Methods, systems, and vehicles for determining traffic status at intersections

Abstract

This invention provides a method, system, and vehicle for determining the traffic status at intersections based on traffic flow. The method includes the following steps: acquiring vehicle information to establish an intersection coordinate system; projecting the traffic flow onto the projection axes of the coordinate system and calculating the sum of projections in each direction; calculating the confidence level of the current traffic status based on the ratio of the sum of projections in the vehicle's passable directions to the sum of projections in the total number of vehicles. A mutual exclusion judgment is performed based on the traffic light recognition results and the confidence level of the current traffic status to make a decision. Accordingly, this invention also provides a system and a vehicle equipped with the system. This invention provides more accurate results by dynamically analyzing traffic flow to predict traffic light status. This prediction method based on traffic flow perception not only improves the accuracy of perception but also provides an effective supplement to the traffic light recognition module, eliminating reliance on visual recognition results and adapting to un-upgraded networked areas.

Description

Technical Field

[0001] This invention relates to the field of autonomous driving technology, specifically to a method, system, and vehicle for determining the traffic status of intersections based on traffic flow. Background Technology

[0002] In the development of autonomous vehicles, the traffic light perception module is a crucial component, enabling the vehicle to recognize traffic lights at intersections and make autonomous decisions to pass through them. At the same time, this module also needs to consider traffic efficiency.

[0003] Currently, there is a wealth of research in the field of traffic light control on the road traffic side. For example, Chinese patent application CN114419906A discloses a method, device, and storage medium for intelligent transportation based on big data. This method collects vehicle images on the road, identifies whether a vehicle is in front or behind in the image, and determines the direction of traffic flow based on the identification result. It detects the traffic flow in the transverse lanes near the traffic light intersection and the longitudinal lanes intersecting with the transverse lanes. It calculates the time required for vehicles in the transverse and longitudinal lanes to pass through the traffic light intersection. Based on the time required for vehicles in the transverse and longitudinal lanes to pass through the traffic light intersection, it allocates the traffic light cycles for the transverse and longitudinal lanes. However, on the vehicle side, traffic light recognition still mainly relies on visual recognition, with only a small portion achieving online information acquisition through improved road traffic IoT.

[0004] However, despite numerous optimizations and improvements, existing traffic light perception modules continue to face many technical challenges, and their performance in vehicle systems remains unstable under different environmental conditions. As mentioned earlier, Chinese patent application CN112053578A discloses a traffic light warning method and detection device based on machine vision technology, which uses machine vision technology to determine the traffic light status of the current traffic light image information. Although it can be used in autonomous vehicles, in complex lighting environments, the perception module may misjudge the traffic light status due to strong sunlight or low light conditions. Simultaneously, dirt on the vehicle's camera lens can also affect the accuracy of traffic light recognition. Furthermore, in rainy, foggy weather or when obstructed by objects, the module's accuracy drops significantly, and traffic light recognition may even fail. Regarding scenarios where the Internet of Things (IoT) is used to obtain traffic light status online, Chinese patent CN114519936B discloses a V2X-based intelligent engine start-stop control method and system at traffic light intersections. It utilizes a 5G V2X vehicle-mounted unit to receive real-time road environment information sent from the cloud to the vehicle, which can be used to obtain traffic light information. In particular, online functionality has been achieved through this network in some areas. However, its main problems are unstable network connections and limited current application coverage. It requires simultaneous upgrades to traffic facilities and vehicle-mounted equipment. Furthermore, it cannot address the lack of V2X technology applications in most areas. Therefore, without relying entirely on vision and the Internet of Things, technologies that rely on artificial intelligence to assist in judgment have emerged. For example, Chinese patent application CN119091612A discloses a method for estimating traffic light cycles based on machine learning. This method can accurately estimate the traffic light cycle by comprehensively analyzing vehicle trajectory data and handle cycle changes. It can detect changes in the traffic light cycle and use Fourier transform and CUSUM methods to identify the time of cycle switching and the parameters of the old and new cycles. Although the above technical solution is used to understand the changing cycle of traffic lights that are not connected to the network, it can still provide vehicles with a relatively accurate basis for judging the traffic light status through wireless networks. However, it is worth noting that if this solution is used on the vehicle side, it clearly contradicts the current mainstream technology of adjusting traffic lights based on artificial intelligence (as mentioned above), and therefore cannot provide reliable judgment results.

