Traffic state data processing method and device based on artificial intelligence

Abstract

The present application relates to a kind of traffic state data processing method and device based on artificial intelligence, method includes: obtaining multiple videos;Multiple videos are the video of multiple predetermined positions on target road;Based on each video, determine the actual passing time of each vehicle in each predetermined position and actual speed;Based on the actual speed and actual passing time of each vehicle in initial predetermined position, determine the predicted passing time of each vehicle in other predetermined positions except initial predetermined position in multiple predetermined positions;Based on the actual passing time, actual speed and predicted passing time of each vehicle in multiple predetermined positions, determine the index difference distance of each vehicle in each predetermined position and the predicted time interval between passing through every two adjacent predetermined positions, and whether target road is congested based on index difference distance and / or predicted time interval is determined;Exponential difference distance and predicted time interval are sent to target vehicle and traffic control server.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and in particular to a method and apparatus for processing traffic status data based on artificial intelligence. Background Technology

[0002] To achieve intelligent infrastructure, a large number of cameras have been installed in highway tunnels and municipal underground roads to detect traffic flow. However, current technologies only monitor vehicles and their speeds in traffic flow. But in the vision of vehicle-road cooperation, simply providing this information is far from enough.

[0003] Therefore, in the field of vehicle-road cooperation, a current research direction is to generate information from the roadside perspective that is of urgent need for future traffic driving systems (including drivers or autonomous vehicles) and tunnel operators, as well as to generate data that is of great significance to future traffic control systems. Summary of the Invention

[0004] This invention provides a traffic state data processing method and apparatus based on artificial intelligence to solve the problem of limited information provided by existing technologies in the direction of vehicle-road cooperation. It enables the generation of information from a roadside perspective that is urgently needed by future traffic driving systems (including drivers or autonomous vehicles) and tunnel operation managers, as well as the generation of data that is of great significance to future traffic control systems.

[0005] An artificial intelligence-based traffic state data processing method includes: acquiring multiple videos; the multiple videos are videos corresponding to multiple predetermined locations on a target road, and each video contains at least one vehicle; based on each video, determining the actual passage time and actual speed of each vehicle at each predetermined location; based on the actual speed and actual passage time of each vehicle at an initial predetermined location, determining the predicted passage time of each vehicle at other predetermined locations besides the initial predetermined location; based on the actual passage time, actual speed, and predicted passage time of each vehicle at the other predetermined locations, determining the exponential difference distance of each vehicle at each predetermined location and the predicted time interval between each vehicle passing through every two adjacent predetermined locations on the target road, and determining whether the target road is congested based on the exponential difference distance and / or the predicted time interval between each vehicle passing through every two adjacent predetermined locations on the target road; sending the exponential difference distance and the predicted time interval between each vehicle passing through every two adjacent predetermined locations on the target road to a target vehicle and a traffic control server; wherein, the target vehicle is a vehicle traveling behind the at least one vehicle.

[0006] In one embodiment, determining the exponential difference distance of each vehicle at each predetermined location and the predicted time interval between each vehicle passing through each of the target road based on the actual passage time and actual speed of each vehicle at multiple predetermined locations and the predicted passage time of each vehicle at other predetermined locations includes: determining the exponential difference distance of each vehicle at each predetermined location based on the actual passage time and actual speed of each vehicle at multiple predetermined locations and the predicted passage time of each vehicle at other predetermined locations; and determining the predicted time interval between each vehicle passing through each of the target road based on the actual passage time of each vehicle at multiple predetermined locations, the predicted passage time of each vehicle at other predetermined locations, the actual time interval between each vehicle passing through the target road, and a pre-trained first model.

[0007] In one embodiment, the number of the plurality of predetermined positions is n, and the distance between any two adjacent predetermined positions is equal. Then, the (i+1)th predetermined position D... i+1 Prediction via time T i ' +1 The corresponding formula is: T i ' +1 =T i +ΔT i ,in, T i For each vehicle at the i-th predetermined position D i The actual transit time, v i For each vehicle at the i-th predetermined position D i The actual vehicle speed, L / (n-1) represents the road distance between two adjacent predetermined locations.

[0008] In one embodiment, each vehicle at the (i+1)th predetermined position D i+1 The expression for the exponential difference distance is: Among them, M i =T i+1 -T′ i+1 =T i+1 -(T i +ΔT i ), T i+1 For each vehicle at the (i+1)th predetermined position D i+1 The actual transit time, T i ' +1 For each vehicle at the (i+1)th predetermined position D i+1 The prediction through time, v i+1 For each vehicle at the (i+1)th predetermined position D i+1 The actual speed of the vehicle.

[0009] In one embodiment, determining whether a target road is congested based on the exponential difference distance includes: summing the exponential difference distances of each vehicle at multiple predetermined locations to obtain the total exponential difference distance of the target road. If the total index difference distance of the target road is greater than or equal to a first preset threshold, the target road is determined to be in a congested state; or, if the total index difference distance of the target road is less than the first preset threshold, the target road is determined to be in a smooth state.

[0010] In one embodiment, the first model is used to determine the allocation coefficient W for the time required to traverse any two adjacent predetermined locations among a plurality of predetermined locations. i Accordingly, the training process of the first model is as follows: The actual passage time of each vehicle at multiple predetermined locations, the predicted passage time of each vehicle at the other predetermined locations, and the actual time interval of each vehicle passing through the target road are input into the first model to train the first model; wherein, the training objective of the first model is: to make the sum of the predicted passage time intervals between every two adjacent predetermined locations of each vehicle equal to the actual time interval of each vehicle passing through the target road; specifically, the training objective function expression of the first model is: min(∑ i,car_id M i ·W i -(T n-1 -T0)), where i = 0,…n-1, car_id is the identifier for each vehicle, M i ·W i This indicates that each vehicle is at the i-th predetermined position D. i The predicted transit time interval between its adjacent predetermined locations, ∑ i,car_id M i ·W i This represents the summation of predictions for each vehicle at every two adjacent predetermined locations across multiple predetermined positions, T. n-1 -T0 represents the actual time interval between each vehicle passing through the target road.

