Vehicle parking early warning method, model and storage medium

Abstract

The invention relates to a vehicle parking early warning method, a model and a storage medium. The method is applied to a vehicle parking early warning model, and comprises the following steps: according to operation data of each vehicle in a plurality of vehicles in a target moment set area, respectively predicting a parking behavior of each vehicle in the plurality of vehicles; the operation data comprises at least one of track data, business data, traffic flow data and freight land data; according to the parking behavior, respectively acquiring a risk score of the parking behavior of each of the plurality of vehicles; according to the risk scores, the risk level of the parking behavior of each vehicle in the multiple vehicles is obtained; and according to the risk level, carrying out early warning on the parking behavior of at least one vehicle in the plurality of vehicles. According to the scheme provided by the invention, the multi-source data can be integrated, the parking behavior of the vehicle can be accurately predicted, the risk grade is evaluated by adopting the risk score of the quantitative parking behavior, and the parking behavior of the vehicle can be timely and accurately pre-warned according to the risk grade.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of intelligent transportation and artificial intelligence, in particular to a vehicle stay early warning method, a model and a storage medium. BACKGROUND

[0002] With the continuous development of economy, various transportation networks are increasingly perfect, and cross-provincial and cross-city transportation has become the norm. As the focus of traffic supervision departments, various key attention vehicles, especially out-of-town trucks, are highly mobile and widely distributed, which brings great difficulty to the supervision work of traffic supervision departments. It is difficult to comprehensively monitor the operation state of the vehicle, the driving behavior of the driver, and the risk situation of the vehicle enterprise, and it is easy to have supervision loopholes, which leads to the difficulty in timely discovering and handling some illegal and irregular behaviors.

[0003] The vehicle supervision of the related technology integrates multiple types of data (for example, static information, vehicle real-time information, road and environment real-time information) from different information sources, uses a vehicle monitoring model to monitor the vehicle in real time, and obtains a vehicle monitoring result, including illegal behavior, dangerous behavior, etc. However, the vehicle supervision of the related technology is based on the analysis of the trajectory and the judgment of the behavior that has occurred, which belongs to post-event monitoring and lacks more comprehensive pre-event risk prediction. Moreover, the vehicle monitoring model of the related technology is based on the loss between the predicted trajectory and the real driving trajectory to train and optimize the vehicle monitoring model, and the accuracy of the vehicle monitoring model in predicting the vehicle stay behavior is not high, which cannot timely and accurately warn the vehicle behavior.

[0004] Therefore, the vehicle supervision of the related technology is to supervise the behavior of the vehicle that has occurred, which belongs to post-event supervision and cannot timely and accurately warn the vehicle behavior. SUMMARY

[0005] To solve or partially solve the problems in the related technology, the present application provides a vehicle stay early warning method, a model and a storage medium, which can integrate multi-source data, accurately predict the stay behavior of the vehicle, use the risk score of the quantitative stay behavior to evaluate the risk level, and timely and accurately warn the stay behavior of the vehicle according to the risk level.

[0006] The first aspect of the present application provides a vehicle stay early warning method, which is applied to a vehicle stay early warning model, and includes: According to the running data of each vehicle in the plurality of vehicles, the stay behavior of each vehicle in the plurality of vehicles is predicted respectively; wherein the running data is the running data of each vehicle in the plurality of vehicles in a target time in a set region, including at least one of trajectory data, business data, traffic flow data, and freight use site data; According to the stay behavior, the risk score of the stay behavior of each vehicle in the plurality of vehicles is obtained respectively; According to the risk score, the risk level of the parking behavior of each vehicle in the plurality of vehicles is obtained respectively; According to the risk level, the parking behavior of at least one vehicle in the plurality of vehicles is warned.

[0007] In an embodiment, the parking behavior of each vehicle in the plurality of vehicles is predicted respectively according to the operation data of each vehicle in the plurality of vehicles, comprising: According to the operation data, the parking behavior characteristics of each vehicle in the plurality of vehicles are predicted respectively, and the parking behavior characteristics are used to represent the parking behavior of each vehicle in the plurality of vehicles, wherein the parking behavior characteristics include at least one of parking time, distance from cargo use land, cargo property weight, and use land property weight.

[0008] In an embodiment, the risk score of the parking behavior of each vehicle in the plurality of vehicles is obtained respectively according to the parking behavior, comprising: According to the parking time of each vehicle in the plurality of vehicles, the minimum parking time and the maximum parking time of the parking time of each vehicle in the plurality of vehicles, the parking time score of each vehicle in the plurality of vehicles is obtained respectively; According to the distance from cargo use land of each vehicle in the plurality of vehicles, the minimum distance from cargo use land and the maximum distance from cargo use land of the distance from cargo use land of each vehicle in the plurality of vehicles, the distance from cargo use land score of each vehicle in the plurality of vehicles is obtained respectively; According to the cargo property weight of each vehicle in the plurality of vehicles, the minimum cargo property weight and the maximum cargo property weight of the cargo property weight of each vehicle in the plurality of vehicles, the cargo property weight score of each vehicle in the plurality of vehicles is obtained respectively; According to the use land property weight of each vehicle in the plurality of vehicles, the minimum use land property weight and the maximum use land property weight of the use land property weight of each vehicle in the plurality of vehicles, the use land property weight score of each vehicle in the plurality of vehicles is obtained respectively; According to at least one of the parking time score, the distance from cargo use land score, the cargo property weight score, and the use land property weight score, the risk score of each vehicle in the plurality of vehicles is obtained respectively.

[0009] In an embodiment, the risk score of each vehicle in the plurality of vehicles is obtained respectively according to at least one of the parking time score, the distance from cargo use land score, the cargo property weight score, and the use land property weight score, comprising: According to the stay time score, the distance from the freight destination score, the freight property weight score, and the stay time score weight, the distance from the freight destination score weight, and the freight property weight score weight, a risk score of each vehicle in the plurality of vehicles is obtained.

[0010] In an embodiment, the obtaining of the risk score of each vehicle in the plurality of vehicles according to the stay time score, the distance from the freight destination score, the freight property weight score, and the stay time score weight, the distance from the freight destination score weight, and the freight property weight score weight further comprises: According to the stay behavior characteristics of each vehicle in the plurality of vehicles, a violation probability of the stay behavior of each vehicle in the plurality of vehicles is predicted; If the risk score matches the violation probability, the risk score matching the violation probability is determined as the risk score of each vehicle in the plurality of vehicles; If the risk score does not match the violation probability, the parameters of the pre-trained vehicle stay early warning model are adjusted, and the step of predicting the stay behavior characteristics of each vehicle in the plurality of vehicles according to the operation data is continued until the risk score matches the violation probability, and the risk score matching the violation probability is determined as the risk score of the stay behavior of each vehicle in the plurality of vehicles.

[0011] In an embodiment, the obtaining of the risk level of the stay behavior of each vehicle in the plurality of vehicles according to the risk score comprises: If the risk score is less than or equal to a first set score threshold, the risk level of the stay behavior is determined as a low risk level; If the risk score is greater than the first set score threshold and less than or equal to a second set score threshold, the risk level of the stay behavior is determined as a medium risk level; If the risk score is greater than the second set score threshold, the risk level of the stay behavior is determined as a high risk level.

[0012] In an embodiment, the obtaining of the risk level of the stay behavior of each vehicle in the plurality of vehicles according to the risk score further comprises: If the risk level matches the violation probability, the risk level matching the violation probability is determined as the risk level of each vehicle in the plurality of vehicles; If the risk level does not match the violation probability, parameters of the pre-trained vehicle stay early warning model are adjusted, and the step of predicting the stay behavior characteristics of each vehicle in the plurality of vehicles according to the operation data is continuously performed until the risk level matches the violation probability, and the risk level matching the violation probability is determined as the risk level of each vehicle in the plurality of vehicles.

[0013] In an embodiment, the step of predicting the violation probability of the stay behavior of each vehicle in the plurality of vehicles according to the stay behavior characteristics of each vehicle in the plurality of vehicles comprises: The stay points of the stay behavior of each vehicle in the plurality of vehicles are analyzed by using a clustering algorithm, and the stay behavior of each vehicle in the plurality of vehicles is divided into normal stay, abnormal stay, and high-risk stay. According to the violation judgment rules of the normal stay, the abnormal stay, and the high-risk stay, the stay time, the distance to the freight use site, and the weight of the nature of the goods of each vehicle in the plurality of vehicles, the violation probability of the stay behavior of each vehicle in the plurality of vehicles is predicted.

