Intersection scene identification method and computer readable storage medium

Abstract

The embodiment of the invention provides an intersection scene recognition method and a computer readable storage medium, and relates to the technical field of traffic signal control. The method comprises the following steps: for a target lane in each direction of a target intersection, obtaining target vehicle passing data of the target lane during a target green light period in a to-be-controlled time period; based on the target vehicle passing data, obtaining a first to-be-identified feature, a second to-be-identified feature and a third to-be-identified feature; determining a traffic flow discrete state of the target lane based on the first to-be-identified feature, determining a traffic demand state of the target lane based on the second to-be-identified feature, and determining a traffic operation state of the target lane based on the third to-be-identified feature; and determining the scene type of the target intersection in the to-be-controlled time period according to the traffic flow discrete state, the traffic demand state and the traffic operation state of the target lane in each direction. Therefore, scene recognition can be performed on the intersection according to the vehicle passing data of the intersection.

Description

Technical Field

[0001] The present application relates to the technical field of traffic signal control, and in particular to an intersection scene recognition method and a computer-readable storage medium. Background Art

[0002] In the field of traffic signal control technology, the traffic status of the same intersection may be different at different times of the day due to the influence of human life patterns. For example, on urban roads, the number of vehicles traveling at each intersection during the morning rush hour (such as 7:00-9:00) is usually greater than the number of vehicles traveling at each intersection during other time periods (such as 9:00-11:00).

[0003] However, if the same traffic signal is used for intersections with different traffic conditions, there may be a mismatch between the traffic signal and the traffic conditions, resulting in vehicle jams due to a short green light time, or idling due to a long green light time.

[0004] Therefore, there is an urgent need for an intersection scene recognition method to perform scene recognition on the intersection based on the vehicle passing data at the intersection. Subsequently, a suitable traffic signal control strategy can be adopted at the intersection according to the scene of the intersection. Summary of the invention

[0005] The purpose of the embodiment of the present application is to provide a scene recognition method for an intersection and a computer-readable storage medium, so as to perform scene recognition on the intersection according to the vehicle passing data at the intersection. The specific technical solution is as follows:

[0006] In a first aspect of an embodiment of the present application, a method for identifying an intersection scene is first provided, the method comprising:

[0007] For the target lane in each direction of the target intersection, the target vehicle passing data of the target lane during the target green light period within the time period to be controlled is obtained;

[0008] Based on the target vehicle passing data, a first feature to be identified that represents the probability distribution of the headway time in the target lane during the target green light period, a second feature to be identified that represents the change in the number of vehicles released by the target lane in each statistical time period during the target green light period, and a third feature to be identified that represents the probability distribution of the vehicle driving speed in the target lane during the target green light period are obtained;

[0009] Determine the traffic flow discrete state of the target lane based on the first feature to be identified, determine the traffic demand state of the target lane based on the second feature to be identified, and determine the traffic operation state of the target lane based on the third feature to be identified;

[0010] According to the traffic flow discrete state, traffic demand state, and traffic operation state of the target lanes in each direction, the scene type of the target intersection within the time period to be controlled is determined.

[0011] In some embodiments, during the time period to be controlled, the target lane has the largest number of released vehicles among the lanes of the road section in any direction of the target intersection.

[0012] In some embodiments, before determining the traffic flow discrete state of the target lane based on the first feature to be identified, the method further includes: for each preset traffic flow discrete state, obtaining first sample vehicle passing data of the first sample lane during a first sample green light period when the first sample lane is in the preset traffic flow discrete state; based on the first sample vehicle passing data, obtaining a first sample feature representing the probability distribution of the headway time in the first sample lane during the first sample green light period;

[0013] The method of determining the traffic flow discrete state of the target lane based on the first feature to be identified includes: for each preset traffic flow discrete state, calculating the similarity between the first sample feature of the preset traffic flow discrete state and the first feature to be identified by using a first similarity algorithm; wherein the first similarity algorithm includes at least one of the following: a bulldozer distance algorithm and a Mahalanobis distance algorithm; and determining the preset traffic flow discrete state with the largest corresponding similarity as the traffic flow discrete state of the target lane.

[0014] In some embodiments, the obtaining of the first sample feature representing the probability distribution of the headway time in the first sample lane during the first sample green light period based on the first sample vehicle passing data includes: determining each headway time at a designated position in the first sample lane during the first sample green light period as the sample headway time based on the first sample vehicle passing data; determining a preset headway time interval to which each sample headway time interval belongs; wherein each preset headway time interval is adjacent and has the same length; for each preset headway time interval, calculating the ratio of the sample headway time of the sample headway time belonging to the preset headway time interval to all sample headway time intervals, to obtain the first sample feature;

[0015] The method of obtaining a first feature to be identified that represents the probability distribution of the vehicle headway in the target lane during the target green light period based on the target vehicle passing data includes: determining, based on the target vehicle passing data, each vehicle headway at a designated position in the target lane during the target green light period as the vehicle headway to be processed; determining a preset vehicle headway interval to which each vehicle headway to be processed belongs; and for each preset vehicle headway interval, calculating the ratio of the vehicle headway to be processed that belongs to the preset vehicle headway interval to all vehicle headways to be processed, to obtain the first feature to be identified.

[0016] In some embodiments, before determining the traffic demand state of the target lane based on the second feature to be identified, the method further includes: for each preset traffic demand state, obtaining second sample vehicle passing data of the second sample lane during a second sample green light period when the second sample lane is in the preset traffic demand state; based on the second sample vehicle passing data, obtaining a second sample feature representing a change in the number of vehicles released by the second sample lane in each statistical time period during the second sample green light period;

[0017] The method of determining the traffic demand state of the target lane based on the second feature to be identified includes: for each preset traffic demand state, calculating the similarity between the second sample feature of the preset traffic demand state and the second feature to be identified; and determining the preset traffic demand state with the greatest corresponding similarity as the traffic demand state of the target lane.

[0018] In some embodiments, the second sample green light period includes a plurality of sample green light cycles; the target green light period includes a plurality of target green light cycles;

[0019] The second sample feature indicating the change of the number of vehicles released by the second sample lane in each statistical time period during the second sample green light period is obtained based on the second sample vehicle passing data, including: for each sample green light cycle, based on the second sample vehicle passing data of the sample green light cycle, determining the number of vehicles released by the second sample lane in each statistical time period of the sample green light cycle as the number of vehicles released in each statistical time period of the sample green light cycle; for any statistical time period, calculating the mean of the number of vehicles released in the statistical time period of each sample green light cycle, to obtain the second sample feature indicating the change of the number of vehicles released by the second sample lane in each statistical time period during the second sample green light period;

[0020] The second feature to be identified that represents the change in the number of vehicles released by the target lane in each statistical time period during the target green light period based on the target vehicle passing data is obtained, including: for each target green light cycle, based on the target vehicle passing data of the target green light cycle, determining the number of vehicles released by the target lane in each statistical time period of the target green light cycle as the number of vehicles released in each statistical time period of the target green light cycle; for any statistical time period, calculating the mean of the number of vehicles released in the statistical time period of each target green light cycle, to obtain the second feature to be identified that represents the change in the number of vehicles released by the target lane in each statistical time period during the target green light period.

[0021] In some embodiments, before determining the traffic operation state of the target lane based on the third feature to be identified, the method further includes: for each preset traffic operation state, obtaining third sample vehicle passing data of the third sample lane during a third sample green light period when the third sample lane is in the preset traffic operation state; based on the third sample vehicle passing data, obtaining a third sample feature representing the probability distribution of the vehicle driving speed in the third sample lane during the third sample green light period;

[0022] The method of determining the traffic operation state of the target lane based on the third feature to be identified includes: for each preset traffic operation state, calculating the similarity between the third sample feature of the preset traffic operation state and the third feature to be identified by a second similarity algorithm; wherein the second similarity algorithm includes at least one of the following: a bulldozer distance algorithm and a Mahalanobis distance algorithm; and determining the preset traffic operation state with the largest corresponding similarity as the traffic operation state of the target lane.

[0023] In some embodiments, the obtaining of the third sample feature representing the probability distribution of the vehicle driving speed in the third sample lane during the third sample green light period based on the third sample vehicle passing data includes: determining the driving speed of each vehicle in the third sample lane during the third sample green light period as the sample driving speed based on the third sample vehicle passing data; determining the preset driving speed interval to which each sample driving speed belongs; wherein each preset driving speed interval is adjacent and has the same length; for each preset driving speed interval, calculating the ratio of the sample driving speed belonging to the preset driving speed interval to all the sample driving speeds to obtain the third sample feature;

[0024] The method of obtaining a third feature to be identified that represents the probability distribution of the vehicle driving speeds in the target lane during the target green light period based on the target vehicle passing data includes: determining the driving speed of each vehicle in the target lane during the target green light period as the driving speed to be processed based on the target vehicle passing data; determining the preset driving speed interval to which each driving speed to be processed belongs; and for each preset driving speed interval, calculating the ratio of the driving speed to be processed that belongs to the preset driving speed interval to all driving speeds to be processed, to obtain the third feature to be identified.

[0025] In some embodiments, the traffic flow discrete state is: discrete, weak following, or strong following; the traffic demand state is: low traffic demand or high traffic demand; the traffic operation state is: slow or smooth; the scene type of the target intersection in the time period to be controlled is determined according to the traffic flow discrete state, traffic demand state, and traffic operation state of the target lanes in each direction, including:

[0026] If at least one target lane in the target lanes in each direction of the target intersection satisfies the first filtering condition, it is determined that the scene type of the target intersection in the time period to be controlled is a queuing balance type; wherein the first filtering condition indicates that the discrete state of traffic flow is strong following, the traffic demand state is high traffic demand, and the traffic operation state is slow movement; if there is no target lane in the target lanes in each direction of the target intersection that satisfies the first filtering condition, and there is at least one target lane that satisfies the second filtering condition, it is determined that the scene type of the target intersection in the time period to be controlled is a capacity type; wherein the second filtering condition indicates that the discrete state of traffic flow is strong following or weak following, the traffic demand state is high traffic demand, and the traffic operation state is unobstructed; if there is no target lane in the target lanes in each direction of the target intersection that satisfies the first filtering condition, and there is no target lane that satisfies the second filtering condition, it is determined that the scene type of the target intersection in the time period to be controlled is an efficiency improvement type.

[0027] In some embodiments, the method further comprises:

[0028] If the scenario type of the target intersection during the time period to be controlled is a queue balancing type, a control strategy of a first type of traffic signal is adopted at the target intersection; wherein the control strategy of the first type of traffic signal is used to shorten the queue length of vehicles at the target intersection; if the scenario type of the target intersection during the time period to be controlled is a capacity type, a control strategy of a second type of traffic signal is adopted at the target intersection; wherein the control strategy of the second type of traffic signal is used to improve the vehicle capacity of the target intersection; if the scenario type of the target intersection during the time period to be controlled is an efficiency improvement type, a control strategy of a third type of traffic signal is adopted at the target intersection; wherein the control strategy of the third type of traffic signal is used to reduce the delay duration of vehicles at the target intersection.

[0029] According to a second aspect of an embodiment of the present application, a device for identifying an intersection scene is provided, the device comprising:

[0030] A target vehicle passing data acquisition module is used to obtain the target vehicle passing data of the target lane during the target green light period within the time period to be controlled for the target lane in each direction of the target intersection; a feature to be identified acquisition module is used to obtain, based on the target vehicle passing data, a first feature to be identified that represents the probability distribution of the headway time in the target lane during the target green light period, a second feature to be identified that represents the change in the number of vehicles released by the target lane in each statistical time period during the target green light period, and a third feature to be identified that represents the probability distribution of the vehicle speed in the target lane during the target green light period; a state determination module is used to determine the traffic flow discrete state of the target lane based on the first feature to be identified, determine the traffic demand state of the target lane based on the second feature to be identified, and determine the traffic operation state of the target lane based on the third feature to be identified; a scene determination module is used to determine the scene type of the target intersection within the time period to be controlled according to the traffic flow discrete state, traffic demand state, and traffic operation state of the target lanes in each direction.