[0011] In one embodiment, after determining the predicted time interval between each vehicle passing through every two adjacent predetermined locations in the target road, the method includes: determining a target predicted time interval for passing through every two adjacent predetermined locations based on the predicted time interval between each vehicle passing through every two adjacent predetermined locations.

[0012] In one embodiment, determining the target predicted time interval for passing through each pair of adjacent predetermined positions based on the predicted time interval for passing through each pair of adjacent predetermined positions for multiple vehicles includes: inputting the predicted time interval for passing through each pair of adjacent predetermined positions for multiple vehicles into a pre-trained second model to determine the target predicted time interval for passing through each pair of adjacent predetermined positions; or, based on the actual speed of multiple vehicles at the initial predetermined position or the travel time of multiple vehicles, inputting the predicted time interval for passing through each pair of adjacent predetermined positions for multiple vehicles into the corresponding pre-trained second model to determine the target predicted time interval for passing through each pair of adjacent predetermined positions; wherein, the second model is used for clustering.

[0013] In one embodiment, determining the congestion status of the target road based on the predicted time interval between each vehicle passing through every two adjacent predetermined locations on the target road includes: determining that the two adjacent predetermined locations are in a congested state if the predicted time interval between passing through the two adjacent predetermined locations is greater than or equal to a second preset threshold; or determining that the two adjacent predetermined locations are in a normal traffic state if the predicted time interval between passing through the two adjacent predetermined locations is less than the second preset threshold.

[0014] An artificial intelligence-based traffic state data processing device includes: an acquisition module for acquiring multiple videos; the multiple videos are videos corresponding to multiple predetermined locations on a target road, and each video contains at least one vehicle; a first determination module for determining, based on each video, the actual transit time and actual speed of each vehicle at each predetermined location; a second determination module for determining, based on the actual speed and actual transit time of each vehicle at an initial predetermined location, the predicted transit time of each vehicle at other predetermined locations besides the initial predetermined location; and a third determination module for determining, based on the actual transit time and actual speed of each vehicle at the multiple predetermined locations, the predicted transit time of each vehicle at other predetermined locations besides the initial predetermined location; and a third determination module for determining, based on the actual transit time and actual speed of each vehicle at the multiple predetermined locations, the predicted transit time of each vehicle at each predetermined location. The system calculates the speed and predicted transit time of each vehicle at the other predetermined locations, determines the exponential difference distance of each vehicle at each predetermined location and the predicted time interval between each vehicle passing through every two adjacent predetermined locations on the target road, and determines whether the target road is congested based on the exponential difference distance and / or the predicted time interval between each vehicle passing through every two adjacent predetermined locations on the target road; a sending module is used to send the exponential difference distance of each vehicle at each predetermined location and the predicted time interval between each vehicle passing through every two adjacent predetermined locations on the target road to the target vehicle and the traffic control server; wherein, the target vehicle is a vehicle traveling behind the at least one vehicle.

[0015] A computer device includes a memory and a processor, the memory storing computer-readable instructions, which, when executed by the processor, cause the processor to perform the steps of the artificial intelligence-based traffic state data processing method described above.

[0016] A storage medium storing computer-readable instructions, which, when executed by one or more processors, cause the one or more processors to perform the steps of the above-described artificial intelligence-based traffic state data processing method.

[0017] The aforementioned AI-based traffic status data processing method and apparatus further determine the exponential differential distance of each vehicle at each predetermined location and the predicted time interval between each vehicle passing through every two adjacent predetermined locations on the target road based on traffic video of the target road. This provides data support for determining the traffic conditions of the target road and provides data for drivers and traffic control centers on the target road, enabling them to take appropriate countermeasures as soon as possible. Attached Figure Description

[0018] Figure 1 An environment map for acquiring initial traffic state data is provided in one embodiment;

[0019] Figure 2 This is one of the flowcharts illustrating an AI-based traffic state data processing method provided in one embodiment;

[0020] Figure 3 This is a second flowchart illustrating an artificial intelligence-based traffic state data processing method provided in one embodiment.

[0021] Figure 4 This is a schematic diagram illustrating the relationship between a predetermined location, the actual vehicle speed corresponding to the predetermined location, the actual passage time, and the predicted passage time in one embodiment.

[0022] Figure 5 This is a schematic diagram of the framework of an artificial intelligence-based traffic state data processing device provided in one embodiment. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0024] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0025] To achieve intelligent infrastructure, numerous cameras have been installed in highway tunnels and municipal underground roads to detect traffic flow. Current technologies only monitor vehicles and their speeds within the traffic flow, but this information is far from sufficient for the vision of vehicle-road cooperation. Therefore, to provide more suitable and detailed information for future traffic and driving systems (including drivers or autonomous vehicles), this invention proposes an artificial intelligence-based traffic state data processing method and apparatus.

[0026] This invention presents an AI-based traffic state data processing method. By analyzing vehicle data from n cameras in one direction at the tunnel entrance, exit, and middle section (including transition and basic sections), a realistic statistical model of bidirectional traffic flow input and output for a given tunnel is established. Furthermore, a real-time prediction model is used to analyze tunnel travel time, capacity, and traffic status. This prediction model possesses learning capabilities, meaning it examines the sources of prediction errors and adjusts parameters based on actual values. Compared to existing AI methods that merely monitor vehicles and speeds, the method provided by this invention generates information from a roadside perspective—information urgently needed by future traffic and driving systems (including drivers or autonomous vehicles) and tunnel operators—that aligns with the vision of vehicle-road cooperation. This information is also crucial data input for future traffic control systems.

[0027] The following is combined with Figures 1-5 The present invention describes an artificial intelligence-based traffic state data processing method and apparatus.