[0014] The second aspect of the present application provides a vehicle stay early warning model, comprising: a processor; and a memory having executable codes stored thereon, when the executable codes are executed by the processor, the processor executes the method as described above.

[0015] The third aspect of the present application provides a computer readable storage medium having executable codes stored thereon, when the executable codes are executed by a processor, the processor executes the method as described above.

[0016] The fourth aspect of the present application provides a computer program product, comprising computer instructions, when the computer instructions are executed by a processor, the method as described above is implemented.

[0017] The technical solution provided by the present application can include the following beneficial results: The technical scheme of the present application comprises the following steps: predicting the staying behavior of each vehicle in the plurality of vehicles according to the running data of each vehicle in the plurality of vehicles; wherein the running data is the running data of each vehicle in the plurality of vehicles in a target time in a set region, and comprises at least one of trajectory data, business data, traffic flow data and freight use data; obtaining the risk score of the staying behavior of each vehicle in the plurality of vehicles according to the staying behavior of each vehicle in the plurality of vehicles; obtaining the risk level of the staying behavior of each vehicle in the plurality of vehicles according to the risk score of the staying behavior of each vehicle in the plurality of vehicles; and giving a warning to the staying behavior of at least one vehicle in the plurality of vehicles according to the risk level of the staying behavior of each vehicle in the plurality of vehicles. The technical scheme can integrate multi-source data, accurately predict the staying behavior of the vehicle, evaluate the risk level by using the risk score of the staying behavior, give a warning to the staying behavior of the vehicle in time and accurately according to the risk level, and provide data support for improving the traffic management efficiency, ensuring the road safety, and optimizing the urban resource allocation.

[0018] It should be understood that the foregoing general description and the following detailed description are only exemplary and explanatory, and are not limiting to the present application. BRIEF DESCRIPTION OF DRAWINGS

[0019] The above and other objects, features and advantages of the present application will become more apparent from the following detailed description when taken in conjunction with the accompanying drawings in which like reference characters refer to like parts throughout the figures, and wherein:

[0020] Figure 1 FIG. 1 is a flow diagram of a vehicle staying warning method according to an embodiment of the present application; Figure 2 FIG. 2 is a flow diagram of a vehicle staying warning method according to another embodiment of the present application; Figure 3 FIG. 3 is a structural diagram of a vehicle staying warning model according to an embodiment of the present application. DETAILED DESCRIPTION

[0021] Embodiments of the present application will be described in more detail by making reference to the accompanying drawings. Although embodiments of the present application are shown in the drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided to make the present application more thorough and complete, and to fully convey the scope of the present application to those skilled in the art.

[0022] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting. As used in this application and the appended claims, the singular forms "a," "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will also be understood that the term "and / or," as used herein, refers to and encompasses any and all possible combinations of one or more of the associated listed items.

[0023] It should be understood that although the terms "first," "second," "third," etc. can be used in this application to describe various information, the information should not be limited to these terms. These terms are only used to distinguish one piece of information from another piece of information. For example, the first information can also be referred to as the second information, and similarly, the second information can also be referred to as the first information without departing from the scope of the application. Therefore, the features defined with "first," "second" can explicitly or implicitly include one or more of the features. In the description of the application, the meaning of "a plurality of" is two or more, unless otherwise specifically limited.

[0024] The embodiment of the application provides a vehicle staying early warning method, which can integrate multi-source data, accurately predict the staying behavior of the vehicle, evaluate the risk level by using the risk score of quantified staying behavior, and timely and accurately warn the staying behavior of the vehicle according to the risk level.

[0025] The technical scheme of the embodiment of the application is described in detail below with reference to the drawings.

[0026] Figure 1 is a flowchart of a vehicle staying early warning method shown in the embodiment of the application.

[0027] Referring to Figure 1 A vehicle staying early warning method applied to a vehicle staying early warning model, comprising: Step 101, predicting the staying behavior of each vehicle in a plurality of vehicles according to the running data of each vehicle in the plurality of vehicles; wherein the running data is the running data of each vehicle in the plurality of vehicles in a target time in a set region, including at least one of trajectory data, business data, traffic flow data, and freight use data.

[0028] In an embodiment, the pre-trained vehicle staying early warning model can obtain the running data of each vehicle of all set types of vehicles in a set region at a current time. The running data of the vehicle can include at least one of the trajectory data, the business data, the traffic flow data, and the freight use data of the vehicle. The target time can be the current time. The set region can be a set administrative region.

[0029] In an embodiment, the pre-trained vehicle stay early warning model can predict the stay behavior characteristics of each vehicle of all set types of vehicles respectively according to the operation data of each vehicle of all set types of vehicles, and represent the stay behavior of each vehicle of all set types of vehicles by using the stay behavior characteristics; the stay behavior characteristics include at least one of stay time, distance from the freight use site, freight property weight, and use site property weight.

[0030] In step 102, the risk score of the stay behavior of each vehicle in the plurality of vehicles is obtained respectively according to the stay behavior of each vehicle in the plurality of vehicles.

[0031] In an embodiment, the pre-trained vehicle stay early warning model can obtain the stay behavior characteristic score of each vehicle of all set types of vehicles respectively according to the stay behavior characteristics representing the stay behavior of each vehicle of all set types of vehicles; and obtain the risk score of the stay behavior of each vehicle of all set types of vehicles respectively according to the stay behavior characteristic score of each vehicle of all set types of vehicles.

[0032] In an embodiment, the pre-trained vehicle stay early warning model can obtain the stay time score of the stay time of each vehicle of all set types of vehicles respectively according to the stay time of each vehicle of all set types of vehicles; obtain the distance from the freight use site score of the distance from the freight use site of each vehicle of all set types of vehicles respectively according to the distance from the freight use site of each vehicle of all set types of vehicles; obtain the freight property weight score of the freight property weight of each vehicle of all set types of vehicles respectively according to the freight property weight of each vehicle of all set types of vehicles; obtain the use site property weight score of the use site property weight of each vehicle of all set types of vehicles respectively according to the use site property weight of each vehicle of all set types of vehicles; and obtain the risk score of the stay behavior of each vehicle of all set types of vehicles respectively according to at least one of the stay time score, the distance from the freight use site score, the freight property weight score, and the use site property weight score of each vehicle of all set types of vehicles.

[0033] In step 103, the risk level of the stay behavior of each vehicle in the plurality of vehicles is obtained respectively according to the risk score of the stay behavior of each vehicle in the plurality of vehicles.

[0034] In an embodiment, the pre-trained vehicle stay early warning model can obtain the risk level of the stay behavior of each vehicle of all set types of vehicles according to the risk score of the stay behavior of each vehicle of all set types of vehicles; wherein the risk level can include a low risk level, a medium risk level, and a high risk level.

[0035] In an embodiment, the pre-trained vehicle stay warning model can obtain the risk level of each vehicle stay behavior of each vehicle of all set types of vehicles according to the risk score of each vehicle stay behavior of each vehicle of all set types of vehicles and a set score threshold. If the risk score of the vehicle stay behavior is less than or equal to a first set score threshold, the risk level of the vehicle stay behavior is determined as a low risk level. If the risk score of the vehicle stay behavior is greater than the first set score threshold and less than or equal to a second set score threshold, the risk level of the vehicle stay behavior is determined as a medium risk level. If the risk score of the vehicle stay behavior is greater than the second set score threshold, the risk level of the vehicle stay behavior is determined as a high risk level.

[0036] In step 104, the stay behavior of at least one vehicle of the plurality of vehicles is warned according to the risk level of each vehicle stay behavior of the plurality of vehicles.

[0037] In an embodiment, the pre-trained vehicle stay warning model can edit warning information corresponding to the risk level of the vehicle stay behavior according to the risk level of each vehicle stay behavior of all set types of vehicles, and send the warning information to the terminal device of the relevant personnel to warn the stay behavior of at least one vehicle of all set types of vehicles.

[0038] In an embodiment, the pre-trained vehicle stay warning model can edit warning information corresponding to the risk level of the vehicle stay behavior according to the risk level of each vehicle stay behavior of all set types of vehicles, and send the warning information to the terminal device of the relevant personnel. The relevant personnel can timely process the stay behavior of the vehicle according to the warning information to avoid the risk caused by the stay behavior of the vehicle.

