A vehicle queue length dynamic estimation method and system based on unmanned aerial vehicle cruising

Abstract

The application discloses a kind of vehicle queuing length dynamic estimation method and system based on unmanned aerial vehicle cruise, belong to traffic state monitoring technical field, for solving technical problem: how to realize the accurate identification of queuing vehicle, tracking and panoramic scene construction, and solve the vehicle position deviation problem caused by time difference under dynamic traffic flow, improve the accuracy and stability of vehicle queuing length estimation.Method includes: through the cruise video data of congestion section of inspection unmanned aerial vehicle collection;Image processing is carried out to cruise video data, and the vehicle characteristic parameter of congestion section is extracted;Lane queuing vehicle discrimination model based on vehicle speed and straight line vehicle spacing two indexes is constructed;The ordered set of queuing vehicle in congestion section is obtained by the model;According to the ordered set of queuing vehicle, the initial queuing length of each lane is determined;Based on space-time compensation correction algorithm, the initial queuing length is dynamically corrected, and the real-time queuing length of each lane in congestion section is obtained.

Description

Technical Field

[0001] This invention relates to the field of traffic condition monitoring technology, and in particular to a method and system for dynamically estimating vehicle queue length based on unmanned aerial vehicle (UAV) patrols. Background Technology

[0002] In modern urban traffic management and the construction of intelligent transportation systems, accurate monitoring of vehicle driving status and real-time analysis of vehicle queuing conditions on long-distance road sections are key to improving road traffic efficiency and optimizing traffic scheduling decisions. However, current mainstream traffic monitoring methods still have many technical limitations and are insufficient to meet the refined management needs of complex traffic scenarios.

[0003] Traditional road traffic monitoring equipment, due to its fixed installation location, suffers from inherent limitations in coverage and observation angle, making it difficult to capture panoramic vehicle movement over long distances or complex intersections. For vehicle queue analysis, fixed monitoring equipment can only acquire vehicle information for localized sections, failing to provide complete queue scene data. This significantly reduces the accuracy of analysis of key parameters such as long queue length and vehicle spatial distribution. Although drone patrol technology, with its flexible mobility, is increasingly being applied to traffic monitoring, the coverage of a single-frame drone image is still limited by shooting altitude and angle. Furthermore, during dynamic patrols, factors such as changing lighting conditions, diverse vehicle types, and complex motion states cause interference, leading to inaccurate detection and tracking interruptions in traditional vehicle recognition and tracking algorithms. Simultaneously, the fusion processing of multiple frames lacks effective algorithmic support, making it difficult to synthesize high-quality wide-angle panoramic images. This still cannot overcome the limitations of single-frame images, hindering the realization of panoramic traffic scene analysis.

[0004] In the field of vehicle queue length estimation, existing technologies face severe challenges in terms of accuracy and stability when dealing with dynamic traffic flow. Accurate queue length estimation relies on the accurate acquisition of vehicle location information. However, in multi-frame image fusion analysis, the actual vehicle position changes over time due to differences in keyframe sampling times. Traditional methods often employ static position mapping, failing to consider temporal position shifts, leading to inaccuracies in vehicle position estimation. This is particularly problematic at intersections, highways, and sections with complex traffic conditions, where vehicles are frequently slow-moving and stopping, with speeds and accelerations constantly changing. The limitations of static mapping methods become even more pronounced, easily causing misjudgments of the lead vehicle's position in the queue. This results in significant deviations between the queue length estimation and the actual traffic flow, failing to meet the demands of traffic management for precise state data.

[0005] In summary, how to overcome the limitations of existing monitoring methods in terms of coverage and viewing angle, achieve accurate vehicle identification, tracking, and panoramic scene construction, and on this basis solve the problem of vehicle position deviation caused by time difference in dynamic traffic flow, and improve the accuracy and stability of vehicle queue length estimation, has become an urgent technical challenge in the field of intelligent traffic monitoring. Summary of the Invention

[0006] This invention provides a method and system for dynamic estimation of vehicle queue length based on unmanned aerial vehicle (UAV) patrol, which addresses the following technical problems: how to achieve accurate identification, tracking, and panoramic scene construction of queued vehicles, and on this basis, solve the problem of vehicle position deviation caused by time difference under dynamic traffic flow, thereby improving the accuracy and stability of vehicle queue length estimation.

[0007] The embodiments of the present invention adopt the following technical solutions:

[0008] On the one hand, embodiments of the present invention provide a method for dynamically estimating vehicle queue length based on drone patrol, the method including: collecting patrol video data of congested road sections through inspection drones;

[0009] The cruise video data is processed to extract vehicle feature parameters of the congested road section; wherein, the vehicle feature parameters include at least one or more of the following: the front coordinates, rear coordinates, vehicle type, vehicle speed, acceleration, vehicle color, and straight-line distance between adjacent vehicles in the same lane for each vehicle target.