[0028] Figure 1An initial traffic state data acquisition environment map provided in one embodiment includes n cameras positioned in one direction at the tunnel entrance, exit, and middle section (including transition and basic sections). Each tunnel entrance and exit may have one camera, and the middle section may have at least one camera. In this invention, as a preferred embodiment, one camera is used to capture traffic flow in one lane, assuming that each vehicle on that lane follows the tunnel's traffic rules (i.e., does not change lanes). Subsequent data analysis based on the video captured by the cameras yields tunnel travel time, capacity, and traffic status. It is understood that the method provided by this invention is also applicable to other scenarios in practical applications, such as when one camera can capture traffic flow on two or more roads traveling in the same direction, and each vehicle may change lanes.

[0029] Figure 2 This application proposes an AI-based traffic state data processing method as one embodiment. This AI-based traffic state data processing method can be executed by an AI-based traffic state data processing device. Specifically, as shown... Figure 2 As shown, the traffic state data processing method based on artificial intelligence provided by the present invention may include the following steps:

[0030] Step 210: Acquire multiple videos; the multiple videos are videos corresponding to multiple predetermined locations on the target road, and each of the multiple videos contains at least one vehicle.

[0031] Each video is a video taken within a certain time period.

[0032] It is understood that there are multiple cameras on the target road, each positioned at a predetermined location on the target road. Specifically, the predetermined locations can be found in [reference needed]. Figure 4 And its related descriptions.

[0033] Optionally, in tunnel applications, as a preferred implementation, one camera is used to capture vehicle data for one lane. This avoids the problem of low vehicle detection success rate due to poor lighting in tunnels. In other words, each video contains data for one lane, and the multiple videos mentioned above correspond to video data for one lane. In other application scenarios, the aforementioned single video includes data for multiple lanes.

[0034] Step 220: Based on each video, determine the actual transit time and actual speed of each vehicle at each predetermined location.

[0035] One possible approach is to determine the actual transit time and actual speed of each vehicle at multiple predetermined locations based on each video and its corresponding pre-trained detection and tracking model.

[0036] The pre-trained detection and tracking model can be, for example, a YOLOv5 or DeepSORT model. Specifically, the pre-trained detection and tracking model can be trained based on image frames from each video within a historical time period; the specific training process can refer to existing technologies. After obtaining the pre-trained detection and tracking model, detection and tracking can be performed on image frames in each video to determine the vehicle information appearing at the corresponding time for each image frame, thereby determining the actual passage time of each vehicle at each predetermined location. Furthermore, since the predetermined location is usually a fixed position while the vehicle is in motion, the speed of the vehicle passing through the predetermined location can be determined based on the distance the vehicle travels within a certain time interval before arriving at the predetermined location, and this speed is called the actual speed of the vehicle at the predetermined location.

[0037] It is understandable that, since each video is a video of a different location on a road, and the lighting and other factors are different on different sections of a road, in order to improve the accuracy of vehicle detection, a detection and tracking model can be trained for each video.

[0038] Step 230: Based on the actual speed and actual passage time of each vehicle at the initial predetermined position, determine the predicted passage time of each vehicle at other predetermined positions among multiple predetermined positions besides the initial predetermined position.

[0039] The initial predetermined location is the starting point for a vehicle to travel on a section of road, such as the location corresponding to a tunnel entrance.

[0040] It is understandable that, after determining the actual speed and actual transit time of each vehicle at the initial predetermined position, it can be assumed that the vehicle continues to travel at a constant speed at the same speed. Since the distance from the initial predetermined position to its corresponding next predetermined position is fixed, the time interval to the next predetermined position can be calculated, thus determining the predicted transit time. Similarly, the predicted transit time for each vehicle at each of the multiple predetermined positions other than the initial predetermined position can be determined.

[0041] Specifically, if the number of predetermined positions is n, and the distance between any two adjacent predetermined positions is equal, then the (i+1)th predetermined position D i+1 Prediction via time T i ' +1 The corresponding formula is: T i ' +1 =T i +ΔT i ,in, T i For each vehicle at the i-th predetermined position Di The actual transit time, v i For each vehicle at the i-th predetermined position D i The actual vehicle speed is given by L / (n-1), which represents the road distance between two adjacent predetermined locations. For details regarding the predetermined locations and their corresponding actual transit times and actual vehicle speeds, please refer to [link to relevant documentation]. Figure 4 And its related descriptions.

[0042] Step 240: Based on the actual passage time and actual speed of each vehicle at multiple predetermined locations, and the predicted passage time of each vehicle at the other predetermined locations, determine the exponential difference distance of each vehicle at each predetermined location and the predicted time interval between each vehicle passing through two adjacent predetermined locations on the target road, and determine whether the target road is congested based on the exponential difference distance and / or the predicted time interval between each vehicle passing through every two adjacent predetermined locations on the target road.

[0043] The exponential difference distance of each vehicle at each predetermined location is used to reflect the distance change caused by the change in vehicle speed between two adjacent predetermined locations on the target road, and can reflect the traffic status between two adjacent predetermined locations on the target road to a certain extent.

[0044] Step 250: The exponential difference distance of each vehicle at each predetermined location and the predicted time interval between each vehicle passing through every two adjacent predetermined locations on the target road are sent to the target vehicle and the traffic control server; wherein, the target vehicle is the vehicle traveling behind the at least one vehicle.

[0045] Step 250 enables the driver of the target vehicle and the traffic control center to understand the communication status between every two predetermined locations on the target road in a timely manner. This allows the driver of the target vehicle to prepare mentally and take appropriate actions in a timely manner, and also allows the traffic control center to understand the situation in a timely manner and make corresponding response strategies.

[0046] The traffic status data processing method based on artificial intelligence provided by this invention further determines the exponential differential distance of each vehicle at each predetermined location and the predicted time interval between each vehicle passing through every two adjacent predetermined locations on the target road based on traffic video of the target road. This provides data support for determining the traffic conditions of the target road and provides data for drivers and traffic control centers on the target road, which helps them to take corresponding measures as soon as possible.