[0039] The vehicle stay warning method of the embodiments of the present application can predict the stay behavior of each vehicle of the plurality of vehicles according to the operation data of each vehicle of the plurality of vehicles, respectively; wherein the operation data is the operation data of each vehicle of the plurality of vehicles in a target time and a set region, including at least one of trajectory data, business data, traffic flow data, and freight use data; obtain the risk score of each vehicle stay behavior of the plurality of vehicles according to the stay behavior of each vehicle of the plurality of vehicles; obtain the risk level of each vehicle stay behavior of the plurality of vehicles according to the risk score of each vehicle stay behavior of the plurality of vehicles; and warn the stay behavior of at least one vehicle of the plurality of vehicles according to the risk level of each vehicle stay behavior of the plurality of vehicles. The vehicle stay warning method can integrate multiple sources of data, accurately predict the stay behavior of the vehicle, evaluate the risk level by quantifying the risk score of the stay behavior, timely and accurately warn the stay behavior of the vehicle according to the risk level, and provide data support for improving traffic management efficiency, ensuring road safety, and optimizing urban resource allocation.

[0040] Figure 2 is a flowchart of a vehicle stay early warning method according to another embodiment of the present application. Figure 2 Figure 1 The scheme of the present application is described in more detail.

[0041] Referring to Figure 2 A vehicle stay early warning method can be applied to a vehicle stay early warning model, comprising: Step 201, collecting the running data of each vehicle in a plurality of vehicles in a set region at a target time.

[0042] In an embodiment, the target time can be the current time. The set region can be a set administrative region. The plurality of vehicles can be a plurality of key concern vehicles supervised by a traffic supervision department, including out-of-town trucks, local trucks, passenger cars, and vehicles with high accident rates. The out-of-town truck is a truck with a different administrative region from the administrative region where the vehicle stay early warning model is deployed. The local truck is a truck with the same administrative region as the administrative region where the vehicle stay early warning model is deployed. The administrative region includes but is not limited to a provincial administrative region, a prefectural administrative region, and a county-level administrative region.

[0043] For example, the administrative region of a truck is A, the administrative region where the vehicle stay early warning model is deployed is B, and the administrative regions A and B are both provincial administrative regions. The truck is an out-of-town truck relative to the administrative region B.

[0044] In an embodiment, the running data of the vehicle can be collected through the GPS (Global Positioning System, Global Positioning System) of the vehicle, the electronic waybill system, the roadside sensing device, and the city GIS (Geographic Information System, Geographic Information System). The running data of the vehicle is multi-source data, including the trajectory data, the business data, the traffic flow data, and the freight use data of the vehicle.

[0045] In an embodiment, the trajectory data such as the position information, the speed, and the driving direction of the vehicle can be collected through the GPS of the vehicle at a set frequency (every 2 seconds).

[0046] In an embodiment, the electronic waybill of the vehicle can be obtained in real time through the electronic waybill system through the logistics platform API (Application Programming Interface, Application Programming Interface), and the business data such as the type of goods transported by the vehicle, the loading and unloading location, the loading and unloading time, and the destination can be obtained according to the electronic waybill.

[0047] In an embodiment, the traffic flow data of the road where the vehicle travels can be collected in real time through the roadside sensing device (for example, radar, camera). ​

[0048] In an embodiment, the freight land data of the logistics park, industrial park and the like can be obtained through the city GIS, and the freight land data includes location information of the freight land.

[0049] In an embodiment, the multi-source data such as the trajectory data, the business data, the traffic flow data and the freight land data can be standardized stored to obtain the running data with unified data format and timestamp.

[0050] In an embodiment, the OpenTS DB database (Open Time Series Database, distributed time series database) and the Ceph (CErn PHysics, distributed storage system) storage system can be used to standardize store the multi-source data such as the trajectory data, the business data, the traffic flow data and the freight land data, and ensure that the data such as the trajectory data, the business data, the traffic flow data and the freight land data have unified data format and accurate (millisecond level) timestamp.

[0051] Step 202, data cleaning is performed on the running data to obtain the running data after data cleaning.

[0052] In an embodiment, the running data of the vehicle with unified data format and timestamp can be sorted by timestamp, and the running data sorted by timestamp can be data cleaned, and the data cleaning includes but is not limited to abnormal data filtering, trajectory drift correction, repeated data deletion and missing data filling.

[0053] In an embodiment, the trajectory data of the vehicle can be filtered for abnormal data, and the trajectory data corresponding to the speed greater than a first set speed threshold or the speed less than a second set speed threshold in the trajectory data of the vehicle is deleted. For example, the trajectory data corresponding to the speed greater than 120 kilometers per hour or the speed less than 0 in the trajectory data of the vehicle is deleted.

[0054] In an embodiment, Kalman filtering can be used to correct the trajectory drift of the trajectory data of the vehicle.

[0055] In an embodiment, the trajectory data can be processed for repeated data deletion according to the timestamp of the trajectory data, the position of the trajectory point of the trajectory data, and the set spatial distance threshold and the set time difference threshold, and the repeated trajectory points of the trajectory data are deleted. For example, one of the two trajectory points with spatial distance less than 5 meters or time difference less than 10 seconds is deleted.

[0056] In an embodiment, the missing data of the trajectory data can be filled through map matching.

[0057] Step 203, the running data after data cleaning is aligned to obtain the running data after alignment.

[0058] In an embodiment, the H3 grid coding system (a spatial indexing system based on a hexagonal grid) and the space-time matching algorithm can be used to associate and fuse the cleaned trajectory data, business data, traffic flow data, freight land use data and other multi-source data according to the time window (± 30 minutes) and the spatial threshold (500 meters), to realize the accurate alignment of the multi-source data such as trajectory data, business data, traffic flow data, freight land use data, and obtain the running data aligned in time and space. The time window is the time range. The spatial threshold is the distance threshold of the spatial position.

[0059] In step 204, according to the running data of each vehicle in the plurality of vehicles after alignment, the stay behavior of each vehicle in the plurality of vehicles is respectively predicted, as well as the violation probability of the stay behavior of each vehicle in the plurality of vehicles.

[0060] In an embodiment, the pre-trained vehicle stay early warning model can be a pre-trained MBLF-NET (Multi-Branch Learning Fusion Network) model. The pre-trained MBLF-NET model fuses the PAM (Partitioning Around Medoids) algorithm and the CART (Classification And Regression Tree) algorithm, analyzes the stay of each vehicle in the plurality of vehicles according to the running data of each vehicle in the plurality of vehicles after alignment, jointly predicts the stay behavior and the violation probability of each vehicle in the plurality of vehicles through a multi-task learning framework, and dynamically evaluates and warns the stay behavior of each vehicle in the plurality of vehicles according to the stay behavior and the violation probability of each vehicle in the plurality of vehicles.

[0061] In an embodiment, the pre-trained MBLF-NET model can predict the stay behavior features of each vehicle in the plurality of vehicles according to the running data of each vehicle in the plurality of vehicles after alignment, and use the stay behavior features to represent the stay behavior of each vehicle in the plurality of vehicles. The stay behavior features include at least one of stay time, distance from freight land, freight property weight, and land property weight.

[0062] In an embodiment, the pre-trained MBLF-NET model can use the PAM-CART algorithm to cluster and analyze the stay points of each vehicle in the plurality of vehicles according to the stay behavior of each vehicle in the plurality of vehicles, extract the stay behavior features of the stay behavior of each vehicle in the plurality of vehicles according to the clustering analysis results, and accurately identify and classify the stay behavior of each vehicle in the plurality of vehicles according to the stay behavior features, and predict the violation probability of the stay behavior of each vehicle in the plurality of vehicles.

[0063] In an embodiment, the pre-trained MBLF-NET model can adopt the PAM algorithm to perform clustering analysis on the stay points of each vehicle in the plurality of vehicles, and divide the stay behavior of each vehicle in the plurality of vehicles into stays of different stay behavior types; wherein the stays of different stay behavior types include normal stay, abnormal stay and high-risk stay. The normal stay is the stay of the vehicle for loading and unloading goods, for example, the stay for loading and unloading goods at the destination. The abnormal stay is the stay of the vehicle at a compliance site, for example, the stay at a site such as a gas station, a service area, a parking lot, etc. The high-risk stay is the stay of the vehicle at a non-compliant site, for example, the illegal occupation stay or overtime stay of the vehicle at a non-designated driving or parking road area such as a non-motor vehicle lane, a sidewalk, an emergency lane, etc., the stay in a city forbidden area or the stay of a dangerous goods vehicle in violation of regulations.

