ETC data-based service area congestion prediction method and information release guidance system

Abstract

The invention provides an ETC data-based service area congestion prediction method and an information issuing guidance system, and the method comprises the steps: carrying out the standardization processing of key features extracted from images captured by cameras at a ramp entrance and an exit, and mapping the key features to a corresponding discrete data range, thereby forming discrete data points; data points are regarded as mass points to move towards a set initial center position under the action of gravitational force to form a circular structure, the radius of a circle is determined by calculating the average distance, the circle is divided according to data point distribution, data point distribution modes in a circular area corresponding to the snapshot data of an entrance and an exit are compared, and a first ID code is generated. According to the method, the ETC portal data and the multi-source information are fused, the congestion trend is accurately predicted, the traffic efficiency and the management level of a service area can be improved, and the traffic congestion condition is relieved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a service area congestion prediction method and an information release induction system based on ETC data. Background Art

[0002] As an important part of the highway, the service area not only provides basic rest services for drivers and passengers, effectively relieves long-distance driving fatigue and improves driving safety, but also undertakes the functions of information release and traffic induction, helping to alleviate road congestion and optimize road network traffic efficiency. It is a key node to ensure the safe and efficient operation of the highway.

[0003] However, at present, the monitoring of vehicle carrying capacity in highway service areas mainly relies on visual solutions, that is, collecting vehicle data through camera capture images. Although this solution is feasible in service areas with complete hardware equipment, its high hardware investment and maintenance costs limit its popularity and it is difficult to be widely used in all service areas. In addition, the visual solution is also limited by weather and lighting environment factors, and the stability and accuracy of data collection are difficult to fully guarantee. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide a service area congestion prediction method and an information release induction system based on ETC data. By integrating ETC gantry data with multi-source information, the congestion trend can be accurately predicted, which helps to improve the traffic efficiency and management level of the service area and alleviate traffic congestion.

[0005] In order to solve the above technical problems, the technical solution of the present invention is as follows:

[0006] In a first aspect, a service area congestion prediction method based on ETC data, the method comprising:

[0007] Standardize the key features extracted from the images captured by the cameras at the entrance and exit of the ramp, and map them to the corresponding discrete data range to form discrete data points. Treat the data points as particles and let them move to the set initial center position under the action of gravity to form a circular structure. Determine the radius of the circle by calculating the average distance and divide the circle according to the distribution of data points. Compare the distribution patterns of data points in the circular area corresponding to the entrance and exit capture data, and generate a first ID code;

[0008] Based on the first ID code, the captured data of the front and rear gantries of the service area are associated to generate a second ID code, and the vehicle information is screened by comparison, and the interval travel speed is calculated according to the gantries' travel time difference and spacing to obtain the predicted vehicles entering the service area;

[0009] The camera captures real-time data of vehicles entering the service area, and dynamically determines the current real-time congestion index of the service area based on the predicted number of vehicles entering the service area;

[0010] The estimated arrival time is calculated based on the remaining distance of the vehicle to the service area entrance, and the flow contribution value of different types of vehicles at the same section of the service area entrance is calculated based on the real-time congestion index;

[0011] Based on the historical traffic behavior data of vehicles, a vehicle portrait library is established, and the traffic behavior of vehicles on the road section is predicted through a neural network model to generate traffic behavior prediction results for future time periods.

[0012] Furthermore, the key features include the exterior color, model, license plate number and vehicle outline of the vehicle; the key features extracted from the images captured by the ramp entrance and exit checkpoint cameras are standardized and mapped to the corresponding discrete data range to form discrete data points, the data points are regarded as particles and moved to the set initial center position under the action of gravity to form a circular structure, the radius of the circle is determined by calculating the average distance and the circle is divided according to the distribution of data points, the distribution pattern of data points in the circular area corresponding to the entrance and exit capture data is compared, and a first ID code is generated, including:

[0013] Each key feature is standardized to obtain standardized key features, including: converting the exterior color into a fixed color classification code, converting the vehicle model into the corresponding vehicle model number, converting the license plate number into a digital code, and converting the vehicle outline into a numerical feature through shape feature extraction;

[0014] Based on the standardized key features, the discrete data range corresponding to each key feature is determined, and the specific value of each key feature is mapped to the corresponding discrete data range to form corresponding discrete data points, so that each entry snapshot data and exit snapshot data can be represented as a set of discrete data points;

[0015] The discrete data points are regarded as particles with gravity, an initial center position is set, and each discrete data point is allowed to move toward the center position under the action of gravity, eventually forming a circular structure;

[0016] The radius of the circle is determined by calculating the average distance from all data points to the center of the circle, and the circle is divided into corresponding parts according to the distribution of data points of different key features;

[0017] The distribution of data points in the circular area corresponding to each entry capture data is compared with the distribution of data points in the circular area corresponding to all exit capture data. If the data point distribution patterns of the two in the divided areas of the circle are consistent, that is, the number and position characteristics of data points in the corresponding areas are similar, then it is determined that the entry and exit capture data are the same vehicle, and the corresponding first ID code is generated.

[0018] Furthermore, based on the first ID code, the captured data of the front and rear gantries of the service area are associated to generate a second ID code, and the vehicle information is screened by comparison, and the interval travel speed is calculated according to the gantries' travel time difference and the distance, so as to obtain the predicted vehicles entering the service area, including:

[0019] The captured data of the front and rear gantries of the service area are correlated and matched, and a second ID code is generated according to the captured time, location and feature information of the vehicle at the front and rear gantries of the service area to identify the vehicle passing through the front and rear gantries of the service area;

[0020] Compare the first ID code with the second ID code to obtain filtered target vehicle information;

[0021] Based on the filtered target vehicle information, the travel time difference and the distance between the gantries at the front and rear of the vehicle in the service area are calculated; the interval travel speed of the vehicle is calculated based on the travel time difference and the distance between the gantries;

[0022] Based on the vehicle's interval travel speed and preset threshold, predict the list of vehicles entering the service area.