[0010] Based on the vehicle characteristic parameters, a lane-level queuing vehicle discrimination model based on two indicators, vehicle speed and straight-line vehicle spacing, is constructed; and through the lane-level queuing vehicle discrimination model, an ordered set of queuing vehicles in the congested road section is obtained.

[0011] Based on the orderly assembly of the queued vehicles, the initial queue length of each lane in the congested road section is determined;

[0012] Based on the spatiotemporal compensation correction algorithm, the initial queue length is dynamically corrected to obtain the real-time queue length of each lane in the congested road segment.

[0013] In one feasible implementation, patrol drones are used to collect patrol video data of congested road sections, specifically including:

[0014] Real-time monitoring of road traffic operation status information based on electronic maps;

[0015] When a traffic jam is detected on a certain road section and the duration of the traffic jam reaches the preset judgment condition, the drone patrol mission is automatically triggered.

[0016] The system locates the head of the traffic jam and controls the inspection drone to reach above the head of the traffic jam. Then, it activates a reverse cruise mode, flying in the opposite direction of the traffic flow to collect cruise video data and calibrate the flight trajectory in real time. The flight parameters of the inspection drone are set to a vertical ground-view perspective.

[0017] In one feasible implementation, image processing is performed on the cruise video data to extract vehicle feature parameters of the congested road section, specifically including:

[0018] Continuous image frames are extracted from the cruise video data, and vehicle target recognition is performed on each frame using the YOLOv8 vehicle detection model to identify several vehicle targets.

[0019] The vehicle targets are continuously tracked using the DeepSORT motion detection model to obtain the vehicle number and vehicle identification box coordinates of each vehicle target.

[0020] Based on the vehicle identification frame coordinates, obtain the front coordinates, rear coordinates, vehicle type, and straight-line distance between adjacent vehicles in the same lane for each vehicle target;

[0021] Based on the flight speed of the inspection drone, the vehicle detection speed identified in the image is corrected to obtain the actual vehicle speed and vehicle acceleration of the target vehicle.

[0022] Extract the RGB values ​​of all pixels within the vehicle recognition box and take the average value as the vehicle color of the target vehicle.

[0023] In one feasible implementation, the vehicle identification frame coordinates are used to obtain the front and rear coordinates, vehicle type, and straight-line distance between adjacent vehicles in the same lane for each vehicle target. Specifically, this includes:

[0024] Determine the traffic flow direction of each lane in the congested road section; wherein, the traffic flow direction includes west to east, east to west, south to north, and north to south;

[0025] Based on the vehicle identification frame coordinates of the vehicle template and the traffic flow direction, the position coordinates of the center point of the front line are determined to obtain the front coordinates, and the position coordinates of the center point of the rear line are determined to obtain the rear coordinates.

[0026] The length of the vehicle recognition frame is determined as the vehicle length, and the vehicle type is matched in a preset vehicle type determination table according to the vehicle length to obtain the vehicle type of each vehicle target.

[0027] The distance between the rear coordinates of the preceding vehicle and the front coordinates of the following vehicle is calculated, and then converted using a conversion coefficient between pixel distance and actual distance to obtain the straight-line distance between adjacent vehicles in the same lane.

[0028] In one feasible implementation, based on the vehicle characteristic parameters, a lane-level queuing vehicle discrimination model is constructed based on two indicators: vehicle speed and straight-line vehicle spacing, specifically including:

[0029] Based on the vehicle speed and a first preset speed threshold, a preliminary judgment model for queuing vehicles based on vehicle speed index is constructed for preliminary judgment of queuing vehicles; wherein, the first preset speed threshold is related to the lane free flow speed;

[0030] Based on the initial determination of the queued vehicles, using the rear coordinates of each queued vehicle as a reference, a neighborhood search algorithm is used to search for the set of upstream vehicles located within the same lane neighborhood; wherein, when the front coordinates of an upstream vehicle are within the neighborhood, the upstream vehicle is included in the set of upstream vehicles.

[0031] A re-discrimination model for queuing vehicles based on the straight-line distance between the queuing vehicles and each upstream vehicle in the upstream vehicle set, and a preset distance threshold, is constructed; wherein, the preset distance threshold is related to the vehicle speed and deceleration of the upstream vehicles.

[0032] The preliminary queuing vehicle discrimination model and the re-queuing vehicle discrimination model are combined to form the lane-level queuing vehicle discrimination model.