[0064] In an embodiment, the stay time is the time of the vehicle in a non-driving state at the stay point. The freight use land distance is the distance between the stay point of the vehicle and the loading and unloading site of the vehicle, and the loading and unloading site is the destination in the business data, which is the pre-set loading and unloading site of the vehicle for this transportation. The freight property weight is a parameter for evaluating the freight property of the vehicle for loading and unloading goods, and the freight property can include the weight and type of the goods. The use land property weight is a parameter for evaluating the use land property of the stay point.

[0065] In an embodiment, the pre-trained MBLF-NET model can obtain city use land data through city GIS, filter according to use land property, extract A-class comprehensive use land, B-class commercial use land, M-class industrial use land, W-class logistics use land, and city use land with a construction state of "under construction" as freight use land, set the use land property weight of the freight use land to 1, and set the use land property weight of other city use land to 0.

[0066] In an embodiment, the pre-trained MBLF-NET model can obtain the trajectory data of each vehicle in the plurality of vehicles according to the aligned operation data of each vehicle in the plurality of vehicles, group the trajectory data in the operation data according to vehicle ID, and obtain the trajectory data of each vehicle in the plurality of vehicles respectively, wherein the trajectory data includes vehicle ID; adopt a vehicle stay domain search algorithm, and determine a stay point set region with a stay domain time search radius greater than a set stay domain time threshold value and a stay domain space search radius less than or equal to a set stay domain space threshold value as the stay domain of each vehicle in the plurality of vehicles according to the trajectory data of each vehicle in the plurality of vehicles; obtain the stay behavior points of the stay domain of each vehicle in the plurality of vehicles according to the stay domain of each vehicle in the plurality of vehicles, take the stay behavior points of the stay domain of each vehicle in the plurality of vehicles as the stay points of each vehicle in the plurality of vehicles, and perform clustering analysis on the stay points of the vehicle by using the PAM algorithm.

[0067] In an embodiment, the pre-trained MBLF-NET model can use a vehicle stay domain search algorithm according to the running data of each vehicle in the aligned multi-vehicle, and use spatio-temporal data analysis to identify the stay point and stay time of the vehicle. The stay domain time search radius refers to the time difference between the last point (stay point) searched in the stay domain search and the starting point (stay point); the stay domain space search radius refers to the straight-line distance between each point (stay point) searched during the stay and the starting point (stay point). The pre-trained MBLF-NET model can set the stay domain time threshold to 30 minutes and the stay domain space threshold to 0.5 kilometers to retain the vehicle freight behavior while eliminating road congestion, eating and refueling, and temporary rest.

[0068] In an embodiment, the step of obtaining the stay behavior points of the stay domain of each vehicle in the plurality of vehicles includes: Step 2401, obtaining the trajectory data sorted by time of each vehicle in the plurality of vehicles according to the vehicle ID of each vehicle in the plurality of vehicles.

[0069] In an embodiment, the pre-trained MBLF-NET model can preprocess the trajectory data obtained by the vehicle GPS, and the preprocessing includes effective field selection, abnormal data elimination, data grouping and sorting, etc., to form trajectory data grouped by vehicle ID, obtain trajectory data of each vehicle in the plurality of vehicles, the trajectory data includes vehicle ID, sort the trajectory data of each vehicle in the plurality of vehicles by timestamp, and obtain trajectory data sorted by time of each vehicle in the plurality of vehicles.

[0070] Step 2402, obtaining the adjacent trajectory point space spherical distance of the trajectory point of each vehicle in the plurality of vehicles according to the trajectory data sorted by time of each vehicle in the plurality of vehicles.

[0071] In an embodiment, the adjacent trajectory point space spherical distance refers to the shortest path length between the longitude and latitude (GPS coordinate point) of two adjacent trajectory points on the earth's surface, i.e. the length of the minor arc of the great circle (the circle obtained by cutting the spherical surface with a plane passing through the center of the sphere) passing through the two points. This distance is used to accurately measure the actual spatial interval between geographical positions. The pre-trained MBLF-NET model can calculate the adjacent trajectory point space spherical distance of the trajectory point of each vehicle according to the longitude and latitude of the trajectory point of the trajectory data sorted by time of each vehicle.

[0072] Step 2403, obtaining the stay point of each vehicle in the plurality of vehicles in the stay domain according to the adjacent trajectory point space spherical distance of the trajectory point of each vehicle in the plurality of vehicles.

[0073] In an embodiment, the pre-trained MBLF-NET model can determine whether the vehicle is moving or staying according to the adjacent trajectory point space spherical distance of the trajectory points of the vehicle; a first point of the staying action of the vehicle is recorded as a staying starting point, and a staying point set region with a staying domain time search radius greater than a set staying domain time threshold and a staying domain space search radius less than or equal to a set staying domain space threshold is searched backward along the time sequence, which is the staying domain of the vehicle; the geometric center point of the staying domain of the vehicle is taken as the staying behavior point of the vehicle, the latitude and longitude of the geometric center point of the staying domain of the vehicle are taken as the latitude and longitude of the staying behavior point, and the staying behavior point of the vehicle in the staying domain is taken as the staying point of the vehicle.

[0074] Step 2404, the staying time of the staying point of each vehicle in the staying domain is obtained according to the time of each vehicle in the staying domain.

[0075] In an embodiment, the pre-trained MBLF-NET model can calculate the time difference between the timestamp of the last staying point and the timestamp of the first staying point according to the timestamp of the last staying point and the timestamp of the first staying point of the vehicle in the staying domain, and predict the staying time of each vehicle in the staying point.

[0076] In an embodiment, the pre-trained MBLF-NET model can predict the violation probability of the staying behavior of each vehicle in the plurality of vehicles according to the staying behavior features of each vehicle in the plurality of vehicles. The violation probability includes a first violation probability and a second violation probability. If the violation probability of the staying behavior is the first violation probability (for example, 0), it means that the staying behavior is not illegal; if the violation probability of the staying behavior is the second violation probability (for example, 1), it means that the staying behavior is illegal.

[0077] In an embodiment, the MBLF-NET model adopts the CART algorithm, and the steps of predicting the violation probability of the staying behavior of the vehicle include: Step 2405, for each staying behavior clustering subset, extract the staying behavior features and labels, and construct a decision tree according to the staying behavior features (staying time, distance from goods land, goods property weight).

[0078] Step 2406, the best split point is selected using the Gini index (Gini Index, classification task) to generate violation judgment tree rules.

[0079] For example, the violation judgment tree rule can be “if the staying time of the staying behavior is greater than 30 minutes and the distance from the goods land is less than 500 meters, then mark the violation probability of the staying behavior as 1, and the staying behavior is a violation staying behavior”).

[0080] The calculation formula of the Gini index is:

[0081] where D represents the current data set, C k represents the kth class of stay behavior clustering subset in the current data set, |C k | represents the number of samples of the kth class of stay behavior clustering subset, and K represents the total number of classes (equal to 3).

[0082] Gini index calculation of feature division: When the data set D is divided into two subsets D1 and D2 according to a certain value a of the feature A, the Gini index after division is:

[0083] The feature A and the division point a that make Gini(D, A=a) minimum are selected to maximize the purity of the subsets.

[0084] Step 2407, integration prediction strategy, the prediction results of multiple CART algorithms are integrated into the final output, the violation probability predicted by each CART algorithm is counted by hard voting (classification task), and the violation probability with the most votes is selected as the violation probability of the stay behavior.

[0085] In an embodiment, the pre-trained MBLF-NET model can use the PAM algorithm, the clustering cluster number K=3, and the stay points of each of the 8 vehicles to perform clustering analysis on the stay behavior of each of the 8 vehicles, and obtain the clustering analysis results shown in Table 1 below; according to the clustering analysis results, the stay behavior characteristics of the stay behavior of each of the 8 vehicles, and the violation judgment rule of each cluster, the violation probability of the stay behavior of each of the 8 vehicles is predicted, and the violation probability of the stay behavior of each of the 8 vehicles is obtained as shown in Table 2 below.