[0023] Furthermore, the real-time data of vehicles entering the service area is captured by the bayonet camera, and the current real-time congestion index of the service area is dynamically determined based on the predicted number of vehicles entering the service area, including:

[0024] According to the vehicle data captured by the bayonet camera, the entrance capture data and exit capture data are recorded;

[0025] Calculate the number of vehicles entering and leaving the service area per unit time;

[0026] Calculate the number of vehicles in the service area, and determine the real-time congestion index of the service area based on the number of vehicles in the area and the carrying capacity of the service area.

[0027] Furthermore, the estimated arrival time is calculated based on the remaining distance of the vehicle to the service area entrance, and the flow contribution values ​​of different types of vehicles at the same section of the service area entrance are calculated based on the real-time congestion index, including:

[0028] Calculate the time when the vehicle arrives at the entrance of the service area based on the vehicle's journey from the node to the entrance of the service area;

[0029] Get all different types of vehicles at the same section at the entrance of the service area;

[0030] Obtain the equivalent number and cross-sectional time interval of each type of vehicle relative to the standard vehicle in the flow;

[0031] Combine the equivalent number of each type of vehicle relative to the standard vehicle in the traffic flow with the time interval of the section to obtain the contribution of each type of vehicle to the traffic flow relative to the standard vehicle in unit time, where the contribution represents the vehicle density;

[0032] According to historical actual data, the contribution of each type of vehicle to traffic flow relative to the standard vehicle is adjusted to obtain the traffic contribution value;

[0033] Add up the flow contribution values ​​of each type of vehicle to obtain the total flow of all vehicles arriving at the same section at the service area entrance at the same time.

[0034] Furthermore, based on the historical traffic behavior data of vehicles, a vehicle portrait library is established, and the traffic behavior of vehicles on the road section is predicted through a neural network model to generate traffic behavior prediction results for future time periods, including:

[0035] Integrate the basic information and traffic records of vehicles into static and dynamic databases;

[0036] Build a portrait tag based on the vehicle's historical traffic behavior, and dynamically update the vehicle portrait library when new traffic data is received;

[0037] Extract key features from vehicle portraits and determine the weight of different features on traffic behavior;

[0038] When the vehicle enters the highway again, the service area it passes through is predicted based on historical data, and the traffic behavior prediction results are output.

[0039] Furthermore, key features are extracted from the vehicle portrait to determine the weights of different features on traffic behavior, including:

[0040] Extract key features from vehicle portraits, including vehicle type, vehicle size, vehicle weight, driving speed, mileage, driver behavior patterns, time period, weather conditions, and historical traffic records;

[0041] For key features, quantification is performed to obtain numerical features;

[0042] Construct a random forest based on the extracted and quantified features. For each tree in the random forest, traverse each node in the tree. For each node, determine the number of samples contained in the node and calculate the Gini impurity of the node.

[0043] Split the features, divide the samples in the node into left child nodes and right child nodes, and calculate the Gini impurity of the left child nodes and the right child nodes respectively;

[0044] Calculate the reduction of Gini impurity based on the Gini impurity of the node and the Gini impurity of the left and right child nodes;

[0045] For each feature, the sum of the reduction in Gini impurity brought by splitting in all trees is counted to get the importance score of the feature;

[0046] The importance scores of each feature are averaged to obtain the final importance score of the feature in the random forest;

[0047] According to the final importance score of the feature in the random forest, the influence weight of different features on the traffic behavior is determined.

[0048] Secondly, the information release and guidance system for service area congestion based on ETC data includes:

[0049] The checkpoint information module is used to standardize the key features extracted from the images captured by the checkpoint cameras at the entrance and exit of the ramp, and map them to the corresponding discrete data range to form discrete data points. The data points are regarded as particles and moved to the set initial center position under the action of gravity to form a circular structure. The radius of the circle is determined by calculating the average distance and the circle is divided according to the distribution of data points. The distribution pattern of data points in the circular area corresponding to the entrance and exit capture data is compared, and a first ID code is generated; based on the first ID code, the capture data of the front and rear gantries of the service area are associated to generate a second ID code, and the vehicle information is screened by comparison, and the interval travel speed is calculated according to the travel time difference and the distance between the gantries to obtain the predicted vehicles entering the service area;

[0050] The multi-source identification module is used to capture real-time data of vehicles entering the service area through the bayonet camera, and dynamically determine the current real-time congestion index of the service area based on the predicted vehicles entering the service area; calculate the estimated arrival time based on the remaining distance of the vehicle to the service area entrance, and calculate the flow contribution value of different types of vehicles at the same section of the service area entrance based on the real-time congestion index;

[0051] The vehicle trajectory prediction module is used to establish a vehicle portrait library based on the vehicle's historical traffic behavior data, and predict the vehicle's traffic behavior on the road section through a neural network model to generate traffic behavior prediction results for future time periods.

[0052] According to a third aspect, a computing device includes:

[0053] one or more processors;

[0054] The storage device is used to store one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors implement the method described.

[0055] In a fourth aspect, a computer-readable storage medium stores a program, and when the program is executed by a processor, the method described is implemented.

[0056] The above solution of the present invention includes at least the following beneficial effects:

[0057] By standardizing the key features extracted from the images captured by the ramp entrance and exit cameras and mapping them to a discrete data range to form data points, and using gravity to form a circular structure for comparison, it is possible to more accurately identify and track vehicles and generate a unique first ID code, effectively solving the errors and missed detection problems in traditional vehicle identification methods. Based on the first ID code, the captured data of the front and rear gantries of the service area are associated to generate a second ID code. By comparing and filtering vehicle information, and combining the time difference and spacing between the gantries to calculate the interval speed, it is possible to accurately predict the vehicles entering the service area. The real-time data of vehicles entering the service area is captured by the bayonet camera, and combined with the predicted information of vehicles entering the service area, the current real-time congestion index of the service area can be dynamically determined, which helps to take effective diversion measures. The estimated arrival time is calculated based on the remaining distance of the vehicle to the service area entrance, and the flow contribution value of different types of vehicles in the same section of the service area entrance is calculated in combination with the real-time congestion index. This can more scientifically evaluate the impact of different vehicles on traffic flow, establish a vehicle portrait library based on the vehicle's historical traffic behavior data, and predict the vehicle's traffic behavior on the road section through a neural network model to generate traffic behavior prediction results for future time periods, which will help plan traffic flow in advance and reduce congestion and traffic accidents.

