Traffic operation risk real-time identification method based on low-altitude unmanned aerial vehicle

Abstract

The invention discloses a traffic operation risk real-time identification method based on a low-altitude unmanned aerial vehicle, relates to the technical field of unmanned aerial vehicle traffic monitoring, and solves the technical problem that unmanned aerial vehicle resources are wasted or the coverage of data acquisition is insufficient due to unreasonable arrangement of unmanned aerial vehicles in the prior art, so that the comprehensive identification efficiency of traffic risks is reduced. Comprising the steps that road section basic data of all road sections in a to-be-monitored area are acquired, and the road section basic data comprise road section types, historical accident data and historical traffic flow data; setting a road section danger score of each road section based on the road section basic data; setting an unmanned aerial vehicle distribution scheme based on the road section risk score; acquiring monitoring video data of the unmanned aerial vehicle, and analyzing based on the monitoring video data to obtain an accident risk score of each traffic subject of the road section; giving an alarm based on the accident risk score; by performing targeted analysis on different road sections and setting the number of the unmanned aerial vehicles used for monitoring the different road sections, the comprehensive efficiency of traffic risk identification is improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of unmanned aerial vehicle traffic monitoring, and specifically relates to a traffic operation risk real-time identification method based on low-altitude unmanned aerial vehicles. BACKGROUND

[0002] In recent years, low-altitude unmanned aerial vehicle technology has made remarkable development in the field of traffic management. With the maturity of unmanned aerial vehicle technology and the expansion of application scenarios, low-altitude unmanned aerial vehicles are gradually used in real-time traffic monitoring, traffic accident detection, and emergency handling and other aspects.

[0003] The prior art (invention patent with publication number CN 118918732 B) discloses a traffic operation risk real-time identification method based on low-altitude unmanned aerial vehicles, relating to the technical field of unmanned aerial vehicle traffic monitoring, which comprises the following steps: a dispatch center assigns a number and a working area to a low-altitude unmanned aerial vehicle, collects unmanned aerial vehicle working environment data, analyzes low-altitude unmanned aerial vehicle working time length, and arranges a low-altitude unmanned aerial vehicle working schedule; the low-altitude unmanned aerial vehicle collects traffic data through a camera, generates a multi-mode traffic data OD matrix, analyzes traffic behavior based on the multi-mode traffic data OD matrix, and detects traffic accidents; the unmanned aerial vehicle implements processing measures according to the traffic accident detection results; the low-altitude unmanned aerial vehicle sends traffic data to the dispatch center in real time for display, storage, and real-time uploading to the cloud.

[0004] The above-mentioned low-altitude unmanned aerial vehicle risk identification system effectively improves the safety and efficiency of unmanned aerial vehicle operation by reasonably setting the schedule of unmanned aerial vehicles and planning the return route of unmanned aerial vehicles for unmanned aerial vehicle alternation. In fact, the probability of traffic accidents occurring on different road sections is different, and the structure of different road sections, building obstructions, etc. also have different effects on traffic accidents on the road section. Using the same unmanned aerial vehicle arrangement method for each road section will result in waste of unmanned aerial vehicle resources or insufficient coverage of data collection, which will reduce the comprehensive efficiency of traffic risk identification. Therefore, a traffic operation risk real-time identification method based on low-altitude unmanned aerial vehicles is needed. SUMMARY

[0005] The application provides a traffic operation risk real-time identification method and device based on low-altitude unmanned aerial vehicles, which solves the technical problem of unreasonable arrangement of unmanned aerial vehicles in the prior art, resulting in waste of unmanned aerial vehicle resources or insufficient coverage of data collection, and further reducing the comprehensive efficiency of traffic risk identification.

[0006] To achieve the above-mentioned purpose, the application adopts the following technical solutions: In a first aspect, a traffic operation risk real-time identification method based on low-altitude unmanned aerial vehicles is provided, comprising: obtain road section basic data of each road section in the to-be-monitored area, the road section basic data including a road section type, historical accident data, and historical traffic flow data; the road section type including a structure type and a speed type of the road section, the structure type including a straight road section, a crossroad section, and an uphill road section, and the speed type including a fast road section and a slow road section; the historical accident data including an accident type and a position corresponding to the accident type, and a number of the accident type; and the historical traffic flow data including a time period traffic flow and a time period speed of the corresponding road section in each time period; set a road section danger score of each road section based on the road section basic data, and set a UAV distribution scheme based on the road section danger score; obtain monitoring video data of the UAV, and analyze the monitoring video data to obtain an accident risk score of each traffic subject of the road section; issue an alarm based on the accident risk score.