[0005] As a result, these problems seriously affect the traffic efficiency and safety of vehicles, becoming key obstacles to the reliable operation of driverless cars.

[0006] In summary, it is also necessary to develop a fallback algorithm that is not easily affected by natural factors and can autonomously make judgments based on traffic scenarios, so as to provide basic-level traffic light recognition assurance when machine learning models make misjudgments or fail. Summary of the Invention

[0007] This invention addresses the problems existing in the prior art by providing an effective and rapid method for determining the status of traffic lights with relatively high accuracy.

[0008] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0009] On the one hand, the present invention provides a method for determining the traffic status based on intersection traffic flow, which includes the following steps:

[0010] Obtain vehicle information and establish a coordinate system for the current intersection based on the vehicle information. The coordinate system has multiple vehicle movement projection axes.

[0011] The vehicle projections on each vehicle's operational projection axis are calculated using the projection method.

[0012] Calculate the sum of the projections of vehicles traveling in the same direction, which is the sum of the projections of vehicles on the same direction as the vehicle's travel direction.

[0013] Calculate the confidence level of the current traffic status, which is the ratio of the sum of the projections of the vehicles traveling in the same direction to the sum of the total vehicle projections in the coordinate system; the total vehicle projection is the sum of the vehicle projections on each vehicle travel projection axis.

[0014] Based on the confidence level of the current traffic status and the current traffic light recognition results at the intersection, a traffic decision is made through mutual exclusion judgment.

[0015] Optionally, the passage decision-making process based on mutual exclusion includes:

[0016] Read the current traffic light recognition result at the intersection. If the recognition result is green, verify the confidence level of the current traffic status:

[0017] If the confidence level of the current passage status is greater than or equal to the passage threshold, passage is permitted.

[0018] If the confidence level of the current traffic status is less than the threshold, a parking instruction is triggered, and data is reported to request manual intervention;

[0019] Read the current traffic light recognition result at the intersection. If the recognition result is a recognition failure or the recognition confidence level is lower than a preset threshold, then verify the confidence level of the current traffic status.

[0020] If the confidence level of the current passage status is greater than or equal to the passage threshold, passage is permitted under the green light status.

[0021] If the confidence level of the current traffic status is less than the threshold, stop the vehicle as if it were in a red light state.

[0022] Optionally, the vehicle information includes the vehicle's position, orientation, and driving status; the driving status includes at least the vehicle's speed.

[0023] The establishment of the coordinate system for the current intersection includes the following steps:

[0024] The positive Y-axis direction is selected from the lane direction that is closest to the direction of the vehicle, and the positive X-axis direction is selected from the lane direction that has the largest angle with the Y-axis.

[0025] The center point of the current intersection is the center of the intersection where the dividing lines of the straight lanes intersect or the center of the set of lane intersections.

[0026] Multiple vehicle operation projection axes are set to adapt to the positive and negative directions of the X and Y axes, as well as the bisector directions of the angles in each quadrant.

[0027] Optionally, before calculating the vehicle projections on each vehicle's operational projection axis, data cleaning of observable vehicles at the intersection is performed, including the following steps:

[0028] Based on the observable vehicle operating parameters in the coordinate system of the current intersection, vehicles with low reference value are eliminated.

[0029] The operating parameters of the observable vehicles at the intersection include the vehicle speed, position quadrant, speed orientation, and distance from the center of the intersection.

[0030] The low reference value vehicles include vehicles turning right, vehicles making U-turns, and slow-moving vehicles at the edge of the road.

[0031] Optionally, excluding vehicles with low reference value includes the following steps:

[0032] Vehicles whose distance from the center of the intersection exceeds 1 / 3 of the intersection radius and whose speed orientation is in the quadrant of the vehicle position plus 1 are identified as right-turning vehicles.

[0033] Vehicles whose distance from the center of the intersection exceeds 1 / 2 of the intersection radius and whose speed is in the opposite direction to the straight-going direction of their lane are identified as U-turn vehicles.

