect the overall traffic demand status of all traffic flows (vehicle or pedestrian) in the current stage, the stage signal value of the target stage at the current moment can be determined based on the stage vehicle state value and stage pedestrian state value of the target stage at the current moment. Furthermore, the stage decision value of the target stage at the current moment can be determined based on the stage signal value of the target stage at the current moment. Target vehicles refer to special vehicles with priority right-of-way, such as ambulances, fire trucks, or police cars.

[0055] The minimum decision time for a phase refers to the earliest time the system is allowed to make a decision to end the current phase after the start of a signal phase. Before this moment, the current phase must continue regardless of traffic demand to ensure basic passage time. The phase advance duration refers to the maximum time window during which a phase can start earlier than planned, i.e., the duration by which the green light can illuminate earlier than the preset cycle phase, such as 3 seconds or 6 seconds. The first preset duration is a fixed safety buffer time (e.g., 4 or 5 seconds) to ensure the stability and safety of the decision. The maximum decision time for a phase refers to the latest time the system is allowed to make a decision to end the current phase within a signal phase. Beyond this time, the system will forcibly end or maintain the phase according to rules to prevent the phase from being extended indefinitely. The phase delay duration refers to the maximum time window during which a phase can be delayed from its planned end, i.e., the duration by which the green light can be delayed from the preset cycle phase, such as 4 seconds or 6 seconds. The second preset duration is a fixed processing buffer time (e.g., 4 or 5 seconds) to ensure that control commands are issued and executed before the phase truly ends. The phase downgrade duration is a fixed backup green light duration, such as 20 seconds or 30 seconds. The priority request value is a signal used to identify whether an emergency or priority passage request is issued during a non-target phase. For example, it is triggered when a 4D radar detects an ambulance, fire truck, or other special vehicle. The target vehicle signal priority status value refers to the instantaneous state for a single light group, determining whether a target vehicle (such as a special vehicle) exists within its entrance detection zone. The seventh, eighth, ninth, and tenth preset values ​​can be set according to actual needs; for example, the seventh preset value can be set to 1, the eighth preset value to 0, the ninth preset value to 3, and the tenth preset value to 1, etc.

[0056] Optionally, if the minimum decision time for the target stage has not been reached at the current time, it indicates that the target stage cannot be adjusted, and the stage decision value for the target stage at the current time is determined to be the seventh preset value. If the minimum decision time for the target stage has been reached at the current time but the maximum decision time for the target stage has not been reached, it indicates that the target stage can be adjusted, and the stage decision value for the target stage at the current time is determined based on one or more of the following: the stage signal value of the target stage at the current time, the stage pedestrian state value of the target stage at the current time, the elapsed duration of the target stage, and the target vehicle signal priority request value of the non-target stage.

[0057] In one embodiment, the operation plan further includes a threshold for the number of exit vehicles corresponding to each light group in the target phase, and the detection result also includes the number of vehicles in the exit vehicle detection area corresponding to the light group; based on the phase decision value of the target phase at the current time, control instructions for controlling the traffic lights at the target intersection are generated, including: When the target stage's stage decision value at the current moment is the seventh preset value, if there is a target lamp group whose overflow signal value at the current moment is the preset target value, then a lamp group overflow early interruption instruction for the target lamp group is generated. When the target stage's stage decision value at the current moment is the eighth preset value, a step instruction to end the target stage is generated. Specifically, when the number of vehicles in the exit vehicle detection area corresponding to the target light group is greater than or equal to the exit vehicle number threshold, the overflow state value of the target light group at the current moment is marked as the first state value; otherwise, it is marked as the second state value. When the number of times the overflow state value of the target light group is the first state value in the k consecutive moments traced back from the current moment reaches the preset overflow number, the overflow signal value of the target light group at the current moment is marked as the preset target value.

[0058] It should be noted that when the target stage's current stage signal value is the fifth preset value, and the target stage's runtime is greater than or equal to the difference between the target stage's stage degradation duration and the fifth preset duration threshold, it indicates that the sensing device corresponding to the target light group is offline, and the target stage's runtime meets the stage end condition. However, since the target stage's current stage pedestrian status value is the second preset value, meaning there are still pedestrians in the pedestrian detection area corresponding to the target light group, the target stage needs to continue to be executed to ensure that pedestrians crossing the street when the green light is about to end or those crossing at a slower speed can cross the street safely, greatly reducing the occurrence of pedestrian-vehicle conflicts.

[0059] Optionally, when the stage decision value of the target stage at the current moment is the seventh preset value, it indicates that the target stage can continue to be executed. However, when the overflow signal value of the target light group (i.e., any light group in the target stage) at the current moment is a preset target value such as 1, it indicates that there are many vehicles in the exit vehicle detection area corresponding to the target light group at the current moment, and it is necessary to stop vehicles from entering the exit vehicle detection area corresponding to the target light group. Therefore, when the target light group is in green light condition, a light group overflow early interruption instruction can be generated for the target light group to adjust the target light group's light condition to red light condition. If the target light group's light condition is red light condition, no control instruction for the target light group is generated. When the stage decision value of the target stage at the current moment is the eighth preset value, it indicates that the execution of the target stage needs to be terminated. In this case, a step instruction for terminating the target stage is generated to quickly terminate the target stage.