[0007] In the traffic operation risk real-time identification method based on low-altitude UAVs provided in the present application, the road section basic data of each road section in the to-be-monitored area is obtained, the road section danger score of each road section is set according to the road section basic data, the UAV distribution scheme is set according to the road section danger score, the monitoring video data of the UAV is obtained, the accident risk score of each traffic subject of the road section is obtained according to the monitoring video data, and the alarm is issued according to the accident risk score. Different road sections are analyzed and set for the number of monitoring UAVs, so as to reduce the waste of UAV resources, ensure the coverage of data collection, and improve the comprehensive efficiency of traffic risk identification.

[0008] In combination with the first aspect, in a possible implementation manner, the setting of the road section danger score of each road section based on the road section basic data includes: extracting the road section type, the historical accident data, and the historical traffic flow data in the road section basic data, extracting the number of accidents in the historical accident data, and extracting the time period traffic flow and the time period speed of each time period in the historical traffic flow data; finding a road section type influence factor corresponding to the road section type in a road section type influence factor lookup table based on the road section type; the road section type influence factor is an influence of the structure and the present of the road section on the accident; obtaining the basic danger score corresponding to the road section by substituting the number of accidents and the road section type influence factor into a set road section object danger quantification function; one form of the road section object danger quantification function is:

[0009] wherein KP is the basic danger score, LY is the road section type influence factor, and SC is the number of accidents of the corresponding road section; The period traffic flow and the period traffic speed are substituted into a set road section main body danger quantification function to obtain a period danger score of the road section corresponding to the period; the road section main body danger quantification function is:

[0010] Wherein, ZP is the period danger score, SCL is the period traffic flow, BCL is a set standard period traffic flow; SCS is the period traffic speed, BCS is a set standard period traffic speed; The product of the basic danger score and the period danger score is taken as a road section danger score of the road section in the period.

[0011] In combination with the above first aspect, in a possible implementation manner, one construction manner of the road section type influence factor lookup table includes: The number of accidents in historical accident data of a plurality of road sections in a to-be-monitored area is obtained; the number of accidents of road sections of the same type is summed to obtain a total number of accidents corresponding to the road section type; The ratio of the total number of accidents to a set standard number of accidents is taken as a road section type influence factor corresponding to the road section type; Each road section type and the corresponding road section type influence factor are integrated into a road section type influence factor lookup table.

[0012] In combination with the above first aspect, in a possible implementation manner, the unmanned aerial vehicle distribution scheme is set based on the road section danger score, including: The road section danger score corresponding to each road section in the to-be-monitored area in the same period is obtained; the road section danger score of each road section is normalized to obtain a normalized road section danger score; The number of available unmanned aerial vehicles is obtained; the number of available unmanned aerial vehicles is divided based on the road section danger score corresponding to each road section to obtain the number of unmanned aerial vehicles for each road section; An unmanned aerial vehicle distribution scheme is generated based on the number of unmanned aerial vehicles for each road section.

[0013] In combination with the above first aspect, in a possible implementation manner, the accident risk score of each traffic main body of the road section is analyzed based on the monitoring video data, including: The monitoring video is frame-extracted to obtain a plurality of monitoring frame images; the monitoring frame images are input into a traffic main body identification model to obtain a main body label and a main body position of a plurality of traffic main bodies; the traffic main body label includes a main body category label and a main body feature label; the traffic main body identification model is obtained through artificial intelligence model training; obtaining subject labels and subject positions of a plurality of traffic subjects in each monitoring frame image in sequence; marking traffic subjects corresponding to the same traffic subject label as the same traffic subject; fitting a plurality of subject positions corresponding to the same traffic subject in the order of frame time of the monitoring frame image into a trajectory graph corresponding to the traffic subject; the trajectory graph comprising position coordinates and time coordinates; predicting a predicted trajectory in a future period of time through the trajectory graph; obtaining a predicted trajectory corresponding to each traffic subject; and analyzing the predicted trajectory of each traffic subject to obtain an accident risk score of each traffic subject.