[0033] In one feasible implementation, the lane-level queuing vehicle discrimination model is used to obtain an ordered set of queuing vehicles in the congested road segment, specifically including:

[0034] The vehicle speed of each vehicle target in the same frame image is input into the preliminary queuing vehicle determination model to obtain the preliminary queuing vehicle sequence in each lane;

[0035] Based on the initial queue of vehicles, obtain the corresponding upstream vehicle set;

[0036] The straight-line distance between each queued vehicle and its corresponding upstream vehicle in the initial queued vehicle sequence, as well as the vehicle speed and deceleration of the upstream vehicle, are input into the queued vehicle re-discrimination model to obtain the complete queued vehicle sequence for each lane.

[0037] Combine the complete queuing vehicle sequences of all lanes into a queuing vehicle set;

[0038] The collection of all queuing vehicles in each frame of the cruise video data is integrated into an ordered collection of queuing vehicles for the congested road section.

[0039] In one feasible implementation, the initial queue length of each lane in the congested road segment is determined based on the orderly assembly of the queued vehicles, specifically including:

[0040] The key frames of the cruise video data are extracted based on the frame difference method, and the scale-invariant feature transformation algorithm SIFT and the fade-in / fade-out weighted average fusion algorithm are used to fuse and stitch the key frames into a panoramic image of the congested road section.

[0041] Within the orderly queue of vehicles, obtain the vehicle sequence chain in each lane of the panoramic image of the congested road section;

[0042] The initial queue length for each lane is determined based on the coordinates of the front of the first vehicle and the rear of the last vehicle in the vehicle sequence chain.

[0043] In one feasible implementation, the initial queue length of each lane is determined based on the head coordinates of the first vehicle and the tail coordinates of the last vehicle in the vehicle sequence chain, specifically including:

[0044] If the congested section is a straight road, the straight-line distance between the coordinates of the front of the first vehicle and the coordinates of the rear of the last vehicle is directly calculated and determined as the initial queue length.

[0045] If the congested road section is a curved road, then the road curve is fitted based on the coordinates of the front of the vehicle, the left side of the rear of the vehicle, and the left side of the front of the vehicle in each pair of adjacent queued vehicles, using the positions of these three points as the curve fitting reference points.

[0046] Based on the characteristic parameters of the fitted road curve, calculate the curve spacing between each pair of adjacent queued vehicles.

[0047] By using the summation method, the distance between vehicles on the curve and the length of each vehicle are summed to obtain the initial queue length of each lane in the congested road section.

[0048] In one feasible implementation, the initial queue length is dynamically corrected based on a spatiotemporal compensation correction algorithm to obtain the real-time queue length of each lane in the congested road segment, specifically including:

[0049] In the cruise video data, key frame sequences are identified for the appearance of the lead vehicle, the lead vehicle's imminent departure, and the appearance of the tail vehicle; and in the key sequence frames, the operating parameters of the lead vehicle and the tail vehicle are extracted; wherein, the operating parameters of the lead vehicle include at least the time of its appearance, the time of its imminent departure, and the instantaneous speed and acceleration of the lead vehicle in the key frame image; the operating parameters of the tail vehicle include at least the time of its appearance.

[0050] The estimated speed of the lead vehicle is calculated based on the instantaneous speed and acceleration of the lead vehicle in the keyframe image and the time difference between the sampling times of adjacent keyframes.

[0051] Based on the estimated speed and the second preset speed threshold, it is determined whether the lead vehicle has left the queue of vehicles.

[0052] A dynamic correction model for real-time queue length is constructed based on the moving distance between the first and last vehicles.

[0053] The running parameters of the lead vehicle and the running parameters of the tail vehicle are input into the dynamic correction model to obtain the real-time queue length of each lane in the congested road segment.

[0054] On the other hand, embodiments of the present invention also provide a dynamic estimation system for vehicle queue length based on unmanned aerial vehicle (UAV) patrols, the system comprising:

[0055] The parameter extraction module is used to collect patrol video data of congested road sections through patrol drones; perform image processing on the patrol video data to extract vehicle feature parameters of the congested road sections; wherein, the vehicle feature parameters include at least one or more of the following: the front coordinates, rear coordinates, vehicle type, vehicle speed, acceleration, vehicle color, and straight-line distance between adjacent vehicles in the same lane for each vehicle target.

[0056] The queuing vehicle identification module is used to construct a lane-level queuing vehicle discrimination model based on two indicators: vehicle speed and straight-line vehicle spacing, based on the vehicle feature parameters; and to obtain an ordered set of queuing vehicles in the congested road section through the lane-level queuing vehicle discrimination model.

[0057] The queue length estimation module is used to determine the initial queue length of each lane in the congested road segment based on the ordered set of queued vehicles; and to dynamically correct the initial queue length based on the spatiotemporal compensation correction algorithm to obtain the real-time queue length of each lane in the congested road segment.