[0060] It should be noted that when the traffic signal controller is not integrated into the traffic control intelligent device, the traffic control intelligent device can send detector instructions corresponding to different light groups to the traffic signal controller according to the control instructions, so that the traffic signal controller can perform corresponding control operations on the traffic lights at the target intersection. In this embodiment, the communication connection between the traffic signal controller and the traffic control intelligent device can be, for example, through protocol messages. The traffic signal controller may include a local fixed-cycle control mode and an inductive control mode. In the local fixed-cycle control mode, the traffic signal controller will control the traffic lights to execute different traffic schemes according to a preset cycle. In the inductive control mode, the traffic signal controller will control the traffic lights according to the control instructions sent by the traffic control intelligent device. For example, after receiving the inductive control instruction sent by the traffic control intelligent device, the traffic signal controller will enter the inductive control mode accordingly, and send the current operating data to the traffic control intelligent device periodically or in real time. At the same time, it can control the light status of each light group of the traffic lights at the target intersection according to the control instructions sent by the traffic control intelligent device. In addition, traffic control intelligent devices can also communicate with intelligent traffic control platforms (such as cloud servers) to report their own status information to the intelligent traffic control platform at regular or irregular intervals. When the intelligent traffic control platform detects that the traffic control intelligent device is offline (such as not receiving status information sent by the traffic control intelligent device for a certain period of time), it can send a timed control timing scheme (such as a daily plan) to the traffic signal controller, so that the traffic signal controller exits the induction control mode and controls the traffic lights at the target intersection based on the timed control traffic scheme.

[0061] Optionally, the cross-sensing control method provided in this embodiment may also include displaying a parameter configuration page, where users can configure all the parameters mentioned in this embodiment according to actual needs to adapt to different control scenarios.

[0062] It should be noted that in this embodiment, the vehicle status value, pedestrian status value, signal value, and decision value of other stages other than the target stage can also be obtained at the current time based on the traffic flow data of the target intersection at the current time and the operation data of the corresponding traffic lights, referring to the method for obtaining the parameters of the target stage in this embodiment.

[0063] In summary, the cross-traffic sensing control method provided in the above embodiments, on the one hand, determines the comprehensive traffic demand of the target light group by integrating the sensing data collected by the sensing devices deployed in the target light group in its own direction and the sensing data collected by the sensing devices deployed in the opposite direction. This allows for mutual verification and supplementation of sensing data from different perspectives (especially the opposite perspective) when facing low visibility conditions such as nighttime, rain, or fog that reduce the quality of video perception in one direction, or when the sensing devices deployed in the target light group in its own direction are offline. This significantly reduces the missed detection rate and ensures accurate and reliable perception of traffic demand at the intersection. On the other hand, by using the running time of the target phase and the preset correspondence between each light group in the target phase and the vehicle number threshold for different running time intervals, the gradient vehicle number threshold corresponding to each light group in the target phase is determined. This establishes a gradient strategy that dynamically increases the vehicle number threshold over time. Thus, at the beginning of the green light, a lower vehicle number threshold can quickly respond to the initial queue of vehicles, while as the green light time extends, a higher vehicle number threshold requires a sufficient number of vehicles to arrive before it is worthwhile to continue allowing traffic. This effectively avoids the problem of "empty runs and low efficiency" caused by sporadic vehicle arrivals during the later stages of a green light. It forces the sensing control system to terminate the current phase promptly when the efficiency of green light usage decreases, allocating valuable green light time to other more demanding traffic directions. This reduces the average delay for all vehicles at the intersection and improves overall traffic efficiency. On the other hand, by acquiring and fusing multi-source heterogeneous data, and then making decisions based on a comparative analysis of real-time traffic demand and dynamic gradient thresholds, control commands are ultimately generated and executed. This closed-loop architecture transforms signal control from a static, pre-set scheme into a dynamic, real-time traffic demand-driven intelligent response system. It can adapt to the dynamic changes in intersection traffic flow and make intelligent decision-making and control, forming a complete closed-loop automatic control from perception, decision-making to execution. Furthermore, the traffic flow data in this application can be obtained from existing traffic monitoring equipment at the intersection, such as electronic police systems, checkpoints, and integrated radar-visual systems. This allows for the direct use of existing and widely available sensing devices as data sources, significantly reducing the hardware deployment and maintenance costs of the sensing control system and laying the foundation for the large-scale promotion of sensing control technology. In other words, the cross-sensing control method provided in this application solves several technical problems, such as unreliable perception under low visibility, low efficiency of green light time utilization, and poor system adaptability, through the organic combination of cross-sensing architecture, gradient threshold decision mechanism and closed-loop control system. Ultimately, it achieves a comprehensive technical effect of improving the traffic efficiency and reliability of intersections in harsh environments and significantly reduces the application cost of sensing control.

[0064] Based on the same inventive concept as the foregoing embodiments, the cross-sensing control method provided in this embodiment will be specifically described below through a specific example. In this example, the sensing device is the sensing device, the signal control intelligent device is the signal control intelligent agent, the target vehicle is a special vehicle, the first detection area is a long-distance detection area, the second detection area is a medium-distance detection area, the third detection area is a short-distance detection area, and the inlet detection area is an inlet coil.