[0014] In combination with the first aspect, in a possible implementation manner, the traffic subject identification model is obtained through artificial intelligence model training, and comprises: obtaining a plurality of monitoring frame images, marking a plurality of traffic subjects in the monitoring frame images, and setting subject labels and subject positions corresponding to the traffic subjects; the traffic subjects are subjects participating in traffic, including people and vehicles; the subject labels comprise subject type labels and subject feature labels; the subject type labels are types of the traffic subjects, including pedestrians, buses, cyclists and small cars; the subject features are features of the traffic subjects, including license plate numbers of the traffic subjects or clothing features of pedestrians; integrating the plurality of monitoring frame images, the subject labels and the subject positions into a plurality of training data and verification data; training the artificial intelligence model using the training data, verifying the artificial intelligence model after training using the verification data, and finally obtaining the subject labels and the subject positions of the plurality of traffic subjects in the monitoring frame images as the input and the output; wherein the artificial intelligence model comprises a recurrent neural network model.

[0015] In combination with the first aspect, in a possible implementation manner, the accident risk score of each traffic subject is obtained based on the predicted trajectory of each traffic subject, and comprises: obtaining the predicted trajectory of each traffic subject, and generating a plurality of time coordinate groups based on the predicted trajectory of each traffic subject; calculating the Euclidean distance between each coordinate point in the same time coordinate group, setting the traffic subjects corresponding to two coordinate points with an Euclidean distance less than a set distance threshold as risk subjects, and marking the coordinate points as risk coordinates of the corresponding risk subjects, and marking the collection time of the risk coordinates as the risk collection time of the corresponding risk subjects; obtaining a plurality of risk coordinates corresponding to the risk subject and a risk collection time corresponding to the risk coordinates; marking a minimum value of a time difference between each risk collection time and a current time as FT; obtaining a distance between the risk coordinates and a current coordinate of the analysis subject, and marking the distance as FL; substituting the minimum value FT and the distance FL into a set risk quantification function to obtain an accident risk score corresponding to the risk subject; the risk quantification function is:

[0016] wherein SGP is the accident risk score corresponding to the risk subject; JT is a set interval duration, and ZL is a shortest reaction distance; the shortest reaction distance is set according to expert experience.

[0017] In combination with the first aspect, in a possible implementation manner, the generating a plurality of time coordinate groups based on the predicted trajectories of the traffic subjects comprises: obtaining a set interval duration, generating a plurality of collection times according to the interval duration; collecting on the predicted trajectories according to the collection times to obtain subject coordinates corresponding to the traffic subjects at the collection times, and dividing the subject coordinates of each traffic subject at the same collection time into the same time coordinate group.

[0018] In combination with the first aspect, in a possible implementation manner, the generating a plurality of time coordinate groups based on the predicted trajectories of the traffic subjects comprises: obtaining an accident risk score of each risk subject, and sequentially performing an alarm according to an order of the accident risk scores from large to small, wherein the alarm comprises obtaining a subject label of a risk subject corresponding to the accident risk score; extracting a subject feature label and a subject type label in the subject label; generating a voice alarm signal based on the subject feature label and the subject type label, and performing an alarm using the voice alarm signal.

[0019] In a second aspect, the present application provides a traffic operation risk real-time identification device based on a low-altitude unmanned aerial vehicle, comprising a processor and a storage medium; the storage medium comprises instructions, and the processor is configured to run the instructions to implement the method described in the first aspect and any possible implementation manner of the first aspect. The traffic operation risk real-time identification device based on the low-altitude unmanned aerial vehicle can be an electronic device, or a chip in the electronic device.

[0020] In a third aspect, the present application provides a computer readable storage medium, and the computer readable storage medium stores instructions, when the instructions run on the traffic operation risk real-time identification device based on the low-altitude unmanned aerial vehicle, the traffic operation risk real-time identification device based on the low-altitude unmanned aerial vehicle executes the method described in the first aspect and any possible implementation manner of the first aspect.

[0021] In a fourth aspect, the present application provides a computer program product comprising instructions which, when the computer program product is executed on the low-altitude unmanned aerial vehicle-based traffic operation risk real-time identification device, cause the low-altitude unmanned aerial vehicle-based traffic operation risk real-time identification device to perform the method described in the first aspect and any possible implementation manner of the first aspect.