[0058] Compared with the prior art, the vehicle queue length dynamic estimation method and system based on UAV patrol provided in this embodiment of the invention has the following beneficial effects:

[0059] (1) This invention uses two indicators, vehicle speed and straight-line vehicle spacing, to iteratively determine queuing vehicles. Combining neighborhood search and angular deviation constraints, it proposes an automatic search algorithm for queuing vehicles in the same lane, which can accurately obtain an ordered set of queuing vehicles. At the same time, it uses the three-point positioning method of adjacent two vehicles to fit the road alignment and performs curve correction on the straight-line distance between vehicles, which effectively solves the error problem of traditional straight-line distance measurement under complex road alignment conditions and improves the estimation accuracy of queue length in complex road scenarios.

[0060] (2) In view of the problem of vehicle position deviation caused by different sampling times of key frames in image fusion, this invention proposes a dynamic correction method for queue length based on spatiotemporal compensation, which effectively improves the adaptability of the method to scenarios of slow vehicle movement or frequent start and stop, and significantly improves the real-time performance and reliability of queue length estimation in the dynamic evolution of congestion. It is especially suitable for scenarios such as urban expressway weaving areas and long-distance queues on highways, providing accurate situational information for traffic control. Attached Figure Description

[0061] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings:

[0062] Figure 1 A flowchart of a method for dynamically estimating vehicle queue length based on unmanned aerial vehicle (UAV) patrol, provided in an embodiment of the present invention;

[0063] Figure 2 A schematic diagram of a drone cruise scenario provided in an embodiment of the present invention;

[0064] Figure 3 A schematic diagram of road curve fitting provided in an embodiment of the present invention;

[0065] Figure 4 This is a schematic diagram of a vehicle queue length dynamic estimation system based on UAV patrol provided in an embodiment of the present invention. Detailed Implementation

[0066] To enable those skilled in the art to better understand the technical solutions of this invention, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this invention, and not all embodiments. Based on the embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this invention.

[0067] This invention provides a method for dynamically estimating vehicle queue length based on unmanned aerial vehicle (UAV) patrols, such as... Figure 1 As shown, the dynamic estimation method for vehicle queue length based on drone patrol specifically includes steps S101-S105:

[0068] S101. Use inspection drones to collect patrol video data of congested road sections.

[0069] Specifically, road traffic status information is monitored in real time based on electronic maps. When congestion is detected on a certain road segment and the congestion duration reaches a preset judgment condition, a drone patrol mission is automatically triggered.

[0070] Furthermore, the system locates the head of the traffic jam and controls the inspection drone to reach above the head of the traffic jam. Then, it activates a reverse cruise mode, flying in the opposite direction of the traffic flow to collect cruise video data and calibrate the flight trajectory in real time. The flight parameters of the inspection drone are set to a vertical ground-view perspective.

[0071] As a feasible implementation method, Figure 2 This is a schematic diagram of a drone cruise scenario provided by an embodiment of the present invention, such as... Figure 2 As shown, based on real-time monitoring of road traffic status using electronic maps, a drone patrol mission is automatically triggered when the congestion duration of a certain road segment reaches a preset threshold. The drone accurately locates the head of the congested section, activates a reverse patrol mode, and flies against the direction of traffic flow, collecting and calibrating its flight trajectory in real time. The drone's flight parameters are set to a vertical ground-view perspective (90° angle between the lens optical axis and the ground), a flight altitude maintained at 80-120 meters above the ground, and a flight speed controlled at 10-15 km / h on normal road sections.

[0072] S102. Perform image processing on the cruise video data to extract vehicle characteristic parameters for congested road sections.

[0073] Vehicle characteristic parameters include at least one or more of the following: the front coordinates, rear coordinates, vehicle type, vehicle speed, acceleration, vehicle color, and straight-line distance between adjacent vehicles in the same lane for each vehicle target.

[0074] Specifically, continuous image frames are extracted from the cruise video data, and the YOLOv8 vehicle detection model is used to identify vehicle targets in each frame, thus identifying several vehicle targets. The DeepSORT motion detection model is then used to continuously track the vehicle targets, obtaining the vehicle number and vehicle bounding box coordinates for each target.

[0075] In one embodiment, continuous tracking of the vehicle is achieved through the motion detection model in the DeepSORT algorithm, obtaining the vehicle number i (i∈Z). + ) and vehicle recognition box coordinates (x c ,y c ,w,h), where x c Here are the coordinates of the bounding box center point along the image width direction; y c(0,0) represents the coordinates of the center point of the bounding box along the height direction of the image; (w) represents the width of the bounding box; (h) represents the height of the bounding box; (0,0) represents the coordinates of the top left corner of the image; and (1,1) represents the coordinates of the bottom right corner of the image.

