A method for predicting the minimum speed range of vehicles at intersections

Abstract

This invention relates to the field of intelligent transportation technology and specifically to a method for predicting the minimum speed range of vehicles at intersections. The method includes: S10, acquiring vehicle and intersection scene information based on a camera and a 3D laser sensor; S20, acquiring behavioral and dynamic features based on a following behavior model; S30, establishing a hierarchical structure to predict the minimum speed range of a target vehicle traveling from its current position to the intersection; S40, establishing a five-level residual neural network for each layer of the hierarchical structure to obtain the probability that the target vehicle's minimum speed falls within each range; and S50, the intersection speed guidance system selecting a suitable minimum speed prediction range based on a predetermined probability threshold to improve the reliability and accuracy of the intersection speed guidance system. This invention is simple in process and easy to operate, achieving accurate prediction of the minimum speed range of vehicles traveling from their current position to the intersection, thus improving the traffic communication efficiency at intersections.

Description

Technical Field

[0001] This invention belongs to the field of intelligent transportation technology, and in particular relates to a method for predicting the minimum driving speed range of vehicles at intersections. Background Technology

[0002] With the rapid development of vehicle-road cooperative driving systems, the demand for various intelligent driving assistance systems and autonomous driving systems is also gradually increasing. By acquiring the status information of the current vehicle, other vehicles in the current lane, and traffic lights, it is possible to effectively optimize the driving strategy of the current vehicle, improve the driving comfort and energy efficiency of the vehicle. Furthermore, by predicting the minimum speed of vehicles traveling at the intersection over a period of time, it is possible to provide the most critical influencing factors for optimizing the target driving speed and the lane selection strategy of the vehicle. In addition, combining micro-traffic flow and intersection information is also conducive to optimizing the timing strategy of traffic lights and improving the traffic efficiency of the intersection.

[0003] In fact, the existing technical solutions only predict the time it takes for a vehicle to arrive at an intersection. If there are other vehicles in front of the target vehicle, it is necessary to recursively predict the arrival time of the vehicles in front and calculate the time it takes for the target vehicle to arrive at the intersection based on the arrival time of the vehicles in front, the distance between the target vehicle and the vehicles in front, and the relative speed. For example, the patent with announcement number CN111681413B provides a method and device for real-time prediction of the time it takes for a motor vehicle to pass through a traffic light-controlled intersection. The method includes: acquiring the driving status information of a motor vehicle traveling on a traffic light-controlled intersection section, including current driving position information and current speed information, and acquiring the traffic light information set at this traffic light-controlled intersection; determining whether there are other motor vehicles traveling in front of the predicted motor vehicle and whether there are traffic lights ahead; when there are motor vehicles traveling in front of the predicted motor vehicle but no traffic lights, determining whether the speed of the motor vehicle in front is zero; when it is not zero, determining the time it takes for the predicted motor vehicle to arrive at the predicted position based on the distance between the predicted motor vehicle and the motor vehicle in front, the speed of the predicted motor vehicle, and the time it takes for the motor vehicle in front of the predicted motor vehicle to arrive at the predicted position. This patent's prediction process is complex and inefficient, and it only predicts the time it takes for a motor vehicle to reach the predicted location, without describing a technical solution for calculating the minimum speed of the motor vehicle.

[0004] Therefore, how to accurately and efficiently predict the minimum speed of vehicles traveling at intersections over a period of time in order to improve traffic efficiency at intersections is a problem that urgently needs to be solved by those in this technical field. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the purpose of this invention is to provide a method for predicting the minimum speed range of vehicles at intersections, thereby solving the problem of low traffic efficiency caused by the inability to accurately predict the minimum speed range of vehicles at intersections in existing technologies.

[0006] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0007] This invention provides a method for predicting the minimum speed range of vehicles at intersections, comprising the following steps:

[0008] S10. Acquire vehicle and intersection scene information based on cameras and 3D laser sensors;

[0009] S20. Obtain behavioral and dynamic features based on the vehicle following behavior model;

[0010] S30. Establish a hierarchical structure for predicting the minimum speed range of a target vehicle from its current location to the intersection;

[0011] S40. Establish a five-level residual neural network for each layer of the hierarchical model to obtain the probability that the minimum speed of the target vehicle is in each interval.