[0086] Table 1:

[0087] In Table 1, the sample index represents a vehicle. Short stay time indicates that the stay time is less than a first set stay time threshold; medium stay time indicates that the stay time is greater than or equal to the first set stay time threshold and less than or equal to a second set stay time threshold; long stay time indicates that the stay time is greater than the second set stay time threshold. Near distance indicates that the distance from the freight use site is less than a first set distance threshold; medium distance indicates that the distance from the freight use site is greater than or equal to the first set distance threshold and less than or equal to a second set distance threshold; far distance indicates that the distance from the freight use site is greater than the second set distance threshold. Low cargo property weight indicates that the cargo property weight is less than a first set cargo property weight threshold; medium cargo property weight indicates that the cargo property weight is greater than or equal to the first set cargo property weight threshold and less than or equal to a second set cargo property weight threshold; high cargo property weight indicates that the cargo property weight is greater than the second set cargo property weight threshold.

[0088] In an embodiment, the violation determination rule of each cluster is as follows: Violation determination rule of cluster 1: If the stay time of the stay behavior is greater than a first set stay time threshold (e.g., 30 minutes) and the cargo property weight is less than a first set cargo property weight threshold (e.g., 0.4), the violation probability of the stay behavior is 0, and the stay behavior is not in violation.

[0089] If the stay time of the stay behavior is greater than the first set stay time threshold and the cargo property weight is greater than or equal to the first set cargo property weight threshold, the violation probability of the stay behavior is 1, and the stay behavior is in violation.

[0090] Violation determination rule of cluster 2: If the distance from the freight use site of the stay behavior is less than a second set distance threshold (e.g., 600 meters), the violation probability of the stay behavior is 0, and the stay behavior is not in violation.

[0091] If the distance from the freight use site of the stay behavior is greater than the second set distance threshold, the violation probability of the stay behavior is 1, and the stay behavior is in violation.

[0092] Violation determination rule of cluster 3: If the stay time of the stay behavior is greater than a second set stay time threshold (e.g., 100 minutes), the violation probability of the stay behavior is 1, and the stay behavior is in violation.

[0093] If the stay time of the stay behavior is less than the second set stay time threshold, the violation probability of the stay behavior is 0, and the stay behavior is not in violation.

[0094] Table 2:

[0095] In an embodiment, a federated learning mechanism can be employed to pre-train the MBLF-NET model according to a loss function of the MBLF-NET model using a dataset comprising operational data of each of a plurality of vehicles, to obtain a pre-trained MBLF-NET model.

[0096] In an embodiment, the loss function of the MBLF-NET model for predicting the stay behavior feature is a first loss function; the loss function of the MBLF-NET model for predicting the violation probability is a second loss function; and the loss function of the MBLF-NET model is obtained according to the first loss function, the second loss function, a first loss function weight of the first loss function, and a second loss function weight of the second loss function.

[0097] In an embodiment, the first loss function can be a mean square error (MSE) of the MBLF-NET model for predicting the stay time, and the first loss function is L duration wherein,

[0098] In the formula, N is the number of samples, i is the i-th sample, is the predicted stay time according to the i-th sample, is the true stay time of the i-th sample.

[0099] In an embodiment, the second loss function can be a cross-entropy loss function of the MBLF-NET model for predicting the violation probability of the stay behavior according to the stay behavior feature, and the second loss function is L violation wherein,

[0100] In the formula, N is the number of samples, i is the i-th sample, p i is the predicted violation probability of the stay behavior i, is the actual violation probability of the stay behavior i.

[0101] In an embodiment, the loss function of the MBLF-NET model is Loss, wherein, Loss= L1×L duration +L2×L violation ; In the formula, L1 is the first loss function weight, and L2 is the second loss function weight.

[0102] In an embodiment, during the pre-training process of the MBLF-NET model, a federated learning mechanism and a dynamic weight distribution mechanism are adopted to make the MBLF-NET model simultaneously predict the stay time and the violation probability, and dynamically distribute the first loss function weight and the second loss function weight, optimize the first loss function weight and the second loss function weight, and make the MBLF-NET model output accurate stay time and violation probability.

[0103] In a specific embodiment, the first loss function weight is 0.6, and the second loss function weight is 0.4.

[0104] In an embodiment, the MBLF-NET model can adopt the CART algorithm, independently train the CART algorithm on each stay behavior cluster subset obtained by clustering analysis using the PAM algorithm, and predict the stay behavior characteristics and the violation probability of the stay behavior through an ensemble strategy.

[0105] Step 205, according to the stay behavior of each vehicle in the plurality of vehicles, respectively obtaining the risk score of the stay behavior of each vehicle in the plurality of vehicles.

[0106] In an embodiment, the pre-trained MBLF-NET model can obtain the stay time score of each vehicle in the plurality of vehicles according to the stay time of each vehicle in the plurality of vehicles, the minimum stay time and the maximum stay time of the stay time of each vehicle in the plurality of vehicles; obtain the distance to the freight use land score of each vehicle in the plurality of vehicles according to the distance to the freight use land of each vehicle in the plurality of vehicles, the minimum distance to the freight use land and the maximum distance to the freight use land of the distance to the freight use land of each vehicle in the plurality of vehicles; obtain the cargo property weight score of each vehicle in the plurality of vehicles according to the cargo property weight of each vehicle in the plurality of vehicles, the minimum cargo property weight and the maximum cargo property weight of the cargo property weight of each vehicle in the plurality of vehicles; obtain the use land property weight score of each vehicle in the plurality of vehicles according to the use land property weight of each vehicle in the plurality of vehicles, the minimum use land property weight and the maximum use land property weight of the use land property weight of each vehicle in the plurality of vehicles; and obtain the risk score of each vehicle in the plurality of vehicles according to at least one of the stay time score, the distance to the freight use land score, the cargo property weight score, and the use land property weight score.

[0107] In an embodiment, the pre-trained MBLF-NET model can obtain the risk score of each vehicle in the plurality of vehicles according to the stay time score, the distance to the freight use land score, and the cargo property weight score of each vehicle in the plurality of vehicles, and the stay time score weight, the distance to the freight use land score weight, and the cargo property weight score weight.

[0108] Taking the dwelling behavior characteristics of vehicles 1-8 shown in Table 2—dwelling time, distance from freight land, and weight of cargo nature—as an example, the dwelling time T for vehicle 1 is 25 minutes, and the minimum dwelling time T0 is... min The maximum stay time is 20 minutes, T. max The dwell time of vehicle 1 is rated S for 100 minutes. T ,in, S T =(TT) min ) ÷ (T max -T min = (25-20) ÷ (100-20) = 0.0625.

[0109] The distance J between vehicle 1 and the freight yard is 100 meters. The minimum distance J between vehicle 1 and the freight yard is... min The maximum distance from the freight transport area is 100 meters. max The distance score between vehicle 1 and the freight area is 2500, and the score is S. J ,in, S J =(JJ min )÷(J max -J min = (100-100) ÷ (2500-100) = 0.

[0110] The weight Q of the cargo nature of vehicle 1 is 0.2, and the minimum weight Q of the cargo nature is... min The maximum weight of cargo nature is 0.2. max The weight score for the cargo nature of vehicle 1 is S, which is 1. Q ,in, S Q = (QQ) min )÷(Q max -Q min = (0.2-0.2) ÷ (1-0.2) = 0.

[0111] In one embodiment, a score S can be assigned based on the dwell time of each of the multiple vehicles. T Distance to freight land rating S J Weighted score of cargo nature S Q And the weighting of the stay time score Q T Distance to freight land scoring weight Q J Weighting of Goods Nature in Scoring Q Q The risk score F for each of the multiple vehicles is obtained, where, F = Q T ×S T +Q J ×S J +QQ xS Q ,Q T +Q J +Q Q =1.

[0112] In an embodiment, the AHP-Delphi method can be used to set the stay time score weight Q T , the freight use site distance score weight Q J , and the freight property weight score weight Q Q , Q T +Q J +Q Q =1.

[0113] In a specific embodiment, the stay time score weight Q T is 0.5, the freight use site distance score weight Q J is 0.3, and the freight property weight score weight Q Q is 0.2.

[0114] Step 206, determining the risk score of each vehicle in the plurality of vehicles according to the risk score and the violation probability of the stay behavior of each vehicle in the plurality of vehicles.

