Road section traffic flow parameter estimation method based on aerial photography data of mobile unmanned aerial vehicle

Abstract

The invention discloses a road section traffic flow parameter estimation method based on aerial data of a mobile unmanned aerial vehicle, and the method obtains traffic information through the mobile unmanned aerial vehicle, proposes a traffic state estimation model based on the unmanned aerial vehicle, considers the traffic flow operation state in a queuing scene, achieves the precise conversion of flow and density data in time duration, and improves the estimation precision. Further calculating to obtain a road section traffic flow parameter estimation value; on the basis, a traffic flow parameter error distribution model suitable for various flight parameters is constructed and used for guiding unmanned aerial vehicle parameter setting scheme decisions under different road section types, so that the road section traffic flow parameter estimation precision is improved. According to the method, the limitation of flow estimation based on a fixed detector in a traditional method is made up, the method has a promoting effect on the aspects of improving traffic operation efficiency, relieving traffic congestion and the like, and auxiliary decision making information and reference basis are provided for a traffic management department to formulate and optimize a traffic signal timing scheme and reduce traffic management cost.

Description

Technical Field

[0001] The present invention relates to the field of traffic information and control, and particularly relates to a method for estimating road traffic flow parameters based on mobile unmanned aerial vehicle (UAV) aerial photography data. Background Art

[0002] With the continuous breakthroughs in new technologies such as artificial intelligence, big data, and autonomous driving, the transportation industry is undergoing a revolutionary transformation, profoundly changing the connotation and methodology of traffic engineering disciplines, and a large number of cutting-edge hot scientific and technological issues have emerged.

[0003] Traffic flow estimation is a key component of effective traffic management and congestion mitigation. Accurate and timely traffic flow data is crucial for optimizing traffic signal timing, vehicle rerouting, and implementing dynamic traffic control strategies. Most existing research on dynamic traffic flow estimation is based on fixed detectors. Although traffic data for a certain section can be observed for a long time, there are usually problems such as less coverage and low flexibility in data collection.

[0004] For example, fixed detectors can only monitor traffic conditions at specific locations, and data cannot be obtained for areas outside the coverage of the detectors. It is also difficult to flexibly adjust the monitoring location and time according to actual needs. Although Global Positioning System (GPS) data can provide continuous spatio-temporal observations of vehicle movement and the speed estimation is relatively accurate, due to its low penetration rate, it cannot represent the true distribution of the entire traffic flow. Roadside fixed sensors such as Bluetooth, WiFi, cameras, RFID, etc. can provide traffic data with high penetration rate, but their spatial coverage is limited.

[0005] The development of unmanned aerial vehicle (UAV) technology provides a new solution for traffic flow estimation. It can effectively solve the limitations of ground sensors and provide a larger sample size and a wider spatial coverage. However, UAV observations have limitations in time coverage. There are time discontinuities in the observations of ground vehicle trajectories during its cruise, and the time interval between adjacent flights may be relatively large, resulting in the lack of vehicle trajectory data for some time periods. If the UAV observation data is directly used for simple calculations to estimate the road section flow, large errors will occur. Therefore, how to effectively utilize UAV observation data, solve its time discontinuity problem, and achieve high-precision road network flow estimation is an urgent problem to be solved in the current traffic field. Summary of the Invention

[0006] Objective of the present invention: to provide a method for estimating traffic flow parameters of road sections based on mobile UAV aerial photography data. This method aims to improve the reliability of UAV mobile aerial photography data, which is beneficial to improving the accuracy of estimating the traffic flow of large-scale road network road sections and OD traffic flow in subsequent research. Specifically, the present invention uses a traffic state estimation model based on UAVs (TSE-UAV) to handle the discontinuity of UAV cruise data, and converts the traffic flow parameter values estimated based on UAV observation data in a short time period into long time period data suitable for dynamic traffic assignment. This method can fill in the gaps in the data and represent the traffic flow more continuously.

[0007] The present invention solves the problems existing in the existing traffic flow estimation methods, such as low flexibility of data collection, limited coverage, and high collection costs. It also solves the problem of large estimation errors caused by fragmented observation data when estimating traffic flow parameters based on UAV observation data, bringing significant economic benefits to the whole process of traffic management, assisting in guiding dynamic and static traffic management decisions, providing auxiliary decision-making information and reference basis for traffic management departments to formulate and optimize measures such as traffic restrictions, vehicle number restrictions, staggered peak travel, congestion charging, and demand management related policies, and having practical significance for subsequent related research.