[0012] S50, the intersection speed guidance system selects a suitable minimum speed prediction range based on a predetermined probability threshold to improve the reliability and accuracy of the intersection speed guidance system.

[0013] Furthermore, the specific steps of step S10 are as follows:

[0014] S101. Install cameras and 3D laser sensors on streetlights;

[0015] S102. Based on traffic light cameras and vehicle recognition algorithms, identify and segment vehicle type, vehicle color, vehicle position, and two-dimensional bounding shape in images;

[0016] S103. By installing traffic light cameras and near-3D laser sensors, obtain all 3D point clouds of the intersection traffic scene. By using the matching algorithm between the vehicle camera and the 3D laser sensor to obtain the 3D point cloud of each vehicle at the intersection, obtain the vehicle's wheelbase information, speed information, three-dimensional shape information, and distance information from the intersection.

[0017] Furthermore, the specific steps of step S20 are as follows:

[0018] S201. Based on the collected data over a period of time and data analysis, obtain the type, color, position, two-dimensional enclosing shape, wheelbase information, speed information, three-dimensional shape information, distance from the intersection, vehicle acceleration information, and deceleration information for each vehicle at each moment at the intersection.

[0019] S202. Based on the location information of each vehicle at every moment at the intersection, perform parameter estimation for the following behavior model of vehicles of different types, different axle lengths and different queuing positions.

[0020] S203. Based on the following behavior model of each vehicle after parameter estimation and the intersection scene information, multiple simulations are performed to obtain the dynamic characteristics of the target vehicle's journey to the intersection.

[0021] Furthermore, in step S202, the parameters of the following behavior model include: maximum acceleration, maximum deceleration, following target distance, following target headway, stopping target distance, maximum driving speed, driver reaction time, and vehicle start-stop time.

[0022] Furthermore, in step S203, the dynamic features include the average, maximum, and minimum values ​​of the speed and time of the target vehicle to the intersection, the 25th percentile, 50th percentile, and 75th percentile values, and the average, maximum, and minimum values ​​of the speed and time of the target vehicle to the intersection, the 25th percentile, 50th percentile, and 75th percentile values.

[0023] Furthermore, the specific steps of step S30 are as follows:

[0024] S301. Divide the minimum speed of the target vehicle during its journey to the intersection into 9 intervals and assign corresponding category labels;

[0025] S302. Based on the definition of congestion and data analysis, two adjacent speed ranges are manually merged each time to form a hierarchical structure with a minimum speed range.

[0026] Furthermore, in step S301, the nine intervals of minimum speed are [0, 5 km / h), [5 km / h, 10 km / h), [10 km / h, 15 km / h), [15 km / h, 20 km / h), [20 km / h, 25 km / h), [25 km / h, 30 km / h), [35 km / h, 40 km / h), [40 km / h, 45 km / h), and [45 km / h, 120 km / h].

[0027] Furthermore, in step S302, the hierarchical structure of the minimum speed range is as follows:

[0028] The minimum speed range for the 9th layer is [0, 5 km / h), [5 km / h, 10 km / h), [10 km / h, 15 km / h), [15 km / h, 20 km / h), [20 km / h, 25 km / h), [25 km / h, 30 km / h), [35 km / h, 40 km / h), [40 km / h, 45 km / h) and [45 km / h, 120 km / h];

[0029] The minimum speed range for the 8th layer is [0, 5 km / h), [5 km / h, 15 km / h), [15 km / h, 20 km / h), [20 km / h, 25 km / h), [25 km / h, 30 km / h), [35 km / h, 40 km / h), [40 km / h, 45 km / h) and [45 km / h, 120 km / h];

[0030] The minimum speed range for the 7th layer is [0, 5 km / h), [5 km / h, 20 km / h), [20 km / h, 25 km / h), [25 km / h, 30 km / h), [35 km / h, 40 km / h), [40 km / h, 45 km / h) and [45 km / h, 120 km / h];