[0058] The invention can accurately predict the congestion of service areas in real time, provide scientific decision-making support for service area managers, and help improve the management efficiency and service quality of service areas. By calculating the flow contribution value and predicting traffic behavior, the invention can optimize the distribution of traffic flow, reduce congestion during peak hours, and improve road traffic capacity. The provision of real-time congestion index and estimated arrival time helps drivers plan their trips in advance, avoid congested periods and sections, and enhance driving experience. By reducing the occurrence of congestion and traffic accidents, it can reduce the economic losses and social costs caused by traffic congestion and improve the utilization efficiency of road resources. Accurate traffic behavior prediction and congestion warning can help improve the level of public safety, reduce secondary disasters and casualties caused by congestion and traffic accidents, and help optimize traffic flow distribution, reduce energy consumption and environmental pollution caused by vehicle idling and frequent start-stop, and promote sustainable development in the transportation field. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 It is a flowchart of a service area congestion prediction method based on ETC data provided by an embodiment of the present invention.

[0060] Figure 2 It is a schematic diagram of a service area congestion information release and guidance system based on ETC data provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0061] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in a form and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0062] like Figure 1 As shown, an embodiment of the present invention proposes a service area congestion prediction method based on ETC data, and the method includes the following steps:

[0063] Step 11, standardize the key features extracted from the images captured by the ramp entrance and exit checkpoint cameras, and map them to the corresponding discrete data range to form discrete data points, regard the data points as particles and let them move to the set initial center position under the action of gravity to form a circular structure, determine the radius of the circle by calculating the average distance and divide the circle according to the distribution of data points, compare the distribution patterns of data points in the circular area corresponding to the entrance and exit capture data, and generate a first ID code;

[0064] Step 12: Based on the first ID code, the captured data of the front and rear gantries of the service area are associated to generate a second ID code, and the vehicle information is screened by comparison, and the interval travel speed is calculated according to the gantries' travel time difference and spacing to obtain the predicted vehicles entering the service area;

[0065] Step 13, using the camera to capture and collect real-time data of vehicles entering the service area, and dynamically determine the current real-time congestion index of the service area based on the predicted number of vehicles entering the service area;

[0066] Step 14, calculate the estimated arrival time according to the remaining distance of the vehicle to the service area entrance, and calculate the flow contribution value of different types of vehicles in the same section of the service area entrance according to the real-time congestion index;

[0067] Step 15, based on the historical traffic behavior data of vehicles, a vehicle portrait library is established, and the traffic behavior of vehicles on the road section is predicted through a neural network model to generate traffic behavior prediction results for future time periods.

[0068] In an embodiment of the present invention, the key features extracted from the images captured by the ramp entrance and exit checkpoint cameras are standardized and mapped to a discrete data range to form discrete data points, which helps to eliminate the dimensional differences between the data and improve the accuracy and efficiency of data processing. The data points are regarded as particles and form a circular structure, which further simplifies the data representation. By comparing the distribution pattern of the data points in the circular area corresponding to the entrance and exit capture data, a first ID code is generated, and accurate identification and tracking of the vehicle is achieved. The capture data of the front and rear gantries of the service area are associated with the first ID code to generate a second ID code, and the vehicle information is filtered by comparison. The interval travel speed is calculated based on the time difference and spacing between the gantries, which helps to understand the driving status of the vehicle on the road section. The real-time data of the vehicle entering the service area is captured by the checkpoint camera, and the current real-time congestion index of the service area is determined based on the predicted dynamics of the vehicle entering the service area, which realizes real-time monitoring and early warning of the congestion situation in the service area. Calculating the estimated arrival time based on the remaining distance of the vehicle to the service area entrance helps to understand the specific time when the vehicle arrives at the service area. Calculating the flow contribution values ​​for different types of vehicles at the same section of the service area entrance based on the real-time congestion index helps to understand the impact of different types of vehicles on the service area flow. Establishing a vehicle portrait library based on the vehicle's historical traffic behavior data helps to understand the vehicle's traffic habits and preferences.

[0069] In a preferred embodiment of the present invention, the above step 11 may include:

[0070] Step 111, performing standardization processing on each key feature to obtain standardized key features, specifically including: converting the appearance color into a fixed color classification code, converting the vehicle model into the corresponding vehicle model number, converting the license plate number into a digital code, and converting the vehicle outline into a numerical feature through shape feature extraction;

[0071] Step 112, based on the standardized key features, determine the discrete data range corresponding to each key feature, map the specific value of each key feature to the corresponding discrete data range, and form corresponding discrete data points, so that each entry capture data and exit capture data can be represented as a set of discrete data points;

[0072] Step 113, treating the discrete data points as mass points with gravity, setting an initial center position, and allowing each discrete data point to move toward the center position under the action of gravity, and finally forming a circular structure;

[0073] Step 114, determining the radius of the circle by calculating the average distance from all data points to the center of the circle, and dividing the circle into corresponding parts according to the distribution of data points of different key features;

[0074] Step 115, compare the distribution of data points in the circular area corresponding to each entry snapshot data with the distribution of data points in the circular area corresponding to all exit snapshot data. If the data point distribution patterns of the two in the divided areas of the circle are consistent, that is, the number and position characteristics of data points in the corresponding areas are similar, then it is determined that the entry and exit snapshot data are the same vehicle, and a corresponding first ID code is generated.