[0115] In an embodiment, the risk score and the violation probability of the stay behavior of each vehicle in the plurality of vehicles can be respectively determined according to whether the risk score and the violation probability of the stay behavior of each vehicle in the plurality of vehicles match; if the risk score and the violation probability of the stay behavior of each vehicle in the plurality of vehicles match, the risk score of the stay behavior of each vehicle in the plurality of vehicles that matches the violation probability is respectively determined as the risk score of the stay behavior of each vehicle in the plurality of vehicles; if the risk score and the violation probability of the stay behavior of each vehicle in the plurality of vehicles do not match, the parameters of the pre-trained MBLF-NET model are adjusted according to the loss function of the stay behavior of each vehicle in the plurality of vehicles and the violation probability of the stay behavior of each vehicle in the plurality of vehicles predicted by the pre-trained MBLF-NET model, and the step of predicting the stay behavior characteristics of each vehicle in the plurality of vehicles according to the operation data of each vehicle in the plurality of vehicles is continuously executed until the risk score and the violation probability of the stay behavior of each vehicle in the plurality of vehicles match, and the risk score that matches the violation probability is respectively determined as the risk score of each vehicle in the plurality of vehicles.

[0116] In an embodiment, the risk score of the vehicle stay behavior matches the violation probability of the vehicle stay behavior if the risk score of the vehicle stay behavior is less than or equal to a first set score threshold (the risk level of the vehicle stay behavior is a low risk level), the violation probability of the vehicle stay behavior is 0 (the stay behavior is not in violation), or the risk score of the vehicle stay behavior is greater than the first set score threshold (the risk level of the vehicle stay behavior is a medium risk level or a high risk level), and the violation probability of the vehicle stay behavior is 1 (the stay behavior is in violation).

[0117] In an embodiment, the risk score of the vehicle stay behavior does not match the violation probability of the vehicle stay behavior if the risk score of the vehicle stay behavior is less than or equal to a first set score threshold (the risk level of the vehicle stay behavior is a low risk level), the violation probability of the vehicle stay behavior is 1 (the stay behavior is in violation), or the risk score of the vehicle stay behavior is greater than the first set score threshold (the risk level of the vehicle stay behavior is a medium risk level or a high risk level), and the violation probability of the vehicle stay behavior is 0 (the stay behavior is not in violation).

[0118] At step 207, according to the risk score of the stay behavior of each vehicle in the plurality of vehicles, the risk level of the stay behavior of each vehicle in the plurality of vehicles is obtained respectively.

[0119] In an embodiment, according to the risk score of each vehicle in the plurality of vehicles and a set score threshold, the stay behavior of each vehicle in the plurality of vehicles can be divided into different risk levels, and the different risk levels include a low risk level, a medium risk level, and a high risk level.

[0120] In an embodiment, if the risk score of the vehicle stay behavior is less than or equal to a first set score threshold (for example, 0.4), the vehicle stay behavior is a low risk level; if the risk score of the vehicle stay behavior is greater than the first set score threshold and less than or equal to a second set score threshold (for example, 0.6), the vehicle stay behavior is a medium risk level; and if the risk score of the vehicle stay behavior is greater than the second set score threshold, the vehicle stay behavior is a high risk level.

[0121] At step 208, according to the risk level and the violation probability of each vehicle in the plurality of vehicles, the risk level of the stay behavior of each vehicle in the plurality of vehicles is determined.

[0122] In an embodiment, according to the risk level and the violation probability of the stay behavior of each vehicle in the plurality of vehicles, whether the risk level of the stay behavior of each vehicle in the plurality of vehicles matches the violation probability of the stay behavior of each vehicle in the plurality of vehicles is determined respectively; if the risk level matches the violation probability, the risk level that matches the violation probability is determined as the risk level of the stay behavior of each vehicle in the plurality of vehicles.

[0123] In an embodiment, if the risk level does not match the violation probability, a loss function of the stay behavior and the violation probability of each vehicle in the plurality of vehicles is predicted according to the pre-trained MBLF-NET model, the parameters of the pre-trained MBLF-NET model are adjusted, and the step of predicting the stay behavior characteristics of each vehicle in the plurality of vehicles is continued until the risk level matches the violation probability, and the risk level that matches the violation probability is determined as the risk level of the stay behavior of each vehicle in the plurality of vehicles.

[0124] In an embodiment, if the risk level of the vehicle stay behavior is a low risk level, the violation probability of the vehicle stay behavior is 0 (the stay behavior is not in violation), or the risk level of the vehicle stay behavior is a medium risk level or a high risk level, and the violation probability of the vehicle stay behavior is 1 (the stay behavior is in violation), the risk level of the vehicle stay behavior matches the violation probability.

[0125] In an embodiment, if the risk level of the vehicle stay behavior is a low risk level, the violation probability of the vehicle stay behavior is 1 (the stay behavior is in violation), or the risk level of the vehicle stay behavior is a medium risk level or a high risk level, and the violation probability of the vehicle stay behavior is 0 (the stay behavior is not in violation), the risk level of the vehicle stay behavior does not match the violation probability.

[0126] In an embodiment, if the risk level of the vehicle stay behavior is a low risk level, the violation probability of the vehicle stay behavior is 1 (the stay behavior is in violation), or the risk level of the vehicle stay behavior is a medium risk level or a high risk level, and the violation probability of the vehicle stay behavior is 0 (the stay behavior is not in violation), the risk level of the vehicle stay behavior does not match the violation probability.

[0127] In an embodiment, a three-level warning mechanism can be adopted, a first-level warning corresponding to a high risk level, a second-level warning corresponding to a medium risk level, and a third-level warning corresponding to a low risk level; wherein, The first-level warning (high risk level) corresponds to a stay time exceeding 3 times the loading and unloading time declared in the electronic waybill, and a high-risk stay (e.g., stay in a city restricted area or stay of a dangerous goods vehicle in violation of regulations); The second-level warning (medium risk level) corresponds to a stay time of 2-3 times the loading and unloading time declared in the electronic waybill, and an abnormal stay (e.g., stay in a highway service area or stay of a general goods vehicle in a sensitive area); The third-level warning (low risk level) corresponds to a stay time of 1-2 times the loading and unloading time declared in the electronic waybill, and a normal stay (e.g., stay in a non-sensitive area).

[0128] In an embodiment, the sensitive area can be a parking lot, a service area, a gas station, etc., which is an area that can be stopped but does not belong to the range of loading and unloading locations. The non-sensitive area can be a conventional loading and unloading area, such as a loading and unloading location, around a logistics park, etc.

[0129] In an embodiment, a three-level early warning mechanism can be adopted, early warning information corresponding to the risk level of the vehicle stay behavior of each vehicle in the plurality of vehicles is edited according to the risk level of the vehicle stay behavior of each vehicle in the plurality of vehicles, and the early warning information can include the stay time of the vehicle stay behavior, the distance from the freight use site, the weight of the nature of the goods, the stay point, the vehicle license plate number, the first-level early warning or the second-level early warning or the third-level early warning. Through the blockchain technology, the data security sharing of the departments of traffic, public security, urban management and the like is realized, the Hyperledger Fabric (Hyperledger) consortium chain is adopted, each node is deployed in different departments, the data is ensured to be tamper-proof, the early warning information is sent to the terminal device of the relevant personnel, and the vehicle stay behavior of each vehicle in the plurality of vehicles is early warned. The relevant personnel can include the vehicle driver, the traffic management personnel.

[0130] In an embodiment, the vehicle stay behavior of each vehicle in the plurality of vehicles can be early warned according to the risk level of the vehicle stay behavior of each vehicle in the plurality of vehicles, and the following early warning rules can be adopted, the early warning rules including: The risk level of the vehicle stay behavior is a low risk level: a third-level early warning, no intervention is needed, and normal operation is performed.

[0131] The risk level of the vehicle stay behavior is a medium risk level: a second-level early warning (yellow early warning) is triggered, the vehicle driver is reminded to pay attention to the vehicle stay specification (for example, the stay time is shortened, and the dangerous area is avoided).

[0132] The risk level of the vehicle stay behavior is a high risk level: a first-level early warning (red early warning) is triggered, the vehicle driver is reminded to immediately take measures (for example, the operation is suspended and the stay is stopped, the route is adjusted, and the forced rest is performed).