[0022] The present application provides a low-altitude unmanned aerial vehicle-based traffic operation risk real-time identification method, which can obtain road segment basic data of each road segment in a to-be-monitored area, set road segment danger scores of each road segment according to the road segment basic data, set an unmanned aerial vehicle allocation scheme according to the road segment danger scores, obtain monitoring video data of the unmanned aerial vehicle, analyze the monitoring video data to obtain accident risk scores of each traffic subject of the road segment, perform an alarm according to the accident risk scores, set the number of unmanned aerial vehicles used for monitoring for different road segments through targeted analysis of the different road segments, so as to reduce waste of unmanned aerial vehicle resources, ensure the coverage of data collection, and further improve the comprehensive efficiency of traffic risk identification.

[0023] It should be understood that the description of technical features, technical solutions, beneficial effects or similar language in the present application does not imply that all features and advantages can be achieved in any single embodiment. On the contrary, it can be understood that the description of a feature or a beneficial effect means that the specific technical feature, technical solution or beneficial effect is included in at least one embodiment. Therefore, the description of technical features, technical solutions or beneficial effects in the specification does not necessarily refer to the same embodiment. Further, the technical features, technical solutions and beneficial effects described in the embodiments can be combined in any appropriate manner. Those skilled in the art will understand that the embodiments can be implemented without one or more specific technical features, technical solutions or beneficial effects of a specific embodiment. In other embodiments, additional technical features and beneficial effects can be identified in specific embodiments that do not embody all embodiments. BRIEF DESCRIPTION OF DRAWINGS

[0024] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the description of the embodiments or the prior art will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and those skilled in the art can obtain other drawings according to these drawings without creating any creative labor.

[0025] Fig. 1 A schematic diagram of the steps of the traffic operation risk real-time identification method in the present application; Fig. 2 A schematic diagram of the steps of generating the unmanned aerial vehicle allocation scheme in the present application. DETAILED DESCRIPTION

[0026] The technical solutions of the present application will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work belong to the scope of protection of the present application.

[0027] Please refer to Figs. 1-2 , a low-altitude unmanned aerial vehicle-based traffic operation risk real-time identification method is provided, comprising: Obtaining road section basic data of each road section in a to-be-monitored area, the road section basic data comprising road section types, historical accident data and historical traffic flow data; the road section types comprising structural types and speed types of the road section, the structural types comprising straight road sections, intersection road sections and uphill road sections, etc., and the speed types comprising fast road sections and slow road sections, etc.; the historical accident data comprising accident types and positions corresponding to accidents occurring in the road section, and the number of accident types; and the historical traffic flow data comprising time period traffic flow and time period speed of the road section in each time period; Setting road section danger scores of each road section based on the road section basic data; the road section danger score is the risk of accidents occurring in the road section in a time period, and the greater the risk of accidents, the higher the corresponding road section danger score; setting an unmanned aerial vehicle distribution scheme based on the road section danger score; the unmanned aerial vehicle distribution scheme is to distribute unmanned aerial vehicles to each road section according to the road section danger score, and the greater the danger score of the road section, the more unmanned aerial vehicles corresponding to the road section, so as to ensure the monitoring effect of the unmanned aerial vehicles; Obtaining monitoring video data of the unmanned aerial vehicles, the monitoring video data being video data collected by a camera arranged on the unmanned aerial vehicle; analyzing the road section accident risk scores of each traffic subject based on the monitoring video data; the traffic subject being a subject participating in traffic, including people and cars, etc.; the accident risk score being a score of the risk of accidents of the traffic subject according to the running track of the traffic subject in the monitoring video data, and the greater the accident risk score, the greater the risk of accidents of the traffic subject in a short time; Performing an alarm based on the accident risk score.

[0028] Based on the above technical scheme, in the traffic operation risk real-time identification method based on low-altitude unmanned aerial vehicle provided in the application, the road section basic data of each road section in the to-be-monitored area is acquired, the road section danger score of each road section is set according to the road section basic data, the unmanned aerial vehicle distribution scheme is set according to the road section danger score, the monitoring video data of the unmanned aerial vehicle is acquired, the accident risk score of each traffic subject of the road section is analyzed according to the monitoring video data, and the alarm is performed according to the accident risk score. The number of unmanned aerial vehicles used for monitoring different road sections is set through targeted analysis of different road sections, so that the waste of unmanned aerial vehicle resources is reduced, the coverage of data acquisition is ensured, and the comprehensive efficiency of traffic risk identification is improved.