[0075] In an embodiment of the present invention, key features such as appearance color, vehicle model, license plate number and vehicle outline are converted into a unified code or numerical form, eliminating the differences between different data sources or formats, improving the consistency and comparability of data, and making standardized data easier to be processed and analyzed by computer programs, thereby improving the efficiency and accuracy of data processing. Continuous or complex feature values ​​are mapped to discrete values ​​or categories, simplifying the representation of data, and representing data points as circular structures, providing an intuitive data visualization method, which is helpful for observing and analyzing the distribution and characteristics of data. By calculating the radius of the circle, the distribution range of the data points can be quantified, and by comparing the distribution pattern of the data points in the circular area of ​​the entrance and exit capture data, it is possible to accurately identify whether it is the same vehicle, thereby improving the accuracy and reliability of vehicle identification. A corresponding first ID code is generated for the data determined to be the same vehicle, providing a unique identification for the vehicle, and by accurately identifying the identity of the vehicle, it is possible to more effectively perform traffic flow statistics, illegal behavior monitoring and traffic incident processing, thereby improving the efficiency and level of traffic management.

[0076] In the embodiment of the present invention, the specific steps include:

[0077] Step 111, convert the exterior color of the vehicle into a fixed color classification code. For example, red may correspond to code "R", blue to code "B", and so on. Convert the vehicle model to the corresponding model number. For example, a sedan may correspond to number "1", an SUV to number "2", a truck to number "3", etc. This numbering system helps to unify the representation of vehicle model information. Convert the license plate number into a digital code. This can be achieved by mapping each character in the license plate number to a number, such as mapping letters AZ to 1-26, and numbers 0-9 remain unchanged, and then combining these numbers into a long integer or encoded string. By identifying the key points or features of the vehicle outline, shape feature extraction is performed to convert the vehicle outline into a numerical feature.

[0078] Step 112, for each standardized key feature, determine the value range, and divide this range into multiple discrete intervals or categories. For example, for the vehicle model number, the value range is 1 to 10, which can be divided into 10 discrete categories, each category corresponds to a vehicle model number, and the specific value of each key feature is mapped to its corresponding discrete data range to form a discrete data point. For example, if the vehicle model number of a car is "3", it is mapped to the category corresponding to the discrete data point "3". For each entry capture data and exit capture data, the values ​​of all its key features are mapped to the corresponding discrete data points to form a set of discrete data points.

[0079] Step 113, each discrete data point is regarded as a mass point with gravity. A mass point is a point without size and shape, but only with mass. Here, the mass point represents an abstract representation of vehicle feature information. An initial center position is set. The center position is the geometric center of the data point set or a position determined according to a certain rule. Under the action of gravity, each discrete data point is moved toward the center position until the data points form a stable circular structure.

[0080] Step 114, by calculating the average distance from all data points to the center of the circle, the radius of the circle is determined, and the radius represents the distribution range of the data point set. According to the distribution of data points of different key features, the circle is divided into corresponding parts. For example, if a key feature has 10 possible values, the circle can be divided into 10 sectors, each sector corresponding to a value.

[0081] Step 115, the distribution of data points in the circular area corresponding to each entry snapshot data is compared with the distribution of data points in the circular area corresponding to all exit snapshot data. If the data point distribution patterns in the divided areas of the circles are consistent, that is, the number and position characteristics of the data points in the corresponding areas are similar, then it is determined that the entry and exit snapshot data are the same vehicle. For the data determined to be the same vehicle, a corresponding first ID code is generated.

[0082] In a preferred embodiment of the present invention, the above step 12 may include:

[0083] Step 121, the captured data of the front and rear gantries of the service area are correlated and matched, and a second ID code is generated according to the captured time, location and feature information of the front and rear gantries of the vehicle in the service area to identify the vehicle passing through the front and rear gantries of the service area;

[0084] Step 122, comparing the first ID code with the second ID code to obtain filtered target vehicle information;

[0085] Step 123, based on the filtered target vehicle information, calculate the travel time difference and the mast distance between the front and rear masts of the vehicle in the service area; calculate the interval travel speed of the vehicle according to the travel time difference and the mast distance;

[0086] Step 124, predicting a list of vehicles entering the service area based on the vehicle's interval travel speed and a preset threshold.

[0087] In an embodiment of the present invention, by associating and matching the entrance snapshot data with the exit snapshot data to generate a first ID code, and associating and matching the front gantry snapshot data with the rear gantry snapshot data of the service area to generate a second ID code, vehicles passing through the ramp checkpoint and the front and rear gantries of the service area can be accurately tracked and identified. By comparing the first ID code with the second ID code, vehicle information that is not recorded in the ramp checkpoint snapshot data but appears in the front and rear gantry snapshot data of the service area can be screened out, indicating abnormal passage behavior. By calculating the passage time difference and the gantry spacing of the vehicle at the front and rear gantries of the service area, the vehicle's interval passage speed can be accurately obtained, and then whether the vehicle enters the service area can be predicted, which will help the service area management department to better understand the traffic conditions of the service area, optimize the resource allocation of the service area, and improve the overall management efficiency and service quality of the service area.

[0088] In the embodiment of the present invention, the specific steps include:

[0089] Step 121, pre-process the collected image data, and use license plate recognition technology to extract the vehicle license plate number from the entrance and exit snapshot images as a key feature. The entrance capture data is associated and matched with the exit capture data to generate a first ID code to identify the vehicle passing through the ramp checkpoint, wherein: is the association matching function, It is the entrance capture data. It is the exit capture data, which is matched according to the vehicle feature information to generate a unique identification code; the association matching function will extract the vehicle feature information in E and X, where the association matching function is a feature vector similarity calculation formula. Through the feature vector similarity calculation, comparison and analysis are performed to find the capture records of the same vehicle at the entrance and exit, and generate a unique first ID code to achieve the association matching of the license plate number in the entrance capture image with the license plate number in the exit capture image. For successfully matched vehicles, a unique first ID code is generated to identify the vehicle passing through the ramp checkpoint, and the information is associated and stored.