[0065] The cross-sensing control method provided in this embodiment can be implemented by a cross-sensing control system. This system may include a sensing device, a signal controller, and a signal control agent communicatively connected to both the sensing device and the signal controller, as well as an intelligent signal control platform communicatively connected to both the signal controller and the signal control agent. The functions of each component of this system will be described below: 1) Sensing devices: including but not limited to 4D radar, electronic police, checkpoints, traffic cameras and radar vision devices, etc., to provide stable video streams and radar detection data for the information control intelligent agent.

[0066] 2) The signal control intelligent agent (also known as an edge computing box) comprises four main functional modules or functions: detection and analysis, scheme reading, control decision-making, and command output. Among these, 21) Detection and Analysis: This involves identifying and extracting detection parameters from video footage or receiving structured data obtained from radar detection data to determine the number of motor vehicles and / or pedestrians in the target detection area. The target detection area includes both vehicle and / or pedestrian detection areas.

[0067] 211) Generate structured data based on secondary analysis of video stream: ① Full recognition: Full recognition of traffic objects in video stream; ② Data statistics: Set up an arbitrary octagonal target detection area in the video annotation range, and select a motor vehicle or pedestrian recognition algorithm to count the number of motor vehicles or pedestrians in the detection area respectively.

[0068] 212) Receive structured data: Receive the number of motor vehicles or pedestrians in the target detection area detected by radar-type equipment in accordance with the agreement.

[0069] 22) Scheme reading: ① Light status reading: The signal control agent reads the signal message through the protocol or intelligent signal control platform to obtain the current operation scheme number, stage number and light status; ② Parameter configuration: Initialize the global general parameters based on the current operation timing scheme, intersection channelization and traffic organization.

[0070] 23) Control Decision: Analysis and decision under gradient threshold sensing control mode: Based on the light status, traffic status, the running time of the current stage and the gradient vehicle number threshold or pedestrian number threshold of the inlet coil, determine whether the current stage should continue.

[0071] 24) Stepping command output: The signal control agent sends a stepping command to the signal controller to quickly end the current stage and enter the next stage.

[0072] 3) The traffic signal includes modules for scheme configuration, control mode setting, output light status, and receiving execution commands. Specifically: ① Scheme configuration: This includes the light group settings, phase settings, scheme settings, time period settings, and schedule settings required by the traffic signal in "local fixed cycle" mode; ② Control mode settings: At least "local fixed cycle" and "local sensing" modes are provided; ③ Output light status: This provides the traffic control agent with the current scheme number, phase number, and light group status; ④ Receiving execution commands: The traffic signal receives commands from the traffic control agent and executes operations such as phase extension, phase termination, early light group shutdown, and phase control mode downgrade.

[0073] Based on the above-described sensing control system, this embodiment provides a cross-sensor control method capable of handling low visibility conditions, comprising the following steps: S1) Model Training: The YOLO model developed by the TinyML team of Alibaba DAMO Academy's Data Analysis and Intelligence Lab is used. This model has been trained on large datasets including COCO, Objects365, and OpenImage, and can effectively identify the target classifications (car, person, motorcycle, truck, bus, bicycle) that need to be labeled. YOLO introduces some new technologies, including a Neural Architecture Search (NAS) backbone network, an efficient reparameterized generalized FPN (RepGFPN), a lightweight head that supports AlignedOTA label assignment, and neural network distillation enhancement technology. The specific labeling process is divided into three steps: S11) For the collected low-visibility video streams, extract and save the image dataset at 1 frame per second and use the YOLO model for preliminary annotation; S12) Use a program to eliminate labeled objects that are irrelevant to the target classification; S13) Manually remove incorrectly labeled objects, add unlabeled target objects, and use the resulting dataset to train the YOLO model.

[0074] S2) The intelligent signal control platform connects the signal controller and the intelligent signal control agent.

[0075] S21) New schemes and daily plans: Timing schemes based on target intersections, new schemes and daily plans.

[0076] S22) The intelligent traffic control platform connects to the traffic signal controller to synchronize the intersection configuration parameters.

[0077] Among them, the intelligent traffic control platform synchronizes the traffic control intelligent agents at intersections with light groups, phases, stages, schemes, daily plans, and scheduling.

[0078] In addition, based on the timing scheme of the target intersection, the traffic signal controller can be configured with light group settings, phase settings, scheme settings, time period settings, and plan settings. Specifically: Light group settings: In the light group setting interface, configure the corresponding sensing device number, minimum green light duration, maximum green light duration, and unit green light parameters for each light group. Phase settings: In the phase setting interface, configure the corresponding light group, flashing yellow, all red, delay, and early termination parameters for each phase. Scheme settings: In the scheme setting interface, add target phases one by one and configure the corresponding green light duration parameters. Time period settings: In the time period setting interface, configure the scheme for each time period and select the control mode in the corresponding operating mode. Plan settings: Set the operating scheme in the plan setting interface; if the scheme is the same every day, select the operating scheme from Sunday to Saturday.

[0079] S23) Signal controllers actively report real-time status: Signal controllers actively send real-time status information to the host computer, i.e., the intelligent traffic control platform and the traffic control intelligent agent. Each message is sent at a 1-second interval and the message should include at least the status information such as the color of each light group, the intersection stage, and the intersection plan.