[0029] In a possible implementation manner, the setting of the road section danger score of each road section based on the road section basic data comprises: extracting the road section type, historical accident data and historical traffic flow data in the road section basic data; extracting the accident frequency in the historical accident data and the time period traffic flow and time period speed of each time period in the historical traffic flow data; and the accident frequency is the accident frequency of the road section in a set time, for example, the accident frequency in one year or three years. In this embodiment, one day is divided into 24 time periods, one hour for one time period, the traffic flow and average speed of each time period in a set number of days are counted, the traffic flow is the number of vehicles passing through the time period, and the average speed is the average value of the speed of each vehicle passing through the road section under the green light; the maximum value of the traffic flow in the same time period in each day in a set number of days is taken as the time period traffic flow of the time period, and the maximum value of the average speed in the same time period in each day in a set number of days is taken as the time period speed of the time period; The road section type influence factor corresponding to the road section type is found in the road section type influence factor lookup table based on the road section type; the road section type influence factor is the influence of the structure of the road section and the present on the occurrence of accidents; The accident frequency and the road section type influence factor are substituted into a set road section object danger quantification function to obtain the basic danger score corresponding to the road section; one form of the road section object danger quantification function is:

[0030] In the formula, KP is the basic danger score, LY is the road section type influence factor, and SC is the accident frequency of the corresponding road section. In this embodiment, the function quantifies the accident occurrence probability caused by the road section type and the structure of the road section itself. When the road section type influence factor of the road section type is larger, it is indicated that the result of the road section type causes the road section to be prone to accidents, and therefore the basic danger score value of the road section is larger. When the accident in the history of the road section is more, it is indicated that the road section is prone to accidents, and the monitoring needs to be strengthened, and therefore the basic danger score value of the road section is larger.

[0031] The period traffic flow and the period traffic speed are substituted into the set road main body danger quantification function to obtain a period danger score of the road in the period; the road main body danger quantification function is:

[0032] In the embodiment, the BCL period traffic flow is set according to the length of the set period, and the standard period traffic flow is 300 in the case of a period of 1 hour; the standard period speed can be the speed limit of the road; in the embodiment, the influence of the historical traffic flow and the historical speed in each period on the accident is quantified, and the more the traffic flow in a period, the more the number of vehicles in the period, and the higher the probability of the accident, so the period danger score in the period is set to be larger; the larger the average speed in a period, the higher the probability of the accident, so the period danger score in the period is set to be larger.

[0033] The product of the basic danger score and the period danger score is taken as a road danger score of the road in the period.

[0034] In a possible implementation, one construction manner of the road type influence factor lookup table includes: obtaining the number of accidents in historical accident data of a plurality of roads in a monitoring area; summing the number of accidents of roads of the same type to obtain a total number of accidents corresponding to the road type; calculating a ratio of the total number of accidents and a set standard number of accidents as a road type influence factor corresponding to the road type; integrating each road type and the corresponding road type influence factor into the road type influence factor lookup table.

[0035] The embodiment obtains the road type influence factor of each type of road by analyzing the number of accidents caused by the structure of different roads, and the more the number of accidents of a certain type of road, the greater the influence of the type of the road on the accident, and the greater the corresponding road type factor.

[0036] In a possible implementation, the unmanned aerial vehicle distribution scheme is set based on the road danger score, including: obtaining a road danger score corresponding to each road in a monitoring area in the same period; performing normalization processing on each road danger score to obtain a normalized road danger score; Obtaining the number of available drones; dividing the number of available drones based on the road segment danger scores corresponding to each road segment to obtain the number of drones for each road segment; for example, multiplying the normalized road segment danger score by the number of available drones, rounding the multiplication result to obtain the number of drones to be set for the road segment, the rounding method can be a method of discarding four and taking five, or a method of rounding using gas, it should be noted that the sum of the number of drones for each road segment after rounding is the same as the number of available drones; generating a drone allocation scheme based on the number of drones for each road segment.