[0047] In one embodiment, such as Figure 3As shown, the method for determining the exponential difference distance of each vehicle at each predetermined location and the predicted time interval between each vehicle passing through every two adjacent predetermined locations on the target road, based on the actual passage time and actual speed of each vehicle at multiple predetermined locations and the predicted passage time of each vehicle at other predetermined locations, includes the following steps:

[0048] Step 310: Based on the actual transit time and actual speed of each vehicle at multiple predetermined locations, and the predicted transit time of each vehicle at the other predetermined locations, determine the exponential differential distance of each vehicle at each predetermined location.

[0049] Each vehicle is at the (i+1)th predetermined position D i+1 The expression for the exponential difference distance can be: Among them, M i =T i+1 -T i ' +1 =T i+1 -(T i +ΔT i ), T i+1 For each vehicle at the (i+1)th predetermined position D i+1 The actual transit time, T i ' +1 For each vehicle at the (i+1)th predetermined position D i+1 The prediction through time, v i+1 For each vehicle at the (i+1)th predetermined position D i+1 The actual vehicle speed. For details regarding the predetermined location and its corresponding actual transit time and actual vehicle speed, please refer to... Figure 4 And its related descriptions.

[0050] It's understandable that in real-world applications, if congestion occurs, at least one target vehicle will appear in the video corresponding to the current predetermined location. However, if, after a preset time interval, no target vehicle is detected in the video corresponding to the next predetermined location, it indicates a high probability of congestion between the current and next predetermined locations. Therefore, the actual vehicle speed v at the next predetermined location will be... i+1 It can be set to 0 or a negative number, and the exponential difference distance calculated based on the actual vehicle speed can be used to determine whether there is congestion.

[0051] Based on the above expression for the exponential difference distance, we know that if the vehicle speed increases, the (i+1)th predetermined position D... i+1 The corresponding velocity v i+1 Not lower than the i-th predetermined position D i Actual vehicle speed v i If Mi is negative, For a positive number less than 1, the exponential difference distance is a small negative number; or,

[0052] If the vehicle speed gradually decreases, the (i+1)th predetermined position D i+1 The corresponding velocity v i+1 Below the i-th predetermined position D i Actual vehicle speed v i If Mi is positive, For positive numbers greater than 1, the exponential difference distance is a large exponential positive number; or

[0053] If the speed difference is 0, then It is 1, and because D i and D i+1 The velocity variation space between them can be very large, and Mi can be positive or negative, so the exponential difference distance will not be 0. The exponential difference distance is a small positive or negative number.

[0054] Therefore, based on the above exponential difference distance, it can be seen that when the vehicle speed gradually decreases, the corresponding exponential difference distance is also larger, which can be clearly distinguished from the cases where the vehicle speed gradually increases or remains constant. Therefore, it is possible to determine whether there is congestion between the target road or between two adjacent predetermined locations on the target road based on the exponential difference distance.

[0055] It's understandable that tunnels aren't very long overall. Therefore, to provide drivers with simpler information, one possible approach is to sum the exponential difference distances of each vehicle at multiple predetermined locations to obtain the total exponential difference distance of the target road. The congestion level of the target road is then determined based on this total exponential difference distance. The total exponential difference distance of the target road can be expressed by the formula:

[0056] Accordingly, the specific process of determining whether a target road is congested based on the exponential difference distance can be, for example:

[0057] If the total index difference distance of the target road is greater than or equal to a first preset threshold, the target road is determined to be in a congested state; or, if the total index difference distance of the target road is less than the first preset threshold, the target road is determined to be in a smooth state.

[0058] Step 320: Based on the actual passage time of each vehicle at multiple predetermined locations, the predicted passage time of each vehicle at the other predetermined locations, the actual time interval of each vehicle passing through the target road, and the pre-trained first model, determine the predicted time interval between every two adjacent predetermined locations of each vehicle passing through the target road.

[0059] In one embodiment, the first model is used to determine the allocation coefficient W for the time required to traverse any two adjacent predetermined locations among a plurality of predetermined locations. i Accordingly, the training process of the first model can be as follows:

[0060] The first model is trained by inputting the actual passage time of each vehicle at multiple predetermined locations, the predicted passage time of each vehicle at the other predetermined locations, and the actual time interval of each vehicle passing through the target road into the first model.

[0061] The training objective of the first model is to make the sum of the predicted passage time intervals between every two adjacent predetermined locations of each vehicle in multiple predetermined locations the same as the actual time interval between each vehicle passing through the target road.

[0062] Specifically, the training objective function expression for the first model is: min(∑ i,car_id M i ·W i -(T n-1 -T0)), where i = 0,…n-1, car_id is the identifier for each vehicle, M i ·W i This indicates that each vehicle is at the i-th predetermined position D. i The predetermined location D adjacent to it i+1 The prediction between them is based on the time interval, ∑ i,car_id M i ·W i This represents the summation of predictions for each vehicle at every two adjacent predetermined locations across multiple predetermined positions, T. n-1 -T0 represents the actual time interval between each vehicle passing through the target road.

[0063] Specifically, ∑ i,car_id M i ·W i You can get it after unfolding. Correspondingly, T n-1 -T0=T n-1,car_id -T 0,car_id .

[0064] In one embodiment, after determining the predicted time interval between every two adjacent predetermined locations in the target road for each vehicle, the method includes:

[0065] Based on the predicted time interval between each two adjacent predetermined locations corresponding to multiple vehicles, the target predicted time interval between each two adjacent predetermined locations is determined.

[0066] It is understandable that since there are multiple vehicles in every two adjacent predetermined positions, that is, there are multiple predicted time intervals corresponding to multiple vehicles in every two adjacent predetermined positions, and only one predicted time interval is needed to represent the predicted time interval for the entire traffic flow in every two adjacent predetermined positions, it is necessary to determine the target predicted time interval for passing through every two adjacent predetermined positions from the predicted time intervals for passing through every two adjacent predetermined positions corresponding to multiple vehicles.