[0080] S24) Query signal status: The host computer sends a real-time status query to the signal. Operation plan number and phase number of time signal and lighting units Upon receiving a command indicating the status of the lights, the signal controller immediately replies with real-time status information.

[0081] S25) Issue control commands: The signal control intelligent agent sends control commands to the signal controller through the intelligent signal control platform, and requires the signal controller to execute them immediately.

[0082] S3) Configure the information control agent system: S31) Integrate multimodal data and convert it into structured data: If it is radar data, directly generate structured data and proceed to step S311; if it is video data, secondary video analysis is required and proceed to step S212.

[0083] S311) Accessing structured data: Receiving structured data directly enables subsequent calculations.

[0084] S312) Access video stream data: Use a trained YOLO model to perform object recognition for the target video stream, and then configure the vehicle detection area (i.e., vehicle detection area) and pedestrian detection area bound to the light group to generate structured data in real time.

[0085] S3121) Video Management: Configure video information. Taking electronic police as an example: Set the electronic police serial number, electronic police number, electronic police IP, online video stream address, electronic police direction, model parameters, whether to enable, video width, and video height information.

[0086] S3122) Object recognition: The trained YOLO model is used for object recognition, and then the vehicle detection area and pedestrian detection area bound to the light group are configured respectively to generate structured data in real time.

[0087] S3123) Classification Configuration Detection Area: Select the target video, screenshot, add a new enclosed polygon such as an octagonal coil and name it, such as East Straight Lane 1.

[0088] S3124) Edit coil information: Configure coil ID, coil name, whether to enable, binding video number, lamp group number, coil type and coil detection target information, and confirm and save.

[0089] S3125) coil configuration: includes inbound lane detection, outbound lane detection and pedestrian detection.

[0090] S31251) Electronic police configuration at opposite entrances: Utilize electronic police screens to configure lane-level vehicle loops and zebra crossing pedestrian loops at opposite entrances as long-range detection zones.

[0091] S31252) Electronic police entrance configuration: The electronic police screen is used to configure lane-level vehicle coils and zebra crossing pedestrian coils in the entrance lane as a short-range detection zone.

[0092] S31253) Checkpoint inbound lane configuration: Utilize the checkpoint screen to configure lane-level vehicle coils for the inbound lane as a mid-range detection area.

[0093] S32) System Configuration: Includes scheme editing, phase editing, lamp group editing, and confirmation rate editing.

[0094] S321) Scheme editing: The operation scheme based on the signal controller configures the parameters of the corresponding stage in the intelligent agent, such as the selection stage, skipping stage, main stage, whether to enable induction coordination, light group difference and stage parameters.

[0095] S3211) Select Stage: Select the stage that has been configured in the stage editor.

[0096] S3212) Skip Stage: If skip stage is enabled, it means that when running the current solution, it supports skipping stages that are not required; otherwise, it runs according to the preset stage order.

[0097] S3213) Main Phase: The most critical phase in the solution. If there is no demand in any of the phases, the main phase will run.

[0098] S3214) Enable Induction Coordination: If induction coordination is enabled, after entering the coordination induction mode, the parameters of lamp group difference, calculation logic, stage duration, stage advance duration and stage delay duration will be called in each stage.

[0099] S32141) Lamp group difference: The calculation interval between the start time of the first stage within the cycle and the unified reference time of the system, in seconds.

[0100] S322) Phase Editing: Configure phase parameters, operation logic, minimum green light duration for the phase, maximum green light duration for the phase, phase downgrade duration, phase delay duration, phase duration, phase advance duration, and phase postpone duration (coordination parameters).

[0101] S3221) Configure stage parameters: Synchronize the signal scheme number and stage to the signal control agent, or configure the stage name, vehicle light group and pedestrian light group according to the stage sequence number; S3222) Operational logic: To adapt to the number of pedestrians and motor vehicles required for the phase termination of the timing scheme under different time periods and the threshold logic judgment, two kinds of logical operations, "tight" and "loose", are provided.

[0102] S3223) Phase Downgrade Duration : The green light duration of the light group during the phased downgrade under the current scheme, in seconds.

[0103] S3224) Minimum Green Light Duration : The minimum green light duration for the next phase of the current plan (excluding flashing yellow and all-red), in seconds.

[0104] S3225) Maximum Green Light Duration The maximum green light duration for the next phase of the current plan (excluding flashing yellow and all-red lights), in seconds.

[0105] S3226) Stage critical value i corresponds to stage runtime The current stage has been running for [duration]. The threshold value is measured in seconds (the system can default to 3 threshold values, all of which are the maximum green light duration of the stage; users can modify them according to their actual situation). The relationship between the stage's running time and the stage threshold value under different stages is shown in Table 1.

[0106] Table 1 S323) Light Group Editing: Configure the light group name, light group type, priority vehicle threshold, import vehicle inspection zone, export vehicle number threshold, and pedestrian number threshold for each light group.

[0107] S3231) Light Group Name: Set the light group name according to the general light group direction, such as East Straight.