[0037] In a possible implementation, the obtaining of the accident risk score of each traffic body based on the monitoring video data analysis includes: frame extraction of the monitoring video to obtain a plurality of monitoring frame images; inputting the monitoring frame images into a traffic body recognition model to obtain body labels and body positions of a plurality of traffic bodies; the traffic body label includes a body category label and a body feature label; the traffic body recognition model is obtained by training an artificial intelligence model; sequentially obtaining the body labels and the traffic body positions of a plurality of traffic bodies in each monitoring frame image; marking the traffic bodies with the same traffic body label as the same traffic body; fitting a plurality of body positions corresponding to the same traffic body into a trajectory graph corresponding to the traffic body according to the order of frame time of the monitoring frame image; the trajectory graph includes position coordinates and time coordinates, for example, the coordinates of each point in the trajectory graph are (X, Y, T), wherein X and Y are position coordinates, and T is a time coordinate; predicting a predicted trajectory in a future period of time through the trajectory graph; the trajectory prediction can be performed by training an artificial intelligence model, and the trajectory in a certain period of time in the future is predicted through the forward direction and speed of the traffic body; obtaining the predicted trajectory corresponding to each traffic body; analyzing the accident risk score of each traffic body based on the predicted trajectory of each traffic body.

[0038] In a possible implementation, the traffic body recognition model is obtained by training an artificial intelligence model, including: obtaining a plurality of monitoring frame images, marking a plurality of traffic bodies in the monitoring frame images, and setting body labels and body positions corresponding to the traffic bodies; the traffic body is a subject participating in traffic, including people and vehicles; the body label includes a body type label and a body feature label; the body type label is the type of the traffic body, including pedestrians, buses, cyclists and small cars; the body feature is the feature of the traffic body, including the license plate number of the traffic body or the clothing feature of the pedestrian; integrating the plurality of monitoring frame images, the body labels and the body positions into a plurality of training data and verification data; The artificial intelligence model is trained using training data, and the trained artificial intelligence model is tested using test data, and finally an input is a monitoring frame image, and an output is subject labels and subject positions of a plurality of traffic subjects in the monitoring frame image; wherein the artificial intelligence model comprises a recurrent neural network model and the like.

[0039] In a possible implementation, the analyzing the predicted trajectory of each traffic subject to obtain an accident risk score of each traffic subject comprises: obtaining the predicted trajectory of each traffic subject, and generating a plurality of time coordinate groups based on the predicted trajectory of each traffic subject; calculating the Euclidean distance between each coordinate point in the same time coordinate group, and setting the two coordinate points corresponding to the traffic subjects as risk subjects if the Euclidean distance is less than a set distance threshold; it can be understood that the distance threshold in this embodiment is less than half the lane width; and marking the coordinate points as risk coordinates of the corresponding risk subjects, and marking the collection time of the risk coordinates as the risk collection time of the corresponding risk subjects; obtaining a plurality of risk coordinates corresponding to the risk subjects and the risk collection time corresponding to the risk coordinates; marking the minimum value of the time difference between each risk collection time and the current time as FT; obtaining the distance between the risk coordinates and the current coordinates of the analysis subject; it can be understood that the coordinates of the traffic subjects in this embodiment are set relative to the road section, for example, the set road section is a crossroad section; at this time, an equal-size rectangular coordinate system is established according to the crossroad section; at this time, the X and Y coordinates in the coordinates are the coordinates in the rectangular coordinate system; and marking the distance as FL; substituting the minimum value FT and the distance FL into a set risk quantification function to obtain the accident risk score corresponding to the risk subject; the risk quantification function is:

[0040] wherein SGP is the accident risk score corresponding to the risk subject; JT is a set interval length, and ZL is the shortest reaction distance; the shortest reaction distance is set according to expert experience, and the shortest reaction distance in this embodiment is set to 10 m; the farther the distance between the risk subject and the risk coordinate, or the longer the time interval to the risk coordinate, the more reaction time is left for the risk subject, and the greater the possibility that the risk subject avoids the accident after receiving the alarm and reacting; at this time, the corresponding accident risk score is set to be smaller.

[0041] In a possible implementation, the generating a plurality of time coordinate groups based on the predicted trajectory of each traffic subject comprises: obtaining a set interval length, generating a plurality of collection times according to the interval length; collecting the subject coordinates of the traffic subjects corresponding to the collection times on the predicted trajectory to obtain the subject coordinates of the traffic subjects corresponding to the collection times; and dividing the subject coordinates of each traffic subject at the same collection time into the same time coordinate group.