[0031] In some embodiments, during the time period to be controlled, the target lane has the largest number of released vehicles among the lanes of the road section in any direction of the target intersection.

[0032] In some embodiments, the device also includes: a first sample vehicle passing data acquisition module, which is used to obtain, for each preset traffic flow discrete state, the first sample vehicle passing data of the first sample lane during the first sample green light period when the first sample lane is in the preset traffic flow discrete state before determining the traffic flow discrete state of the target lane based on the first feature to be identified; a first sample feature acquisition module, which is used to obtain, based on the first sample vehicle passing data, a first sample feature representing the probability distribution of the headway time in the first sample lane during the first sample green light period; the state determination module is specifically used to: for each preset traffic flow discrete state, calculate the similarity between the first sample feature of the preset traffic flow discrete state and the first feature to be identified through a first similarity algorithm; wherein the first similarity algorithm includes at least one of the following: a bulldozer distance algorithm and a Mahalanobis distance algorithm; and determine the preset traffic flow discrete state with the largest corresponding similarity as the traffic flow discrete state of the target lane.

[0033] In some embodiments, the sample feature acquisition module is specifically used to: determine, based on the first sample vehicle passing data, each vehicle headway at a designated position in the first sample lane during the first sample green light period as a sample vehicle headway; determine a preset vehicle headway interval to which each sample vehicle headway belongs; wherein each preset vehicle headway interval is adjacent and has the same length; for each preset vehicle headway interval, calculate the ratio of the sample vehicle headway belonging to the preset vehicle headway interval to all sample vehicle headway intervals to obtain a first sample feature; the to-be-identified feature acquisition module is specifically used to: determine, based on the target vehicle passing data, each vehicle headway at a designated position in the target lane during the target green light period as a to-be-processed vehicle headway; determine the preset vehicle headway interval to which each to-be-processed vehicle headway belongs; for each preset vehicle headway interval, calculate the ratio of the to-be-processed vehicle headway belonging to the preset vehicle headway interval to all to-be-processed vehicle headway intervals to obtain a first to-be-identified feature.

[0034] In some embodiments, the device also includes: a second sample vehicle passing data acquisition module, which is used to obtain, for each preset traffic demand state, second sample vehicle passing data of the second sample lane during the second sample green light period when the second sample lane is in the preset traffic demand state before determining the traffic demand state of the target lane based on the second feature to be identified; a second sample feature acquisition module, which is used to obtain, based on the second sample vehicle passing data, a second sample feature representing the change in the number of vehicles released by the second sample lane in each statistical time period during the second sample green light period; the state determination module is specifically used to: for each preset traffic demand state, calculate the similarity between the second sample feature of the preset traffic demand state and the second feature to be identified; and determine the preset traffic demand state with the largest corresponding similarity as the traffic demand state of the target lane.

[0035] In some embodiments, the second sample green light period includes multiple sample green light cycles; the target green light period includes multiple target green light cycles; the second sample feature acquisition module is specifically used to: for each sample green light cycle, based on the second sample vehicle passing data of the sample green light cycle, determine the number of vehicles released by the second sample lane in each statistical time period of the sample green light cycle as the number of vehicles released in each statistical time period of the sample green light cycle; for any statistical time period, calculate the mean of the number of vehicles released in the statistical time period of each sample green light cycle, and obtain the second sample feature representing the change in the number of vehicles released by the second sample lane in each statistical time period of the second sample green light period; the to-be-identified feature acquisition module is specifically used to: for each target green light cycle, based on the target vehicle passing data of the target green light cycle, determine the number of vehicles released by the target lane in each statistical time period of the target green light cycle as the number of vehicles released in each statistical time period of the target green light cycle; for any statistical time period, calculate the mean of the number of vehicles released in the statistical time period of each target green light cycle, and obtain the second to-be-identified feature representing the change in the number of vehicles released by the target lane in each statistical time period of the target green light period.

[0036] In some embodiments, the device also includes: a third sample vehicle passing data acquisition module, which is used to obtain, for each preset traffic operation state, the third sample vehicle passing data of the third sample lane during the third sample green light period when the third sample lane is in the preset traffic operation state before determining the traffic operation state of the target lane based on the third feature to be identified; a third sample feature acquisition module, which is used to obtain, based on the third sample vehicle passing data, a third sample feature representing the probability distribution of the vehicle driving speed in the third sample lane during the third sample green light period; the state determination module is specifically used to: for each preset traffic operation state, calculate the similarity between the third sample feature of the preset traffic operation state and the third feature to be identified through a second similarity algorithm; wherein the second similarity algorithm includes at least one of the following: a bulldozer distance algorithm and a Mahalanobis distance algorithm; and determine the preset traffic operation state with the largest corresponding similarity as the traffic operation state of the target lane.

[0037] In some embodiments, the third sample feature acquisition module is specifically used to: determine the driving speed of each vehicle in the third sample lane during the third sample green light period as the sample driving speed based on the third sample vehicle passing data; determine the preset driving speed interval to which each sample driving speed belongs; wherein the preset driving speed intervals are adjacent and the lengths of the preset driving speed intervals are the same; for each preset driving speed interval, calculate the ratio of the sample driving speed belonging to the preset driving speed interval to all sample driving speeds to obtain the third sample feature; the to-be-identified feature acquisition module is specifically used to: determine the driving speed of each vehicle in the target lane during the target green light period as the to-be-processed driving speed based on the target vehicle passing data; determine the preset driving speed interval to which each to-be-processed driving speed belongs; for each preset driving speed interval, calculate the ratio of the to-be-processed driving speed belonging to the preset driving speed interval to all to-be-processed driving speeds to obtain the third sample feature.

[0038] In some embodiments, the traffic flow discrete state is: discrete, weak following, or strong following; the traffic demand state is: low traffic demand or high traffic demand; the traffic operation state is: slow or smooth; the scene determination module is specifically used to: if there is at least one target lane in the target lanes in each direction of the target intersection that meets the first screening condition, then determine that the scene type of the target intersection in the time period to be controlled is a queuing balance type; wherein the first screening condition indicates that the traffic flow discrete state is strong following, the traffic demand state is high traffic demand, and the traffic operation state is slow; if the target vehicles in each direction of the target intersection meet the first screening condition, then determine that the scene type of the target intersection in the time period to be controlled is a queuing balance type; wherein the first screening condition indicates that the traffic flow discrete state is strong following, the traffic demand state is high traffic demand, and the traffic operation state is slow; If there is no target lane that meets the first filtering condition among the target lanes in each direction of the target intersection, and there is at least one target lane that meets the second filtering condition, then the scene type of the target intersection in the time period to be controlled is determined to be the capacity type; wherein the second filtering condition represents: the traffic flow discrete state is strong following or weak following, the traffic demand state is high traffic demand, and the traffic operation state is smooth; if there is no target lane that meets the first filtering condition among the target lanes in each direction of the target intersection, and there is no target lane that meets the second filtering condition, then the scene type of the target intersection in the time period to be controlled is determined to be the efficiency improvement type.

[0039] In some embodiments, the device also includes: a first control strategy determination module, which is used to adopt a control strategy of a first type of traffic signal at the target intersection if the scenario type of the target intersection during the time period to be controlled is a queue balancing type; wherein the control strategy of the first type of traffic signal is used to shorten the queue length of vehicles at the target intersection; a second control strategy determination module, which is used to adopt a control strategy of a second type of traffic signal at the target intersection if the scenario type of the target intersection during the time period to be controlled is a capacity type; wherein the control strategy of the second type of traffic signal is used to improve the vehicle capacity of the target intersection; a third control strategy determination module, which is used to adopt a third type of traffic signal control strategy at the target intersection if the scenario type of the target intersection during the time period to be controlled is an efficiency improvement type; wherein the control strategy of the third type of traffic signal is used to reduce the delay duration of vehicles at the target intersection.

[0040] According to a third aspect of the embodiments of the present application, an electronic device is provided, including:

[0041] Memory, used to store computer programs;

[0042] The processor is used to implement any of the above-mentioned intersection scene recognition methods when executing the program stored in the memory.

[0043] An embodiment of the present application further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, any of the above-mentioned intersection scene recognition methods is implemented.

[0044] An embodiment of the present application also provides a computer program product comprising instructions, which, when executed on a computer, enables the computer to execute any of the above-described intersection scene recognition methods.

[0045] Beneficial effects of the embodiments of the present application:

[0046] An embodiment of the present application provides an intersection scene recognition method, the method comprising: for a target lane in each direction of a target intersection, obtaining target vehicle passing data of the target lane during a target green light period within a time period to be controlled; based on the target vehicle passing data, obtaining a first feature to be identified that represents the probability distribution of the headway time in the target lane during the target green light period, a second feature to be identified that represents the change in the number of vehicles released by the target lane during each statistical time period during the target green light period, and a third feature to be identified that represents the probability distribution of the vehicle speed in the target lane during the target green light period; determining the traffic flow discrete state of the target lane based on the first feature to be identified, determining the traffic demand state of the target lane based on the second feature to be identified, and determining the traffic operation state of the target lane based on the third feature to be identified; determining the scene type of the target intersection within the time period to be controlled according to the traffic flow discrete state, traffic demand state, and traffic operation state of the target lanes in each direction.

[0047] Based on the above processing, for the target lane in each direction of the target intersection, the first feature to be identified, the second feature to be identified and the third feature to be identified can be obtained based on the target vehicle passing data of the target lane during the target green light period.

[0048] Since the first feature to be identified can represent the probability distribution of the headway time in the target lane during the target green light period, the probability distribution can reflect the proportion of vehicles in the traffic flow that are constrained to varying degrees by the preceding vehicle, and therefore, the discrete state of the traffic flow in the target lane can be determined based on the first feature to be identified. Since the second feature to be identified can represent the change in the number of vehicles released by the target lane during each statistical time period during the target green light period, the change in the number can reflect the queue length of the vehicles in the target lane, and therefore, the traffic demand state of the target lane can be determined based on the second feature to be identified. Since the third feature to be identified can represent the probability distribution of the vehicle speed in the target lane during the target green light period, the probability distribution can reflect the proportion of vehicles with different driving speeds in the traffic flow, and therefore, the traffic operation state of the target lane can be determined based on the third feature to be identified.

[0049] Furthermore, any target lane can be described from three aspects: traffic flow discrete state, traffic demand state, and traffic operation state. It is also possible to determine the scene type of the target intersection during the time period to be controlled based on the traffic flow discrete state, traffic demand state, and traffic operation state of the target lanes in each direction. In this way, it is possible to obtain the scene type of the target intersection during the time period to be controlled based on the vehicle passing data of the target intersection during the time period to be controlled.

[0050] Of course, implementing any product or method of the present application does not necessarily require achieving all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application, and for ordinary technicians in this field, other embodiments can also be obtained based on these drawings.

[0052] Figure 1 A first flow chart of the intersection scene recognition method provided in an embodiment of the present application;

[0053] Figure 2 A flow chart of obtaining a first feature to be identified based on target vehicle passing data provided in an embodiment of the present application;

[0054] Figure 3 A flow chart of obtaining a second feature to be identified based on target vehicle passing data provided in an embodiment of the present application;

[0055] Figure 4 A flow chart of obtaining a third feature to be identified based on target vehicle passing data provided in an embodiment of the present application;

[0056] Figure 5 A schematic diagram of a flow chart of obtaining a first feature to be identified, a second feature to be identified, and a third feature to be identified of a target road section in each direction provided in an embodiment of the present application;

[0057] Figure 6 A flowchart of determining a discrete state of traffic flow in a target lane based on a first feature to be identified provided in an embodiment of the present application;

[0058] Figure 7 A flow chart of determining the traffic demand state of the target lane based on the second feature to be identified provided in an embodiment of the present application;

[0059] Figure 8 A flow chart of determining the traffic operation state of the target lane based on the third feature to be identified provided in an embodiment of the present application;

[0060] Fig. 9 A second flow chart of the intersection scene recognition method provided in an embodiment of the present application;

[0061] FIG10( a ) is a schematic diagram of lanes in different traffic flow discrete states provided by an embodiment of the present application;

[0062] FIG10( b1 ) is a schematic diagram of a change in the number of vehicles released from a lane with low traffic demand provided by an embodiment of the present application;

[0063] FIG10( b2 ) is a schematic diagram of the change in the number of vehicles released from a lane with high traffic demand provided by an embodiment of the present application;

[0064] FIG10( c ) is a schematic diagram of a road in different traffic operation states provided in an embodiment of the present application;

[0065] Fig.11 A schematic diagram of the structure of a control strategy determination device provided in an embodiment of the present application;

[0066] Fig.12 A structural diagram of an intersection scene recognition device provided in an embodiment of the present application;

[0067] Fig.13 A structural diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0068] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field based on the present application belong to the scope of protection of the present application.