[0034] Vehicles whose speed is below a preset speed threshold and located at the intersection boundary are considered edge slow vehicles.

[0035] Optionally, calculating the vehicle projection includes the following steps:

[0036] The vehicle's velocity vector is projected onto the nearest vehicle travel projection axis, and the absolute value is taken to calculate the projection value; the projection values ​​on the vehicle travel projection axis are added together to obtain the vehicle projection sum;

[0037] For a vehicle in a critical left-turn state, the projection value of the vehicle in a critical left-turn state is evenly distributed to the straight-axis and left-turn axis of the vehicle in a critical left-turn state.

[0038] For lane directions that do not exist at T-junctions, the vehicles belonging to the corresponding vehicle travel projection axis are reclassified to the adjacent vehicle travel projection axis.

[0039] Optionally, calculating the confidence level of the current passage status also includes the following steps:

[0040] Calculate the sum of projections of vehicles traveling in the same direction; the addends to the sum of projections of vehicles traveling in the same direction include the sum of projections in the vehicle's own direction and the sum of projections in the coexisting directions;

[0041] The coexisting directions refer to the directions of straight-going vehicles and left-turning vehicles that are allowed to pass during the same traffic light phase.

[0042] Optionally, the following steps are also included: when a vehicle enters the intersection area, the real-time traffic status confidence is periodically calculated and continued until the vehicle leaves the intersection area, and the calculation results are stored in an information queue;

[0043] When the passage decision is executed, the calculation results of the information queue are extracted and the weighted average is used to obtain the confidence level of the current passage status.

[0044] On the other hand, the present invention also provides a traffic status judgment system based on intersection traffic flow based on the above method;

[0045] The system includes: a traffic light sensing module, a vehicle information reading module, an intersection coordinate system establishment module, an obstacle reading module, and a data processing module;

[0046] The traffic light sensing module is used to identify the current traffic light status at the intersection;

[0047] The vehicle information reading module is used to acquire vehicle information;

[0048] The intersection coordinate system establishment module is used to establish the coordinate system of the current intersection and calculate the vehicle running projection axis of the current intersection;

[0049] The obstacle reading module is used to filter dynamic obstacles;

[0050] The data processing module is used to calculate the confidence level of the current passage status and the execution of passage decisions.

[0051] In another aspect, the present invention also provides a vehicle equipped with the aforementioned traffic flow-based intersection status determination system.

[0052] Compared with the prior art, the present invention has the following beneficial effects:

[0053] This invention provides more accurate results by predicting traffic light status through dynamic analysis of traffic flow. This prediction method based on traffic flow perception not only improves the accuracy of perception but also provides an effective supplement to the traffic light recognition module, eliminating reliance on visual recognition results and adapting to un-upgraded network areas. Attached Figure Description

[0054] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0055] Figure 1 This is a flowchart of a method in a specific embodiment of the present invention;

[0056] Figure 2 This is a flowchart of traffic flow screening in a specific embodiment of the present invention;

[0057] Figure 3 This is a schematic diagram of vehicle projection in a specific embodiment of the present invention.

[0058] In the picture: 1. Vehicle on its own, 2. Vehicle in front, 3. Vehicle going in the same direction, 4. Vehicle going in the opposite direction, 5. Vehicle turning left, 6. Vehicle turning right, 7. Slow-moving vehicle at the edge. Detailed Implementation

[0059] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0060] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0061] It is worth noting that, unless otherwise specified, the methods used in this invention are all conventional methods; and the raw materials and equipment used are all conventional commercially available products, and their sources are not specifically limited.

[0062] It should also be noted that, for ease of understanding, the method steps in the specific embodiments of the present invention are described in a certain order, but those skilled in the art can change the order of the steps according to actual needs, so this should not be used as a limiting condition.

[0063] Please see Figure 1 and Figure 2 As shown in the figure, this embodiment provides a method for determining the traffic status based on intersection traffic flow, which mainly includes the following steps:

[0064] The vehicle system acquires vehicle information in real time from in-vehicle sensors, including vehicle position, vehicle orientation, and vehicle driving status; furthermore, the vehicle driving status includes at least vehicle speed.