[0042] In a possible implementation, the alerting based on the accident risk score comprises: obtaining the accident risk score of each risk subject, and sequentially alerting in descending order of the accident risk score, the alerting comprising obtaining the subject label of the risk subject corresponding to the accident risk score; extracting the subject feature label and the subject type label in the subject label; generating a voice alarm signal based on the subject feature label and the subject type label, and using the voice alarm signal to alert; For example, if the subject feature label in the subject label is "red shirt" and the subject type label is "pedestrian", the voice alarm signal at this time is "red shirt, please be careful of the nearby vehicle"; the voice alarm signal can be obtained by training an artificial intelligence model, and it is worth noting that the direction alarm can be generated based on the positional relationship between the current coordinates of the traffic subject and the risk coordinates, and the positional relationship between the risk coordinates of another traffic subject and the set distance threshold. Specifically, the traffic subject is taken as the broadcast subject, the forward direction of the broadcast subject is taken as the front, and when the risk coordinates of another traffic subject are located in front of the broadcast subject, the direction alarm is front; when the risk coordinates of another traffic subject are located behind the broadcast subject, the direction alarm is rear.

[0043] Some data in the above formula is calculated by removing the dimension, and the formula is obtained by software simulation of a large amount of collected data to obtain a formula closest to the real situation; the preset parameters and the preset threshold in the formula are set by a person skilled in the art according to the actual situation or obtained by a large amount of data simulation.

[0044] Working principle of the present application: By obtaining the road segment basic data of each road segment in the to-be-monitored area, setting the road segment danger score of each road segment according to the road segment basic data, setting the unmanned aerial vehicle distribution scheme according to the road segment danger score, obtaining the monitoring video data of the unmanned aerial vehicle, analyzing the accident risk score of each traffic subject of the road segment according to the monitoring video data, alerting according to the accident risk score, and setting the number of unmanned aerial vehicles used for monitoring for different road segments through targeted analysis of different road segments, the waste of unmanned aerial vehicle resources is reduced, the coverage of data collection is ensured, and the comprehensive efficiency of traffic risk identification is improved.

[0045] The above embodiments are only used to illustrate the technical method of the present application but not limit the present application. Although the present application is described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present application can be modified or replaced equivalently without departing from the spirit and scope of the technical method of the present application.

Claims

1. A method for real-time identification of traffic operation risks based on low-altitude unmanned aerial vehicles (UAVs), characterized in that, include: Acquire basic road segment data for each road segment within the area to be monitored. The basic road segment data includes road segment type, historical accident data, and historical traffic flow data. Based on the basic data of each road segment, a road segment hazard score is set for each road segment; A drone allocation scheme is set based on road segment hazard assessment; Acquire surveillance video data from drones, and analyze the surveillance video data to obtain accident risk scores for each traffic entity on the road segment; Alerts are issued based on accident risk scores.

2. The method for real-time identification of traffic operation risks based on low-altitude unmanned aerial vehicles according to claim 1, characterized in that, The method of setting road segment hazard scores for each road segment based on basic road segment data includes: Extract road segment type, historical accident data, and historical traffic flow data from the basic road segment data; extract the number of accidents from the historical accident data; and extract the traffic flow and speed for each time period from the historical traffic flow data. Based on the road segment type, the road segment type influence factor corresponding to the road segment type is searched in the road segment type influence factor lookup table; the road segment type influence factor is the impact of the road segment's structure and current condition on the occurrence of accidents; Substitute the accident frequency and road segment type influencing factors into the set road segment object hazard quantification function to obtain the basic hazard score corresponding to the road segment; Substituting the traffic flow and speed during the time period into the set road segment main hazard quantification function, the time-period hazard score of the road segment corresponding to the time period is obtained; The product of the basic hazard score and the time-period hazard score is taken as the road segment hazard score for the time period.

3. The method for real-time identification of traffic operation risks based on low-altitude unmanned aerial vehicles according to claim 2, characterized in that, One method for constructing the road segment type influence factor lookup table includes: Obtain the number of accidents from historical accident data for several road sections within the area to be monitored; sum the number of accidents for road sections of the same type to obtain the total number of accidents corresponding to that road section type; The ratio of the total number of accidents to the set standard number of accidents is used as the road segment type influence factor corresponding to the road segment type. Each road segment type and its corresponding road segment type influencing factor are integrated into a road segment type influencing factor lookup table.