[0069] In the field of traffic signal control technology, the traffic status of the same intersection may be different at different times of the day due to the influence of human life patterns. For example, on urban roads, the number of vehicles traveling at each intersection during the morning rush hour (such as 7:00-9:00) is usually greater than the number of vehicles traveling at each intersection during other time periods (such as 9:00-11:00).

[0070] However, if the same traffic signal is used for intersections with different traffic conditions, there may be a mismatch between the traffic signal and the traffic conditions, resulting in vehicle jams due to a short green light time, or idling due to a long green light time.

[0071] Therefore, there is an urgent need for an intersection scene recognition method to perform scene recognition on the intersection based on the vehicle passing data at the intersection. Subsequently, a suitable traffic signal control strategy can be adopted at the intersection according to the scene of the intersection.

[0072] An embodiment of the present application provides an intersection scene recognition method, which can be applied to an electronic device, which may be a traffic signal control system, or the electronic device may communicate with a traffic signal control system, and the electronic device may obtain vehicle passing data at an intersection controlled by the traffic signal control system.

[0073] See also Figure 1 , Figure 1 A first flow chart of the intersection scene recognition method provided in the embodiment of the present application, the method comprises the following steps:

[0074] S101: For a target lane in each direction of a target intersection, target vehicle passing data of the target lane during a target green light period within a time period to be controlled is obtained.

[0075] S102: Based on the target vehicle passing data, a first feature to be identified that represents the probability distribution of the headway time in the target lane during the target green light period, a second feature to be identified that represents the change in the number of vehicles released by the target lane in each statistical time period during the target green light period, and a third feature to be identified that represents the probability distribution of the vehicle driving speed in the target lane during the target green light period are obtained.

[0076] S103: Determine the traffic flow discrete state of the target lane based on the first feature to be identified, determine the traffic demand state of the target lane based on the second feature to be identified, and determine the traffic operation state of the target lane based on the third feature to be identified.

[0077] S104: Determine the scenario type of the target intersection within the time period to be controlled according to the traffic flow discrete state, traffic demand state, and traffic operation state of the target lanes in each direction.

[0078] Based on the above processing, for the target lane in each direction of the target intersection, the first feature to be identified, the second feature to be identified, and the third feature to be identified can be obtained based on the target vehicle passing data of the target lane during the target green light period. Since the first feature to be identified can represent the probability distribution of the headway time in the target lane during the target green light period, the probability distribution can reflect the proportion of vehicles in the traffic flow that are constrained to different degrees by the front vehicle, therefore, the traffic flow discrete state of the target lane can be determined based on the first feature to be identified. Since the second feature to be identified can represent the change in the number of vehicles released by the target lane in each statistical time period during the target green light period, the change in the number can reflect the queue length of the vehicles in the target lane, therefore, the traffic demand state of the target lane can be determined based on the second feature to be identified. Since the third feature to be identified can represent the probability distribution of the vehicle speed in the target lane during the target green light period, the probability distribution can reflect the proportion of vehicles with different driving speeds in the traffic flow, therefore, the traffic operation state of the target lane can be determined based on the third feature to be identified. Furthermore, for any target lane, it can be described from three aspects: the discrete state of traffic flow, the traffic demand state, and the traffic operation state. It is also possible to determine the scene type of the target intersection during the time period to be controlled based on the traffic flow discrete state, traffic demand state, and traffic operation state of the target lanes in each direction. In this way, it is possible to obtain the scene type of the target intersection during the time period to be controlled based on the vehicle passing data of the target intersection during the time period to be controlled.

[0079] For step S101, the target intersection is an intersection that needs to be identified as a scene type. For example, in a city road, an intersection on a city trunk road can be selected as the target intersection.

[0080] Any intersection includes multiple sections in different directions, that is, the sections included in the target intersection (which may be referred to as target sections) may be determined according to the directions of the sections. For example, a common cross intersection includes sections in four directions (e.g., sections in four directions of east, west, south, and north), and a T-intersection includes sections in three directions.

[0081] For each road section, the road section may include at least one lane. Accordingly, for the target road section in each direction in the target intersection, a lane may be determined from the lanes included in the target road section as the target lane in the direction. The process of specifically determining the target lane in the target road section will be described in subsequent embodiments.

[0082] It is understandable that the scene type of the target intersection is not fixed due to the influence of human activities (for example, there are more vehicles on the road during the day, but fewer vehicles on the road at night). Therefore, the scene type of the target intersection in a certain time period (which can be called the time period to be controlled) can be selected according to actual needs.

[0083] In one implementation, a day can be divided into multiple time periods. For example, a day can be divided into multiple time periods with each hour being a time period. Accordingly, each of the above time periods can be used as a time period to be controlled.

[0084] Alternatively, the technicians can also preset a fixed time period as the time period to be controlled according to actual needs. For example, according to the human life pattern, the morning peak time period (such as 7:00-9:00) and / or the evening peak time period (such as 17:00-20:00) can be determined every day, and then the morning peak time period and / or the evening peak time period can be determined as the time period to be controlled.

[0085] It is understandable that when determining the control strategy of the traffic signal adopted in the time period to be controlled on a certain day, the vehicle passing data in the historical time period of the time period to be controlled can be obtained, and the above steps S101-S104 are executed to obtain the scene type of the target intersection in the time period to be controlled. For example, for the target intersection, if the time period to be controlled is the morning rush hour period on Tuesday of a certain week, the vehicle passing data during the morning rush hour period on Tuesday of the previous week can be obtained, or the vehicle passing data during the morning rush hour period on Monday of the week can be obtained, and the above steps S101-S104 are executed, and then, according to the obtained scene type, the control strategy of the traffic signal adopted in the time period to be controlled on Tuesday of the week is determined.

[0086] The specific method of determining the control strategy of the traffic signal will be described in subsequent embodiments.

[0087] In some embodiments, the electronic device can obtain the signal timing data of the target intersection, and the signal timing data includes: the intersection identification, the road section identification, the lane identification, the green light start time, and the green light end time. Among them, the intersections correspond to each other, the road sections in any intersection correspond to each other, and the lanes in any road section also correspond to each other.

[0088] Furthermore, the electronic device can determine all green light cycles (i.e., the target green light cycle in this application) included in the time period to be controlled for each lane in the target section of the target intersection in the signal timing data according to the identification of the target intersection, the identification of the target section, and the identification of each lane in the target section. A green light cycle means: the duration of a green light in a traffic signal cycle.

[0089] For the target lane in each direction of the target intersection, after obtaining all the green light cycles included in the time period to be controlled for the target lane, the time corresponding to one or more green light cycles can be determined as the target green light period.

[0090] The electronic device can also obtain the vehicle passing data of the target intersection during the time period to be controlled, and the vehicle passing data includes: the intersection identification, the road section identification, the lane identification, the lane type, and the corresponding relationship between the vehicle passing time. Among them, the vehicle passing time corresponding to a lane identification indicates: the time for each vehicle released by the lane corresponding to the lane identification to travel to the specified position in the lane (such as the stop line of the lane).

[0091] Furthermore, the electronic device can determine the vehicle passing data of each lane in the target section of the target intersection within the time period to be controlled in the vehicle passing data based on the identification of the target intersection, the identification of the target section, and the identification of each lane in the target section.

[0092] In addition, the vehicle passing data may also include: license plate number and road section travel time. The road section travel time corresponding to a license plate number indicates: the time when the vehicle corresponding to the license plate number enters the road section, and the time when the vehicle corresponding to the license plate number exits the road section.

[0093] Furthermore, the vehicle passing data of the target lane during the target green light period (ie, target vehicle passing data) can be obtained based on the signal timing data of the target intersection and the vehicle passing data of the target intersection during the time period to be controlled.

[0094] In some embodiments, during the time period to be controlled, the target lane has the largest number of released vehicles among all lanes of the road section in any direction of the target intersection.

[0095] In the embodiment of the present application, for each lane in the target road section, the number of vehicle passing times in the vehicle passing data of the lane in the time period to be controlled can be used as the number of vehicles released by the lane in the time period to be controlled (which can be called the release number). Furthermore, the lane corresponding to the largest release number can be determined as the target lane of the target road section, so that the target lanes in different directions can be determined.

[0096] Based on the above processing, when determining the target lanes in each direction, for each lane in the target road section, the lane with the largest number of vehicles released during the time period to be controlled can be determined as the target lane. Therefore, the traffic state of the target lane can accurately reflect the traffic state of the road section to which the target lane belongs. Subsequently, when the scene type of the target intersection is determined according to the traffic state of each target lane and the control strategy of the traffic signal is determined, it can be ensured that the needs of the road sections in each direction in the target intersection are met.

[0097] In some embodiments, for each lane in the target road section, after obtaining the number of released vehicles in the lane, the ratio of the number of released vehicles in the lane to the traffic capacity of the lane can be calculated as the saturation of the lane, and the lane corresponding to the maximum saturation is determined as the target lane. The traffic capacity of any lane is preset by the technician according to the actual conditions of the lane.

[0098] Based on the above processing, when determining the target lanes in each direction, for each lane in the target road section, the lane with the largest saturation can be determined as the target lane. Therefore, the traffic status of the target lane can accurately reflect the traffic status of the road section to which the target lane belongs. Subsequently, when the scene type of the target intersection is determined according to the traffic status of each target lane and the control strategy of the traffic signal is determined, it can be ensured that the needs of the road sections in each direction of the target intersection are met.

[0099] For step S102, for each target lane, the target vehicle passing data of the target lane can be processed to obtain a first feature to be identified that represents the probability distribution of the headway time in the target lane during the target green light period, a second feature to be identified that represents the change in the number of vehicles released by the target lane in each statistical time period during the target green light period, and a third feature to be identified that represents the probability distribution of the vehicle driving speed in the target lane during the target green light period.

[0100] In some embodiments, see Figure 2 , Figure 2A flowchart of obtaining a first feature to be identified based on target vehicle passing data is provided in an embodiment of the present application. The steps of obtaining a first feature to be identified representing the probability distribution of the headway time in the target lane during the target green light period based on the target vehicle passing data include:

[0101] S201: Based on the target vehicle passing data, determine the headway time of each vehicle at a designated position in the target lane during the target green light period as the headway time to be processed.

[0102] S202: Determine the preset headway interval to which each headway to be processed belongs.

[0103] The preset headway intervals are adjacent to each other, and the lengths of the preset headway intervals are the same.

[0104] S203: For each preset headway interval, calculate the ratio of the headway to be processed that belongs to the preset headway interval to all headway to be processed, and obtain a first feature to be identified.

[0105] Headway time means: the time difference between the heads of adjacent vehicles on the same lane passing a specified position in the lane (such as the stop line of the lane).

[0106] With respect to step S201, since the target vehicle passing data includes: the vehicle passing time of the target lane in the time period to be controlled (which may be referred to as the first vehicle passing time), and the vehicle passing time indicates: the time taken for each vehicle released by the target lane to travel to the designated position in the target lane (such as the stop line of the lane). Therefore, the vehicle passing time (which may be referred to as the second vehicle passing time) within the target green light period may be determined in the first vehicle passing time, and the difference between every two adjacent second vehicle passing times may be calculated to obtain the headway time of each vehicle at the designated position in the target lane during the target green light period (i.e., the headway time to be processed).

[0107] For example, the headway time between any two vehicles can be calculated according to the following formula (1):

[0108] T i，i+1 =pass_time i+1 -pass_time i (1)

[0109] Among them, T i,i+1 Indicates the headway time between the i-th vehicle and the i+1-th vehicle; pass_time i+1 Indicates the time when the i+1th vehicle passes the specified position of the target lane, pass_time i It represents the time when the i-th vehicle passes the specified position of the target lane.