[0090] Step 122, collect vehicle image data from the capture system of the front and rear door frames of the service area, pre-process the collected image data to ensure that the image is clear and readable, and also use the license plate recognition technology to extract the vehicle license plate number from the captured images of the front and rear door frames. The captured data of the front gantry of the service area is associated and matched with the captured data of the rear gantry to generate a second ID code to identify the vehicles passing through the front and rear gantry of the service area, wherein: It is the data captured by the front door frame. It is the data captured by the rear door frame. Represents the association matching function; the association matching function extracts the vehicle feature information in A and B, compares and analyzes it through feature vector similarity calculation, finds the capture records of the same vehicle at the front and rear door frames, and generates a unique second ID code to associate and match the license plate number in the front door frame capture image with the license plate number in the rear door frame capture image. For successfully matched vehicles, a unique second ID code is generated to identify the vehicles passing through the front and rear gantries of the service area and store the information in an associated manner.

[0091] Step 123, pass The first ID code is compared with the second ID code to filter out vehicle information that is not recorded in the ramp checkpoint snapshot data but appears in the service area front and rear gantry snapshot data, where: It represents a filtered vehicle information set; the first ID code is associated and compared with the second ID code to find the vehicle information that does not exist in the first ID code list but appears in the second ID code list.

[0092] Step 124, for each vehicle in the second ID code list, calculate the travel time difference between the front and rear gantries in the service area, and use the travel time difference and the gantries spacing to calculate the Calculate the interval speed of the vehicle, set the threshold, and predict whether the vehicle enters the service area based on the threshold. List the vehicles, where: is the interval speed, is the travel time difference between the front and rear gantries, is the mast spacing.

[0093] In a preferred embodiment of the present invention, the above step 13 may include:

[0094] Step 131, recording the entrance capture data and the exit capture data according to the vehicle data captured by the bayonet camera;

[0095] Step 132, calculating the number of vehicles entering the service area and the number of vehicles leaving the service area per unit time;

[0096] Step 133, calculating the number of vehicles in the service area, and determining the real-time congestion index of the service area according to the number of vehicles in the service area and the carrying capacity of the service area.

[0097] In the embodiment of the present invention, by capturing vehicle data with a bayonet camera and recording the captured data at the entrance and exit in detail, it is possible to ensure that each vehicle entering and leaving the service area is accurately tracked and recorded. By counting the number of vehicles entering and leaving the service area per unit time, the traffic flow of the service area can be grasped in real time, which helps to timely discover changes in traffic flow and provide timely and effective information support for traffic management. By calculating the number of vehicles in the service area, the current vehicle occupancy of the service area can be accurately understood, providing a scientific basis for judging the congestion status of the service area.

[0098] In the embodiment of the present invention, the specific steps include:

[0099] Step 131, sorting the captured vehicle data, classifying and storing them according to the entrance and exit, and adding a unique identifier to each record for subsequent data association and query.

[0100] Step 132, pass Count the number of vehicles leaving the service area per unit time, through Count the number of vehicles leaving the service area per unit time, among which, At a certain moment in a unit of time , capture data through the entrance The number of vehicles leaving the service area is counted. It is the entrance capture data. At a certain moment in a unit of time , capture data through the export The number of vehicles leaving the service area is counted. It is the export snapshot data. is the index, is the index, is the number of vehicles leaving the service area per unit time, It is the number of vehicles leaving the service area per unit time.

[0101] According to the above formula, the number of vehicles entering and leaving the service area per unit time is counted. In each time window, the vehicle records entering the service area during this time period are filtered out from the entrance snapshot data, and the number is counted. Similarly, the vehicle records leaving the service area during this time period are filtered out from the exit snapshot data, and the number is counted.

[0102] Step 133, based on the number of vehicles entering and leaving the service area in each time window, the number of vehicles in the service area at the end of the time period is calculated according to the above formula, and the carrying capacity of the service area (such as the maximum number of parking spaces, road capacity) is preset, and the calculated number of vehicles in the service area is compared with the carrying capacity of the service area. If the number of vehicles in the service area is close to or exceeds the carrying capacity of the service area, the service area is judged to be in a congested state, and the congestion state is divided into the following three categories: ,in, It is the real-time congestion index of the service area.

[0103] In a preferred embodiment of the present invention, the above step 14 may include:

[0104] The estimated time of arrival is calculated based on the remaining distance of the vehicle to the entrance of the service area;

[0105] According to the real-time congestion index, different types of vehicles passing through the same section of the service area entrance are Calculate the traffic contribution value, where: is the total amount, are all vehicle categories, It is The converted traffic volume of the vehicle type, It is The time interval between vehicles passing through the section, is the weight factor, is the flow adjustment factor, is the exponential decay factor, is the decay constant, is the index.

[0106] In an embodiment of the present invention, by accurately calculating the estimated arrival time of the vehicle, the service area can make reception preparations in advance, such as arranging parking spaces, preparing catering services, etc., thereby improving service efficiency and reducing vehicle waiting time. By calculating the flow contribution values ​​of different types of vehicles, the composition and changing trends of traffic flow can be more accurately understood, and decision support can be provided to the traffic management department, such as adjusting the duration of signal lights and optimizing lane settings. By analyzing the congestion index and flow contribution value in real time, the service area can timely discover the bottleneck points of traffic congestion and take corresponding measures to guide traffic, such as adding temporary lanes and guiding vehicles to detour, thereby effectively alleviating traffic congestion. Taking into account various factors such as the remaining travel of the vehicle, the real-time congestion index, and the vehicle category, the analysis results are more comprehensive and accurate. By introducing parameters such as weight factors, flow adjustment coefficients, and exponential decay factors, the calculation model can be flexibly adjusted according to actual conditions to adapt to different traffic scenarios and needs.

[0107] In a preferred embodiment of the present invention, the above step 15 may include:

[0108] Step 151, integrating the basic information and traffic records of the vehicle into the static database and the dynamic database;

[0109] Step 152, constructing a portrait tag based on the historical traffic behavior of the vehicle, and dynamically updating the vehicle portrait library when receiving new traffic data;

[0110] Step 153, extracting key features from the vehicle portrait and determining the influence weights of different features on the traffic behavior;

[0111] Step 154, when the vehicle enters the expressway again, the service area it passes through is predicted based on historical data, and the traffic behavior prediction result is output.