[0108] S3232) Light type: Includes three types: motor vehicles, pedestrians and non-motor vehicles; S3233) Threshold for Imported Vehicle Inspection Area : The duration of the stage bound to the light group is less than the stage threshold i, corresponding to the stage runtime. Imported vehicle inspection area The threshold for the number of vehicles is shown in Table 2.

[0109] Table 2 S3234) Threshold for the number of exported vehicles : The exit inspection area bound to the light assembly The threshold for the number of vehicles.

[0110] S3235) Pedestrian Threshold : Pedestrian detection zone linked to the light group The threshold for the number of pedestrians.

[0111] S323) Confirmation Rate Editing: Includes abnormal parking confirmation rate, overflow demand confirmation rate, offline confirmation rate, vehicle vacancy confirmation rate, pedestrian vacancy confirmation rate, and stage demand confirmation rate.

[0112] S3231) Abnormal Parking Confirmation Rate :continuous This time, vehicles in the import lane repeatedly appeared at the same location. Second-rate.

[0113] S3232) Overflow Demand Confirmation Rate :continuous The number of vehicles overflowing the coil exceeds the threshold. The cumulative number of times reached Second-rate.

[0114] S3234) Offline Confirmation Rate :continuous The cumulative number of times frames could not be captured reached [number]. Second-rate.

[0115] S3235) Vehicle Emptying Confirmation Rate :continuous Next, current traffic light group The number of vehicles in the designated imported vehicle inspection area is less than the threshold. The number of times it appears Second-rate.

[0116] S3236) Pedestrian Vacancy Confirmation Rate :continuous Next, current traffic light group The number of pedestrians in the bound pedestrian detection area is less than the threshold. The number of times it appears Second-rate.

[0117] S4) Signal control intelligent agent operation sensing mode: The signal control intelligent agent issues temporary cycle plans to the signal controller, analyzes the preset stage parameters, and makes stage decisions based on the signal controller's operating status and real-time traffic environment.

[0118] S41) When the signal control agent is online and issuing commands, the intelligent signal control platform issues commands to the signal controller to modify the scheduling plan to the daily plan of sensor control; when the signal control agent is offline, the intelligent signal control platform changes the daily plan of sensor control (default 11) back to the original daily plan of fixed control.

[0119] S42) Phase Decision: Based on the phase's running time, equipment online status, exit overflow status, number of vehicles in the inlet detection zone, number of pedestrians in the pedestrian detection zone, and system parameters, a phase decision is made, including phase vehicle and pedestrian status values, phase signal values, phase decisions, and command issuance. The specific steps are as follows: S421) Stage vehicle and pedestrian state values: Determine whether the vehicle and pedestrian demand in the current frame stage meets the dynamic threshold, including calculating vehicle detection state values, vehicle light group state values, and pedestrian detection state values.

[0120] S4211) Calculate the number of vehicles in the entrance inspection area of ​​the light group: The number of vehicles in the entrance inspection area of ​​the light group is the sum of the number of vehicles in the short-distance inspection area, medium-distance inspection area and long-distance inspection area of ​​all lanes of the light group.

[0121] S4212) Vehicle detection status value vds_car: If the number of vehicles in the import vehicle detection area corresponding to the light group in the current frame is greater than number_in_coil... If the number of vehicles in the imported vehicle detection area corresponding to the light group is -1 (due to reasons such as detector disconnection), the value is 2; otherwise, the value is 0.

[0122] S4213) Vehicle light group status value vv_status: Takes data from the past m frames. If the vehicle detection status value is 2 and has accumulated at least n frames, the vehicle light group status value is 2; if the vehicle detection status value is 0 and has accumulated at least n frames, the vehicle light group status value is 0; otherwise, it is 1.

[0123] S4214) Pedestrian detection status value vds_person: If the number of pedestrians in the pedestrian detection area number_pedestrian_coil > 1, the value is 1; otherwise, the value is 0.

[0124] S4215) Stage-level vehicle and personnel status values: The stage-level vehicle and personnel status values ​​are calculated using "tight logic" and "loose logic".

[0125] S42151) "Tight Logic": The stage-level vehicle light group state value is taken as the minimum value of the vehicle detection state value corresponding to the light group within the stage; the stage-level pedestrian light group state value is taken as the minimum value of the pedestrian detection state value corresponding to the light group within the stage.

[0126] S42152) "Loose Logic": If the vehicle detection status value corresponding to the light group in the stage has a value of 1, then the stage-level vehicle light group status value is 1; otherwise, the minimum value of the vehicle detection status value corresponding to the light group in the stage is taken (e.g., 0, 0 is taken as 0, 0, 2 is taken as 2, and only 2, 2 is taken as 2); if the pedestrian detection status value corresponding to the light group in the stage has a value of 1, then the stage-level pedestrian light group status value is taken as 1; otherwise, it is taken as 0.

[0127] S422) Stage signal value: Determines whether there is a passage requirement in the current stage or whether the detector is empty or invalid. The value range is 0, 1, 2, 3.

[0128] (S4221) If the stage-level vehicle light group status value is 0, all light groups in this stage meet the vehicle empty release confirmation rate confirmation_null_car, and the stage-level pedestrian light group status value is 0 and meets the pedestrian empty release confirmation rate confirmation_null_pedestrian, then the stage signal value is 0; if the stage-level vehicle light group status value is 2 and meets the offline (empty data) confirmation rate confirmation_offline, then the stage signal value is 2, otherwise the stage signal value is 1.