[0067] Specifically, the predicted time intervals between each pair of adjacent predetermined positions for multiple vehicles can be input into a pre-trained second model to determine the target predicted time intervals between each pair of adjacent predetermined positions; alternatively, based on the actual speeds of multiple vehicles at the initial predetermined positions or the travel time periods of multiple vehicles, the predicted time intervals between each pair of adjacent predetermined positions for multiple vehicles can be input into their respective pre-trained second models to determine the target predicted time intervals between each pair of adjacent predetermined positions; wherein, the second model is used for clustering. The second model can be, for example, a K-nearest neighbor (Knn) model, used to cluster the predicted time intervals between each pair of adjacent predetermined positions for multiple vehicles.

[0068] The target prediction time interval between each two adjacent predetermined positions can be understood as: the target prediction time interval between each two adjacent predetermined positions determined based on the prediction time interval between multiple vehicles located between each two adjacent predetermined positions, rather than the prediction time interval between a single vehicle passing through each two adjacent predetermined positions.

[0069] It is understandable that traffic flow varies across different time periods. Therefore, other factors affecting the predicted time interval between two adjacent predetermined locations can be incorporated based on the actual situation. For example, during the morning rush hour (7:00-9:00) or holidays, traffic flow is high, resulting in a significantly different predicted time interval between two adjacent predetermined locations compared to other time periods. Therefore, the predicted time intervals between two adjacent predetermined locations for multiple vehicles traveling on the target road during different time periods can be input into the corresponding pre-trained second model for clustering, yielding a predicted time interval M representing different time periods. i ·W iSimilarly, if there are few or many vehicles on the target road, the actual speeds of the vehicles at their initial predetermined positions will vary significantly. Therefore, the actual speeds of the vehicles at their initial predetermined positions can be used as influencing factors to determine the target prediction time interval between each pair of adjacent predetermined positions when the vehicles have different actual speeds at their initial predetermined positions. For example, if the actual speeds of the vehicles at their initial predetermined positions are all approximately 30 km / h, the first target prediction time interval between each pair of adjacent predetermined positions can be determined based on the prediction time intervals between each pair of adjacent predetermined positions and the corresponding clustering model. If the actual speeds of the vehicles at their initial predetermined positions are all approximately 35 km / h, the second target prediction time interval between each pair of adjacent predetermined positions can be determined based on the prediction time intervals between each pair of adjacent predetermined positions and the corresponding clustering model.

[0070] In one embodiment, the congestion status of the target road is determined based on the predicted time interval between every two adjacent predetermined locations of each vehicle passing through the target road. Specifically, this can be achieved by:

[0071] If the target prediction time interval between two adjacent predetermined locations is greater than or equal to a second preset threshold, it is determined that the two adjacent predetermined locations are in a congested state; or, if the target prediction time interval between two adjacent predetermined locations is less than the second preset threshold, it is determined that the two adjacent predetermined locations are in a normal passage state.

[0072] It is understandable that the congestion of a target road can be determined simultaneously based on the exponential difference distance and the predicted time interval between any two adjacent predetermined positions of each vehicle passing through the target road. Specifically, in conjunction with the relevant descriptions above, for example, the target road can be determined to be congested if the total exponential difference distance of the target road is greater than or equal to a first preset threshold and the predicted time interval between any two adjacent predetermined positions of the target road is greater than or equal to a second preset threshold; or, the target road can be determined to be unobstructed if the total exponential difference distance of the target road is less than the first preset threshold and the predicted time interval between any two adjacent predetermined positions of the target road is less than the second preset threshold.

[0073] Figure 4 This is a schematic diagram illustrating the relationship between the predetermined location, the actual vehicle speed corresponding to the predetermined location, the actual passage time, and the predicted passage time provided by the present invention.

[0074] like Figure 4As shown, there are four lanes: West-East Lane 1, West-East Lane 2, East-West Lane 1, and East-West Lane 2. For West-East Lane 1 and West-East Lane 2, the corresponding i-th predetermined position is... Figure 4 D in i The (i+1)th reserved position is Figure 4 D in i+1 The (i+2)th reserved position is Figure 4 D in i+2 The i-th reserved position D i The corresponding actual transit time is T. i The actual vehicle speed is v i The (i+1)th predetermined position D i+1 The corresponding actual transit time is T. i+1 The actual vehicle speed is v i+1 The (i+2)th predetermined position D i+2 The corresponding actual transit time is T. i+2 The actual vehicle speed is v i+2 For West-East Lane 1 and West-East Lane 2, the actual vehicle speed at the initial predetermined position is v. i That is, v0, the (i+1)th predetermined position D i+1 The predicted transit time is: M i ·W i , of which M i Based on the i-th predetermined position D i The corresponding actual transit time is T. i and the (i+1)th predetermined position D i+1 The corresponding actual transit time is T. i+1 Confirmed; the (i+2)th predetermined position D i+2 The predicted transit time is: M i+1 ·W i+1 , of which M i+1 Based on the actual passage time corresponding to the (i+1)th predetermined position, T i+1 and the (i+2)th predetermined position D i+2 The corresponding actual transit time is T. i+2 Sure.

[0075] Similarly, for East-West Lane 1 and East-West Lane 2, the corresponding i-th predetermined position can be... Figure 4 D in i+2 The (i+1)th reserved position can be Figure 4 D in i+1 The (i+2)th reserved position can be Figure 4 D in iThe actual passage time and actual vehicle speed corresponding to each predetermined position are similar to those of West-East Lane 1 and West-East Lane 2, and will not be repeated here for the sake of simplicity.

[0076] The following describes the AI-based traffic state data processing device provided by the present invention, which corresponds to the AI-based traffic state data processing method described above.

[0077] Figure 5 The traffic state data processing device based on artificial intelligence provided by the present invention, such as Figure 5 As shown, the traffic state data processing device based on artificial intelligence provided by the present invention includes:

[0078] The acquisition module 510 is used to acquire multiple videos; the multiple videos are videos corresponding to multiple predetermined locations on the target road, and each video contains at least one vehicle.