[0008] To achieve the above functions, the present invention designs a method for estimating traffic flow parameters of road sections based on mobile UAV aerial photography data. For the target road section with multiple lanes, the following steps S1 - S2 are executed to complete the estimation of traffic flow parameters of the target road section in multiple scenarios:

[0009] Step S1: Use UAVs to observe each lane of the target road section. For different traffic states on the target road section, construct a traffic state estimation model based on fragmented UAV aerial photography data, analyze the traffic flow observation characteristics of free flow scenarios and intersection queue scenarios, study the road section spatio-temporal diagram in combination with UAV and vehicle trajectories, and estimate the traffic flow parameters of the extended time period based on feasible vehicle trajectories.

[0010] Step S2: Divide the target road section into three types: free flow road section, congested road section, and road section queued due to signal intersections. For different road section types, set different UAV flight speeds and directions, and based on the traffic state estimation model, calculate the traffic flow parameter estimation error distribution curves for the three types of road sections: free flow road section, congested road section, and road section queued due to signal intersections.

[0011] Advantages: Compared with the prior art, the advantages of the present invention include:

[0012] Since the drone cruises along a predetermined route during the research period, there are data discontinuity characteristics in the time dimension for a specific section. Through the construction of a refined microscopic model and the analysis of typical characteristics in the queuing scenario, the present invention effectively converts short-time fragmented drone aerial data into long-time data that can be used for dynamic traffic flow estimation, improves the applicability of fragmented drone data in the dynamic traffic flow estimation scenario, and constructs an error distribution model of traffic flow parameter estimation for various types of road sections under different drone flight parameters, which helps to select a more appropriate drone flight mode for different types of road sections and further improves the accuracy of traffic flow parameter estimation based on drones. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 is a flowchart of a method for estimating road traffic flow parameters based on mobile drone aerial data according to an embodiment of the present invention;

[0014] Figure 2 is a spatio-temporal diagram of the TSE-UAV model according to an embodiment of the present invention;

[0015] Figure 3 is a schematic diagram of an optimization method of the TSE-UAV model in a queuing scenario according to an embodiment of the present invention;

[0016] Figure 4 is a summary diagram of traffic flow parameter error distribution curves for various types of road sections based on different drone flight parameters according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] The following further describes the present invention with reference to the drawings. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and should not be used to limit the protection scope of the present invention.

[0018] A method for estimating road traffic flow parameters based on mobile drone aerial data provided by an embodiment of the present invention, for a target road section with multiple lanes, referring to Figure 1 , the following steps S1-S2 are executed to complete the estimation of traffic flow parameters for the target road section in multiple scenarios:

[0019] Step S1: Use a drone (UAV) to observe each lane of the target road section. For different traffic states on the target road section, construct a traffic state estimation model based on fragmented drone aerial data (TSE-UAV model), analyze the traffic flow observation characteristics in the free flow scenario and the intersection queuing scenario, study the spatio-temporal diagram of the road section in combination with the drone and vehicle trajectories, and estimate the traffic flow parameters for an extended time period based on the feasible vehicle trajectories;

[0020] Referring to Figure 2 , the specific steps of step S1 are as follows:

[0021] Step S1.1: For the free flow scenario, the method for constructing the traffic state estimation model is as follows:

[0022] The multi-lane road section l is divided into multiple single lanes, and for one lane s, the drone observation time period θ l The coordinate system for lane s is constructed based on the spatial position of the road section, and the vehicle trajectory observed by the drone is extracted and divided into the spatiotemporal region R x (For simplicity, define x as the period of observation by the drone θ l The index of lane s on inner road segment l) for R x Each vehicle in the vehicle is numbered, and each trajectory segment of each vehicle is numbered, and a feasible trajectory set Ω is constructed using a feasible trajectory recognition algorithm. x , driving distance set D x , travel time set T x , and the area set A x ; Estimate the drone observation time period θ l The flow rate q in lane s x and density k x As follows:

[0023]

[0024] The specific method of the feasible trajectory identification algorithm described in step S1.1 is as follows:

[0025] The drone collects data during the observation period θ l Observed vehicle trajectory in lane s of inner segment l Where n is the vehicle number, n = 1, 2, ..., N, N is the total number of vehicles, i is the vehicle's trajectory segment number, i = 1, 2, ..., I, I is the total number of trajectory segments; based on the observed vehicle trajectory Construct the total vehicle trajectory set Φ x , construct a feasible trajectory set Ω x =Φ x ;