[0076] Furthermore, based on the vehicle recognition box coordinates, the front and rear coordinates, vehicle type, and straight-line distance between adjacent vehicles in the same lane are obtained for each vehicle target.

[0077] As a feasible implementation method, the traffic flow direction of each lane in the congested road section is first determined; the traffic flow direction includes west to east, east to west, south to north, and north to south. Based on the vehicle recognition frame coordinates of the vehicle template and the traffic flow direction, the position coordinates of the center point of the front line are determined to obtain the front coordinates, and the position coordinates of the center point of the rear line are determined to obtain the rear coordinates. Then, the length of the vehicle recognition frame is determined as the vehicle length, and the vehicle is matched against a preset vehicle type determination table to obtain the vehicle type of each target vehicle. The distance between the rear coordinates of the preceding vehicle and the front coordinates of the following vehicle of adjacent vehicles is calculated, and converted using a conversion coefficient between pixel distance and actual distance to obtain the straight-line distance between adjacent vehicles in the same lane.

[0078] In one embodiment, the position of the center point A of the vehicle front line in the vehicle recognition frame is set to (x i ,y i The position of the center point B of the rear line is (X). i ,Y i The representation of points A and B differs depending on the direction of traffic flow, as shown in Table 1. According to Table 1, the present invention calculates and represents the coordinates of the front and rear of vehicles in traffic flows of different directions differently, enabling the identification of congested vehicles in different directions of traffic flow within the same coordinate system.

[0079] Table 1

[0080]

[0081] Taking the vehicle recognition frame length w as the vehicle length, this value is matched with a pre-set vehicle type determination table to obtain the vehicle type determination result, as shown in Table 2. According to Table 2, based on the different length ranges mentioned above, vehicles can be divided into small cars (length less than 4 meters), medium cars (length between 4 and 6 meters), large cars (length between 6 and 12 meters), and extra-large cars (length greater than 12 meters).

[0082] Table 2

[0083] small cars <4 mid-size car [4,6] large vehicles [6,12] super-large vehicle >12

[0084] Then according to d i=λdist((X i ,Y i ),(x i+1 ,y i+1 )), i∈Z + Calculate the straight-line distance d between adjacent vehicles i and i+1 in the same lane. i Where λ is the conversion coefficient between pixel distance and actual distance.

[0085] Furthermore, based on the flight speed of the inspection drone, the detected vehicle speed in the image is corrected to obtain the actual vehicle speed and acceleration. Then, the RGB values ​​of all pixels within the vehicle recognition box are extracted, and the average value is taken as the vehicle color.

[0086] In one embodiment, in traffic videos captured by a drone during patrol, the vehicle speed directly detected is not the actual speed because the drone itself is in motion. In this invention, according to v n,i =V n,i -v, estimates vehicle speed, based on... Estimate the vehicle's acceleration. Where v is the drone's cruising speed; v n,i V represents the instantaneous velocity of vehicle i in the nth frame of the image. n,i Δt represents the vehicle detection speed; Δt represents the unit time between each frame.

[0087] S103. Based on vehicle characteristic parameters, construct a lane-level queuing vehicle discrimination model based on two indicators: vehicle speed and straight-line vehicle spacing; and obtain an ordered set of queuing vehicles in congested road sections through the lane-level queuing vehicle discrimination model.

[0088] Specifically, based on the vehicle speed and a first preset speed threshold, a preliminary judgment model for queuing vehicles based on vehicle speed index is constructed to preliminarily determine queuing vehicles; wherein, the first preset speed threshold is related to the lane free-flow speed.

[0089] As a feasible implementation method, a preliminary vehicle queuing discrimination model is constructed based on the vehicle speed parameters extracted from each frame of the video clip: Among them, v th1 =αV N When S1(i) = 1, vehicle i is determined to be in a low-speed or stationary state; when S1(i) = 0, vehicle i is determined to be in a normal driving state. Where v i v is the instantaneous speed of vehicle i; th1 The first preset speed threshold is α; α is the adjustment coefficient; V N Let N be the free-flow velocity of lane N.

[0090] Furthermore, based on the initially determined queued vehicles, using the rear coordinates of each queued vehicle as a reference, a neighborhood search algorithm is used to search for the set of upstream vehicles located within the neighborhood of the same lane; wherein, when the front coordinates of an upstream vehicle are within the neighborhood, the upstream vehicle is included in the set of upstream vehicles.