[0079] The first determining module 520 is used to determine, based on each video, the actual passing time and actual speed of each vehicle at each predetermined location;

[0080] The second determining module 530 is used to determine the predicted passage time of each vehicle at other predetermined positions among multiple predetermined positions, based on the actual speed and actual passage time of each vehicle at the initial predetermined position.

[0081] The third determining module 540 is used to determine the exponential difference distance of each vehicle at each predetermined location and the predicted time interval between each vehicle passing through each of two adjacent predetermined locations in the target road, based on the actual passage time and actual speed of each vehicle at multiple predetermined locations and the predicted passage time of each vehicle at the other predetermined locations, and to determine whether the target road is congested based on the exponential difference distance and / or the predicted time interval between each vehicle passing through each of two adjacent predetermined locations in the target road.

[0082] The sending module 550 is used to send the exponential differential distance of each vehicle at each predetermined location and the predicted time interval between each vehicle passing through every two adjacent predetermined locations in the target road to the target vehicle and the traffic control server; wherein the target vehicle is a vehicle traveling behind the at least one vehicle.

[0083] The traffic status data processing device based on artificial intelligence provided by this invention further determines the exponential differential distance of each vehicle at each predetermined location and the predicted time interval between each vehicle passing through every two adjacent predetermined locations on the target road by using traffic video of the target road. This provides data support for determining the traffic conditions of the target road and provides data to drivers and traffic control centers on the target road, which helps them to take appropriate measures as soon as possible.

[0084] In one embodiment, the third determining module 540 includes:

[0085] The first determining unit is used to determine the exponential differential distance of each vehicle at each predetermined location based on the actual transit time and actual speed of each vehicle at multiple predetermined locations, and the predicted transit time of each vehicle at the other predetermined locations.

[0086] The second determining unit is used to determine the predicted time interval between every two adjacent predetermined locations of each vehicle on the target road based on the actual passage time of each vehicle at multiple predetermined locations, the predicted passage time of each vehicle at other predetermined locations, the actual time interval of each vehicle passing through the target road, and the pre-trained first model.

[0087] In one embodiment, the number of the plurality of predetermined positions is n, and the distance between any two adjacent predetermined positions is equal. Then, the (i+1)th predetermined position D... i+1 Prediction via time T i ' +1 The corresponding formula is: T i ' +1 =T i +ΔT i ,in, T i For each vehicle at the i-th predetermined position D i The actual transit time, v i For each vehicle at the i-th predetermined position D i The actual vehicle speed, L / (n-1) represents the road distance between two adjacent predetermined locations.

[0088] In one embodiment, each vehicle at the (i+1)th predetermined position D i+1 The expression for the exponential difference distance is: Among them, M i =T i+1 -T i ' +1 =T i+1 -(T i +ΔT i ), Ti+1 For each vehicle at the (i+1)th predetermined position D i+1 The actual transit time, T i ' +1 For each vehicle at the (i+1)th predetermined position D i+1 The prediction through time, v i+1 For each vehicle at the (i+1)th predetermined position D i+1 The actual speed of the vehicle.

[0089] In one embodiment, the first determining unit includes:

[0090] The summation module is used to sum the exponential difference distances of each vehicle at multiple predetermined locations to obtain the total exponential difference distance of the target road.

[0091] The third determining unit is configured to determine that the target road is congested if the total exponential difference distance of the target road is greater than or equal to a first preset threshold; or,

[0092] The fourth determining unit is used to determine that the target road is in a smooth state when the total index difference distance of the target road is less than a first preset threshold.

[0093] In one embodiment, the third determining module further includes:

[0094] The training unit is used to input the actual passage time of each vehicle at multiple predetermined locations, the predicted passage time of each vehicle at the other predetermined locations, and the actual time interval of each vehicle passing through the target road into the first model to train the first model.

[0095] The training objective of the first model is to make the sum of the predicted passage time intervals between every two adjacent predetermined positions of each vehicle in multiple predetermined positions the same as the actual time interval of each vehicle passing through the target road.

[0096] Specifically, the training objective function expression for the first model is: min(∑ i,car_id M i ·W i -(T n-1 -T0)), where i = 0,…n-1, car_id is the identifier for each vehicle, M i ·W i This indicates that each vehicle is at the i-th predetermined position D. i The predicted transit time interval between its adjacent predetermined locations, ∑ i,car_id M i ·W iThis represents the summation of predictions for each vehicle at every two adjacent predetermined locations across multiple predetermined positions, T. n-1 -T0 represents the actual time interval between each vehicle passing through the target road.

[0097] In one embodiment, the device further includes:

[0098] The fourth determining module is used to determine the target prediction time interval for passing through each two adjacent predetermined positions based on the prediction time interval between each two adjacent predetermined positions corresponding to multiple vehicles.

[0099] In one embodiment, the fourth determining module includes:

[0100] The fifth determining unit is used to input the predicted time intervals between every two adjacent predetermined positions for multiple vehicles into the pre-trained second model to determine the target predicted time intervals between every two adjacent predetermined positions; or,

[0101] The sixth determining unit is used to determine the target prediction time interval between each two adjacent predetermined positions by inputting the predicted time interval between each vehicle passing through each two adjacent predetermined positions into the corresponding pre-trained second model according to the actual speed of the multiple vehicles at the initial predetermined position or the travel time of the multiple vehicles; wherein, the second model is used for clustering.

[0102] In one embodiment, the second determining unit includes:

[0103] The seventh determining unit is configured to determine that there is congestion between two adjacent predetermined locations if the target prediction time interval between two adjacent predetermined locations is greater than or equal to a second preset threshold; or,

[0104] The eighth determining unit is used to determine that the two adjacent predetermined locations are in a normal passage state when the target prediction time interval between the two adjacent predetermined locations is less than a second preset threshold.