[0112] In an embodiment of the present invention, the basic information of the vehicle and the traffic record are integrated into a static database and a dynamic database, and the basic information of the vehicle and the traffic record are integrated into the static and dynamic databases, thereby realizing the centralized management of data. The division of the static database and the dynamic database makes the storage and query of data more efficient. The static database stores the basic information of the vehicle, and the dynamic database stores the traffic record of the vehicle. The combination of the two can quickly obtain the historical traffic information of the vehicle. The integrated data supports multi-dimensional analysis, such as vehicle type, travel time, and travel section. By constructing portrait tags, the characteristics of the vehicle can be accurately portrayed, such as the commonly used sections, travel time preferences, and the habit of staying in service areas. When new traffic data is received, the vehicle portrait library is dynamically updated to ensure the real-time and accuracy of the vehicle portrait, making the prediction results more reliable. The construction and update of vehicle portraits help to provide personalized services for vehicles, such as customized traffic suggestions and preferential information, to enhance user experience. Extracting key features from vehicle portraits helps to uncover key factors that affect vehicle traffic behavior. Prediction models are trained using historical data and the weights of different features on traffic behavior are determined. By predicting the service areas that vehicles may pass through, routes can be planned in advance to avoid congestion and delays and improve traffic efficiency. Based on the prediction results, personalized service recommendations can be provided to vehicles, such as recommending service areas and providing discount information to improve user experience and satisfaction.

[0113] In the embodiment of the present invention, the specific steps include:

[0114] Step 151, collect basic information and traffic records of vehicles. Basic vehicle information includes but is not limited to license plate number, vehicle model, vehicle type, and owner information; traffic records include the time, location, speed, and travel direction of each time the vehicle passes through the ETC gantry or bayonet camera. The collected basic vehicle information and traffic records are integrated. For the static database, the basic attributes of the vehicle are mainly stored. For the dynamic database, the traffic conditions of the vehicle are recorded in real time, including the specific information of each time it passes through the ETC gantry or bayonet camera. By establishing a unique identifier for each vehicle, such as the license plate number, the information in the static database and the dynamic database are associated, and the integrated data is stored in the static database and the dynamic database respectively.

[0115] Step 152: Based on the historical traffic behavior of the vehicle, a portrait label is constructed. The portrait label is an abstract description of the vehicle's traffic behavior, including the vehicle's driving habits, frequented locations, and vehicle type preferences. A unique portrait label is constructed for each vehicle. When the system receives new traffic data, the vehicle portrait library is dynamically updated in a timely manner.

[0116] Step 153: extract key features from the vehicle portrait, which can reflect the traffic behavior and preference of the vehicle. Through random forest, determine the influence weight of different features on the traffic behavior.

[0117] Step 154, when the vehicle enters the highway again, the system identifies the vehicle's identity information through the ETC gantry or bayonet camera, and compares this information with the records in the vehicle portrait library to determine the vehicle's historical traffic behavior and portrait label. Based on the vehicle's historical traffic behavior and portrait label, the trained prediction model is used to calculate the service area that the vehicle may pass through. During the prediction process, the system will comprehensively consider factors such as the current traffic conditions and the facilities in the service area to improve the accuracy and practicality of the prediction. The prediction results are output to the driver or the traffic management department. The prediction results may include the name of the service area that the vehicle may pass through, the arrival time, and the driving route information. With this information, the driver can plan the driving route in advance to avoid congestion and delays; the traffic management department can also optimize the resource allocation and management strategy of the service area based on the prediction results.

[0118] In a preferred embodiment of the present invention, the above step 153 may include:

[0119] Step 1531, extracting key features from the vehicle portrait, including vehicle type, vehicle size, vehicle weight, driving speed, driving mileage, driver behavior pattern, time period, weather conditions, and historical traffic records;

[0120] Step 1532, quantizing the key features to obtain numerical features;

[0121] Step 1533, construct a random forest based on the extracted and quantified features. For each tree in the random forest, traverse each node in the tree. If the node is split by the feature, Calculate the reduction in the Gini impurity of the node, where is the number of categories, is a node, Is a node Belongs to category The sample proportion of It is a feature, It is a feature When splitting, the node The reduction in Gini impurity is Is a node The number of samples in and is the number of samples in the left and right child nodes after the split, and Is the left child node and the right child node belonging to the category after the split The sample proportion of is the index;

[0122] Step 1534, for each feature, the sum of the reductions in Gini impurity brought about by splitting in all trees is counted to obtain the importance score of the feature;

[0123] Step 1535, averaging the importance score of each feature to obtain the final importance score of the feature in the random forest;

[0124] Step 1536, based on the final importance score of the feature in the random forest, determine the influence weight of different features on the traffic behavior.

[0125] In the embodiment of the present invention, by extracting key features such as vehicle type, vehicle size, vehicle weight, driving speed, mileage, driver behavior pattern, time period, weather conditions and historical traffic records, the vehicle status and driving environment can be fully reflected. These features cover multiple aspects such as vehicle attributes, driving status, external environment and historical behavior, which helps to analyze vehicle traffic behavior more accurately. Non-numerical features are converted into numerical features so that all features can be input into the model in a unified form. Quantification processing helps to eliminate the dimensional differences between features and improve the convergence speed and prediction accuracy of the model. Random forest can effectively process high-dimensional data. By calculating the Gini impurity reduction of the node, the contribution of each feature to the model prediction ability can be evaluated. By counting the sum of the Gini impurity reduction brought by each feature in all trees through splitting and calculating the average value, the key features that have the greatest impact on traffic behavior can be identified. By determining the influence weights of different features on traffic behavior, the specific influence of each feature on vehicle traffic behavior can be quantified.

[0126] In the embodiment of the present invention, the specific steps include:

[0127] Step 1531, collect detailed information of the vehicle from the traffic management system, ETC system or other data sources, including vehicle type, vehicle size, vehicle weight, driving speed, mileage, driver behavior pattern), time period, weather conditions and historical traffic records, and extract the key features listed above from the collected vehicle portrait data.