[0026] For the feasible trajectory set Ω x Each observed vehicle trajectory in If the corresponding driving time The driving time corresponding to the vehicle in front If there is no intersection (i.e., vehicle n and vehicle n-1 are not in the drone's field of view at the same time), then the observed vehicle trajectory is an infeasible trajectory, in Ω x Remove

[0027] Based on the feasible trajectory set Ω after removing the infeasible trajectories x , for Ω xEach vehicle trajectory within is calculated separately as follows:

[0028] Identify the area enclosed by the vehicle and the vehicle in front Calculate the area The driving distance of the vehicle within Driving time And the area of the area On this basis, construct a driving distance set D for traffic flow and density estimation x A driving time set T x And an area set A of the area x .

[0029] Refer to Figure 3 Step S1.2: For the intersection queuing scenario caused by signal control, the method for constructing a traffic state estimation model is as follows:

[0030] First, identify the queuing area observed by the drone, and divide the intersection queuing scenario into the following two categories:

[0031] The first category: There are parked vehicles within the observation field of view of the drone;

[0032] The second category: There are no parked vehicles within the observation field of view of the drone, but there are additional vehicles;

[0033] The definition of the additional vehicle is as follows:

[0034] When the drone is at the observation start time t0, without considering the intersection queuing vehicles at this time, combine the density estimation value obtained by the method in step S1.1 and the length of the drone position from the stop line at the intersection at t0 to estimate the number of vehicles on this lane at this time. The vehicles before this vehicle fleet are the "additional vehicles"; that is, the additional vehicles are the vehicles that have already been queuing on the lane s outside the observation field of view of the drone at the observation start time t0 of the drone. Limited by the field of view, the drone cannot directly observe the additional vehicles.

[0035] By analyzing the traffic flow characteristics of the two types of intersection queuing scenarios, adopt a traffic flow optimization method and a density optimization method to obtain a traffic flow adjustment value q adjust And a density adjustment value k adjust ;

[0036] The traffic flow optimization method described in step S1.2 is specifically as follows:

[0037] For the first type of intersection queuing scenario, search for the first parked vehicle that appears in the observation field of view of the drone, and record its first observed time t1. Calculate the number of queuing vehicles q at t1 by combining the distance between the parking position of this vehicle and the intersection stop line and the average vehicle length queue; If the drone is still in the red light state when it flies to the intersection, the number of queuing vehicles is directly obtained through observation; when the drone flies to the intersection, record the time t2; combine the time points t1 and t2, as well as the known signal timing plan and the number of queuing vehicles q at t1 queue , calculate the number of additional vehicles q extra and the additional duration t of the queuing vehicle fleet formed by it extra ; For the original traffic flow estimation value q x and the observation duration θ l are adjusted to obtain the optimized total observation duration t of this lane adjust and the traffic flow adjustment value q a dj us t ;

[0038] For the second type of intersection queuing scenario, search for the first additional vehicle that appears within the observation field of the drone, record its first observed time t4 and the observed speed v, use v to infer its parking position and starting time t3, and further calculate the number of queuing vehicles (which is the number of additional vehicles in this scenario) with this additional vehicle as the end of the queuing vehicle fleet. Combine the signal timing plan to adjust the original traffic flow estimation value q x and the observation duration θ l are adjusted to obtain the optimized total observation duration t of this lane adjust and the corresponding traffic flow value q adjust .

[0039] For the traffic flow estimation value q x and the observation duration θ l are adjusted to calculate the optimized total observation duration t of the lane adjust and the corresponding traffic flow value q adjust as follows:

[0040]

[0041] where θ l is the observation duration, q extra represents the number of additional vehicles, t extra represents the duration for the additional vehicles to form a queue, q queue represents the number of queuing vehicles, represents the total number of vehicles including vehicle x0 (the vehicle at the front of the research lane when not considering queuing vehicles at the starting moment) and the vehicles that join the queue later; δ p is a dummy variable for the first type of intersection queuing scenario, δ p =1 indicates that a parked vehicle appears within the observation field of the drone; δ e is a dummy variable for the second type of intersection queuing scenario, δ e= 1 indicates that there are no parked vehicles but there are additional vehicles within the observation field of the UAV, and vice versa for 0; δ r indicates that the signal light is in the red state when the UAV flies to the intersection, and vice versa for 0; T R is the total duration of the red light, T G is the total duration of the green light; the superscripts (1) and (2) of each variable correspond to the first type of queuing scenario and the second type of queuing scenario respectively, and the superscript t a represents the state of the variable corresponding to time a (for example represents the green light value at time a), t a-b represents the duration experienced by the variable from time a to time b (for example represents the duration of the red light experienced from time a to time b).