[0105] In one embodiment, a computer device is provided, the computer device including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the following steps corresponding to an artificial intelligence-based traffic state data processing method: acquiring multiple videos; the multiple videos are videos corresponding to multiple predetermined locations on a target road, and each video contains at least one vehicle; based on each video, determining the actual transit time and actual speed of each vehicle at each predetermined location; and based on the actual speed and actual transit time of each vehicle at an initial predetermined location, determining the predicted transit time of each vehicle at other predetermined locations besides the initial predetermined location. Based on the actual transit time and actual speed of each vehicle at multiple predetermined locations, and the predicted transit time of each vehicle at other predetermined locations, the exponential difference distance of each vehicle at each predetermined location and the predicted time interval between each vehicle passing through every two adjacent predetermined locations on the target road are determined. Based on the exponential difference distance and / or the predicted time interval between each vehicle passing through every two adjacent predetermined locations on the target road, it is determined whether the target road is congested. The exponential difference distance of each vehicle at each predetermined location and the predicted time interval between each vehicle passing through every two adjacent predetermined locations on the target road are sent to the target vehicle and the traffic control server. The target vehicle is the vehicle traveling behind the at least one vehicle.

[0106] On the other hand, the present invention also provides a computer program product, the computer program product comprising a computer program stored on a non-transitory computer-readable storage medium, the computer program comprising program instructions, wherein when the program instructions are executed by a computer, the computer is able to execute the AI-based traffic state data processing method provided by the present invention, wherein the AI-based traffic state data processing method comprises: acquiring multiple videos; the multiple videos being videos corresponding to multiple predetermined locations on a target road, each of the multiple videos containing at least one vehicle; based on each video, determining the actual passage time and actual speed of each vehicle at each predetermined location; based on the actual speed and actual passage time of each vehicle at an initial predetermined location, determining the actual speed of each vehicle at the multiple predetermined locations excluding the initial predetermined location. The predicted passage time to other predetermined locations outside the designated location; based on the actual passage time and actual speed of each vehicle at multiple predetermined locations, and the predicted passage time of each vehicle at the other predetermined locations, the exponential difference distance of each vehicle at each predetermined location and the predicted time interval between each vehicle passing through every two adjacent predetermined locations on the target road are determined, and the congestion of the target road is determined based on the exponential difference distance and / or the predicted time interval between each vehicle passing through every two adjacent predetermined locations on the target road; the exponential difference distance of each vehicle at each predetermined location and the predicted time interval between each vehicle passing through every two adjacent predetermined locations on the target road are sent to the target vehicle and the traffic control server; wherein, the target vehicle is the vehicle traveling behind the at least one vehicle.

[0107] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, is implemented to perform the AI-based traffic state data processing method provided by the present invention. The AI-based traffic state data processing method includes: acquiring multiple videos; the multiple videos being videos corresponding to multiple predetermined locations on a target road, each of the multiple videos containing at least one vehicle; determining, based on each video, the actual transit time and actual speed of each vehicle at each predetermined location; and determining, based on the actual speed and actual transit time of each vehicle at an initial predetermined location, the predicted transit time of each vehicle at other predetermined locations besides the initial predetermined location. Based on the actual transit time and actual speed of each vehicle at multiple predetermined locations, and the predicted transit time of each vehicle at other predetermined locations, the exponential difference distance of each vehicle at each predetermined location and the predicted time interval between each vehicle passing through every two adjacent predetermined locations on the target road are determined. Based on the exponential difference distance and / or the predicted time interval between each vehicle passing through every two adjacent predetermined locations on the target road, it is determined whether the target road is congested. The exponential difference distance of each vehicle at each predetermined location and the predicted time interval between each vehicle passing through every two adjacent predetermined locations on the target road are sent to the target vehicle and the traffic control server. The target vehicle is the vehicle traveling behind the at least one vehicle.

[0108] The system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0109] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0110] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A traffic state data processing method based on artificial intelligence, characterized in that, The method includes: Acquire multiple videos; the multiple videos are videos corresponding to multiple predetermined locations on the target road, and each video contains at least one vehicle; Based on each video, determine the actual transit time and actual speed of each vehicle at each predetermined location; Based on the actual speed and actual transit time of each vehicle at the initial predetermined position, the predicted transit time of each vehicle at other predetermined positions among multiple predetermined positions besides the initial predetermined position is determined. Based on the actual transit time and actual speed of each vehicle at multiple predetermined locations, and the predicted transit time of each vehicle at the other predetermined locations, the exponential difference distance of each vehicle at each predetermined location and the predicted time interval between each vehicle passing through every two adjacent predetermined locations on the target road are determined, and whether the target road is congested is determined based on the exponential difference distance and / or the predicted time interval between each vehicle passing through every two adjacent predetermined locations on the target road. The exponential differential distance of each vehicle at each predetermined location and the predicted time interval between each vehicle passing through every two adjacent predetermined locations on the target road are sent to the target vehicle and the traffic control server; wherein, the target vehicle is the vehicle traveling behind the at least one vehicle. Each vehicle at the (i+1)th predetermined position The expression for the exponential difference distance is: ,in, , For each vehicle at the (i+1)th predetermined position The actual transit time For each vehicle at the (i+1)th predetermined position The prediction time For each vehicle at the (i+1)th predetermined position The actual vehicle speed; If the number of predetermined positions is n, and the distance between any two adjacent predetermined positions is equal, then the (i+1)th predetermined position... Predicted time The corresponding formula is: ,in, , For each vehicle at the i-th predetermined position The actual transit time For each vehicle at the i-th predetermined position The actual vehicle speed This indicates the road distance between two adjacent predetermined locations.