[0091] Furthermore, based on the straight-line distance between the queuing vehicles and each upstream vehicle in the upstream vehicle set and the preset distance threshold, a queuing vehicle re-discrimination model based on the straight-line distance is constructed; wherein, the preset distance threshold is related to the vehicle speed and deceleration of the upstream vehicles.

[0092] Finally, the preliminary queuing vehicle discrimination model and the re-discrimination model of queuing vehicles are combined to construct a lane-level queuing vehicle discrimination model.

[0093] Furthermore, the vehicle speed of each target vehicle in the same frame is input into the preliminary queuing vehicle determination model to obtain the preliminary queuing vehicle sequence for each lane. Based on the preliminary queuing vehicle sequence, the corresponding upstream vehicle set is obtained. The straight-line distance between each queuing vehicle and its corresponding upstream vehicle, as well as the vehicle speed and deceleration of the upstream vehicles, are input into the queuing vehicle re-determination model to obtain the complete queuing vehicle sequence for each lane. The complete queuing vehicle sequences of all lanes are combined into a queuing vehicle set. All queuing vehicle sets in each frame of the cruise video data are integrated into an ordered queuing vehicle set for the congested road segment.

[0094] As a feasible implementation method, for vehicles initially determined to be likely to queue, using the center point B of the rear lane line as a reference, a set R(i) of upstream vehicles located within the neighborhood of the same lane is searched. Only vehicles whose front lane center point A is within this neighborhood are included in the set. The search algorithm is as follows: R(i) = {i|r i ≤3L veh , -15°≤θ(i,i+1)≤15°};where r is the search neighborhood radius (meters); L veh θ is the standard vehicle length; θ(i,i+1) is the angular deviation between vehicle i and its adjacent vehicle i+1.

[0095] Then, obtain the speed of the upstream vehicle adjacent to i in R(i), and determine whether its speed is less than or equal to v. th1 If so, and the straight-line vehicle spacing is less than or equal to a preset vehicle spacing threshold, then the adjacent upstream vehicles are added to the queue, thus obtaining a complete queue of orderly vehicles in the same lane. The formula for re-judging queued vehicles based on straight-line vehicle spacing is as follows: in, When S2(i+1) = 1, vehicle i+1 is determined to be spatially adjacent to vehicle i; when S2(i+1) = 0, vehicle i+1 and vehicle i are determined to not belong to the same queue. i+1 d is the distance between vehicle i+1 and vehicle i; th β is the vehicle spacing threshold; v is the adjustment coefficient; i+1 t represents the speed of vehicle i+1; r For driver reaction time; a d For deceleration.

[0096] Based on the process loop described above, the set q of queued vehicles in the same frame image is determined. Here, q is composed of multiple subsets of queued vehicles in the same frame image, and each subset represents a queued vehicle sequence chain of lane N, abbreviated as P. N The queue of vehicles is determined by consecutive frames q1, q2…q. n The temporal correlation and integration ultimately form an ordered sequence set Q consisting of all queuing vehicles. n The calculation formula is: Q n =Q n-1 ∪q n ,n∈Z + And n≥1; where, In the formula, when n = 1, Q n-1 Let Q0 represent the initial set of vehicles in the queue. n is the number of video frames during the drone's cruise; Q n Let q be an ordered set of all queued vehicles from n drone cruise video frames; n Let n be the ordered set of queued vehicles in the nth frame of the image.

[0097] S104. Determine the initial queue length for each lane in the congested section based on the orderly assembly of queuing vehicles.

[0098] Specifically, key frames of cruise video data are extracted based on the frame difference method, and the scale-invariant feature transformation algorithm SIFT and the fade-in / fade-out weighted average fusion algorithm are used to fuse and stitch the key frames into a panoramic image of the congested road section.

[0099] Furthermore, within the orderly assembly of queuing vehicles, the vehicle sequence chain in each lane is obtained from the panoramic image of the congested road segment. Based on the coordinates of the front and rear of the first and last vehicles in the vehicle sequence chain, the initial queue length for each lane is determined.

[0100] As a feasible implementation method, if the congested road segment is a straight road, the straight-line distance between the coordinates of the front of the first vehicle and the rear of the last vehicle is directly calculated to determine the initial queue length. If the congested road segment is a curved road, the coordinates of the front of the vehicle, the left rear of the vehicle, and the left front of the vehicle behind the vehicle in each adjacent queue are used as curve fitting reference points to fit the road curve. Based on the characteristic parameters of the fitted road curve, the curve spacing between each adjacent queue of vehicles is calculated. Finally, the curve spacing and the length of each vehicle are summed using an accumulation method to obtain the initial queue length for each lane in the congested road segment.