[0065] Optionally, to reduce resource requirements and improve response speed, the system should first determine whether the vehicle has reached the intersection area based on its information, and whether the vehicle is within the intersection and has speed. If the vehicle has not reached the intersection or has speed, it is not necessary to determine the traffic status, so this task is skipped; otherwise, the subsequent steps are continued.

[0066] The coordinate system of the current intersection is established based on the vehicle information. Considering that some intersections are not standard cross-shaped intersections, a non-cartesian coordinate system is adopted to adapt to different intersection requirements. Thus, a planar non-cartesian coordinate system is established according to the actual situation of the intersection; the center point of the current intersection and the vehicle's running projection axis are calculated.

[0067] The establishment of the non-Cartesian coordinate system at the current intersection includes the following steps:

[0068] (1) Read all lanes within the intersection and filter out the straight-ahead lanes, recording their orientation. In particular, orientations that are not significantly different will not be stored repeatedly; that is, after filtering, it will be determined whether the orientation is within the tolerance range of each previously recorded orientation. If none are found, then it will be recorded. If fewer than four orientations are obtained, the same filtering will be performed on the starting and ending orientations of the left-turn lanes. Taking the vehicle's orientation as the positive Y-axis direction, the lane direction with the largest angle to the Y-axis will be selected as the positive X-axis direction; that is, among all orientations, the orientation with the smallest angle (without direction) to the vehicle's orientation will be selected as the positive Y-axis direction of the intersection. Then, the orientation with the largest angle (clockwise) to the positive Y-axis direction will be selected as the positive X-axis direction of the intersection.

[0069] (2) The center point of the intersection is the center of the intersection or the center of the lane intersection point set. Specifically, the center of the four intersections of the leftmost dividing lines of the four directions of straight lanes is taken as the center point of the intersection, and half the difference between the poles of the intersection polygon is taken as the intersection radius. When there is an intersection of the straight lanes in each direction, the center of the intersection point set of the innermost dividing lines of each lane is taken as the center point of the intersection. The length from the center point of the intersection to the side of the intersection polygon minus the boundary value is taken as the traffic flow screening radius. If the vehicle is located in the non-motorized vehicle lane and in the first quadrant of the established coordinate system, it is considered to be in the non-motorized vehicle lane waiting to turn left. The vehicle may be not straight due to obstruction. The module actively resets the x-axis to the new positive y-axis.

[0070] (3) Set multiple vehicle movement projection axes to match the positive and negative directions of each axis, as well as the bisector direction of each quadrant angle; specifically, taking a crossroads as an example: Figure 3 As shown by the dashed line at the midpoint, the angle bisectors of the X and Y axes in both directions and the four quadrants are used as the four straight and four left-turn projection axes, thus forming a star-shaped projection axis.

[0071] Before projection calculation, the data is first cleaned. This process involves eliminating vehicles with low reference value based on the operational parameters of observable vehicles at the intersection in the non-Cartesian coordinate system. Specifically, the vehicle uses sensor data to read all dynamic obstacles from the perception channel, such as vehicles within and near the intersection, and filters out other vehicles within the intersection area as observable vehicles. The filtering criteria include whether they are motor vehicles and whether their speed exceeds a certain value. The operational parameters of observable vehicles at the intersection include their vehicle speed, position quadrant, speed orientation, and distance from the intersection center.

[0072] Optionally, if the above operation of reading and filtering traffic flow within the intersection determines that there is no traffic flow, it means that there is no need to determine the passage status and the vehicle can pass normally or rely on the onboard vision system to recognize traffic lights, i.e., there is no obstruction, so this task is skipped; otherwise, the subsequent steps are continued, such as if vehicle 1 is obstructed by vehicle 2 in front within the intersection range, causing the image sensor to be unable to collect traffic light signal images.

[0073] Vehicles with low reference value include right-turning vehicles (6), U-turning vehicles, and slow-moving vehicles at the edge of the traffic light (7). These vehicles cannot provide effective information for judging the status of traffic lights, so the removal of vehicles with low reference value involves the following steps:

[0074] For vehicles whose distance from the center of the intersection exceeds 1 / 3 of the intersection radius and whose speed direction is in the quadrant of the vehicle position plus 1 (the fourth quadrant is the first quadrant), they are identified as right-turning vehicles 6.