[0110] For step S202, the preset headway interval to which each headway to be processed belongs can be determined, wherein each preset headway interval is adjacent to another, and each preset headway interval has the same length.

[0111] For example, the length of each preset headway interval is 3 seconds, and the preset headway intervals are [0,3), [3,6), [6,9)... Accordingly, when a headway to be processed is 5 seconds, it can be determined that the headway to be processed belongs to the preset headway interval [3,6); when a headway to be processed is 8 seconds, it can be determined that the headway to be processed belongs to the preset headway interval [6,9).

[0112] For step S203, for each preset headway interval, the ratio of the headway to be processed that belongs to the preset headway interval to all the headway to be processed is calculated. Accordingly, the probability distribution of the headway to be processed (i.e., the probability distribution of the headway in the target lane during the target green light period) can be determined based on the ratio corresponding to each preset headway interval.

[0113] For example, each ratio can be calculated according to the following formula (2):

[0114]

[0115] Where P(r) is the ratio of the headway to be processed that belongs to the rth preset headway interval to all the headway to be processed, n r represents the number of headway times to be processed that belong to the rth preset headway time interval, and N represents the number of all headway times to be processed.

[0116] Furthermore, the probability distribution X of the headway to be processed can be expressed as:

[0117] X={r, P(r)}, r=1, 2, 3,...

[0118] Wherein, r represents the rth preset headway interval, and P(r) is the ratio of the headway to be processed belonging to the rth preset headway interval to all headway to be processed.

[0119] In some embodiments, the target green light period includes multiple target green light cycles. Figure 3 , Figure 3 A flowchart of obtaining a second feature to be identified based on target vehicle passing data is provided in an embodiment of the present application. The steps of obtaining a second feature to be identified representing the change in the number of vehicles released by the target lane during each statistical time period during the target green light period based on the target vehicle passing data include:

[0120] S301: For each target green light cycle, based on the target vehicle passing data of the target green light cycle, determine the number of vehicles released by the target lane in each statistical time period of the target green light cycle as the number of vehicles released in each statistical time period of the target green light cycle.

[0121] S302: For any statistical time period, calculate the average of the number of released vehicles in each statistical time period of each target green light cycle to obtain a second feature to be identified that represents the change in the number of released vehicles in the target lane in each statistical time period during the target green light period.

[0122] In the embodiment of the present application, each target green light cycle can be divided according to a preset duration to determine the statistical time periods included in each target green light cycle.

[0123] For example, the preset duration may be 5 seconds. Accordingly, for a target green light cycle, the first statistical time period included in the target green light cycle is 0 to 5 seconds in the target green light cycle, the second statistical time period included in the target green light cycle is 5 to 10 seconds in the target green light cycle, and so on, until all statistical time periods included in the target green light cycle are determined.

[0124] Correspondingly, since the target vehicle passing data includes: the passing time of the target lane within the time period to be controlled (i.e., the first passing time), therefore, for each statistical time period of each target green light cycle, the number of passing times in each statistical time period during the target green light can be determined, that is, the number of vehicles released by the target lane in each statistical time period of the target green light cycle is determined as the release number in each statistical time period of the target green light cycle.

[0125] For any statistical time period, calculate the average number of releases in the statistical time period within each target green light cycle.

[0126] For example, for the first statistical time period mentioned above, the mean of the number of vehicles released in the first statistical time period within each target green light cycle can be calculated (which can be called the mean corresponding to the first statistical time period). Correspondingly, for the second statistical time period, the mean of the number of vehicles released in the second statistical time period within each target green light period can be calculated. And so on, until the mean corresponding to each statistical time period is obtained, the change in the number of vehicles released in the target lane within each statistical time period during the target green light period can be obtained (which can be called the periodic traffic flow rate). Furthermore, the second feature to be identified that represents the change in the number of vehicles released in the target lane within each statistical time period during the target green light period can be obtained.

[0127] For example, the change in the number of vehicles released by the target lane during each statistical time period during the target green light period can be expressed as:

[0128] flow=[flow1,flow2,flow3,…”

[0129] Among them, flow i Represents the mean value corresponding to the i-th statistical time period.

[0130] In some embodiments, see Figure 4 , Figure 4 A flowchart of obtaining a third feature to be identified based on target vehicle passing data is provided in an embodiment of the present application. The steps of obtaining a third feature to be identified representing the probability distribution of the vehicle speed in the target lane during the target green light period based on the target vehicle passing data include:

[0131] S401: Based on the target vehicle passing data, determine the driving speed of each vehicle in the target lane during the target green light period as the driving speed to be processed.

[0132] S402: Determine the preset driving speed range to which each to-be-processed driving speed belongs.

[0133] The preset driving speed intervals are adjacent to each other, and the lengths of the preset driving speed intervals are the same.

[0134] S403: For each preset driving speed interval, calculate the ratio of the to-be-processed driving speeds belonging to the preset driving speed interval to all the to-be-processed driving speeds to obtain a third feature to be identified.

[0135] For step S401, since the target vehicle passing data includes: license plate number and road section travel time. Therefore, for each vehicle, the driving time of the vehicle in the target road section is obtained according to the time when the vehicle corresponding to the license plate number enters the road section and the time when the vehicle corresponding to the license plate number exits the road section. The ratio of the length of the target road section to the driving time is calculated as the driving speed of the vehicle. Furthermore, the driving speed of each vehicle in the target lane during the target green light period can be determined as the driving speed to be processed.

[0136] For step S402, the preset driving speed interval to which each to-be-processed driving speed belongs may be determined, wherein the preset driving speed intervals are adjacent and have the same length.

[0137] For example, the length of each preset driving speed interval is 5 m / s (meters per second), and the preset driving speed intervals are [0,5), [5,10), [10,15), [15,20)... Accordingly, when a to-be-processed driving speed is 19 m / s, it can be determined that the to-be-processed driving speed belongs to the preset driving speed interval [15,20); when a to-be-processed driving speed is 32 m / s, it can be determined that the to-be-processed driving speed belongs to the preset driving speed interval [30,35).

[0138] For step S403, for each preset driving speed interval, the ratio of the to-be-processed driving speed belonging to the preset driving speed interval to all the to-be-processed driving speeds is calculated. Accordingly, the probability distribution of the to-be-processed driving speeds (i.e., the probability distribution of the driving speeds of the vehicles in the target lane during the target green light period) can be determined according to the ratio corresponding to each preset driving speed interval.

[0139] For example, each ratio can be calculated according to the following formula (3):

[0140]

[0141] Wherein, P(j) is the ratio of the to-be-processed driving speed belonging to the jth preset driving speed interval to all the to-be-processed driving speeds, n j represents the number of to-be-processed driving speeds belonging to the j-th preset driving speed interval, and M represents the number of all to-be-processed driving speeds.

[0142] Furthermore, the probability distribution Y of the driving speed to be processed can be expressed as:

[0143] Y={j,P(j)},j=1,2,3,...

[0144] Wherein, j represents the jth preset driving speed interval, and P(j) is the ratio corresponding to the jth preset driving speed interval.

[0145] With respect to step S103 , the traffic state of the target lane may be reflected by the traffic flow discrete state, traffic demand state, and traffic operation state of the target lane.

[0146] The traffic flow discrete state is used to indicate whether the latter vehicle is constrained by the former vehicle among two adjacent vehicles in the same lane. It can be understood that the larger the headway between two adjacent vehicles, the larger the distance between the two adjacent vehicles, and the less the latter vehicle is constrained by the former vehicle. Correspondingly, the smaller the headway between two adjacent vehicles, the smaller the distance between the two adjacent vehicles, and the more the latter vehicle is constrained by the former vehicle.

[0147] The traffic demand state is used to indicate whether the lane can release all queued vehicles within a green light period. It is understandable that in a green light period, if the number of vehicles that need to be released by the lane is small, the vehicles can be completely released in the first several statistical time periods. Therefore, there is an empty release situation in the green light cycle. Correspondingly, it is manifested as follows: according to the sequence of the statistical time periods, in each statistical time period, the number of released vehicles first increases, then tends to be stable, and finally decreases. If the number of vehicles that need to be released by the lane is large, the vehicles cannot be completely released in a green light cycle. Therefore, there is no empty release situation in the green light cycle. Correspondingly, it is manifested as follows: according to the sequence of the statistical time periods, in each statistical time period, the number of released vehicles first increases, then tends to be stable, until the end of the green light cycle.

[0148] Traffic operation status is used to indicate the speed of vehicles on the lane. It is understandable that the less external interference vehicles on the lane are subject to, the faster the vehicles are. Correspondingly, the greater the external interference vehicles on the lane are subject to, the slower the vehicles are. For example, vehicles on highways are generally not disturbed by pedestrians and other vehicles when driving, and their driving speed is relatively fast; vehicles on urban roads need to avoid pedestrians and oncoming vehicles when driving, and their driving speed is relatively slow.

[0149] The specific process of determining the traffic flow discrete state of the target lane based on the first feature to be identified, determining the traffic demand state of the target lane based on the second feature to be identified, and determining the traffic operation state of the target lane based on the third feature to be identified will be described in subsequent embodiments.

[0150] For step S104, since the target intersection includes target lanes in multiple directions, and the traffic state of each target lane may be different. Therefore, after determining the traffic flow discrete state, traffic demand state, and traffic operation state of the target lane in each direction of the target intersection according to the above steps S101-S103, the scene type of the target intersection in the time period to be controlled can be determined in combination with the traffic flow discrete state, traffic demand state, and traffic operation state of the target lanes in each direction. The process of specifically determining the scene type of the target intersection in the time period to be controlled will be described in subsequent embodiments.

[0151] See also Figure 5 , Figure 5 A schematic diagram of a flow chart for obtaining a first feature to be identified, a second feature to be identified, and a third feature to be identified of a target road section in each direction provided in an embodiment of the present application.

[0152] S501: Start.

[0153] S502: Obtain vehicle passing data and signal timing data at the intersection.

[0154] That is, the signal timing data of the target intersection and the vehicle passing data of the target intersection during the time period to be controlled are obtained.

[0155] S503: Traverse the key lanes in each direction.

[0156] That is, for a target road section in any direction of a target intersection, a target lane in the target road section is determined.

[0157] S504: Obtain vehicle passing data within each green light time period.

[0158] That is, the target vehicle passing data of the target lane during the target green light period within the time period to be controlled is obtained, wherein the target green light period includes at least a plurality of target green light cycles.

[0159] S505: Extracting a first feature to be identified, a second feature to be identified, and a third feature to be identified.

[0160] That is, the target vehicle passing data of the target lane is processed to obtain a first feature to be identified that represents the probability distribution of the vehicle headway in the target lane during the target green light period, a second feature to be identified that represents the change in the number of vehicles released by the target lane in each statistical time period during the target green light period, and a third feature to be identified that represents the probability distribution of the vehicle driving speed in the target lane during the target green light period.

[0161] S506: Determine whether all directions have been traversed. If yes, execute step S507, if no, return to execute step S503.

[0162] S507: End.

[0163] In some embodiments, see Figure 6 , Figure 6 A flowchart of determining the discrete state of traffic flow in a target lane based on a first feature to be identified is provided in an embodiment of the present application. Figure 6 middle:

[0164] S601: For each preset traffic flow discrete state, obtain first sample vehicle passing data of the first sample lane during a first sample green light period when the first sample lane is in the preset traffic flow discrete state.

[0165] S602: Based on the first sample vehicle passing data, a first sample feature representing the probability distribution of the headway time of the first sample lane during the first sample green light period is obtained.

[0166] S603: For each preset traffic flow discrete state, calculate the similarity between the first sample feature of the preset traffic flow discrete state and the first feature to be identified by using a first similarity algorithm.

[0167] The first similarity algorithm includes at least one of the following: a bulldozer distance algorithm and a Mahalanobis distance algorithm.

[0168] S604: Determine the preset traffic flow discrete state with the greatest corresponding similarity as the traffic flow discrete state of the target lane.