4. The method for real-time identification of traffic operation risks based on low-altitude unmanned aerial vehicles according to claim 1, characterized in that, The drone allocation scheme based on road segment hazard assessment includes: Obtain the road segment hazard score corresponding to each road segment in the monitoring area during the same time period; normalize each road segment hazard score to obtain the normalized road segment hazard score; Obtain the number of available drones; divide the number of available drones into the number of drones for each road segment based on the road segment hazard score; generate a drone allocation scheme based on the number of drones for each road segment.

5. The method for real-time identification of traffic operation risks based on low-altitude unmanned aerial vehicles according to claim 1, characterized in that, The accident risk scores for each traffic entity on the road segment obtained based on the analysis of surveillance video data include: The surveillance video is frame-by-frame extracted to obtain several surveillance frame images; these images are then input into a traffic subject recognition model to obtain subject labels and subject locations for several traffic subjects; the traffic subject labels include subject category labels and subject feature labels; the traffic subject recognition model is trained using an artificial intelligence model. Sequentially obtain the subject labels and positions of several traffic subjects in each monitoring frame image; mark the traffic subjects corresponding to the same traffic subject label as the same traffic subject; fit the positions of several subjects corresponding to the same traffic subject into a trajectory map corresponding to the traffic subject according to the frame time sequence of the monitoring frame images; Predict the trajectory over a future period using trajectory maps; obtain the predicted trajectory for each traffic entity; and obtain the accident risk score for each traffic entity based on the analysis of the predicted trajectory.

6. The method for real-time identification of traffic operation risks based on low-altitude unmanned aerial vehicles according to claim 5, characterized in that, The traffic entity identification model is obtained through training an artificial intelligence model, including: Acquire several monitoring frame images, mark several traffic entities in the monitoring frame images, and set corresponding entity labels and entity positions for the traffic entities; the entity labels include entity type labels and entity feature labels; the entity type label is the type of traffic entity; the entity feature is the feature of the traffic entity; integrate the several monitoring frame images, entity labels, and entity positions into several training data and validation data; The artificial intelligence model is trained using training data and tested using testing data. The final input is a monitoring frame image, and the output is the subject label and subject position of several traffic subjects in the monitoring frame image.

7. The method for real-time identification of traffic operation risks based on low-altitude unmanned aerial vehicles according to claim 5, characterized in that, The accident risk score for each traffic entity is obtained from the predicted trajectory analysis of each traffic entity, including: Obtain the predicted trajectories of each traffic entity, and generate several time coordinate groups based on the predicted trajectories of each traffic entity; Calculate the Euclidean distance between each coordinate point in the same time coordinate group, and set the traffic entities corresponding to two coordinate points whose Euclidean distance is less than a set distance threshold as risk entities, and mark the coordinate points as the risk coordinates of the corresponding risk entities, and mark the collection time of the risk coordinates as the risk collection time of the corresponding risk entities; Obtain several risk coordinates corresponding to the risk subject, and the risk collection time corresponding to the risk coordinates; mark the minimum time difference between each risk collection time and the current time as FT; obtain the distance between the risk coordinates and the current coordinates of the analysis subject, and mark the distance as FL; substitute the minimum value FT and the distance FL into the set risk quantification function to obtain the accident risk score corresponding to the risk subject.

8. The method for real-time identification of traffic operation risks based on low-altitude unmanned aerial vehicles according to claim 7, characterized in that, The generation of several time coordinate sets based on the predicted trajectories of each traffic entity includes: The set interval duration is obtained, and several collection times are generated according to the interval duration. The coordinates of the traffic subject corresponding to the collection time are obtained by collecting data on the predicted trajectory according to the collection time. The coordinates of each traffic subject at the same collection time are divided into the same time coordinate group.

9. The method for real-time identification of traffic operation risks based on low-altitude unmanned aerial vehicles according to claim 1, characterized in that, The alert based on accident risk scoring includes: Obtain the accident risk score for each risk subject and issue alerts in descending order of accident risk score.

10. The method for real-time identification of traffic operation risks based on low-altitude unmanned aerial vehicles according to claim 9, characterized in that, The alarm includes: obtaining the subject tag of the risk subject corresponding to the accident risk score; extracting the subject feature tag and subject type tag from the subject tag; generating a voice alarm signal based on the subject feature tag and subject type tag, and using the voice alarm signal to issue an alarm.

Images