[0101] In one embodiment, Figure 3 A schematic diagram of road curve fitting provided for an embodiment of the present invention, such as... Figure 3 As shown, the positions of points ABA (two adjacent queued vehicles) are used as the reference points for curve fitting of the road linearity, and linear fitting is performed on the curved road. The curve spacing is... The calculation is as follows: In the formula, r is the radius of the fitted circle for the road alignment; (a,b) are the coordinates of the circle's center; (X... i ,Y i ),(x i+1 ,y i+1 ( ) are the coordinates of point B, downstream of two adjacent queued vehicles, and point A, upstream of two adjacent queued vehicles.

[0102] Based on the curve spacing between each pair of adjacent vehicles, and combined with the target vehicle length, the initial queue length L0 is calculated using the following formula: In the formula, I is the number of the last car in the queue.

[0103] S105. Based on the spatiotemporal compensation correction algorithm, the initial queue length is dynamically corrected to obtain the real-time queue length of each lane in the congested road section.

[0104] Specifically, key frame sequences of the appearance of the lead vehicle, the lead vehicle's impending departure, and the appearance of the tail vehicle are identified in the cruise video data; and the operating parameters of the lead vehicle and the tail vehicle are extracted from the key sequence frames; wherein, the operating parameters of the lead vehicle include at least the time of its appearance, the time of its impending departure, and the instantaneous speed and acceleration of the lead vehicle in the key frame image; the operating parameters of the tail vehicle include at least the time of its appearance.

[0105] Furthermore, based on the instantaneous speed and acceleration of the lead vehicle in the keyframe image, and the time difference between sampling times of adjacent keyframes, the estimated speed of the lead vehicle is calculated. Based on the estimated speed and a second preset speed threshold, it is determined whether the lead vehicle has left the queue of vehicles.

[0106] Further, based on the moving distances of the leading vehicle and the trailing vehicle, a dynamic correction model for the real-time queue length is constructed. The operating parameters of the leading vehicle and the trailing vehicle are input into the dynamic correction model to obtain the real-time queue length of each lane in the congested section.

[0107] As a feasible implementation, during the determination process of the queuing vehicle queue, whether the leading vehicle leaves is a key influencing factor. Set the speed threshold for the leading vehicle to leave the queue as v th2 =γV N , then the determination formula for whether the leading vehicle is in the queuing queue is: where, v i′ =v n′,i′ +a n′,i′ T. When S3(i′)=1, it is determined that the leading vehicle i is in a low speed and is still in the queuing vehicle queue; when S3(i′)=0, it is determined that the leading vehicle i′ has driven normally and left the queuing vehicle queue. Among them, v i′ is the estimated speed of the leading vehicle i′; T is the time difference between adjacent key frame sampling moments; v th2 is the second preset speed threshold required to leave the queuing queue; γ is an adjustment coefficient; V N is the free flow speed of the Nth lane.

[0108] Record the moving distance of the vehicle within the time interval T1=t1 - t0 from the appearance of the leading vehicle to the moment when the leading vehicle is about to leave as L1. Record the time difference between when the leading vehicle is about to leave and the appearance of the trailing vehicle as T2=t2 - t1, and record the time difference between the image when the leading vehicle is about to leave and the moment when the leading vehicle is about to leave the queue as T3. The dynamic correction model of the constructed real-time queue length L s is:

[0109] where, n′ = 1, n′ = 1, 2, 3, i, j ∈ Z + and i ≤ j ≤ I. In the formula, L s is the corrected real-time queue length; when T3 ≥ T2, it means that when the trailing vehicle appears, the leading vehicle has not left the queue. At this time, the displacement of the leading vehicle is L2; when T3 < T2, at the moment of determining the queue tail vehicle, the queue head vehicle has left the queue, and it is necessary to determine the position of the new queue tail vehicle j and the displacement L3 of the leading vehicle after the time T2.

[0110] Through this dynamic correction method for the queue length based on spatio-temporal compensation, the adaptability of the method to scenarios of slow vehicle movement or frequent start and stop is effectively improved, and the real-time performance and reliability of the queue length estimation in the congestion dynamic evolution process are significantly enhanced. It is especially applicable to scenarios such as the weaving area of urban expressways and long-distance queuing situations on expressways, providing accurate situation information for traffic control.

[0111] In addition, embodiments of the present invention also provide a dynamic estimation system for vehicle queue length based on unmanned aerial vehicle (UAV) patrols, such as... Figure 4 As shown, the vehicle queue length dynamic estimation system 400 based on UAV patrol specifically includes:

[0112] The parameter extraction module 410 is used to collect patrol video data of congested road sections through patrol drones; perform image processing on the patrol video data to extract vehicle feature parameters of the congested road sections; wherein, the vehicle feature parameters include at least one or more of the following: the front coordinates, rear coordinates, vehicle type, vehicle speed, acceleration, vehicle color, and straight-line distance between adjacent vehicles in the same lane for each vehicle target.