[0129] (S4222) If the stage signal value is 2, then when the current stage's running time reaches (stage downgrade duration - stage minimum decision time), the pedestrian light group can be controlled to end at that time, while the vehicle light group continues to run until the downgrade time. If the stage-level pedestrian light group state value > 0, the vehicle light group of the current stage continues; otherwise, the current stage ends. In this way, the safety of pedestrians crossing the street with green tails or at slower crossing speeds is ensured, greatly reducing conflicts between pedestrians and vehicles.

[0130] S423) Stage decision: Determine whether to continue or terminate the current stage based on the stage signal value.

[0131] S4231) Stage Continue: If the stage decision value value_stage_decision is 1, the stage continues to run, and the stage decision value is initialized to 1. The stage decision value is 1 in the following cases: (S42311) If the current time is less than the minimum stage decision time (min_stage_decision), then the stage decision value is 1. Wherein, the minimum stage decision time = stage duration - stage advance duration - 5; the maximum stage decision time = stage duration + stage delay duration - 5.

[0132] S42312) The current time is between the stage minimum decision time min_stage_decision and the stage maximum decision time max_stage_decision (closed before open) and satisfies one of the following conditions: The S423121) stage signal value is 1, and the priority request value for non-P stages is not 3 (a non-P stage is any stage other than the current stage). The S423122) stage signal value is 1, and the stage-level pedestrian light group status value is 1; The signal value of stage S423123) is 2, the duration of the current stage is less than (stage downgrade duration - 4), and the signal priority request value of the target vehicle in non-P stages is not 3; non-P stages are any stage other than the current stage (denoted as P stage); The S423124) stage signal value is 2, the current stage running time is >= (stage downgrade time - 4), and the stage-level pedestrian light group status value is 1; The signal value for stage S423125 is 3.

[0133] S4232) Stage End: The stage ends when the stage decision value value_stage_decision is 0. The stage decision value is 0 in the following cases: (S42321) When the current moment is between the minimum stage decision time min_stage_decision and the maximum stage decision time max_stage_decision (closed before open), any of the following conditions must be met: ① The stage signal value value_stage_signal is 0; ② The stage decision value of the previous second is 0; ③ The stage signal value is 2 and the stage duration is greater than the stage degradation duration - 4, and the pedestrian light group status value of stage P is 0; ④ The target vehicle signal priority request value of non-P stage is 3 and the pedestrian light group status value of the current stage is 0.

[0134] S42322) Current moment > maximum decision moment of the stage.

[0135] S424) Issue stage or lamp group end command: Control stage or lamp group status through protocol command.

[0136] S4241) End Phase Instruction: If the current phase continues and the light is green, no instruction is sent; otherwise, a "step" instruction is issued.

[0137] S4242) Early termination of light group instruction: If the current light group continues and is in a green light state, no instruction is sent; otherwise, when the exit overflows, an "overflow" instruction is sent to the corresponding light group to end the green light state of the light group early.

[0138] S425) Update Then return to step S321).

[0139] S5) Evaluate and optimize the basic parameters of the signal control agent.

[0140] In summary, to address the problem that existing video-based perception methods for traffic signalized intersections suffer from low visibility and low target recognition rates at night and in rainy or foggy weather, resulting in insufficient vehicle clearance time, the cross-sensing control method provided in the above embodiments supports image recognition, data analysis, control decision-making, and command output after accessing multiple video streams and real-time signal information. This enables intelligent cross-sensing traffic control under low visibility conditions to improve the traffic efficiency of traffic signalized intersections, while significantly reducing the application cost of sensing control and improving pedestrian crossing safety and vehicle traffic efficiency.

[0141] This application also provides an electronic device, including a processor and a memory, wherein the memory stores computer program instructions for execution on the processor, and when the processor executes the computer program instructions, it implements the cross-sensing control method as described above.

[0142] This application also provides a computer-readable storage medium storing computer instructions that, when executed by a processor, implement the cross-sensing control method described above.

Claims

1. A cross-sensing control method, characterized in that, The method includes: Acquire traffic flow data and corresponding traffic light operation data of the target intersection at the current moment; the traffic flow data is collected and generated by sensing devices deployed at multiple different locations and directions at the target intersection, and the operation data includes the operation plan, the current target stage, the light status of each light group in the target stage, and the running time of the target stage; Based on traffic flow data, the comprehensive traffic demand of each light group in the target phase is determined; specifically, for the target light group, the comprehensive traffic demand of the target light group is determined by combining the perception data collected by the perception devices deployed in the target light group in its own direction and the perception data collected by the perception devices deployed in the opposite direction; the target light group is any light group whose light status is green in the target phase. Based on the running time of the target phase and the pre-defined correspondence between each light group in the target phase and the vehicle number threshold for different running time intervals, determine the current gradient vehicle number threshold for each light group in the target phase. Based on operational data, the overall traffic demand of each traffic light group in the target phase, and the threshold for the number of vehicles at different gradients, control instructions are generated to control the traffic lights at the target intersection.