[0031] The minimum speed range for the 6th layer is [0, 5 km / h), [5 km / h, 20 km / h), [20 km / h, 25 km / h), [25 km / h, 30 km / h), [35 km / h, 40 km / h) and [40 km / h, 120 km / h];

[0032] The minimum speed range for the 5th layer is [0, 5 km / h), [5 km / h, 20 km / h), [20 km / h, 25 km / h), [25 km / h, 30 km / h) and [35 km / h, 120 km / h];

[0033] The minimum speed range for the fourth layer is [0, 5 km / h), [5 km / h, 25 km / h), [25 km / h, 30 km / h), and [35 km / h, 120 km / h].

[0034] The minimum speed range for the third layer is [0, 5 km / h), [5 km / h, 30 km / h), and [35 km / h, 120 km / h].

[0035] The minimum speed range for the second layer is [0, 5 km / h) and [5 km / h, 120 km / h];

[0036] The minimum speed range for the first layer is [0, 120 km / h].

[0037] Furthermore, the specific steps of step S40 are as follows:

[0038] S401. Based on the hierarchical structure of the minimum speed range, design and train a 5-level residual neural network for each layer from layer 2 to layer 9. Each level of the network has 128 nodes, and the last level is the output level, which is of type softMax.

[0039] S402. By collecting data on the status and characteristics of vehicles queuing at intersections, a dataset is established, and eight hierarchical neural networks are trained.

[0040] S403, based on 8 neural networks, predicts the minimum speed range of a target vehicle during its journey to the intersection and the corresponding probability.

[0041] Furthermore, in step S402, the neural network corresponding to the 9th layer model is fully trained during the training process, while the neural networks corresponding to the 8th to 2nd layers are trained only for the last layer, and the other layers reuse the neural network corresponding to the 9th layer model.

[0042] The method for predicting the minimum speed range of vehicles at intersections provided by this invention has at least the following advantages compared with the prior art:

[0043] Existing technologies only predict the arrival time of vehicles at intersections, achieving only the prediction of the time it takes for a vehicle to reach a predicted location. They do not describe technical solutions for calculating the minimum speed of a vehicle, and the prediction process is complex and inefficient. This invention, based on a traffic light camera placed on a streetlight and a 3D laser sensor, can more accurately obtain the driving and traffic conditions of each vehicle at the intersection. Based on these driving and traffic conditions, this invention can more accurately estimate the following dynamic model of each vehicle's driving behavior. The prediction model of this invention not only combines static features of the scene but also incorporates various behavioral features from the following behavior model based on data fitting and dynamic features from multiple simulations, thereby improving the accuracy of predicting the minimum speed range during the journey to the intersection. This patent establishes a hierarchical minimum speed prediction model from coarse to fine. Each layer uses a 5-level neural network with the same structure to predict the probability of the target vehicle's minimum speed falling within each minimum speed range at the current layer during the journey to the intersection. This effectively expands the prediction range, improves prediction accuracy, and selects a suitable minimum speed prediction range based on a predetermined probability threshold, thereby improving the traffic efficiency at the intersection. Attached Figure Description

[0044] To more clearly illustrate the solution of the present invention, a brief introduction will be given to the drawings used in the description of the embodiments below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0045] Figure 1 A flowchart illustrating a method for predicting the minimum speed range of vehicles at intersections, provided as an embodiment of the present invention. Detailed Implementation

[0046] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. Preferred embodiments of the invention are shown in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to provide a thorough and complete understanding of the disclosure of the invention.

[0047] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention.

[0048] This invention provides a method for predicting the minimum speed range of vehicles at intersections. The method is applied to predicting the minimum speed range of vehicles over a period of time. The method includes the following steps:

[0049] S10. Acquire vehicle and intersection scene information based on cameras and 3D laser sensors;

[0050] S20. Obtain behavioral and dynamic features based on the vehicle following behavior model;

[0051] S30. Establish a hierarchical structure for predicting the minimum speed range of a target vehicle from its current location to the intersection;

[0052] S40. Establish a five-level residual neural network for each layer of the hierarchical model to obtain the probability that the minimum speed of the target vehicle is in each interval.