2. The traffic state data processing method based on artificial intelligence as described in claim 1, characterized in that, The determination of the exponential difference distance of each vehicle at each predetermined location and the predicted time interval between each vehicle passing through every two adjacent predetermined locations on the target road, based on the actual transit time and actual speed of each vehicle at multiple predetermined locations and the predicted transit time of each vehicle at other predetermined locations, includes: Based on the actual transit time and actual speed of each vehicle at multiple predetermined locations, and the predicted transit time of each vehicle at the other predetermined locations, the exponential differential distance of each vehicle at each predetermined location is determined. Based on the actual transit time of each vehicle at multiple predetermined locations, the predicted transit time of each vehicle at other predetermined locations, the actual time interval of each vehicle passing through the target road, and a pre-trained first model, the predicted time interval between each vehicle passing through every two adjacent predetermined locations on the target road is determined.

3. The traffic state data processing method based on artificial intelligence as described in claim 1, characterized in that, The process of determining whether a target road is congested based on the exponential difference distance includes: The total exponential difference distance of the target road is obtained by summing the exponential difference distances of each vehicle at multiple predetermined locations. ; If the total exponential difference distance of the target road is greater than or equal to a first preset threshold, the target road is determined to be in a congested state; or, If the total index difference distance of the target road is less than a first preset threshold, the target road is determined to be in a smooth state.

4. The traffic state data processing method based on artificial intelligence as described in claim 3, characterized in that, The first model is used to determine the allocation coefficient of the time required to traverse any two adjacent predetermined locations among multiple predetermined locations. Accordingly, the training process of the first model is as follows: The first model is trained by inputting the actual passage time of each vehicle at multiple predetermined locations, the predicted passage time of each vehicle at the other predetermined locations, and the actual time interval of each vehicle passing through the target road into the first model. The training objective of the first model is to make the sum of the predicted passage time intervals between every two adjacent predetermined positions of each vehicle in multiple predetermined positions the same as the actual time interval of each vehicle passing through the target road. Specifically, the training objective function expression for the first model is: ,in, car_id To identify each vehicle. This indicates that each vehicle is at the i-th predetermined position. The predicted time interval between its adjacent predetermined locations, This means that the predictions for each vehicle at multiple predetermined locations are summed over time intervals between every two adjacent predetermined locations. This represents the actual time interval between each vehicle passing through the target road.

5. The traffic state data processing method based on artificial intelligence as described in claim 4, characterized in that, After determining the predicted time interval between every two adjacent predetermined locations in the target road for each vehicle, the method includes: Based on the predicted time interval between each two adjacent predetermined locations corresponding to multiple vehicles, the target predicted time interval between each two adjacent predetermined locations is determined.

6. The traffic state data processing method based on artificial intelligence as described in claim 5, characterized in that, The determination of the target prediction time interval for each two adjacent predetermined locations based on the prediction time interval between each two adjacent predetermined locations corresponding to multiple vehicles includes: The predicted time intervals between every two adjacent predetermined locations for multiple vehicles are input into a pre-trained second model to determine the target predicted time intervals between every two adjacent predetermined locations; or... Based on the actual speed of multiple vehicles at the initial predetermined position or the travel time of multiple vehicles, the predicted time interval between each two adjacent predetermined positions is input into the corresponding pre-trained second model to determine the target predicted time interval between each two adjacent predetermined positions; wherein, the second model is used for clustering.

7. The traffic state data processing method based on artificial intelligence as described in claim 6, characterized in that, The process of determining the congestion status of the target road based on the predicted time interval between every two adjacent predetermined locations of each vehicle passing through the target road includes: If the time interval between target predictions from two adjacent predetermined locations is greater than or equal to a second preset threshold, it is determined that there is congestion between the two adjacent predetermined locations; or... If the target prediction time interval between two adjacent predetermined locations is less than a second preset threshold, it is determined that the two adjacent predetermined locations are in a normal passage state.

8. A traffic state data processing device based on artificial intelligence, characterized in that, The device includes: The acquisition module is used to acquire multiple videos; the multiple videos are videos corresponding to multiple predetermined locations on the target road, and each video contains at least one vehicle. The first determining module is used to determine, based on each video, the actual transit time and actual speed of each vehicle at each predetermined location; The second determining module is used to determine the predicted passage time of each vehicle at other predetermined positions besides the initial predetermined position based on the actual speed and actual passage time of each vehicle at the initial predetermined position. The third determining module is used to determine the exponential difference distance of each vehicle at each predetermined location and the predicted time interval between each vehicle passing through each of two adjacent predetermined locations on the target road, based on the actual passage time and actual speed of each vehicle at multiple predetermined locations and the predicted passage time of each vehicle at the other predetermined locations. The module is also used to determine whether the target road is congested based on the exponential difference distance and / or the predicted time interval between each vehicle passing through each of two adjacent predetermined locations on the target road. The sending module is used to send the exponential differential distance of each vehicle at each predetermined location and the predicted time interval between each vehicle passing through every two adjacent predetermined locations in the target road to the target vehicle and the traffic control server; wherein, the target vehicle is a vehicle traveling behind the at least one vehicle. Each vehicle at the (i+1)th predetermined position The expression for the exponential difference distance is: ,in, , For each vehicle at the (i+1)th predetermined position The actual transit time For each vehicle at the (i+1)th predetermined position The prediction time For each vehicle at the (i+1)th predetermined position The actual vehicle speed; If the number of predetermined positions is n, and the distance between any two adjacent predetermined positions is equal, then the (i+1)th predetermined position... Predicted time The corresponding formula is: ,in, , For each vehicle at the i-th predetermined position The actual transit time For each vehicle at the i-th predetermined position The actual vehicle speed This indicates the road distance between two adjacent predetermined locations.

9. A computer device comprising a memory and a processor, wherein the memory stores computer-readable instructions, characterized in that, When the computer-readable instructions are executed by the processor, the processor performs the steps of the AI-based traffic state data processing method as described in any one of claims 1 to 7.

10. A storage medium storing computer-readable instructions, characterized in that, When the computer-readable instructions are executed by one or more processors, the one or more processors perform the steps of the AI-based traffic state data processing method as described in any one of claims 1 to 7.

Images