[0042] The density optimization method described in step S1.2 is specifically as follows:

[0043] Based on the total observation duration t adjust obtained by the traffic optimization method, the corresponding traffic value q adjust (i.e., the number of vehicles, unit: vehicle) within it is calculated, and the density adjustment value corresponding to the lane for the total observation duration t adjust is calculated as follows:

[0044]

[0045] In the formula, k adjust is the density adjustment value corresponding to the total observation duration t adjust , l length is the length of the road section.

[0046] Step S1.3: Expand the UAV observation time period θ l and establish a long-term parameter conversion algorithm for the expanded time period:

[0047]

[0048] Among them, Θ l is the total duration of the expanded time period, Q l and K l are the traffic and density corresponding to the total duration respectively, δ is a virtual variable for the intersection queuing scenario, δ = 1 indicates considering the intersection queuing scenario adjustment value, and vice versa for 0;

[0049] Calculate the speed estimation value of the target road section l within Θ l as follows:

[0050]

[0051] In the formula, V l is the speed estimation value of the target road section l within Θ l ;

[0052] The traffic flow parameters are composed of the traffic volume, density, and speed of the vehicles on the lane.

[0053] Step S2: Divide the target road section into three types: free flow road section, congested road section, and queuing road section caused by signal intersections. For different road section types, set different UAV flight speeds and directions, and based on the traffic state estimation model, calculate the traffic flow parameter estimation error distribution curves for the three types of road sections: free flow road section, congested road section, and queuing road section caused by signal intersections, respectively.

[0054] The specific steps of Step S2 are as follows:

[0055] Step S2.1: Divide the target road section into the following three types:

[0056] Type 1: Free flow road section;

[0057] Type 2: Congested road section;

[0058] Type 3: Queuing road section caused by signal intersections;

[0059] By setting different UAV flight speeds and directions, for the first and second type of road sections, use the traffic state estimation model in Step S1.1 to estimate the traffic volume, density, and speed of the vehicles on the lane;

[0060] For the third type of road section, use the traffic state estimation model in Step S1.2 to estimate the traffic volume, density, and speed;

[0061] The specific flight parameters of the UAV are set as follows: the speed is 1 times the average vehicle speed and 3 times the average vehicle speed, and the direction is the same as the vehicle driving direction and the opposite direction of the vehicle driving;

[0062] Step S2.2: Based on the flight parameter settings, use a cross - design form for speed and direction, fit to obtain the estimation error distribution curves of traffic volume, density, and speed for each type of road section under different flight parameter settings, and select the optimal UAV flight parameter setting scheme for each type of road section with the minimum RMSE as the standard.

[0063] The calculation formula of RMSE is as follows:

[0064]

[0065] where N is the total number of observations, y i and are the observed value and estimated value of the road section traffic flow parameters (including traffic volume, density, speed), respectively.

[0066] The following is an application embodiment of the present invention:

[0067] (1) Data collection and extraction

[0068] Road network data: The road network data of the city center at the 300 km quantity level in Xuancheng City, including 189 nodes and 264 road segments;

[0069] UAV aerial photography data: On September 1, 2020, the aerial photography data of 210 UAV flights on 107 road segments in Xuancheng City.

[0070] (2) Experimental design

[0071] Three typical road segments were selected for the experiment to evaluate the effectiveness of the method: 1. Free-flow road segment; 2. Congested road segment; 3. Road segment with queues caused by signal intersections.

[0072] (3) Experimental results

[0073] The traffic state estimation model based on fragmented UAV aerial photography data (TSE-UAV model) proposed by the present invention was used to process UAV observation data, and the estimation error distribution curves of traffic flow parameters for various types of road segments were constructed. The specific experimental results are shown in Figure 4 .

[0074] It can be seen from the results that the method of the present invention has high accuracy in the estimation of road segment density and flow rate, performs well on free-flow road segments and congested road segments, and the optimal UAV setting schemes corresponding to different parameter estimations can be found. However, the accuracy performance on road segments with queues caused by signal intersections is relatively low, but it is still within the acceptable range of RMSE error. Generally speaking, for free-flow road segments, the "three times speed + forward" flight mode is better; for congested road segments, the "three times speed + reverse" flight mode is better; for queuing road segments, the "reverse" flight mode is better. At this time, the influence of the UAV flight speed on the estimation results is not obvious.