[0128] Step 1532: For non-numerical features, quantization processing is required. For example, the vehicle type can be encoded as a numerical value, such as sedan = 1, SUV = 2; the weather conditions can be encoded as numerical values, such as sunny = 1, rainy = 2, snowy = 3; for some features that can be measured by numerical values, their numerical form is directly retained and the data is normalized.

[0129] Step 1533, using the quantized features to construct a random forest, which consists of multiple decision trees, each of which is constructed by randomly extracting samples and features from the original data set. For each tree in the random forest, traverse all its nodes, and for each node, calculate its Gini impurity. Gini impurity is an indicator to measure the uneven distribution of sample categories in a node. For each node, try to use each feature to split and divide the samples in the node into a left child node and a right child node. Calculate the Gini impurity of the left child node and the right child node respectively, and calculate the reduction of Gini impurity by the above formula based on the Gini impurity of the node and the Gini impurity of the left child node and the right child node.

[0130] Step 1534: For each feature, the sum of the reductions in Gini impurity brought about by splitting in all trees is counted. This sum reflects the overall contribution of the feature to the classification task in the random forest. The sum of the reductions in Gini impurity is used as the importance score of the feature.

[0131] Step 1535, average the importance scores of each feature to obtain the final importance score of the feature in the random forest. This score reflects the average importance of the feature in the random forest.

[0132] Step 1536, based on the final importance score of the feature in the random forest, the importance score is normalized so that the sum of the weights of all features is 1, and the influence weights of different features on the traffic behavior are determined by dividing the importance score of each feature by the sum of the importance scores of all features, wherein the higher the weight, the greater the influence of the feature on the traffic behavior.

[0133] like Figure 2 As shown, the embodiment of the present invention also provides an information release guidance system 20 for service area congestion based on ETC data, including:

[0134] The checkpoint information module 21 is used to standardize the key features extracted from the images captured by the checkpoint cameras at the entrance and exit of the ramp, and map them to the corresponding discrete data range to form discrete data points. The data points are regarded as particles and moved to the set initial center position under the action of gravity to form a circular structure. The radius of the circle is determined by calculating the average distance and the circle is divided according to the distribution of data points. The distribution pattern of data points in the circular area corresponding to the entrance and exit capture data is compared, and a first ID code is generated; based on the first ID code, the capture data of the front and rear gantries of the service area are associated to generate a second ID code, and the vehicle information is screened by comparison, and the interval travel speed is calculated according to the gantry travel time difference and the spacing to obtain the predicted vehicle entering the service area.

[0135] The multi-source identification module 22 is used to capture real-time data of vehicles entering the service area through the bayonet camera, and dynamically determine the current real-time congestion index of the service area based on the predicted vehicles entering the service area; calculate the estimated arrival time based on the remaining distance of the vehicle to the service area entrance, and calculate the flow contribution value of different types of vehicles at the same section of the service area entrance based on the real-time congestion index.

[0136] The vehicle trajectory prediction module 23 is used to establish a vehicle profile library based on the historical vehicle traffic behavior data, and predict the vehicle's traffic behavior on the road section through a neural network model to generate traffic behavior prediction results for future time periods.

[0137] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A service area congestion prediction method based on ETC data, characterized in that: The method comprises: Standardize the key features extracted from the images captured by the cameras at the entrance and exit of the ramp, and map them to the corresponding discrete data range to form discrete data points. Treat the data points as particles and let them move to the set initial center position under the action of gravity to form a circular structure. Determine the radius of the circle by calculating the average distance and divide the circle according to the distribution of data points. Compare the distribution patterns of data points in the circular area corresponding to the entrance and exit capture data, and generate a first ID code; Based on the first ID code, the captured data of the front and rear gantries of the service area are associated to generate a second ID code, and the vehicle information is screened by comparison, and the interval travel speed is calculated according to the gantries' travel time difference and spacing to obtain the predicted vehicles entering the service area; The camera captures real-time data of vehicles entering the service area, and dynamically determines the current real-time congestion index of the service area based on the predicted number of vehicles entering the service area; The estimated arrival time is calculated based on the remaining distance of the vehicle to the service area entrance, and the flow contribution value of different types of vehicles at the same section of the service area entrance is calculated based on the real-time congestion index; Based on the historical traffic behavior data of vehicles, a vehicle portrait library is established, and the traffic behavior of vehicles on the road section is predicted through a neural network model to generate traffic behavior prediction results for future time periods.

2. The service area congestion prediction method based on ETC data according to claim 1, characterized in that: The key features include the vehicle's exterior color, model, license plate number and vehicle outline; the key features extracted from the images captured by the ramp entrance and exit checkpoint cameras are standardized and mapped to the corresponding discrete data range to form discrete data points. The data points are regarded as particles and moved to the set initial center position under the action of gravity to form a circular structure. The radius of the circle is determined by calculating the average distance and the circle is divided according to the distribution of data points. The distribution pattern of data points in the circular area corresponding to the entrance and exit capture data is compared, and the first ID code is generated, including: Each key feature is standardized to obtain standardized key features, including: converting the exterior color into a fixed color classification code, converting the vehicle model into the corresponding vehicle model number, converting the license plate number into a digital code, and converting the vehicle outline into a numerical feature through shape feature extraction; Based on the standardized key features, the discrete data range corresponding to each key feature is determined, and the specific value of each key feature is mapped to the corresponding discrete data range to form corresponding discrete data points, so that each entry snapshot data and exit snapshot data can be represented as a set of discrete data points; The discrete data points are regarded as particles with gravity, an initial center position is set, and each discrete data point is allowed to move toward the center position under the action of gravity, eventually forming a circular structure; The radius of the circle is determined by calculating the average distance from all data points to the center of the circle, and the circle is divided into corresponding parts according to the distribution of data points of different key features; The distribution of data points in the circular area corresponding to each entry capture data is compared with the distribution of data points in the circular area corresponding to all exit capture data. If the data point distribution patterns of the two in the divided areas of the circle are consistent, that is, the number and position characteristics of data points in the corresponding areas are similar, then it is determined that the entry and exit capture data are the same vehicle, and the corresponding first ID code is generated.