[0133] In an embodiment, different electronic fence radii can be set according to different risk levels, the electronic fence radius of the first-level early warning can be a first set radius (for example, 12 meters), the electronic fence radius of the second-level early warning can be a second set radius (for example, 8 meters), and the electronic fence radius of the third-level early warning can be a third set radius (for example, 5 meters). Different electronic fence radii are set according to different risk levels, precise early warning is realized, and the vehicle with a high risk level of the stay behavior is more accurately monitored and prevented.

[0134] In an embodiment, the electronic fence radius is a region with a center at the stay point of the vehicle and a set radius. A first set radius of the no-go area is drawn for the vehicle with a high risk level of the stay behavior, and a stay within the electronic fence radius range will continuously automatically trigger a warning. The vehicle with a high risk level of the stay behavior means that the stay within a certain range around the vehicle may have risks, so the electronic fence radius is set to be the largest. Different risk levels are set to correspond to different electronic fence radii, which can be adjusted according to actual needs, to prevent a vehicle with a high risk level from being judged as a low risk level just by driving away a distance, resulting in oversight of supervision. Moreover, this is also associated with the monitoring system, and a larger electronic fence radius means a wider range that can be monitored, which can warn of possible risks around the vehicle and make preparations in advance.

[0135] In an embodiment, according to the warning information, traffic management personnel can be automatically dispatched to investigate the area where the key attention vehicle is located, check whether the stay behavior of the key attention vehicle has a high risk, realize fixed-point investigation, reduce the cost of manual investigation, and improve the efficiency of law enforcement.

[0136] In an embodiment, according to the distance between the stay behavior of the key attention vehicle and the cargo land, the spatial relationship between the stay point and the cargo land (for example, a logistics park) can be analyzed, the utilization rate and compliance of the cargo land can be evaluated, the hotspot area of illegal occupation of the road with a high risk level can be identified, and the expansion or function adjustment of the cargo land can be promoted. The distance between the stay behavior and the cargo land can provide scientific and reliable data support for the urban planning department to optimize land planning and optimize the layout of the cargo land (for example, add temporary loading and unloading points).

[0137] In an embodiment, the key attention vehicle stays for too long at a key road section (main road, bridge) during the traffic peak period, causing traffic congestion. According to the stay time of the stay behavior of the key attention vehicle, the traffic signal light cycle can be dynamically adjusted, the compliant key attention vehicle can be given priority to release, the traffic flow can be regulated and the congestion can be relieved, so that the stay time of the key attention vehicle is shortened and the congestion during the traffic peak period is relieved.

[0138] In an embodiment, the key attention vehicle stays for a long time in a residential area or an environmentally sensitive area (for example, a hospital, a school), causing noise and exhaust pollution. According to the stay time of the stay behavior of the key attention vehicle, the key attention vehicle in the residential area or the environmentally sensitive area can be subjected to directional regulation (for example, limiting the night transportation of the key attention vehicle), the stay time of the key attention vehicle in the residential area or the environmentally sensitive area can be reduced, and the noise complaint rate can be reduced.

[0139] In an embodiment, by focusing on the parking behavior of the vehicle through big data analysis, more accurate management policies (for example, setting of parking restriction areas and parking restriction time periods) are formulated. Based on the parking behavior of the focused vehicle, the management policy is formulated, subjective decision bias is reduced, fixed-point law enforcement is deployed in high-frequency violation areas, and law enforcement efficiency is improved.

[0140] In an embodiment, the focused vehicle is focused on the parking behavior in the accident-prone road section (for example, a curve or a slope), which may cause a traffic accident. The parking behavior of the focused vehicle in the accident-prone road section is supervised, the traffic accident rate of the accident-prone road section is reduced, and the emergency response speed of the accident-prone road section is improved.

[0141] In an embodiment, by analyzing the parking rules of the focused vehicle, the logistics industry supervision and service are optimized, the logistics industry operation efficiency is improved, the compliance of the logistics industry transportation is ensured, the logistics cost is reduced, the transportation timeliness is improved, the enterprise transportation compliance awareness is enhanced, the illegal behavior is reduced, and the legal risk of the enterprise is reduced.

[0142] The vehicle parking early warning method of the embodiment of the application predicts the parking behavior of each vehicle in a plurality of vehicles according to the operation data of each vehicle in the plurality of vehicles, wherein the operation data is the operation data of each vehicle in the plurality of vehicles in a target time period in a set area, and includes at least one of trajectory data, business data, traffic flow data, and freight use data. The risk score of the parking behavior of each vehicle in the plurality of vehicles is obtained according to the parking behavior of each vehicle in the plurality of vehicles. The risk level of the parking behavior of each vehicle in the plurality of vehicles is obtained according to the risk score of the parking behavior of each vehicle in the plurality of vehicles. The parking behavior of at least one vehicle in the plurality of vehicles is early warned according to the risk level of the parking behavior of each vehicle in the plurality of vehicles. The vehicle parking early warning method can integrate multiple sources of data, accurately predict the parking behavior of the vehicle, evaluate the risk level by using the quantitative risk score of the parking behavior, early warn the parking behavior of the vehicle according to the risk level, provide reliable data support for improving the traffic management efficiency, ensuring road safety, and optimizing the allocation of urban resources, effectively improve the discovery ability and disposal efficiency of the traffic supervision department for the illegal parking behavior of the focused vehicle, and reduce the road safety hazards and accident risks.

[0143] Further, the vehicle stay early warning method of the embodiment of the application predicts the stay behavior characteristics of each vehicle in the plurality of vehicles respectively according to the operation data of each vehicle in the plurality of vehicles; obtains the risk score of the stay behavior of each vehicle in the plurality of vehicles respectively according to the stay behavior characteristics of each vehicle in the plurality of vehicles; predicts the rule violation probability of the stay behavior of each vehicle in the plurality of vehicles respectively according to the stay behavior characteristics of each vehicle in the plurality of vehicles; if the risk score matches the rule violation probability, the risk score matching the rule violation probability is determined as the risk score of each vehicle in the plurality of vehicles respectively; if the risk score does not match the rule violation probability, the parameters of the pre-trained vehicle stay early warning model are adjusted, and the step of predicting the stay behavior characteristics of each vehicle in the plurality of vehicles respectively according to the operation data is continued to be executed until the risk score matches the rule violation probability, and the risk score matching the rule violation probability is determined as the risk score of the stay behavior of each vehicle in the plurality of vehicles respectively. The stay behavior characteristics and the rule violation probability of the vehicle stay behavior are jointly predicted, the accuracy of the stay behavior characteristics prediction is improved through the rule violation probability, the stay time and the rule violation probability of the stay behavior can be simultaneously paid attention to, the dynamic risk quantification of the multi-dimensional stay behavior characteristics is realized, the accuracy of the vehicle stay behavior risk early warning is improved through the dynamic risk assessment.

[0144] Further, the vehicle stay early warning method of the embodiment of the application predicts the stay behavior characteristics of each vehicle in the plurality of vehicles respectively according to the operation data of each vehicle in the plurality of vehicles; obtains the risk score of the stay behavior of each vehicle in the plurality of vehicles respectively according to the stay behavior characteristics of each vehicle in the plurality of vehicles; predicts the rule violation probability of the stay behavior of each vehicle in the plurality of vehicles respectively according to the stay behavior characteristics of each vehicle in the plurality of vehicles; if the risk score matches the rule violation probability, the risk score matching the rule violation probability is determined as the risk score of each vehicle in the plurality of vehicles respectively; if the risk score does not match the rule violation probability, the parameters of the pre-trained vehicle stay early warning model are adjusted, and the step of predicting the stay behavior characteristics of each vehicle in the plurality of vehicles respectively according to the operation data is continued to be executed until the risk score matches the rule violation probability, and the risk score matching the rule violation probability is determined as the risk score of the stay behavior of each vehicle in the plurality of vehicles respectively. The stay behavior characteristics and the rule violation probability of the vehicle stay behavior are jointly predicted, the accuracy of the stay behavior characteristics prediction is improved through the rule violation probability, the stay time and the rule violation probability of the stay behavior can be simultaneously paid attention to, the dynamic risk quantification of the multi-dimensional stay behavior characteristics is realized, the dynamic risk assessment is used, and the accuracy of the vehicle stay behavior risk early warning is improved.

[0145] Corresponding to the foregoing application function implementation method embodiment, the application further provides a vehicle stay early warning model and a corresponding embodiment.

[0146] Figure 3 FIG. 1 is a structural schematic diagram of a vehicle stay early warning model according to an embodiment of the application.