[0075] Therefore, under the above experimental settings, using UAV cruise data for road segment traffic flow parameter estimation has high accuracy.

[0076] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those of ordinary skill in the art, various changes can be made without departing from the gist of the present invention.

Claims

1. A method for estimating road traffic flow parameters based on mobile UAV aerial photography data, characterized in that, For the target section with multiple lanes, perform the following steps S1 - S2 to complete the traffic flow parameter estimation of the target section under multiple scenarios: Step S1: Use drones to observe each lane of the target section. For different traffic states on the target section, construct a traffic state estimation model based on fragmented drone aerial data, analyze the traffic flow observation characteristics of free - flow scenarios and intersection queueing scenarios, study the road section spatio - temporal diagram in combination with drone and vehicle trajectories, and estimate traffic flow parameters for an extended time period based on feasible vehicle trajectories; Step S2: Divide the target section into three types: free - flow section, congested section, and queuing section caused by signal intersections. For different section types, set different drone flight speeds and directions. Based on the traffic state estimation model, calculate the traffic flow parameter estimation error distribution curves for the three types of sections: free - flow section, congested section, and queuing section caused by signal intersections, respectively.

2. The method for estimating road traffic flow parameters based on mobile UAV aerial photography data according to claim 1, wherein The specific steps of Step S1 are as follows: Step S1.1: For the free - flow scenario, the method for constructing the traffic state estimation model is as follows: Divide the multi-lane section l into multiple single lanes. For one of the lanes s, construct a coordinate system for lane s based on the UAV observation time period θ l and the spatial position of the section, extract the vehicle trajectories observed by the UAV and divide the spatio-temporal region R x , for each vehicle in R x number each vehicle, and number each trajectory segment of each vehicle, and use the feasible trajectory recognition algorithm to construct the feasible trajectory set Ω x , the travel distance set D x , the travel time set T x , and the regional area set A x ; estimate the flow q l and density k x of lane s within the UAV observation time period θ x as follows: Step S1.2: For the intersection queueing scenario caused by signal control, the method for constructing the traffic state estimation model is as follows: First, identify the queuing area observed by the drone, and divide the intersection queueing scenario into the following two categories: The first category: There are parked vehicles within the drone's observation field of view; The second category: There are no parked vehicles within the drone's observation field of view, but there are additional vehicles; By analyzing the traffic flow characteristics of two types of intersection queuing scenarios, using the flow optimization method and the density optimization method, the flow adjustment value q adjust and the density adjustment value k adjust ; Step S1.3: Expand the observation time period θ of the drone l and establish a long-term parameter conversion algorithm for the extended time period: where, Θ l is the total duration of the extended time period, Q l and K l are the flow rate and density corresponding to the total duration respectively, δ is a dummy variable for the intersection queue scenario, δ = 1 indicates that the adjusted value considering the intersection queue scenario, otherwise it is 0; Calculate Θ l Estimated speed value of the target road section l inside: where V l is the estimated speed of the target road segment l l within Θ Take the flow, density, and speed of vehicles on the lane as traffic flow parameters.

3. The method for estimating road traffic flow parameters based on mobile UAV aerial photography data according to claim 2, wherein The specific method of the feasible trajectory recognition algorithm described in Step S1.1 is as follows: The acquisition UAV within the observation time period θ l The observed vehicle trajectory on lane s of the internal road section l where n is the vehicle number, n = 1, 2, ……, N, N is the total number of vehicles, i is the trajectory segment number of the vehicle, i = 1, 2, ……, I, I is the total number of trajectory segments; based on the observed vehicle trajectory Construct the total vehicle trajectory set Φ x , construct the feasible trajectory set Ω x = Φ x ; For the set of feasible trajectories Ω x For each observed vehicle trajectory If its corresponding travel time and the corresponding travel time of its preceding vehicle have no intersection, then the observed vehicle trajectory is an infeasible trajectory and is removed from Ω x Remove Based on the set of feasible trajectories Ω after removing non-feasible trajectories x , for each vehicle trajectory in Ω x , the following calculations are performed respectively: Identify the area enclosed by the vehicle and the vehicle in front Calculate the area The driving distance of the vehicle inside Driving time And the area ; On this basis, construct a set of driving distances D for traffic flow and density estimation x 、A set of driving times T x , and a set of area areas A x .

4. The method for estimating road traffic flow parameters based on mobile UAV aerial photography data according to claim 2, wherein The definition of the additional vehicles described in Step S1.2 is as follows: Vehicles that are already in the queuing state on lane s outside the drone's observation field of view at the observation start time t0 of the drone.