[0075] Vehicles whose distance from the center of the intersection exceeds 1 / 2 of the intersection radius and whose speed and direction are opposite to the straight-going direction of their lane are identified as U-turn vehicles.

[0076] For edge slow vehicles 7, vehicles that are at the intersection boundary and have a low speed are excluded because their status is difficult to determine; therefore, vehicles with a speed lower than the preset speed threshold and located at the intersection boundary are edge slow vehicles.

[0077] The vehicle projections on each vehicle's travel projection axis are calculated using a projection method. The method of projection summation can be determined based on actual needs, such as calculating the number of projections, weighted summation of projection values, or summation of projection values; therefore, in this embodiment... Figure 3 As shown, this is implemented using a projection summation method, specifically as follows:

[0078] First, the speed of vehicles in the intersection area is collected by sensors, including the speed magnitude and direction, thereby forming a speed vector;

[0079] Next, the velocity vector is projected onto the nearest vehicle's trajectory projection axis, and the projected value is obtained, such as... Figure 3 The black bold line segment;

[0080] Finally, the projection values ​​on each vehicle's running projection axis are summed to obtain the vehicle projection sum for that vehicle's running projection axis.

[0081] Furthermore, the nearest vehicle travel projection axis is the vehicle travel projection axis with the smallest angle to the corresponding vehicle velocity direction.

[0082] Therefore, the confidence level of the current traffic status is calculated as the ratio of the sum of the projections of vehicles traveling in the same direction to the sum of the projections of the total number of vehicles. The total sum of vehicle projections is the sum of the projections of vehicles on the eight vehicle travel projection axes in this embodiment; the sum of the projections of vehicles traveling in the same direction is the sum of the projections of vehicles on the same travel projection axes as the vehicle's own travel direction, and the addend to the sum of the projections of vehicles traveling in the same direction includes the projections of vehicles in the vehicle's own direction and the projections of vehicles in coexisting directions; furthermore, coexisting directions are the directions of straight-ahead vehicles and left-turning vehicles that are allowed to travel in the same traffic light phase; such as... Figure 3 In the middle, there are vehicles 3 traveling in the same direction in the same lane in the direction of the vehicle. The coexisting directions include vehicles in adjacent non-turning lanes, vehicles traveling in the opposite direction 4, and vehicles turning left 5. These types of vehicles are considered as reference vehicles because they enter the intersection area and have a certain speed. They are projected onto the nearest vehicle running projection axis to obtain traffic flow information.

[0083] In this embodiment, the projection method is to project the velocity vector of each vehicle onto its nearest coordinate axis. Because it is a statistical value, the absolute value is taken without a negative sign, and it is considered to belong to the traffic flow of the corresponding axis.

[0084] Furthermore, the critical state for left turn (i.e., just starting or already turning) is determined, and its projection is evenly distributed onto the left turn and straight-ahead axes for calculation.

[0085] Furthermore, although this embodiment is designed for crossroads, T-shaped intersections also exist in reality. Therefore, it is necessary to reclassify the traffic flow that does not actually exist in T-shaped intersections.

[0086] Therefore, the sum of the projections of the vehicles traveling in the same direction as the vehicle is calculated and compared with the sum of the projections of all vehicles, and the ratio is used as the confidence level of the current traffic status.

[0087] The above are the steps for calculating the confidence level of the current traffic status. In this specific embodiment, corresponding activation conditions need to be designed to save system resources. Simultaneously, when the vehicle is at the same intersection and not turning left from a non-motorized vehicle lane, the coordinate system can be reused after a simple check, reducing computational overhead. Therefore, when a vehicle enters the intersection area, the above calculation process is initiated, as follows:

[0088] Based on the confidence level of the current traffic status and the current traffic light recognition results at the intersection, a traffic decision is made and executed through mutual exclusion judgment.

[0089] Among them, decision execution: by calculating the traffic flow status within the intersection, the probability of each passage status is calculated based on mutual exclusivity;

[0090] A. Read the traffic light recognition result at the current intersection through the traffic light sensing module. When the recognition result is green, verify the confidence level of the current traffic status:

[0091] If the current traffic status confidence level is greater than or equal to the traffic threshold, the vehicle is allowed to pass.