[0053] S50, the intersection speed guidance system selects a suitable minimum speed prediction range based on a predetermined probability threshold to improve the reliability and accuracy of the intersection speed guidance system.

[0054] This invention has a simple process and is easy to operate. It can predict the minimum speed range of vehicles traveling from their current position to the intersection. The prediction results are accurate and improve the traffic communication efficiency of the intersection.

[0055] To enable those skilled in the art to better understand the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0056] This invention provides a method for predicting the minimum speed range of vehicles at intersections, applicable to predicting the minimum speed range of vehicles over a period of time, such as... Figure 1 As shown, in this embodiment, the method for predicting the minimum speed range of vehicles at the intersection includes the following steps:

[0057] S10. Acquire vehicle and intersection scene information based on cameras and 3D laser sensors;

[0058] Specifically, the steps of step S10 are as follows:

[0059] S101. Install cameras and 3D laser sensors on streetlights;

[0060] S102. Based on traffic light cameras and vehicle recognition algorithms, identify and segment vehicle type, vehicle color, vehicle position, and two-dimensional bounding shape in images;

[0061] S103. By installing traffic light cameras and near-3D laser sensors, all 3D point clouds of the intersection traffic scene are obtained. By using the matching algorithm of the vehicle camera and 3D laser sensor to obtain the 3D point cloud of each vehicle at the intersection, the wheelbase information, speed information, three-dimensional shape information, and distance information from the intersection of the vehicle are obtained, which can more accurately predict the following behavior model of each vehicle.

[0062] S20. Obtain behavioral and dynamic features based on the vehicle following behavior model;

[0063] Specifically, the steps of step S20 are as follows:

[0064] S201. Based on the collected data over a period of time and data analysis, obtain the type, color, position, two-dimensional enclosing shape, wheelbase information, speed information, three-dimensional shape information, distance from the intersection, vehicle acceleration information, and deceleration information for each vehicle at each moment at the intersection.

[0065] S202. Based on the location information of each vehicle at every moment at the intersection, perform parameter estimation for the following behavior model of vehicles of different types, different axle lengths and different queuing positions.

[0066] Furthermore, in this embodiment, the parameters of the following behavior model include: maximum acceleration, maximum deceleration, target distance to the vehicle being followed, time distance to the front of the target vehicle being followed, target stopping distance, maximum driving speed, driver reaction time, and vehicle start-stop time;

[0067] S203. Based on the following behavior model of each vehicle after parameter estimation and the intersection scene information, perform multiple simulations to obtain the dynamic characteristics of the target vehicle in the process of reaching the intersection.

[0068] Furthermore, in this embodiment, the dynamic features include the average, maximum, and minimum values ​​of the target vehicle's speed and time to the intersection, as well as the 25th, 50th, and 75th percentile values; the average, maximum, and minimum values ​​of the target vehicle's speed and time to the intersection, as well as the 25th, 50th, and 75th percentile values; and these are added to the input features of the neural network. The model input features of this embodiment not only include static features of the intersection scene but also features based on a following behavior model after data fitting, and dynamic features of the target vehicle arriving at the intersection after multiple simulations based on the following behavior model. The output value of the neural network is the probability that the target vehicle's minimum speed falls within each minimum speed range during its journey to the intersection, thus improving the accuracy of predicting the minimum speed range during the journey to the intersection.

[0069] S30. Establish a hierarchical structure for predicting the minimum speed range of a target vehicle from its current location to the intersection;

[0070] Specifically, the steps of step S30 are as follows:

[0071] S301. Divide the minimum speed of the target vehicle during its journey to the intersection into 9 intervals and assign corresponding category labels;

[0072] Furthermore, in this embodiment, the nine intervals of minimum speed are [0, 5 km / h), [5 km / h, 10 km / h), [10 km / h, 15 km / h), [15 km / h, 20 km / h), [20 km / h, 25 km / h), [25 km / h, 30 km / h), [35 km / h, 40 km / h), [40 km / h, 45 km / h), and [45 km / h, 120 km / h). This embodiment of the invention adopts a hierarchical prediction interval of the minimum vehicle speed from coarse to fine. By expanding the prediction range and improving the prediction accuracy, the driving assistance system can select a suitable minimum speed prediction interval based on a predetermined probability threshold (e.g., greater than 95%), and provide a suitable guiding speed for crossing intersections, thereby improving the reliability and accuracy of the intersection speed guidance driving decision system.