[0147] Referring to FIG. 1, the vehicle stay early warning model 1000 includes a memory 1010 and a processor 1020. Figure 3

[0148] ​The processor 1020 can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gates or transistor logic, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can be any conventional processor. The memory 1010 can include various types of storage units, such as a system memory, a read-only memory (ROM), and a permanent storage device. Among them, the ROM can store static data or instructions required by the processor 1020 or other modules of the computer. The permanent storage device can be a readable and writable storage device. The permanent storage device can be a non-volatile storage device that does not lose stored instructions and data even after the computer is powered off. In some embodiments, the permanent storage device uses a mass storage device (such as a magnetic or optical disk, a flash memory) as a permanent storage device. In some other embodiments, the permanent storage device can be a removable storage device (such as a floppy disk, an optical drive). The system memory can be a readable and writable storage device or a volatile readable and writable storage device, such as a dynamic random access memory. The system memory can store some or all of the instructions and data required by the processor during runtime. In addition, the memory 1010 can include a combination of any computer readable storage media, including various types of semiconductor memory chips (such as DRAM, SRAM, SDRAM, flash memory, programmable read-only memory), magnetic disks and / or optical disks. In some embodiments, the memory 1010 can include a readable and / or writable removable storage device, such as a compact disc (CD), a read-only digital versatile disc (such as DVD-ROM, double-layer DVD-ROM), a read-only Blu-ray disc, an ultra-density optical disc, a flash memory card (such as an SD card, a min SD card, a Micro-SD card, etc.), a magnetic floppy disk, etc. The computer readable storage medium does not include a carrier wave and a transient electronic signal transmitted through wireless or wired transmission.

[0149] The memory 1010 stores executable code, which, when processed by the processor 1020, can cause the processor 1020 to perform part or all of the above-mentioned methods.

[0150] In addition, the method according to the present application can also be implemented as a computer program or a computer program product, which comprises computer program code instructions for executing part or all of the steps in the above method according to the present application.

[0151] Alternatively, the present application can also be implemented as a computer readable storage medium (or a non-transitory machine readable storage medium or a machine readable storage medium) having stored executable codes (or computer programs or computer instruction codes) which, when executed by a processor of a vehicle stay warning model (or an electronic device, or a server, etc.), cause the processor to execute part or all of the steps of the above method according to the present application.

[0152] The present application also provides a computer program product comprising computer instructions, which, when executed by a processor, implement the method as described above.

[0153] The above has described the embodiments of the present application, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes are obvious to those skilled in the art without departing from the scope and spirit of the described embodiments. The selection of the terms used herein is intended to best explain the principles of the embodiments, practical application, or improvement to the technology in the market, or to enable other ordinary skilled in the art to understand the embodiments disclosed herein.

Claims

1. A method for vehicle idling warning, characterized in that, The method is applied to a vehicle idling warning model, including: Based on the operational data of each of the multiple vehicles, the dwell behavior of each of the multiple vehicles is predicted; wherein, the operational data is the operational data of each of the multiple vehicles in a designated area at a target time, including at least one of trajectory data, business data, traffic flow data, and freight land use data. Based on the dwelling behavior, a risk score for the dwelling behavior of each of the multiple vehicles is obtained; Based on the risk score, the risk level of the stopping behavior of each of the multiple vehicles is obtained; Based on the risk level, an early warning is issued for the stationary behavior of at least one of the multiple vehicles.

2. The method according to claim 1, characterized in that, The step of predicting the dwell behavior of each of the multiple vehicles based on the operating data of each vehicle includes: Based on the operational data, the dwelling behavior characteristics of each of the multiple vehicles are predicted, and the dwelling behavior characteristics are used to represent the dwelling behavior of each of the multiple vehicles. The dwelling behavior characteristics include at least one of dwelling time, distance from freight land, cargo nature weight, and land use nature weight.

3. The method according to claim 2, characterized in that, The step of obtaining a risk score for the dwelling behavior of each of the multiple vehicles based on the dwelling behavior includes: Based on the dwell time of each vehicle in the plurality of vehicles, the minimum dwell time and the maximum dwell time of each vehicle in the plurality of vehicles, a dwell time score is obtained for each vehicle in the plurality of vehicles. Based on the distance between each of the multiple vehicles and the freight land, the minimum distance between each of the multiple vehicles and the freight land, and the maximum distance between each of the multiple vehicles and the freight land, a distance score between each of the multiple vehicles and the freight land is obtained. Based on the cargo nature weight of each vehicle in the plurality of vehicles, the minimum cargo nature weight and the maximum cargo nature weight of each vehicle in the plurality of vehicles, the cargo nature weight score of each vehicle in the plurality of vehicles is obtained respectively. Based on the land use weight of each vehicle in the plurality of vehicles, the minimum land use weight and the maximum land use weight of each vehicle in the plurality of vehicles, the land use weight score of each vehicle in the plurality of vehicles is obtained respectively. Based on at least one of the following: the dwell time score, the distance to freight land score, the cargo nature weight score, and the land use nature weight score, a risk score is obtained for each of the multiple vehicles.

4. The method according to claim 3, characterized in that, The step of obtaining a risk score for each of the multiple vehicles based on at least one of the following: the dwell time score, the distance to freight land score, the cargo nature weight score, and the land use nature weight score, includes: Based on the dwell time score, the distance to freight land score, the cargo nature weight score, and the dwell time score weight, the distance to freight land score weight, and the cargo nature weight score weight, a risk score is obtained for each of the multiple vehicles.

5. The method according to claim 4, characterized in that, The step of obtaining a risk score for each of the multiple vehicles based on the dwell time score, the distance to freight land score, the cargo nature weight score, and the weights of the dwell time score, the distance to freight land score, and the cargo nature weight score, respectively, further includes: Based on the stopping behavior characteristics of each of the multiple vehicles, predict the probability of violation of the stopping behavior of each of the multiple vehicles. If the risk score matches the violation probability, then the risk score that matches the violation probability is determined as the risk score of each of the multiple vehicles. If the risk score does not match the violation probability, the parameters of the pre-trained vehicle dwelling warning model are adjusted, and the steps continue to be executed to predict the dwelling behavior characteristics of each of the multiple vehicles based on the running data, until the risk score matches the violation probability. The risk score that matches the violation probability is then determined as the risk score for the dwelling behavior of each of the multiple vehicles.

6. The method according to claim 5, characterized in that, The step of obtaining the risk level of the stopping behavior of each of the multiple vehicles based on the risk score includes: If the risk score is less than or equal to the first set scoring threshold, the risk level of the stay behavior is determined to be low risk. If the risk score is greater than the first set scoring threshold and less than or equal to the second set scoring threshold, the risk level of the stay behavior is determined to be medium risk. If the risk score is greater than the second set score threshold, the risk level of the stay behavior is determined to be high risk level.

7. The method according to claim 6, characterized in that, The step of obtaining the risk level of the stopping behavior of each of the multiple vehicles based on the risk score also includes: If the risk level matches the violation probability, then the risk level that matches the violation probability is determined as the risk level of each of the multiple vehicles. If the risk level does not match the violation probability, the parameters of the pre-trained vehicle dwelling warning model are adjusted, and the steps continue to be executed. Based on the running data, the dwelling behavior characteristics of each of the multiple vehicles are predicted until the risk level matches the violation probability. The risk level that matches the violation probability is determined as the risk level of each of the multiple vehicles.

8. The method according to claim 5, characterized in that, The step of predicting the probability of violation of the stopping behavior of each of the multiple vehicles based on the stopping behavior characteristics of each vehicle includes: Clustering algorithms are used to perform cluster analysis on the stopping points of each vehicle among the multiple vehicles, and the stopping behavior of each vehicle among the multiple vehicles is divided into normal stopping, abnormal stopping, and high-risk stopping. Based on the violation judgment rules for normal stay, abnormal stay, and high-risk stay, the probability of violation for the stay behavior of each of the multiple vehicles is predicted by considering the stay time, distance from the freight land, and weight of the cargo nature of each vehicle.

9. A vehicle idling warning model, characterized in that, include: processor; as well as A memory having executable code stored thereon, which, when executed by the processor, causes the processor to perform the method as described in any one of claims 1-8.

10. A computer-readable storage medium, characterized in that: It stores executable code that, when executed by a processor, causes the processor to perform the method as described in any one of claims 1-8.

Images