[0092] If the confidence level of the current traffic status is less than the threshold, a stop command is triggered to decelerate and brake, and the data is reported to the cloud to request manual intervention.

[0093] B. Read the traffic light recognition results at the current intersection through the traffic light sensing module. If the recognition result is a recognition failure or the recognition confidence is lower than a preset threshold, then verify the confidence of the current traffic status:

[0094] If the confidence level of the current passage status is greater than or equal to the passage threshold, proceed as if the light is green.

[0095] If the confidence level of the current traffic status is less than the threshold, stop the vehicle as if it were a red light.

[0096] Optionally, the above process states are calculated periodically every 100ms when the vehicle enters the intersection area (e.g., when the intersection is 30 meters ahead), and this calculation continues until the vehicle leaves the intersection area, and the calculation results are stored in the information queue.

[0097] To improve the accuracy of the judgment, this embodiment also extracts the calculation results of the information queue during the decision execution. The queue length is 30, and the data within 3 seconds is recorded. The final result is the confidence level of the current passage status, which is the average of the data frames within 3 seconds.

[0098] On the other hand, this embodiment configures a corresponding traffic status judgment system based on intersection traffic flow on the basis of the above method. The system includes: a traffic light sensing module, a vehicle information reading module, an intersection coordinate system establishment module, an obstacle reading module, and a data processing module;

[0099] The traffic light sensing module is used to identify the current traffic light status at the intersection. Optionally, the traffic light sensing module in this embodiment adopts a visual recognition module, which can identify the current traffic light status at the intersection under conditions where visibility meets the shooting conditions and there is no obstruction.

[0100] The vehicle information reading module is used to acquire vehicle information. It is connected to the vehicle's onboard sensors, including but not limited to speed sensors and positioning modules (such as Beidou and GPS).

[0101] The intersection coordinate system establishment module is used to establish the non-Cartesian coordinate system of the current intersection and calculate the center point of the current intersection and the vehicle's projection axis.

[0102] The obstacle reading module is used to filter dynamic obstacles and eliminate vehicles with low reference value.

[0103] The data processing module is used to calculate the confidence level of the current passage status and to execute decisions.

[0104] This embodiment also provides a vehicle, which can be an unmanned vehicle or a manned vehicle. The vehicle is equipped with the aforementioned traffic status judgment system based on intersection traffic flow. When it is an unmanned vehicle, the processing method described above is followed; when it is a manned vehicle, the driver is given feedback information.

[0105] Finally, it should be noted that the above content is only used to illustrate the technical solution of the present invention, and is not intended to limit the scope of protection of the present invention. Simple modifications or equivalent substitutions made by those skilled in the art to the technical solution of the present invention do not depart from the essence and scope of the technical solution of the present invention.