[0073] S302. Based on the definition of congestion and data analysis, two adjacent speed ranges are manually merged each time to form a hierarchical structure with a minimum speed range.

[0074] Specifically, in this embodiment, the hierarchical structure of the minimum speed range is as follows:

[0075] The minimum speed range for the 9th layer is [0, 5 km / h), [5 km / h, 10 km / h), [10 km / h, 15 km / h), [15 km / h, 20 km / h), [20 km / h, 25 km / h), [25 km / h, 30 km / h), [35 km / h, 40 km / h), [40 km / h, 45 km / h) and [45 km / h, 120 km / h];

[0076] The minimum speed range for the 8th layer is [0, 5 km / h), [5 km / h, 15 km / h), [15 km / h, 20 km / h), [20 km / h, 25 km / h), [25 km / h, 30 km / h), [35 km / h, 40 km / h), [40 km / h, 45 km / h) and [45 km / h, 120 km / h];

[0077] The minimum speed range for the 7th layer is [0, 5 km / h), [5 km / h, 20 km / h), [20 km / h, 25 km / h), [25 km / h, 30 km / h), [35 km / h, 40 km / h), [40 km / h, 45 km / h) and [45 km / h, 120 km / h];

[0078] The minimum speed range for the 6th layer is [0, 5 km / h), [5 km / h, 20 km / h), [20 km / h, 25 km / h), [25 km / h, 30 km / h), [35 km / h, 40 km / h) and [40 km / h, 120 km / h];

[0079] The minimum speed range for the 5th layer is [0, 5 km / h), [5 km / h, 20 km / h), [20 km / h, 25 km / h), [25 km / h, 30 km / h) and [35 km / h, 120 km / h];

[0080] The minimum speed range for the fourth layer is [0, 5 km / h), [5 km / h, 25 km / h), [25 km / h, 30 km / h), and [35 km / h, 120 km / h].

[0081] The minimum speed range for the third layer is [0, 5 km / h), [5 km / h, 30 km / h), and [35 km / h, 120 km / h].

[0082] The minimum speed range for the second layer is [0, 5 km / h) and [5 km / h, 120 km / h];

[0083] The minimum speed range for the first layer is [0, 120 km / h].

[0084] S40. Establish a five-level residual neural network for each layer of the hierarchical model to obtain the probability that the minimum speed of the target vehicle is in each interval.

[0085] Specifically, the steps of step S40 are as follows:

[0086] S401. Based on the hierarchical structure of the minimum speed range, design and train a 5-level residual neural network for each layer from layer 2 to layer 9. Each level of the network has 128 nodes, and the last level is the output level, which is of type softMax.

[0087] S402. By collecting data on the status and characteristics of vehicles queuing at intersections, a dataset is established, and eight hierarchical neural networks are trained.

[0088] Furthermore, in this embodiment, the neural network corresponding to the 9th layer model is fully trained during the training process, while the neural networks corresponding to the 8th to 2nd layers are trained only for the last layer. Other layers of neural networks reuse the neural network corresponding to the 9th layer model, reducing the complexity of building and training the neural network.

[0089] S403, based on 8 neural networks, predicts the minimum speed range of a target vehicle during its journey to the intersection and the corresponding probability.

[0090] S50, the intersection speed guidance system selects a suitable minimum speed prediction range based on a predetermined probability threshold to improve the reliability and accuracy of the intersection speed guidance system.