2. The method as described in claim 1, characterized in that, Traffic flow data includes video stream data, and comprehensive traffic demand includes the number of traffic objects within the detection area. Based on traffic flow data, the comprehensive traffic demand for each light group in the target phase is determined, including: Object annotation is performed on video stream data using a trained YOLO model; Based on the video stream data after object labeling, the number of traffic objects in the detection area bound to each light group in the target stage is counted. The detection area includes the import vehicle detection area, which includes the first detection area, the second detection area, and the third detection area, which are covered by sensing devices at different locations and target the same import lane. The number of vehicles in the import vehicle detection area is the sum of the number of vehicles identified in the first detection area, the second detection area, and the third detection area.

3. The method as described in claim 1 or 2, characterized in that, Based on the running time of the target phase and the pre-defined correspondence between each light group in the target phase and the vehicle number threshold for different running time intervals, the gradient vehicle number threshold corresponding to each light group in the target phase is determined, including: If the running time of the target phase is within the first running time interval corresponding to the target light group and the first running time interval corresponds to the first threshold, then the gradient vehicle number threshold corresponding to the target light group is determined as the first threshold; the start time of the first running time interval is greater than or equal to the minimum green light duration of the phase corresponding to the target phase; If the running time of the target stage is within the second running time interval corresponding to the target light group and the second running time interval corresponds to the second threshold, then the gradient vehicle number threshold corresponding to the target light group is determined as the second threshold; the second threshold is greater than or equal to the first threshold. If the running time of the target phase is within the third running time interval corresponding to the target light group and the third running time interval corresponds to the third threshold, then the gradient vehicle number threshold corresponding to the target light group is determined as the third threshold; the start time of the third running time interval is less than or equal to the maximum green light duration of the target phase, the third threshold is greater than or equal to the second threshold and the third threshold is greater than the first threshold.

4. The method as described in claim 2, characterized in that, Based on operational data and the overall traffic demand of each traffic light group during the target phase, as well as the gradient vehicle quantity threshold, control instructions are generated for controlling the traffic lights at the target intersection, including: Based on the operational data and the number of traffic objects in the detection area bound to each light group in the target phase, as well as the gradient vehicle number threshold, determine the phase vehicle state value and phase pedestrian state value of the target phase at the current moment. Based on the vehicle state value and pedestrian state value of the target stage at the current moment, determine the stage signal value of the target stage at the current moment; Based on the target stage's stage signal value at the current moment, determine the target stage's stage decision value at the current moment; Based on the target stage's stage decision value at the current moment, generate control instructions for controlling the traffic lights at the target intersection.

5. The method as described in claim 4, characterized in that, The detection area also includes a pedestrian detection area. The operation plan includes computational control logic and pedestrian number thresholds for each light group's corresponding pedestrian detection area in the target phase. Based on the operational data and the number of traffic objects within the detection area bound to each light group in the target phase, as well as the gradient vehicle number threshold, the phase vehicle state value and phase pedestrian state value for the current moment in the target phase are determined, including: At the current moment, if the number of vehicles in the imported vehicle detection area corresponding to the target light group is empty or invalid, the vehicle detection status of the target light group at the current moment is marked as the third preset value; if the number of vehicles in the imported vehicle detection area corresponding to the target light group is greater than or equal to the gradient vehicle number threshold, the vehicle detection status of the target light group at the current moment is marked as the first preset value; otherwise, the vehicle detection status of the target light group at the current moment is marked as the second preset value. If, in the m consecutive moments preceding the current moment, the number of times the vehicle detection status value of the target light group is greater than or equal to ... If the number of pedestrians in the pedestrian detection area corresponding to the target light group is greater than or equal to the pedestrian number threshold, then the pedestrian status value of the target light group at the current time is marked as the first preset value; otherwise, the pedestrian status value of the target light group at the current time is marked as the second preset value. When the operation control logic is the first operation logic, mark the vehicle state value of the target stage at the current time as the minimum value of the vehicle state value of all light groups in the target stage at the current time, and mark the pedestrian state value of the target stage at the current time as the minimum value of the pedestrian state value of all light groups in the target stage at the current time. When the operation control logic is the second operation logic, if at least one light group in the target stage has a vehicle state value of the first preset value at the current time, then mark the target stage vehicle state value of the current time as the first preset value; otherwise, mark the target stage vehicle state value of the current time as the minimum value of the vehicle state values ​​of all light groups in the target stage at the current time. If at least one light group in the target stage has a pedestrian state value of the first preset value, then mark the target stage pedestrian state value of the current time as the first preset value; otherwise, mark the target stage pedestrian state value of the current time as the second preset value.