Claims

1. A method for judging the traffic state of an intersection based on traffic flow, comprising the following steps: obtaining ego vehicle information; establishing a coordinate system of the current intersection based on the ego vehicle information, the coordinate system being provided with a plurality of vehicle running projection axes; calculating vehicle projection sums on each vehicle running projection axis by projection method; calculating a same-direction running vehicle projection sum, which is the sum of vehicle projection sums on vehicle running projection axes in the same direction as the ego vehicle; calculating a current traffic state confidence, which is the ratio of the same-direction running vehicle projection sum to a total vehicle projection sum; the total vehicle projection sum being the sum of vehicle projection sums on each vehicle running projection axis; calculating a same-direction running vehicle projection sum; the addends of the same-direction running vehicle projection sum including a projection sum in the ego vehicle direction and a projection sum in a coexisting direction; the coexisting direction being a straight running vehicle direction and a left turning vehicle direction allowed to run in the same traffic light phase; making a traffic decision based on the current traffic state confidence and a current intersection traffic light recognition result by mutual exclusivity judgment; the ego vehicle information including the ego vehicle position, the ego vehicle orientation, and the ego vehicle driving state; the ego vehicle driving state being provided with at least the ego vehicle speed; the establishment of the coordinate system of the current intersection comprising the following steps: selecting a lane direction closest to the ego vehicle orientation as the positive direction of the Y-axis and selecting a lane direction with the largest included angle with the Y-axis as the positive direction of the X-axis; selecting the center of intersection points of straight running lane boundaries or the center of lane intersection points as the center point of the current intersection; setting a plurality of vehicle running projection axes to adapt to the positive and negative directions of the X-axis and the Y-axis and the directions of angle bisectors of each quadrant; calculating the vehicle projection sum comprising the following steps: projecting the speed vector of a vehicle onto the nearest vehicle running projection axis and taking the absolute value to obtain a projection value; and adding the projection values on the vehicle running projection axes to obtain the vehicle projection sum; for a left turning critical state vehicle, dividing the projection value of the left turning critical state vehicle equally between the straight running axis and the left turning axis of the left turning critical state vehicle; and for a lane direction not existing in a T-shaped intersection, reclassifying the vehicles belonging to the corresponding vehicle running projection axis to adjacent vehicle running projection axes; the traffic decision making based on mutual exclusivity judgment comprising: reading the current intersection traffic light recognition result; when the recognition result is a green light, checking the current traffic state confidence; if the current traffic state confidence is greater than or equal to a traffic threshold, allowing traffic; if the current traffic state confidence is less than the threshold, triggering a stop instruction and reporting data for manual takeover; reading the current intersection traffic light recognition result; when the recognition result is a recognition failure or a recognition confidence lower than a preset threshold, checking the current traffic state confidence; if the current traffic state confidence is greater than or equal to a traffic threshold, running in the green light state; if the current traffic state confidence is less than the threshold, stopping in the red light state; before calculating the vehicle projection sum on each vehicle running projection axis, performing data cleaning on observable vehicles at the intersection, comprising the following steps: ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ 2. The intersection traffic flow-based traffic light phase determination method of claim 1, wherein: ​ ​ ​ ​ ​ ​ ​ 3. The intersection traffic flow-based traffic light phase determination method of claim 1, wherein: ​ Eliminating low reference value vehicles based on operation parameters of the intersection observable vehicles in the coordinate system of the current intersection; The operation parameters of the intersection observable vehicles include vehicle speed, position quadrant, speed orientation and distance to intersection center of the intersection observable vehicles; The low reference value vehicles include right-turn vehicles, U-turn vehicles and edge slow-speed vehicles.

4. The intersection traffic flow-based traffic light phase determination method of claim 3, wherein: Eliminating low reference value vehicles includes the following steps: For a vehicle whose distance to intersection center exceeds 1 / 3 of the intersection radius and whose speed orientation is in the quadrant which is one more than the vehicle's position quadrant, the vehicle is determined as the right-turn vehicle; For a vehicle whose distance to intersection center exceeds 1 / 2 of the intersection radius and whose speed orientation is opposite to the straight direction of the lane, the vehicle is determined as the U-turn vehicle; For a vehicle whose speed is lower than the preset speed threshold and which is located at the edge of the intersection, the vehicle is determined as the edge slow-speed vehicle.

5. The intersection traffic flow-based traffic light phase determination method of claim 1, wherein: Further comprising the following steps: periodically calculating real-time traffic state confidence when the ego vehicle enters the intersection range, continuing until the ego vehicle leaves the intersection range, and storing the calculation results in the information queue; When performing the traffic decision execution, extracting the calculation results of the information queue and obtaining the current traffic state confidence by weighted average.

6. A system for judging the traffic flow state at an intersection based on traffic flow, characterized by: A system for implementing the intersection traffic flow based traffic state judgment method of any one of claims 1-5; The system comprises a traffic light perception module, an ego vehicle information reading module, an intersection coordinate system establishment module, an obstacle reading module and a data processing module; The traffic light perception module is used to identify the traffic light state of the current intersection; The ego vehicle information reading module is used to obtain ego vehicle information; The intersection coordinate system establishment module is used to establish the coordinate system of the current intersection and calculate the vehicle operation projection axis of the current intersection; The obstacle reading module is used to screen dynamic obstacles; The data processing module is used to calculate the current traffic state confidence and traffic decision execution.

7. A vehicle characterized by: Deploying the intersection traffic flow based traffic state judgment system of claim 6.

Images