[0091] The method for predicting the minimum speed range of vehicles at intersections described in the above embodiments, compared with existing technologies, only predicts the time it takes for a vehicle to arrive at the intersection, and only achieves the prediction of the time it takes for a motor vehicle to arrive at the predicted location. It does not describe a technical solution for calculating the minimum speed of a motor vehicle, and the prediction process is complex and inefficient. This invention, based on traffic light cameras and 3D laser sensors placed on streetlights, can more accurately obtain the driving and traffic states of each vehicle at the intersection. Based on the driving and traffic states, this invention can more accurately estimate the following dynamic model of each vehicle's driving behavior. The prediction model of this invention not only combines the static features of the scene but also incorporates various behavioral features based on the following behavior model after data fitting and dynamic features from multiple simulations, thereby improving the accuracy of predicting the minimum speed range during the process of driving to the intersection. This patent establishes a hierarchical minimum speed prediction model from coarse to fine. Each layer uses a 5-level neural network with the same structure to predict the probability of the target vehicle's minimum speed falling within each minimum speed range at the current layer during the process of driving to the intersection, effectively expanding the prediction range, improving the prediction accuracy, and selecting a suitable minimum speed prediction range according to a predetermined probability threshold, thereby improving the traffic efficiency of the intersection.

[0092] Obviously, the embodiments described above are merely preferred embodiments of the present invention, and not all embodiments. The accompanying drawings illustrate preferred embodiments of the present invention, but do not limit the scope of the patent. The present invention can be implemented in many different forms; rather, these embodiments are provided to provide a more thorough and complete understanding of the disclosure of the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing specific embodiments, or make equivalent substitutions for some of the technical features. Any equivalent structures made using the content of this specification and drawings, directly or indirectly applied to other related technical fields, are similarly within the scope of patent protection of this invention.

Claims

1. A method for predicting the minimum speed range of vehicles at intersections, characterized in that, Includes the following steps: S10. Acquire vehicle and intersection scene information based on cameras and 3D laser sensors; S20. Obtain behavioral and dynamic features based on the vehicle following behavior model; S30. Establish a hierarchical structure for predicting the minimum speed range of a target vehicle from its current location to the intersection; The specific steps are as follows: S301. Divide the minimum speed of the target vehicle during its journey to the intersection into 9 intervals and assign corresponding category labels; S302. Based on the definition of congestion and data analysis, two adjacent speed ranges are manually merged each time to form a hierarchical structure with a minimum speed range. The hierarchical structure of the minimum speed range is as follows: The minimum speed range for the 9th layer is [0, 5 km / h), [5 km / h, 10 km / h), [10 km / h, 15 km / h), [15 km / h, 20 km / h), [20 km / h, 25 km / h), [25 km / h, 30 km / h), [35 km / h, 40 km / h), [40 km / h, 45 km / h) and [45 km / h, 120 km / h]; The minimum speed range for the 8th layer is [0, 5 km / h), [5 km / h, 15 km / h), [15 km / h, 20 km / h), [20 km / h, 25 km / h), [25 km / h, 30 km / h), [35 km / h, 40 km / h), [40 km / h, 45 km / h) and [45 km / h, 120 km / h]; The minimum speed range for the 7th layer is [0, 5 km / h), [5 km / h, 20 km / h), [20 km / h, 25 km / h), [25 km / h, 30 km / h), [35 km / h, 40 km / h), [40 km / h, 45 km / h) and [45 km / h, 120 km / h]; The minimum speed range for the 6th layer is [0, 5 km / h), [5 km / h, 20 km / h), [20 km / h, 25 km / h), [25 km / h, 30 km / h), [35 km / h, 40 km / h) and [40 km / h, 120 km / h]; The minimum speed range for the 5th layer is [0, 5 km / h), [5 km / h, 20 km / h), [20 km / h, 25 km / h), [25 km / h, 30 km / h) and [35 km / h, 120 km / h]; The minimum speed range for the fourth layer is [0, 5 km / h), [5 km / h, 25 km / h), [25 km / h, 30 km / h), and [35 km / h, 120 km / h]. The minimum speed range for the third layer is [0, 5 km / h), [5 km / h, 30 km / h), and [35 km / h, 120 km / h]. The minimum speed range for the second layer is [0, 5 km / h) and [5 km / h, 120 km / h]; The minimum speed range for the first layer is: [0, 120 km / h]; S40. Establish a five-level residual neural network for each layer of the hierarchical model to obtain the probability that the minimum speed of the target vehicle is in each interval. S50, the intersection speed guidance system selects a suitable minimum speed prediction range based on a predetermined probability threshold to improve the reliability and accuracy of the intersection speed guidance system.