6. The method as described in claim 4 or 5, characterized in that, Based on the vehicle state value and pedestrian state value of the target stage at the current moment, determine the stage signal value of the target stage at the current moment, including: If the vehicle status value of the target stage at the current moment is the first preset value and all light groups in the target stage meet the vehicle idle release confirmation rate, and the pedestrian status value of the target stage at the current moment is the second preset value and all light groups in the target stage meet the pedestrian idle release confirmation rate, then mark the stage signal value of the target stage at the current moment as the fourth preset value. If the vehicle status value of the target stage at the current moment is the third preset value and all light groups in the target stage meet the offline confirmation rate, then mark the stage signal value of the target stage at the current moment as the fifth preset value. Otherwise, the stage signal value of the target stage at the current moment is the sixth preset value; Specifically, the vehicle vacancy confirmation rate of a light group is such that, in a consecutive n-timeframe from the current time, the number of times the number of vehicles in the import vehicle detection area corresponding to that light group is less than a preset number of vehicles is greater than or equal to a first preset occurrence number; the pedestrian vacancy confirmation rate of a light group is such that, in a consecutive n-timeframe from the current time, the number of times the number of pedestrians in the pedestrian detection area corresponding to that light group is less than a preset number of pedestrians is greater than or equal to a second preset occurrence number; and the offline confirmation rate of a light group is such that, in a consecutive n-timeframe from the current time, the number of times the number of vehicle objects in the import vehicle detection area corresponding to that light group is empty or invalid is greater than or equal to a third preset occurrence number.

7. The method as described in claim 6, characterized in that, Based on the target stage's stage signal value at the current moment, determine the target stage's stage decision value at the current moment, including: When the first objective condition is met, the stage decision value of the target stage at the current moment is determined to be the seventh preset value; the first objective condition includes any one of the following conditions: the current moment has not reached the minimum decision moment of the target stage, where the minimum decision moment is the difference between the stage duration of the target stage and the first preset duration threshold, and the first preset duration threshold is the sum of the stage advance duration of the target stage and the first preset duration; the current moment has reached the minimum decision moment of the target stage but has not reached the maximum decision moment of the target stage, and the second objective condition is met, where the maximum decision moment is the difference between the stage duration of the target stage and the second preset duration threshold, and the second preset duration threshold is the difference between the stage delay duration and the second preset duration; wherein, the second objective condition includes any one of the following conditions: target stage The current stage signal value is the sixth preset value and there is a non-target stage target vehicle signal priority request value that is not the ninth preset value; the target stage's current stage signal value is the sixth preset value and the target stage's current stage pedestrian status value is the first preset value; the target stage's current stage signal value is the fifth preset value, the target stage's elapsed duration is less than the difference between the target stage's stage downgrade duration and the third preset duration threshold, and there is a non-target stage target vehicle signal priority request value that is not the ninth preset value; the target stage's current stage signal value is the fifth preset value, the target stage's elapsed duration is greater than or equal to the difference between the target stage's stage downgrade duration and the fourth preset duration threshold, and the target stage's current stage pedestrian status value is the first preset value; When the third objective condition is met, the stage decision value of the target stage at the current moment is determined to be the eighth preset value; wherein, the third objective condition includes any one of the following conditions: the current moment has reached the maximum stage decision time of the target stage; the current moment has reached the minimum stage decision time of the target stage but has not reached the maximum stage decision time of the target stage, and the fourth objective condition is met; the fourth objective condition includes any one of the following conditions: the stage signal value of the target stage at the current moment is the fourth preset value; the stage decision value of the target stage at the previous moment is the eighth preset value; the stage signal value of the target stage at the current moment is the fifth preset value, the running length of the target stage is greater than or equal to the difference between the stage degradation time of the target stage and the fifth preset time threshold, and the stage pedestrian state value of the target stage at the current moment is the second preset value; there exists a target vehicle signal priority request value that is not in the target stage and the stage pedestrian state value of the target stage at the current moment is the second preset value; Specifically, if, up to the current moment, the duration of the target vehicle signal priority status value of a light group in the non-target phase is greater than or equal to the third preset duration, then the target vehicle signal priority request value in the non-target phase at the current moment is marked as the ninth preset value; if, at the current moment, the number of target vehicles in the import vehicle detection area corresponding to that light group in the non-target phase is greater than the preset number of target vehicles, then the target vehicle signal priority status value of the target light group at the current moment is marked as the tenth preset value; the target vehicle is a special vehicle with priority right-of-way.

8. The method as described in claim 7, characterized in that, The operational plan also includes the threshold number of exit vehicles corresponding to each light group in the target phase; Based on the target stage's stage decision value at the current moment, control instructions are generated for controlling the traffic lights at the target intersection, including: When the target stage's stage decision value at the current moment is the seventh preset value, if there is a target lamp group whose overflow signal value at the current moment is the preset target value, then a lamp group overflow early interruption instruction for the target lamp group is generated. When the target stage's stage decision value at the current moment is the eighth preset value, a step instruction to end the target stage is generated. Specifically, when the number of vehicles in the exit vehicle detection area corresponding to the target light group is greater than or equal to the exit vehicle number threshold, the overflow state value of the target light group at the current moment is marked as the first state value; otherwise, it is marked as the second state value. When the number of times the overflow state value of the target light group is the first state value in the k consecutive moments traced back from the current moment reaches the preset overflow number, the overflow signal value of the target light group at the current moment is marked as the preset target value.

9. An electronic device, characterized in that, It includes a processor and a memory, the memory storing computer program instructions for execution on the processor, wherein when the processor executes the computer program instructions, it implements the cross-sensing control method as described in any one of claims 1-8.

10. A computer-readable storage medium, characterized in that, It stores computer instructions, which, when executed by a processor, implement the cross-sensing control method as described in any one of claims 1-8.

Images