5. A method for estimating road traffic flow parameters based on aerial photography data of mobile drones according to claim 2, characterized in that, The specific flow optimization method described in Step S1.2 is as follows: For the first type of intersection queuing scenario, search for the first parked vehicle that appears in the observation field of view of the drone, and record the time t1 when it is first observed. Based on the distance between the parking position of this vehicle and the stop line of the intersection, combined with the average vehicle length, calculate the number of queuing vehicles q at t1 queue ; If the traffic light is still red when the drone flies to the intersection, the number of queuing vehicles is directly obtained through observation; when the drone flies to the intersection, record the time t2; combined with the time points t1 and t2, as well as the known signal timing plan and the number of queuing vehicles q at t1 queue , calculate the additional number of vehicles q extra and the additional duration t of the queuing vehicle fleet formed extra ; Adjust the original traffic flow estimation value q x and the observation duration θ l to obtain the optimized total observation duration t adjust and the traffic flow adjustment value q a dj us t ; For the queuing scenario at the second type of intersection, search for the first additional vehicle that appears within the observation field of the UAV, record its first observed time t4 and observed speed v, use v to estimate its stopping position and starting time t3, further calculate the number of queuing vehicles with this additional vehicle as the end of the queuing fleet, and combine the signal timing plan to adjust the original traffic flow estimate q x and the observation duration θ l to obtain the optimized total observation duration t for this lane adjust and the corresponding traffic flow value q adjust .

6. The method for estimating road traffic flow parameters based on aerial photography data of mobile drones according to claim 5, wherein For the estimated flow value q x and the observation duration θ l Make adjustments to calculate the total observation duration t after lane optimization adjust and the corresponding flow value q adjust As shown in the following formula: where, θ l is the observation duration, q extra represents the number of additional vehicles, t extra represents the duration taken for the additional vehicles to form a queue, q queue represents the number of queuing vehicles, represents the total number of vehicles that joined the queue after vehicle x0 among the queuing vehicles; δ p is a dummy variable for the queuing scenario at the first type of intersection, δ p = 1 indicates that there are parked vehicles within the observation field of view of the UAV; δ e is a dummy variable for the queuing scenario at the second type of intersection, δ e = 1 indicates that there are no parked vehicles within the observation field of view of the UAV but there are additional vehicles, and vice versa; δ r indicates that the signal is red when the UAV flies to the intersection, and vice versa; T R is the total red light duration, T G is the total green light duration; the superscripts (1) and (2) of each variable correspond to the first type of queuing scenario and the second type of queuing scenario respectively, and the superscript t a represents the state of the variable at time a, t a-b represents the duration experienced by the variable from time a to time b.

7. The method for estimating road traffic flow parameters based on mobile UAV aerial photography data according to claim 2, wherein The specific density optimization method described in Step S1.2 is as follows: The total observation duration t obtained based on the traffic optimization method adjust corresponding traffic value q within adjust , calculate the density adjustment value corresponding to the total observation duration t of the lane adjust : where k adjust is the density adjustment value corresponding to the total observation duration t adjust , and l length is the road section length.

8. The method for estimating road traffic flow parameters based on aerial photography data of mobile drones according to claim 2, wherein, The specific steps of Step S2 are as follows: Step S2.1: Divide the target section into the following three types: The first category: Free - flow section; The second category: Congested section; The third category: Queuing section caused by signal intersections; By setting different drone flight speeds and directions, for the first and second types of sections, use the traffic state estimation model of Step S1.1 to estimate the flow, density, and speed of vehicles on the lane; For the third type of section, use the traffic state estimation model of Step S1.2 to estimate the flow, density, and speed; The specific flight parameters of the drone are set as follows: The speed is 1 times the average vehicle speed and 3 times the average vehicle speed, and the direction is the same as the vehicle driving direction and the opposite of the vehicle driving direction; Step S2.2: Based on the flight parameter settings, use a cross - design form for speed and direction to fit the estimation error distribution curves of flow, density, and speed for different types of sections with different flight parameter settings, and select the optimal drone flight parameter setting scheme for each type of section based on the minimum value of RMSE.

9. A method for estimating road traffic flow parameters based on mobile drone aerial photography data according to claim 8, characterized in that, The calculation formula of RMSE in Step S2.2 is as follows: where N is the total number of observations, and y i and are the observed value and the estimated value of the traffic flow parameters of the road section, respectively.

Images