[0169] In the embodiment of the present application, the technician can determine based on actual experience that the first sample lane is in a preset traffic flow discrete state during the first sample time period, and then obtain the vehicle passing data of the first sample lane during the first sample green light period (i.e., the first sample vehicle passing data) as the first sample feature of the preset traffic flow discrete state. The process of determining the first sample green light period in the first sample time period can refer to the process of determining the target green light period in the time period to be controlled, which will not be repeated here. The length of the first sample time period is consistent with the length of the time period to be controlled.

[0170] Furthermore, based on the first sample vehicle passing data, a first sample feature representing the probability distribution of the headway in the first sample lane during the first sample green light period can be obtained. The specific process of obtaining the first sample feature will be described in the subsequent embodiments.

[0171] For each preset traffic flow discrete state, the similarity between the first sample feature of the preset traffic flow discrete state and the first feature to be identified can be obtained by a first similarity algorithm. The first similarity algorithm includes at least one of the following: a bulldozer distance algorithm, a Mahalanobis distance algorithm, and other algorithms for calculating the similarity between two probability distributions.

[0172] When the first similarity algorithm only includes one of the above algorithms for calculating the similarity between two probability distributions, the first similarity algorithm can be used to calculate the similarity between the first sample feature of the preset traffic flow discrete state and the first feature to be identified.

[0173] When the first similarity algorithm includes multiple algorithms for calculating the similarity between two probability distributions, for each first similarity algorithm, the first similarity between the first sample feature of the preset traffic flow discrete state and the first feature to be identified can be calculated by the first similarity algorithm, and then the similarity between the first sample feature of the preset traffic flow discrete state and the first feature to be identified can be obtained by combining the calculated multiple first similarities. For example, the weighted sum of the calculated multiple first similarities can be calculated to obtain the similarity between the first sample feature of the preset traffic flow discrete state and the first feature to be identified.

[0174] In one implementation, if the preset traffic flow discrete state has multiple first sample features, the similarity between each first sample feature of the preset traffic flow discrete state and the first feature to be identified can be calculated, and based on the average of the multiple similarities, the similarity between the first feature to be identified and the first sample feature of the preset traffic flow discrete state is obtained. In this way, the accuracy of the determined traffic flow discrete state of the target lane can be further improved.

[0175] The greater the similarity between the first feature to be identified and the first sample feature of the preset traffic flow discrete state, the closer the traffic flow discrete state of the target lane is to the preset traffic flow discrete state. Therefore, the preset traffic flow discrete state with the greatest corresponding similarity can be determined as the traffic flow discrete state of the target lane.

[0176] Based on the above processing, the traffic flow discrete state of the target lane can be determined by calculating the similarity between the first feature to be identified and the first sample feature of each preset traffic flow discrete state, thereby ensuring the accuracy of the determined traffic flow discrete state of the target lane.

[0177] In some embodiments, the first feature to be identified can be input into a pre-trained neural network model to obtain the discrete state of the traffic flow of the target lane. The neural network model is trained based on the first sample feature and the corresponding sample label, and the sample label can represent the preset discrete state of the traffic flow corresponding to the first sample feature. In this way, the first feature to be identified is processed based on the neural network model to determine the discrete state of the traffic flow of the target lane, which can improve the efficiency of intersection scene recognition.

[0178] In addition, in the related art, when determining the discrete state of traffic flow in a lane, the discrete state of traffic flow in the lane can be determined by judging whether the headway of vehicles in the lane is greater than a preset threshold. However, due to the different actual conditions of different roads, the preset threshold may also be different, making it difficult to set the threshold.

[0179] In some embodiments, a traffic flow discrete state sample library may be pre-established to store the first sample features of the above-mentioned preset traffic flow discrete states. That is, the above-mentioned steps S601-S604 may be performed based on the first sample features of the preset traffic flow discrete states recorded in the traffic flow discrete state sample library.

[0180] Accordingly, after determining the traffic flow discrete state of the target lane, the first feature to be identified corresponding to the target lane can be recorded in the traffic flow discrete state sample library, and the corresponding traffic flow discrete state of the target lane can be marked for subsequent scene type identification processes at other intersections. In this way, the traffic flow discrete state sample library can be continuously updated, and the accuracy of determining the traffic flow discrete state of the lane can be improved.

[0181] It can be understood that, compared with the above-mentioned method of determining the discrete state of traffic flow of the target road section based on the neural network model, the method of determining the discrete state of traffic flow of the lane provided in the embodiment of the present application does not require pre-labeling of a large amount of sample data, which can reduce the cost of manual labeling.

[0182] In addition, compared with the method of determining the discrete state of traffic flow in a lane through a threshold, the method of determining the discrete state of traffic flow in a lane provided in the embodiment of the present application can combine macro data and micro data to jointly determine the discrete state of traffic flow in the lane. In this way, the accuracy of the determined discrete state of traffic flow in the lane can be improved, and then, the accuracy of the scene type of the target intersection can be improved.

[0183] In some embodiments, step S602 includes:

[0184] Step 1: Based on the first sample vehicle passing data, determine the headway time of each vehicle at a specified position in the first sample lane during the first sample green light period as the sample headway time.

[0185] Step 2: Determine the preset headway interval to which each sample headway belongs.

[0186] The preset headway intervals are adjacent to each other, and the lengths of the preset headway intervals are the same.

[0187] Step 3: For each preset headway interval, calculate the ratio of the sample headway that belongs to the preset headway interval to all sample headway, and obtain the first sample feature.

[0188] In the embodiment of the present application, the process of obtaining the first sample feature based on the first sample vehicle passing data is similar to the process of obtaining the first feature to be identified based on the target vehicle passing data in the above steps S201-S203, and will not be repeated here.

[0189] In this way, the method of determining the first sample feature of the preset traffic flow discrete state is ensured to be consistent with the method of determining the first feature to be identified. When the similarity between the first sample feature of the preset traffic flow discrete state and the first feature to be identified is subsequently calculated, the similarity between the target road section and the first sample road section in each preset traffic flow discrete state can be accurately reflected, thereby ensuring the accuracy of recognition.

[0190] In some embodiments, see Figure 7 , Figure 7 A flowchart of determining the traffic demand state of the target lane based on the second feature to be identified is provided in an embodiment of the present application. Figure 7 middle:

[0191] S701: For each preset traffic demand state, obtain second sample vehicle passing data of the second sample lane during a second sample green light period when the second sample lane is in the preset traffic demand state.

[0192] S702: Based on the second sample vehicle passing data, obtain a second sample feature representing a change in the number of vehicles released by the second sample lane in each statistical time period during the second sample green light period.

[0193] The step of determining the traffic demand state of the target lane based on the second feature to be identified includes:

[0194] S703: For each preset traffic demand state, calculate the similarity between the second sample feature of the preset traffic demand state and the second feature to be identified.

[0195] S704: Determine the preset traffic demand state with the greatest corresponding similarity as the traffic demand state of the target lane.

[0196] In the embodiment of the present application, the technician can determine, based on actual experience, that the second sample lane is in a preset traffic demand state during the second sample time period, and then obtain the vehicle passing data of the second sample lane during the second sample green light period (i.e., the second sample vehicle passing data) as the second sample feature of the preset traffic demand state. The process of determining the second sample green light period in the second sample time period can refer to the process of determining the target green light period in the time period to be controlled, which will not be repeated here. The length of the second sample time period is consistent with the length of the time period to be controlled.

[0197] Furthermore, based on the second sample vehicle passing data, a second sample feature representing the change in the number of vehicles released by the second sample lane in each statistical time period during the second sample green light period can be obtained. The specific process of obtaining the second sample feature will be described in the subsequent embodiments.

[0198] For each preset traffic demand state, the similarity between the second sample feature of the preset traffic demand state and the second feature to be identified can be calculated based on a preset algorithm for calculating the similarity between two sequences. The preset algorithm can be a dynamic time warping (DTW) algorithm.

[0199] In one implementation, if the preset traffic demand state has multiple second sample features, the similarity between each second sample feature of the preset traffic demand state and the second feature to be identified can be calculated, and then the average of the multiple similarities can be calculated, and the traffic demand state of the target lane can be determined based on the average. In this way, the accuracy of the determined traffic flow discrete state of the target lane can be further improved.

[0200] The greater the similarity between the second feature to be identified and the second sample feature of the preset traffic demand state, the closer the traffic demand state of the target lane is to the preset traffic demand state. Therefore, the preset traffic demand state with the greatest corresponding similarity can be determined as the traffic demand state of the target lane.

[0201] Based on the above processing, the traffic demand state of the target lane can be determined by calculating the similarity between the second feature to be identified and the second sample feature of each preset traffic demand state, thereby ensuring the accuracy of the determined traffic demand state of the target lane.

[0202] In some embodiments, the second feature to be identified can be input into a pre-trained neural network model to obtain the traffic demand state of the target lane. The neural network model is trained based on the second sample feature and the corresponding sample label, and the sample label can represent the preset traffic demand state corresponding to the second sample feature. In this way, the traffic demand state of the target lane is determined by processing the second feature to be identified based on the neural network model, which can improve the efficiency of intersection scene recognition.

[0203] In addition, in the related art, when determining the traffic demand state of a lane, the traffic demand state of the lane can be determined by judging whether the length of the queue of vehicles in the lane is greater than a preset threshold. However, due to the different actual conditions of different roads, the preset threshold may also be different, and it is difficult to set the threshold.

[0204] In some embodiments, a traffic demand state sample library may be pre-established to store the second sample characteristics of the above-mentioned preset traffic demand states. That is, the above steps S701-S704 may be performed based on the second sample characteristics of the preset traffic demand states recorded in the traffic demand state sample library.

[0205] Accordingly, after determining the traffic demand state of the target lane, the second feature to be identified corresponding to the target lane can be recorded in the traffic demand state sample library, and the corresponding traffic demand state of the target lane can be marked for subsequent scene type identification processes at other intersections. In this way, the traffic demand state sample library can be continuously updated, and the accuracy of determining the traffic demand state of the lane can be improved.

[0206] It can be understood that, compared with the above-mentioned method of determining the traffic demand status of the target lane based on the neural network model, the method of determining the traffic demand status of the lane provided in the embodiment of the present application does not require pre-labeling of a large amount of sample data, which can reduce the cost of manual labeling.

[0207] In addition, compared with the method of determining the traffic demand state of a lane through a threshold, the method of determining the traffic demand state of a lane provided in the embodiment of the present application can combine macro data and micro data to jointly determine the traffic demand state of the lane. In this way, the accuracy of the determined traffic demand state of the lane can be improved, and then, the accuracy of the scene type of the target intersection can be improved.

[0208] In some embodiments, the second sample green light period includes multiple sample green light cycles. The above step S702 includes:

[0209] Step (1): For each sample green light cycle, based on the second sample vehicle passing data of the sample green light cycle, determine the number of vehicles released by the second sample lane in each statistical time period of the sample green light cycle as the number of vehicles released in each statistical time period of the sample green light cycle.

[0210] Step (2): For any statistical time period, calculate the mean of the number of vehicles released in each sample green light cycle in the statistical time period, and obtain a second sample feature representing the change in the number of vehicles released in the second sample lane in each statistical time period during the second sample green light period.

[0211] In the embodiment of the present application, the process of obtaining the second sample feature based on the second sample vehicle passing data is similar to the process of obtaining the second feature to be identified based on the target vehicle passing data in the above steps S301-S302, and will not be repeated here.

[0212] In this way, the method of determining the second sample characteristics of the preset traffic demand state is ensured to be consistent with the method of determining the second characteristic to be identified. When the similarity between the second sample characteristics of the preset traffic demand state and the second characteristic to be identified is subsequently calculated, the similarity between the target road section and the second sample lane in each preset traffic demand state can be accurately reflected, thereby ensuring the accuracy of identification.

[0213] In some embodiments, see Figure 8 , Figure 8 A flowchart for determining the traffic operation status of the target lane based on a third feature to be identified is provided in an embodiment of the present application. Figure 8 middle:

[0214] S801: For each preset traffic operation state, obtain third sample vehicle passing data of the third sample lane during a third sample green light period when the third sample lane is in the preset traffic operation state.