3. The service area congestion prediction method based on ETC data according to claim 2, characterized in that: Based on the first ID code, the captured data of the front and rear gantries of the service area are associated to generate a second ID code, and the vehicle information is screened by comparison. The interval travel speed is calculated according to the gantries' travel time difference and spacing to obtain the predicted vehicles entering the service area, including: The captured data of the front and rear gantries of the service area are correlated and matched, and a second ID code is generated according to the captured time, location and feature information of the vehicle at the front and rear gantries of the service area to identify the vehicle passing through the front and rear gantries of the service area; Compare the first ID code with the second ID code to obtain filtered target vehicle information; Based on the filtered target vehicle information, the travel time difference and the distance between the gantries at the front and rear of the vehicle in the service area are calculated; the interval travel speed of the vehicle is calculated based on the travel time difference and the distance between the gantries; Based on the vehicle's interval travel speed and preset threshold, predict the list of vehicles entering the service area.

4. The service area congestion prediction method based on ETC data according to claim 3 is characterized in that: The camera captures and collects real-time data of vehicles entering the service area, and dynamically determines the current real-time congestion index of the service area based on the predicted number of vehicles entering the service area, including: According to the vehicle data captured by the bayonet camera, the entrance capture data and exit capture data are recorded; Calculate the number of vehicles entering and leaving the service area per unit time; Calculate the number of vehicles in the service area, and determine the real-time congestion index of the service area based on the number of vehicles in the area and the carrying capacity of the service area.

5. The service area congestion prediction method based on ETC data according to claim 4, characterized in that: The estimated arrival time is calculated based on the remaining distance of the vehicle to the service area entrance, and the traffic contribution values ​​of different types of vehicles at the same section of the service area entrance are calculated based on the real-time congestion index, including: Calculate the time when the vehicle arrives at the entrance of the service area based on the vehicle's journey from the node to the entrance of the service area; Get all different types of vehicles at the same section at the entrance of the service area; Obtain the equivalent number and cross-sectional time interval of each type of vehicle relative to the standard vehicle in the flow; Combine the equivalent number of each type of vehicle relative to the standard vehicle in the traffic flow with the time interval of the section to obtain the contribution of each type of vehicle to the traffic flow relative to the standard vehicle in unit time, where the contribution represents the vehicle density; According to historical actual data, the contribution of each type of vehicle to traffic flow relative to the standard vehicle is adjusted to obtain the traffic contribution value; Add up the flow contribution values ​​of each type of vehicle to obtain the total flow of all vehicles arriving at the same section at the service area entrance at the same time.

6. The service area congestion prediction method based on ETC data according to claim 5, characterized in that: Based on the historical traffic behavior data of vehicles, a vehicle portrait library is established, and the traffic behavior of vehicles on the road section is predicted through a neural network model to generate traffic behavior prediction results for future time periods, including: Integrate the basic information and traffic records of vehicles into static and dynamic databases; Build a portrait tag based on the vehicle's historical traffic behavior, and dynamically update the vehicle portrait library when new traffic data is received; Extract key features from vehicle portraits and determine the weight of different features on traffic behavior; When the vehicle enters the highway again, the service area it passes through is predicted based on historical data, and the traffic behavior prediction results are output.

7. The service area congestion prediction method based on ETC data according to claim 6, characterized in that: Extract key features from vehicle portraits and determine the weights of different features on traffic behavior, including: Extract key features from vehicle portraits, including vehicle type, vehicle size, vehicle weight, driving speed, mileage, driver behavior patterns, time period, weather conditions, and historical traffic records; For key features, quantification is performed to obtain numerical features; Construct a random forest based on the extracted and quantified features. For each tree in the random forest, traverse each node in the tree. For each node, determine the number of samples contained in the node and calculate the Gini impurity of the node. Split the features, divide the samples in the node into left child nodes and right child nodes, and calculate the Gini impurity of the left child nodes and the right child nodes respectively; Calculate the reduction of Gini impurity based on the Gini impurity of the node and the Gini impurity of the left and right child nodes; For each feature, the sum of the reduction in Gini impurity brought by splitting in all trees is counted to get the importance score of the feature; The importance scores of each feature are averaged to obtain the final importance score of the feature in the random forest; According to the final importance score of the feature in the random forest, the influence weight of different features on the traffic behavior is determined.

8. A service area congestion information release guidance system based on ETC data, the system implementing the method as claimed in any one of claims 1 to 7, characterized in that: include: The checkpoint information module is used to standardize the key features extracted from the images captured by the checkpoint cameras at the entrance and exit of the ramp, and map them to the corresponding discrete data range to form discrete data points, regard the data points as particles and let them move to the set initial center position under the action of gravity to form a circular structure, determine the radius of the circle by calculating the average distance and divide the circle according to the distribution of data points, compare the distribution patterns of data points in the circular area corresponding to the entrance and exit capture data, and generate a first ID code; Based on the first ID code, the captured data of the front and rear gantries of the service area are associated to generate a second ID code, and the vehicle information is screened by comparison, and the interval travel speed is calculated according to the gantries' travel time difference and spacing to obtain the predicted vehicles entering the service area; The multi-source identification module is used to capture real-time data of vehicles entering the service area through the bayonet camera, and dynamically determine the current real-time congestion index of the service area based on the predicted vehicles entering the service area; calculate the estimated arrival time based on the remaining distance of the vehicle to the service area entrance, and calculate the flow contribution value of different types of vehicles at the same section of the service area entrance based on the real-time congestion index; The vehicle trajectory prediction module is used to establish a vehicle portrait library based on the vehicle's historical traffic behavior data, and predict the vehicle's traffic behavior on the road section through a neural network model to generate traffic behavior prediction results for future time periods.

9. A computing device, characterized in that include: one or more processors; A storage device for storing one or more programs, when the one or more programs are executed by the one or more processors, the one or more processors implement the method as claimed in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a program, which, when executed by a processor, implements the method according to any one of claims 1 to 7.

Images