2. The method for predicting the minimum speed range of vehicles at an intersection according to claim 1, characterized in that, The specific steps of step S10 are as follows: S101. Install cameras and 3D laser sensors on streetlights; S102. Based on traffic light cameras and vehicle recognition algorithms, identify and segment vehicle type, vehicle color, vehicle position, and two-dimensional bounding shape in images; S103. By installing traffic light cameras and near-3D laser sensors, obtain all 3D point clouds of the intersection traffic scene. By using the matching algorithm between the vehicle camera and the 3D laser sensor to obtain the 3D point cloud of each vehicle at the intersection, obtain the vehicle's wheelbase information, speed information, three-dimensional shape information, and distance information from the intersection.

3. The method for predicting the minimum speed range of vehicles at an intersection according to claim 1, characterized in that, The specific steps of step S20 are as follows: S201. Based on the collected data over a period of time and data analysis, obtain the type, color, position, two-dimensional enclosing shape, wheelbase information, speed information, three-dimensional shape information, distance from the intersection, vehicle acceleration information, and deceleration information for each vehicle at each moment at the intersection. S202. Based on the location information of each vehicle at every moment at the intersection, perform parameter estimation for the following behavior model of vehicles of different types, different axle lengths and different queuing positions. S203. Based on the following behavior model of each vehicle after parameter estimation and the intersection scene information, multiple simulations are performed to obtain the dynamic characteristics of the target vehicle's journey to the intersection.

4. The method for predicting the minimum speed range of vehicles at an intersection according to claim 3, characterized in that, In step S202, the parameters of the following behavior model include: maximum acceleration, maximum deceleration, target following distance, time distance to the front of the target vehicle, target stopping distance, maximum driving speed, driver reaction time, and vehicle start-stop time.

5. The method for predicting the minimum driving speed range of vehicles at an intersection according to claim 3, characterized in that, In step S203, the dynamic features include the average, maximum, and minimum values ​​of the speed and time of the target vehicle to the intersection, the 25th percentile, 50th percentile, and 75th percentile values, and the average, maximum, and minimum values ​​of the speed and time of the target vehicle to the intersection, the 25th percentile, 50th percentile, and 75th percentile values.

6. The method for predicting the minimum speed range of vehicles at an intersection according to claim 1, characterized in that, In step S301, the nine intervals of minimum speed are [0, 5 km / h), [5 km / h, 10 km / h), [10 km / h, 15 km / h), [15 km / h, 20 km / h), [20 km / h, 25 km / h), [25 km / h, 30 km / h), [35 km / h, 40 km / h), [40 km / h, 45 km / h), and [45 km / h, 120 km / h].

7. The method for predicting the minimum vehicle speed range at an intersection according to claim 1, characterized in that, The specific steps of step S40 are as follows: S401. Based on the hierarchical structure of the minimum speed range, design and train a 5-level residual neural network for each layer from layer 2 to layer 9. Each level of the network has 128 nodes, and the last level is the output level, which is of type softMax. S402. By collecting data on the status and characteristics of vehicles queuing at intersections, a dataset is established, and eight hierarchical neural networks are trained. S403, based on 8 neural networks, predicts the minimum speed range of a target vehicle during its journey to the intersection and the corresponding probability.

8. The method for predicting the minimum speed range of vehicles at an intersection according to claim 7, characterized in that, In step S402, the neural network corresponding to the 9th layer model is fully trained during the training process, while the neural networks corresponding to the 8th to 2nd layers are trained only for the last layer, and the other layers reuse the neural network corresponding to the 9th layer model.

Images