[0215] S802: Based on the third sample vehicle passing data, obtain a third sample feature representing the probability distribution of the vehicle driving speed in the third sample lane during the third sample green light period.

[0216] S803: For each preset traffic operation state, calculate the similarity between the third sample feature of the preset traffic operation state and the third feature to be identified by using a second similarity algorithm.

[0217] The second similarity algorithm includes at least one of the following: a bulldozer distance algorithm and a Mahalanobis distance algorithm.

[0218] S804: Determine the corresponding preset traffic operation state with the greatest similarity as the traffic operation state of the target lane.

[0219] In the embodiment of the present application, the technician can determine, based on actual experience, that the third sample lane is in a certain preset traffic operation state within the third sample time period, and then obtain the vehicle passing data of the third sample lane during the third sample green light period (i.e., the third sample vehicle passing data) as the third sample feature of the preset traffic operation state. The process of determining the third sample green light period in the third sample time period can refer to the above-mentioned process of determining the target green light period in the time period to be controlled, and will not be repeated here. The length of the third sample time period is consistent with the length of the time period to be controlled.

[0220] For each preset traffic operation state, the similarity between the third sample feature of the preset traffic operation state and the third feature to be identified can be obtained by a second similarity algorithm. The second similarity algorithm includes at least one of the following: a bulldozer distance algorithm, a Mahalanobis distance algorithm, and other algorithms for calculating the similarity between two probability distributions.

[0221] When the second similarity algorithm only includes one of the above algorithms for calculating the similarity between two probability distributions, the similarity between the third sample feature of the preset traffic operation state and the third feature to be identified can be calculated by the second similarity algorithm.

[0222] When the second similarity algorithm includes multiple algorithms for calculating the similarity between two probability distributions, for each second similarity algorithm, the second similarity between the third sample feature of the preset traffic operation state and the third feature to be identified can be calculated by the second similarity algorithm, and then the similarity between the third sample feature of the preset traffic operation state and the third feature to be identified can be obtained by combining the calculated multiple second similarities. For example, the weighted sum of the calculated multiple second similarities can be calculated to obtain the similarity between the third sample feature of the preset traffic operation state and the third feature to be identified.

[0223] In one implementation, if the preset traffic operation state has multiple third sample features, the similarity between each third sample feature of the preset traffic operation state and the third feature to be identified can be calculated, and based on the average of the multiple similarities, the similarity between the third feature to be identified and the third sample feature of the preset traffic operation state is obtained. In this way, the accuracy of the determined traffic operation state of the target lane can be further improved.

[0224] The greater the similarity between the third feature to be identified and the third sample feature of the preset traffic operation state, the closer the traffic operation state of the target lane is to the preset traffic operation state. Therefore, the preset traffic operation state with the greatest corresponding similarity can be determined as the traffic operation state of the target lane.

[0225] Based on the above processing, the traffic operation state of the target lane can be determined by calculating the similarity between the third feature to be identified and the third sample feature of each preset traffic operation state, thereby ensuring the accuracy of the determined traffic operation state of the target lane.

[0226] In some embodiments, the third feature to be identified can be input into a pre-trained neural network model to obtain the traffic operation state of the target lane. The neural network model is trained based on the third sample feature and the corresponding sample label, and the sample label can represent the preset traffic operation state corresponding to the third sample feature. In this way, the traffic operation state of the target lane is determined by processing the third feature to be identified based on the neural network model, which can improve the efficiency of intersection scene recognition.

[0227] In addition, in the related art, when determining the traffic operation state of a lane, the traffic operation state of the lane can be determined by judging whether the driving speed of the vehicle in the lane is greater than a preset threshold. However, due to the different actual conditions of different roads, the preset threshold may also be different, and it is difficult to set the threshold.

[0228] In some embodiments, a traffic operation state sample library may be pre-established to store the third sample features of the above-mentioned preset traffic operation states. That is, the above steps S801-S804 may be performed based on the third sample features of the preset traffic operation states recorded in the traffic operation state sample library.

[0229] Accordingly, after determining the traffic operation state of the target lane, the third feature to be identified corresponding to the target lane can be recorded in the traffic operation state sample library, and the corresponding traffic operation state of the target lane can be marked for subsequent scene type identification processes at other intersections. In this way, the traffic operation state sample library can be continuously updated, and the accuracy of determining the traffic operation state of the lane can be improved.

[0230] It can be understood that, compared with the above-mentioned method of determining the traffic operation status of the target lane based on the neural network model, the method of determining the traffic operation status of the lane provided in the embodiment of the present application does not require pre-labeling of a large amount of sample data, which can reduce the cost of manual labeling.

[0231] In addition, compared with the method of determining the traffic operation status of a lane through a threshold, the method of determining the traffic operation status of a lane provided in the embodiment of the present application can combine macro data and micro data to jointly determine the traffic operation status of the lane. In this way, the accuracy of the determined traffic operation status of the lane can be improved, and then, the accuracy of the scene type of the target intersection can be improved.

[0232] In some embodiments, step S802 includes:

[0233] Step 1: Based on the third sample vehicle passing data, determine the driving speed of each vehicle in the third sample lane during the third sample green light period as the sample driving speed.

[0234] Step 2: Determine the preset driving speed range to which each sample driving speed belongs.

[0235] The preset driving speed intervals are adjacent to each other, and the lengths of the preset driving speed intervals are the same.

[0236] Step 3: For each preset driving speed interval, calculate the ratio of the sample driving speeds belonging to the preset driving speed interval to all the sample driving speeds to obtain the third sample feature.

[0237] In the embodiment of the present application, the process of obtaining the third sample feature based on the third sample vehicle passing data is similar to the process of obtaining the third feature to be identified based on the target vehicle passing data in the above steps S401-S403, and will not be repeated here.

[0238] In this way, the method of determining the third sample feature of the preset traffic operation state is consistent with the method of determining the third feature to be identified. When the similarity between the third sample feature of the preset traffic operation state and the third feature to be identified is subsequently calculated, the similarity between the target road section and the third sample road section in each preset traffic operation state can be accurately reflected, thereby ensuring the accuracy of recognition.

[0239] In some embodiments, the traffic flow discrete state is: discrete, weak following, or strong following; the traffic demand state is: low traffic demand or high traffic demand; the traffic operation state is: slow or smooth. Fig. 9 As shown, Fig. 9 A second flow chart of the intersection scene recognition method provided in the embodiment of the present application, Figure 1 On the basis of this, step S104 comprises:

[0240] S1041: If at least one target lane in the target lanes in each direction of the target intersection satisfies the first screening condition, it is determined that the scene type of the target intersection in the time period to be controlled is a queuing balance type.

[0241] Among them, the first screening condition indicates that: the traffic flow discrete state is strong following, the traffic demand state is high traffic demand, and the traffic operation state is slow movement.

[0242] S1042: If there is no target lane that meets the first filtering condition among the target lanes in each direction of the target intersection, and there is at least one target lane that meets the second filtering condition, it is determined that the scene type of the target intersection in the time period to be controlled is a capacity type.

[0243] Among them, the second screening condition indicates that: the traffic flow discrete state is strong following or weak following, the traffic demand state is high traffic demand, and the traffic operation state is smooth.

[0244] S1043: If there is no target lane that meets the first screening condition and no target lane that meets the second screening condition among the target lanes in each direction of the target intersection, it is determined that the scene type of the target intersection in the time period to be controlled is an efficiency improvement type.

[0245] Among them, each traffic flow discrete state is predetermined according to data rules and traffic knowledge models, wherein the traffic knowledge model is obtained by technicians based on experience. As shown in FIG10(a), FIG10(a) is a schematic diagram of lanes with different traffic flow discrete states provided in an embodiment of the present application, in which a gray rectangular block represents a vehicle on the lane.

[0246] For any lane, if the traffic flow discrete state of the lane is discrete in a time period, it means that most vehicles in the traffic flow are not restricted by the front vehicle in this time period, and the headway distribution of vehicles leaving the stop line is discrete, such as lane 1 in Figure 10(a).

[0247] For any lane, if the traffic flow discrete state of the lane is weak following state in a time period, it means that part of the traffic flow is restricted by the front vehicle, and part is not restricted by the front vehicle, and a certain headway is maintained. This is shown in lane 2 in Figure 10(a).

[0248] For any lane, if the traffic flow discrete state of the lane is a strong follow-up state within a time period, it means that within this time period, most vehicles in the traffic flow are constrained by the front vehicle, and the headway time between vehicles leaving the stop line is relatively concentrated and small, such as lane 3 in Figure 10(a).

[0249] Each traffic demand state is predetermined according to data rules and traffic knowledge models. As shown in Figure 10 (b1) and Figure 10 (b2), Figure 10 (b1) is a schematic diagram of the change in the number of vehicles released from a lane with low traffic demand provided in an embodiment of the present application. In Figure 10 (b1), the X-axis represents the duration, T represents a green light cycle, and the Y-axis represents the number of vehicles released from the lane in each statistical time period. Since the traffic demand state of the lane is low traffic demand, that is, the number of vehicles required to be released by the lane is small, and the vehicles can be fully released in the first multiple statistical time periods. Therefore, there is an empty release situation in the green light cycle. Correspondingly, it is manifested as: according to the order of the statistical time periods, in each statistical time period, the number of released vehicles first increases, then tends to be stable, and finally decreases.

[0250] Figure 10 (b2) is a schematic diagram of the change in the number of vehicles released in a lane with high traffic demand provided by an embodiment of the present application. Since the traffic demand state of the lane is high traffic demand, that is, the number of vehicles required to be released in the lane is large, and the vehicles cannot be fully released in one green light cycle. Therefore, there is no empty release in the green light cycle. Correspondingly, it is manifested as follows: according to the sequence of the statistical time periods, in each statistical time period, the number of released vehicles first increases, then tends to be stable, until the end of the green light cycle is reached.

[0251] Each traffic operation state is predetermined according to data rules and traffic knowledge models. As shown in FIG10( c ), FIG10( c ) is a schematic diagram of a road with different traffic operation states provided by an embodiment of the present application, in which a gray rectangular block represents a vehicle on a lane.

[0252] For any lane, within a time period, if the traffic operation state of the lane is slow, it means that the speed of most vehicles in the traffic flow is distributed in a lower speed space, and the driving speed is relatively slow. Generally, there is a congested lane or there is a large interference in the lane, such as serious interference from pedestrians and motor vehicles. For example, lane 1 in Figure 10(c).

[0253] For any lane, within a time period, if the traffic status of the lane is unblocked, it means that the speeds of most vehicles in the traffic flow are distributed in a higher speed space and the driving speed is faster. This usually occurs in lanes with smaller traffic flow or lanes with larger traffic flow but less interference on the road, such as lane 2 in Figure 10(c).

[0254] In an embodiment of the present application, when there is at least one target lane that satisfies the first screening condition in the target lanes in each direction, the scene type of the target intersection in the time period to be controlled is preferentially determined to be a queuing balance class. When the scene type of the target intersection in the time period to be controlled is not a queuing balance class (i.e., there is no target lane that satisfies the first screening condition), and there is at least one target lane that satisfies the second screening condition, it is determined whether the scene type of the target intersection in the time period to be controlled is a capacity class. And when the scene type of the target intersection in the time period to be controlled is not a queuing balance class, and is not a capacity class (i.e., there is neither a target lane that satisfies the first screening condition nor a target lane that satisfies the second screening condition), it is determined that the scene type of the target intersection in the time period to be controlled is an efficiency improvement class.

[0255] Based on the above processing, when determining the scene type of the target intersection, for the target lanes in each direction, when there is a target lane that meets the first screening condition, the scene type of the target intersection in the time period to be controlled can be determined as a queue balancing type. When the scene type of the target intersection is determined according to the traffic status of each target lane and the control strategy of the traffic signal is determined, it can be ensured that the needs of the sections in each direction of the target intersection are met.

[0256] In some embodiments, the intersection scene recognition method provided by the present application further includes:

[0257] If the scene type of the target intersection during the time period to be controlled is a queue balance type, the control strategy of the first type of traffic signal is adopted at the target intersection.

[0258] Among them, the control strategy of the first type of traffic signal is used to shorten the queue length of vehicles at the target intersection.

[0259] If the scenario type of the target intersection during the time period to be controlled is the traffic capacity type, the control strategy of the second type of traffic signal is adopted at the target intersection.

[0260] Among them, the control strategy of the second type of traffic signal is used to improve the vehicle capacity of the target intersection.

[0261] If the scenario type of the target intersection during the time period to be controlled is efficiency improvement, the control strategy of the third type of traffic signal is adopted at the target intersection.

[0262] Among them, the control strategy of the third type of traffic signal is used to reduce the delay time of vehicles at the target intersection.

[0263] In the embodiment of the present application, if the scene type of the target intersection in the time period to be controlled is a queue balance type, that is, in each target section of the target intersection, at least one target lane has a traffic flow discrete state of strong following, a traffic demand state of high traffic demand, and a traffic operation state of slow movement, in order to avoid too many vehicles queuing in the at least one target section, it can be determined to adopt the control strategy of the first type of traffic signal at the target intersection. The control strategy of the first type of traffic signal is used to shorten the queue length of vehicles at the target intersection.

[0264] If the scenario type of the target intersection in the time period to be controlled is the capacity type, that is, there is at least one target lane with a high traffic demand and a smooth traffic operation state, and there is no target road section with a strong following traffic flow discrete state, a high traffic demand, and a slow traffic operation state, then in order to avoid congestion in each target road section of the target intersection, it can be determined to adopt the second type of traffic signal control strategy at the target intersection. Among them, the second type of traffic signal control strategy is used to improve the vehicle capacity of the target intersection.

[0265] If the scene type of the target intersection in the time period to be controlled is efficiency improvement, it means that the traffic status of each target lane in the target intersection is good. At this time, in order to reduce the number of vehicle stops or reduce the delay time of the vehicle, it can be determined to adopt the third type of traffic signal control strategy at the target intersection. Among them, the third type of traffic signal control strategy is used to reduce the delay time of vehicles at the target intersection.

[0266] In this way, it is possible to adopt a suitable traffic signal control strategy at the target intersection according to the scenario type of the target intersection within the time period to be controlled, and recommend the optimal control strategy for the target intersection.

[0267] Fig.11 A schematic diagram of the structure of a control strategy determination device provided in an embodiment of the present application. Fig.11 In the process, the data of the intersection to be controlled (i.e., the target intersection) can be input into the feature extraction and mining module, wherein the data of the target intersection includes: the vehicle passing data and the signal timing data of the target intersection. The feature extraction and mining module can obtain the headway probability distribution (i.e., the first feature to be identified that represents the probability distribution of the headway in the target lane during the target green light period), the periodic traffic flow rate (i.e., the second feature to be identified that represents the change in the number of vehicles released by the target lane in each statistical time period during the target green light period), and the vehicle speed probability distribution (i.e., the third feature to be identified that represents the probability distribution of the vehicle speed in the target lane during the target green light period) according to the above step S102.

[0268] The feature extraction and mining module can input the obtained first feature to be identified, the second feature to be identified, and the third feature to be identified into the scene recognition module, and the scene recognition module can obtain the first sample feature of each preset traffic flow discrete state, the second sample feature of each preset traffic demand state, and the third sample feature of each preset traffic operation state stored in the scene library construction module. Among them, each preset traffic flow discrete state includes: strong following, weak following, and discrete; each preset traffic demand state includes: high traffic demand and low traffic demand; each preset traffic operation state includes: smooth and slow.

[0269] Correspondingly, the scene recognition module can measure the similarity between the first feature to be identified and the first sample feature of each preset traffic flow discrete state, obtain the similarity between the first feature to be identified and the first sample feature of each preset traffic flow discrete state, classify the obtained similarity based on the preset classification method, and determine the traffic flow discrete state corresponding to the first feature to be identified. Among them, the preset classification method can be: the nearest neighbor class mean classification method, or other classification methods based on distance measurement (such as center clustering method). Similarly, the second feature to be identified and the second sample feature of each preset traffic demand state are measured for similarity, and the similarity between the second feature to be identified and the second sample feature of each preset traffic demand state is obtained. The obtained similarity is classified based on the preset classification method to determine the traffic demand state corresponding to the second feature to be identified. The third feature to be identified and the third sample feature of each preset traffic operation state are measured for similarity, and the similarity between the third feature to be identified and the third sample feature of each preset traffic operation state is obtained. The obtained similarity is classified based on the preset classification method to determine the traffic operation state corresponding to the third feature to be identified. In this way, the traffic status of the target lanes in different directions of the target intersection can be obtained, and then the scene recognition result can be obtained (i.e., the scene type of the target intersection in the time period to be controlled is determined according to step S104).

[0270] In addition, after determining the scene type of the target intersection within the time period to be controlled, the scene library (ie, the sample library in the above embodiment) may be updated using the first feature to be identified, the second feature to be identified, and the third feature to be identified.

[0271] The intersection scene output module is used to output the scene type of the determined target intersection within the time period to be controlled. Subsequently, the control strategy of the traffic signal to be adopted can be determined according to the scene type of the target intersection within the time period to be controlled.

[0272] In the technical solution of the present application, operations such as acquisition, storage, use, processing, transmission, provision and disclosure of vehicle passing data are all carried out with the user's authorization.

[0273] Based on the same inventive concept, the embodiment of the present application provides an intersection scene recognition device. Fig.12 , Fig.12 A structural diagram of an intersection scene recognition device provided in an embodiment of the present application, the device comprising:

[0274] The target vehicle passing data acquisition module 1201 is used to acquire the target vehicle passing data of the target lane during the target green light period within the time period to be controlled for the target lane in each direction of the target intersection;

[0275] The to-be-recognized feature acquisition module 1202 is used to obtain, based on the target vehicle passing data, a first to-be-recognized feature representing the probability distribution of the headway time in the target lane during the target green light period, a second to-be-recognized feature representing the change in the number of vehicles released by the target lane in each statistical time period during the target green light period, and a third to-be-recognized feature representing the probability distribution of the vehicle speed in the target lane during the target green light period;

[0276] A state determination module 1203, configured to determine a traffic flow discrete state of the target lane based on the first feature to be identified, determine a traffic demand state of the target lane based on the second feature to be identified, and determine a traffic operation state of the target lane based on the third feature to be identified;

[0277] The scene determination module 1204 is used to determine the scene type of the target intersection within the time period to be controlled according to the traffic flow discrete state, traffic demand state, and traffic operation state of the target lanes in each direction.

[0278] In some embodiments, during the time period to be controlled, the target lane has the largest number of released vehicles among the lanes of the road section in any direction of the target intersection.

[0279] In some embodiments, the apparatus further comprises:

[0280] A first sample vehicle passing data acquisition module is used to acquire, for each preset traffic flow discrete state, first sample vehicle passing data of the first sample lane during a first sample green light period when the first sample lane is in the preset traffic flow discrete state before determining the traffic flow discrete state of the target lane based on the first feature to be identified;

[0281] A first sample feature acquisition module, configured to obtain, based on the first sample vehicle passing data, a first sample feature representing a probability distribution of a headway time of a vehicle in the first sample lane during the first sample green light period;

[0282] The state determination module 1203 is specifically used to: for each preset traffic flow discrete state, calculate the similarity between the first sample feature of the preset traffic flow discrete state and the first feature to be identified through a first similarity algorithm; wherein the first similarity algorithm includes at least one of the following: a bulldozer distance algorithm and a Mahalanobis distance algorithm; and determine the preset traffic flow discrete state with the largest corresponding similarity as the traffic flow discrete state of the target lane.

[0283] In some embodiments, the sample feature acquisition module is specifically used to: determine, based on the first sample vehicle passing data, each headway at a designated position in the first sample lane during the first sample green light period as a sample headway; determine a preset headway interval to which each sample headway belongs; wherein each preset headway interval is adjacent and has the same length; for each preset headway interval, calculate the ratio of the sample headway belonging to the preset headway interval to all sample headway intervals to obtain a first sample feature;

[0284] The module 1202 for acquiring features to be identified is specifically used to: determine, based on the target vehicle passing data, each headway at a designated position in the target lane during the target green light period as the headway to be processed; determine the preset headway interval to which each headway to be processed belongs; and for each preset headway interval, calculate the ratio of the headway to be processed that belongs to the preset headway interval to all headway to be processed, to obtain the first feature to be identified.

[0285] In some embodiments, the apparatus further comprises:

[0286] A second sample vehicle passing data acquisition module is used to acquire, for each preset traffic demand state, second sample vehicle passing data of the second sample lane during a second sample green light period when the second sample lane is in the preset traffic demand state before determining the traffic demand state of the target lane based on the second feature to be identified;

[0287] A second sample feature acquisition module, configured to obtain, based on the second sample vehicle passing data, a second sample feature representing a change in the number of vehicles released by the second sample lane in each statistical time period during the second sample green light period;

[0288] The state determination module 1203 is specifically used to: for each preset traffic demand state, calculate the similarity between the second sample feature of the preset traffic demand state and the second feature to be identified; and determine the preset traffic demand state with the largest similarity as the traffic demand state of the target lane.

[0289] In some embodiments, the second sample green light period includes multiple sample green light cycles; the target green light period includes multiple target green light cycles; the second sample feature acquisition module is specifically used to: for each sample green light cycle, based on the second sample vehicle passing data of the sample green light cycle, determine the number of vehicles released by the second sample lane in each statistical time period of the sample green light cycle as the number of vehicles released in each statistical time period of the sample green light cycle; for any statistical time period, calculate the mean of the number of vehicles released in each statistical time period of each sample green light cycle, and obtain the number of vehicles released in each statistical time period of the second sample green light period. The second sample feature of the change in the number of vehicles released by the second sample lane; the to-be-identified feature acquisition module 1202 is specifically used to: for each target green light cycle, based on the target vehicle passing data of the target green light cycle, determine the number of vehicles released by the target lane in each statistical time period of the target green light cycle as the number of vehicles released in each statistical time period of the target green light cycle; for any statistical time period, calculate the mean of the number of vehicles released in the statistical time period of each target green light cycle, and obtain the second to-be-identified feature representing the change in the number of vehicles released by the target lane in each statistical time period during the target green light period.

[0290] In some embodiments, the device also includes: a third sample vehicle passing data acquisition module, which is used to obtain, for each preset traffic operation state, the third sample vehicle passing data of the third sample lane during the third sample green light period when the third sample lane is in the preset traffic operation state before determining the traffic operation state of the target lane based on the third feature to be identified; a third sample feature acquisition module, which is used to obtain, based on the third sample vehicle passing data, a third sample feature representing the probability distribution of the vehicle driving speed in the third sample lane during the third sample green light period; the state determination module 1203 is specifically used to: for each preset traffic operation state, calculate the similarity between the third sample feature of the preset traffic operation state and the third feature to be identified through a second similarity algorithm; wherein the second similarity algorithm includes at least one of the following: a bulldozer distance algorithm and a Mahalanobis distance algorithm; and determine the preset traffic operation state with the largest corresponding similarity as the traffic operation state of the target lane.

[0291] In some embodiments, the third sample feature acquisition module is specifically used to: determine the driving speed of each vehicle in the third sample lane during the third sample green light period based on the third sample passing data as the sample driving speed; determine the preset driving speed interval to which each sample driving speed belongs; wherein each preset driving speed interval is adjacent and has the same length; for each preset driving speed interval, calculate the ratio of the sample driving speed belonging to the preset driving speed interval to all sample driving speeds to obtain the third sample feature;

[0292] The module 1202 for acquiring features to be identified is specifically used to: determine the driving speed of each vehicle in the target lane during the target green light period based on the target vehicle passing data as the driving speed to be processed; determine the preset driving speed interval to which each driving speed to be processed belongs; for each preset driving speed interval, calculate the ratio of the driving speed to be processed belonging to the preset driving speed interval to all driving speeds to be processed, and obtain the third feature to be identified.

[0293] In some embodiments, the discrete state of the traffic flow is: discrete, weak following, or strong following; the traffic demand state is: low traffic demand or high traffic demand; the traffic operation state is: slow or unblocked; the scene determination module 1204 is specifically used to: if there is at least one target lane in the target lanes in each direction of the target intersection that meets the first screening condition, then determine that the scene type of the target intersection in the time period to be controlled is a queuing balance type; wherein the first screening condition indicates that the discrete state of the traffic flow is strong following, the traffic demand state is high traffic demand, and the traffic operation state is slow; if the target lanes in each direction of the target intersection If there is no target lane that meets the first filtering condition among the marked lanes, and there is at least one target lane that meets the second filtering condition, then the scene type of the target intersection during the time period to be controlled is determined to be the capacity type; wherein the second filtering condition represents: the traffic flow discrete state is strong following or weak following, the traffic demand state is high traffic demand, and the traffic operation state is smooth; if there is no target lane that meets the first filtering condition among the target lanes in all directions of the target intersection, and there is no target lane that meets the second filtering condition, then the scene type of the target intersection during the time period to be controlled is determined to be the efficiency improvement type.

[0294] In some embodiments, the apparatus further comprises:

[0295] The first control strategy determination module is used to adopt the control strategy of the first type of traffic signal at the target intersection if the scenario type of the target intersection during the time period to be controlled is a queue balancing type; wherein the control strategy of the first type of traffic signal is used to shorten the queue length of vehicles at the target intersection; the second control strategy determination module is used to adopt the control strategy of the second type of traffic signal at the target intersection if the scenario type of the target intersection during the time period to be controlled is a capacity type; wherein the control strategy of the second type of traffic signal is used to improve the vehicle capacity of the target intersection; the third control strategy determination module is used to adopt the control strategy of the third type of traffic signal at the target intersection if the scenario type of the target intersection during the time period to be controlled is an efficiency improvement type; wherein the control strategy of the third type of traffic signal is used to reduce the delay time of vehicles at the target intersection.

[0296] The present application also provides an electronic device, such as Fig.13 As shown, including:

[0297] Memory 1301, used for storing computer programs;

[0298] The processor 1302 is used to implement the steps of any intersection scene recognition method in the above-mentioned embodiments when executing the program stored in the memory 1301.

[0299] Furthermore, the electronic device may further include a communication bus and / or a communication interface, and the processor 1302, the communication interface, and the memory 1301 communicate with each other via the communication bus.

[0300] The communication bus mentioned in the above electronic device can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The communication bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus.

[0301] The communication interface is used for communication between the above electronic device and other devices.

[0302] The memory may include a random access memory (RAM) or a non-volatile memory (NVM), such as at least one disk memory. Optionally, the memory may also be at least one storage device located away from the aforementioned processor.

[0303] The above-mentioned processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0304] In another embodiment provided in the present application, a computer-readable storage medium is provided, in which a computer program is stored. When the computer program is executed by a processor, the steps of any of the above-mentioned intersection scene recognition methods are implemented.

[0305] In another embodiment provided in the present application, a computer program product including instructions is also provided, which, when executed on a computer, enables the computer to execute any intersection scene recognition method in the above embodiments.

[0306] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions may be transmitted from a website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server or data center. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more available media integration. The available medium may be a magnetic medium, (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a solid-state hard disk (SSD), etc.

[0307] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0308] Each embodiment in this specification is described in a related manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device, electronic device, and computer-readable storage medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiments.

[0309] The above description is only a preferred embodiment of the present application and is not intended to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application are included in the protection scope of the present application.

Claims

1. A method for identifying intersection scenes, characterized in that: The method comprises: For the target lane in each direction of the target intersection, the target vehicle passing data of the target lane during the target green light period within the time period to be controlled is obtained; Based on the target vehicle passing data, a first feature to be identified that represents the probability distribution of the headway time in the target lane during the target green light period, a second feature to be identified that represents the change in the number of vehicles released by the target lane in each statistical time period during the target green light period, and a third feature to be identified that represents the probability distribution of the vehicle driving speed in the target lane during the target green light period are obtained; Determine the traffic flow discrete state of the target lane based on the first feature to be identified, determine the traffic demand state of the target lane based on the second feature to be identified, and determine the traffic operation state of the target lane based on the third feature to be identified; According to the traffic flow discrete state, traffic demand state, and traffic operation state of the target lanes in each direction, the scene type of the target intersection within the time period to be controlled is determined.

2. The method according to claim 1, characterized in that During the time period to be controlled, the target lane has the largest number of released vehicles among the lanes of the road section in any direction of the target intersection.

3. The method according to claim 1, characterized in that Before determining the traffic flow discrete state of the target lane based on the first feature to be identified, the method further includes: For each preset traffic flow discrete state, obtaining first sample vehicle passing data of the first sample lane during a first sample green light period when the first sample lane is in the preset traffic flow discrete state; Based on the first sample vehicle passing data, a first sample feature representing a probability distribution of a headway time of a vehicle in the first sample lane during the first sample green light period is obtained; The determining the traffic flow discrete state of the target lane based on the first feature to be identified includes: For each preset traffic flow discrete state, the similarity between the first sample feature of the preset traffic flow discrete state and the first feature to be identified is calculated by using a first similarity algorithm; wherein the first similarity algorithm includes at least one of the following: a bulldozer distance algorithm and a Mahalanobis distance algorithm; The preset traffic flow discrete state with the largest corresponding similarity is determined as the traffic flow discrete state of the target lane.

4. The method according to claim 3, characterized in that The obtaining, based on the first sample vehicle passing data, a first sample feature representing a probability distribution of a headway time in the first sample lane during the first sample green light period includes: Based on the first sample vehicle passing data, determining the headway time of each vehicle at a designated position in the first sample lane during the first sample green light period as the sample headway time; Determine the preset headway interval to which each sample headway belongs; wherein the preset headway intervals are adjacent and have the same length; For each preset headway interval, calculate the ratio of the sample headway of the preset headway interval to all sample headway, and obtain a first sample feature; The obtaining, based on the target vehicle passing data, a first feature to be identified that represents a probability distribution of the headway time of the target vehicle in the target lane during the target green light period includes: Based on the target vehicle passing data, determining each headway of each vehicle at a designated position in the target lane during the target green light period as the headway of each vehicle to be processed; Determine the preset headway interval to which each headway to be processed belongs; For each preset headway interval, the ratio of the headway to be processed that belongs to the preset headway interval to all the headway to be processed is calculated to obtain the first feature to be identified.

5. The method according to claim 1, characterized in that Before determining the traffic demand state of the target lane based on the second feature to be identified, the method further includes: For each preset traffic demand state, obtaining second sample vehicle passing data of the second sample lane during a second sample green light period when the second sample lane is in the preset traffic demand state; Based on the second sample vehicle passing data, a second sample feature is obtained which indicates a change in the number of vehicles released by the second sample lane in each statistical time period during the second sample green light period; The determining the traffic demand state of the target lane based on the second feature to be identified includes: For each preset traffic demand state, calculating the similarity between the second sample feature of the preset traffic demand state and the second feature to be identified; The preset traffic demand state with the greatest corresponding similarity is determined as the traffic demand state of the target lane.

6. The method according to claim 5, characterized in that The second sample green light period includes a plurality of sample green light cycles; the target green light period includes a plurality of target green light cycles; The second sample feature indicating the change in the number of vehicles released by the second sample lane in each statistical time period during the second sample green light period is obtained based on the second sample vehicle passing data, including: For each sample green light cycle, based on the second sample vehicle passing data of the sample green light cycle, determine the number of vehicles released by the second sample lane in each statistical time period of the sample green light cycle as the number of vehicles released in each statistical time period of the sample green light cycle; For any statistical time period, the mean value of the number of released vehicles in each sample green light cycle in the statistical time period is calculated to obtain a second sample feature representing the change in the number of released vehicles in the second sample lane in each statistical time period during the second sample green light period; The second feature to be identified, which indicates the change in the number of vehicles released by the target lane in each statistical time period during the target green light period, is obtained based on the target vehicle passing data, including: For each target green light cycle, based on the target vehicle passing data of the target green light cycle, determine the number of vehicles released by the target lane in each statistical time period of the target green light cycle as the number of vehicles released in each statistical time period of the target green light cycle; For any statistical time period, the mean value of the number of vehicles released in each target green light cycle in the statistical time period is calculated to obtain a second feature to be identified that represents the change in the number of vehicles released in the target lane in each statistical time period during the target green light period.

7. The method according to claim 1, characterized in that Before determining the traffic running state of the target lane based on the third feature to be identified, the method further includes: For each preset traffic operation state, obtaining third sample vehicle passing data of the third sample lane during a third sample green light period when the third sample lane is in the preset traffic operation state; Based on the third sample vehicle passing data, a third sample feature representing the probability distribution of the vehicle driving speed in the third sample lane during the third sample green light period is obtained; The determining the traffic running state of the target lane based on the third feature to be identified includes: For each preset traffic operation state, the similarity between the third sample feature of the preset traffic operation state and the third feature to be identified is calculated by a second similarity algorithm; wherein the second similarity algorithm includes at least one of the following: a bulldozer distance algorithm and a Mahalanobis distance algorithm; The corresponding preset traffic operation state with the greatest similarity is determined as the traffic operation state of the target lane.

8. The method according to claim 7, characterized in that The method of obtaining, based on the third sample vehicle passing data, a third sample feature representing the probability distribution of the vehicle speed in the third sample lane during the third sample green light period, comprises: Based on the third sample vehicle passing data, determining the driving speed of each vehicle in the third sample lane during the third sample green light period as a sample driving speed; Determine a preset driving speed interval to which each sample driving speed belongs; wherein each preset driving speed interval is adjacent and has the same length; For each preset driving speed interval, calculate the ratio of the sample driving speeds belonging to the preset driving speed interval to all the sample driving speeds to obtain a third sample feature; The method of obtaining, based on the target vehicle passing data, a third feature to be identified that represents a probability distribution of the vehicle speed in the target lane during the target green light period, comprises: Based on the target vehicle passing data, determining the driving speed of each vehicle in the target lane during the target green light period as the driving speed to be processed; Determine the preset driving speed interval to which each driving speed to be processed belongs; For each preset driving speed interval, the ratio of the to-be-processed driving speeds belonging to the preset driving speed interval to all the to-be-processed driving speeds is calculated to obtain a third feature to be identified.

9. The method according to claim 1, characterized in that: The traffic flow discrete state is: discrete, weak following, or strong following; the traffic demand state is: low traffic demand or high traffic demand; the traffic operation state is: slow or smooth; Determining the scene type of the target intersection within the time period to be controlled according to the traffic flow discrete state, traffic demand state, and traffic operation state of the target lanes in each direction includes: If at least one of the target lanes in each direction of the target intersection satisfies the first screening condition, it is determined that the scene type of the target intersection in the time period to be controlled is a queuing balance type; wherein the first screening condition indicates that the traffic flow discrete state is strong following, the traffic demand state is high traffic demand, and the traffic operation state is slow moving; If there is no target lane that meets the first screening condition among the target lanes in each direction of the target intersection, and there is at least one target lane that meets the second screening condition, it is determined that the scene type of the target intersection in the time period to be controlled is a capacity type; wherein the second screening condition indicates that: the traffic flow discrete state is strong following or weak following, the traffic demand state is high traffic demand, and the traffic operation state is unblocked; If there is no target lane that meets the first screening condition and no target lane that meets the second screening condition among the target lanes in each direction of the target intersection, it is determined that the scene type of the target intersection in the time period to be controlled is efficiency improvement type.

10. The method according to claim 9, characterized in that The method further comprises: If the scene type of the target intersection in the time period to be controlled is a queue balance type, a control strategy of a first type of traffic signal is adopted at the target intersection; wherein the control strategy of the first type of traffic signal is used to shorten the queue length of vehicles at the target intersection; If the scenario type of the target intersection in the time period to be controlled is a traffic capacity type, a control strategy of a second type of traffic signal is adopted at the target intersection; wherein the control strategy of the second type of traffic signal is used to improve the vehicle traffic capacity of the target intersection; If the scenario type of the target intersection within the time period to be controlled is efficiency improvement, a control strategy for the third type of traffic signal is adopted at the target intersection; wherein the control strategy for the third type of traffic signal is used to reduce the delay duration of vehicles at the target intersection.

11. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 10 is implemented.

Images