[0113] The queuing vehicle identification module 420 is used to construct a lane-level queuing vehicle discrimination model based on two indicators: vehicle speed and straight-line vehicle spacing, based on the vehicle feature parameters; and to obtain an ordered set of queuing vehicles in the congested road section through the lane-level queuing vehicle discrimination model.

[0114] The queue length estimation module 430 is used to determine the initial queue length of each lane in the congested road segment based on the ordered set of queued vehicles; and to dynamically correct the initial queue length based on the spatiotemporal compensation correction algorithm to obtain the real-time queue length of each lane in the congested road segment.

[0115] The various embodiments in this invention are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the embodiments of apparatus, devices, and non-volatile computer storage media are basically similar to the method embodiments, so the descriptions are relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0116] The foregoing has described specific embodiments of the present invention. Furthermore, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0117] The above description is merely an embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, various modifications and variations can be made to the embodiments of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of the embodiments of the present invention should be included within the protection scope of the present invention.

Claims

1. A dynamic estimation method for vehicle queue length based on unmanned aerial vehicle (UAV) patrol, characterized in that, The method includes: By using inspection drones, patrol video data of congested road sections can be collected; The cruise video data is processed to extract vehicle feature parameters for the congested road sections. These vehicle feature parameters include at least one or more of the following: the front and rear coordinates, vehicle type, vehicle speed, acceleration, vehicle color, and the straight-line distance between adjacent vehicles in the same lane for each vehicle target. Specifically, they include: Continuous image frames are extracted from the cruise video data, and vehicle target recognition is performed on each frame using the YOLOv8 vehicle detection model to identify several vehicle targets. The vehicle targets are continuously tracked using the DeepSORT motion detection model to obtain the vehicle number and vehicle identification box coordinates of each vehicle target. Based on the vehicle recognition frame coordinates, the front and rear coordinates, vehicle type, and straight-line distance between adjacent vehicles in the same lane are obtained for each vehicle target. Specifically, this includes: The system determines the traffic flow direction of each lane in the congested road segment; wherein the traffic flow direction includes west to east, east to west, south to north, and north to south; based on the vehicle identification frame coordinates of the vehicle template and the traffic flow direction, it determines the position coordinates of the center point of the front line to obtain the front coordinates, and determines the position coordinates of the center point of the rear line to obtain the rear coordinates; it determines the length of the vehicle identification frame as the vehicle length, and matches it with a preset vehicle type determination table according to the vehicle length to obtain the vehicle type of each vehicle target; it calculates the distance between the rear coordinates of the preceding vehicle and the front coordinates of the following vehicle of adjacent vehicles, and converts it through a conversion coefficient between pixel distance and actual distance to obtain the straight-line distance between adjacent vehicles in the same lane. Based on the flight speed of the inspection drone, the vehicle detection speed identified in the image is corrected to obtain the actual vehicle speed and vehicle acceleration of the target vehicle. Extract the RGB values ​​of all pixels within the vehicle recognition box and take the average value as the vehicle color of the target vehicle; Based on the vehicle characteristic parameters, a lane-level queuing vehicle discrimination model based on two indicators, vehicle speed and straight-line vehicle spacing, is constructed; and through the lane-level queuing vehicle discrimination model, an ordered set of queuing vehicles in the congested road section is obtained. Based on the orderly assembly of the queued vehicles, the initial queue length for each lane in the congested road section is determined, specifically including: The key frames of the cruise video data are extracted based on the frame difference method, and the scale-invariant feature transformation algorithm SIFT and the fade-in / fade-out weighted average fusion algorithm are used to fuse and stitch the key frames into a panoramic image of the congested road section. Within the orderly queue of vehicles, obtain the vehicle sequence chain in each lane of the panoramic image of the congested road section; Based on the coordinates of the front of the first vehicle and the rear of the last vehicle in the vehicle sequence chain, the initial queue length for each lane is determined, specifically including: If the congested section is a straight road, the straight-line distance between the coordinates of the front of the first vehicle and the coordinates of the rear of the last vehicle is directly calculated and determined as the initial queue length. If the congested road section is a curved road, then the road curve is fitted based on the coordinates of the front of the vehicle, the left side of the rear of the vehicle, and the left side of the front of the vehicle in each pair of adjacent queued vehicles, using the positions of these three points as the curve fitting reference points. Based on the characteristic parameters of the fitted road curve, calculate the curve spacing between each pair of adjacent queued vehicles. By accumulating the distance between vehicles on the curve and the length of each vehicle, the initial queue length of each lane in the congested road section is obtained. Based on a spatiotemporal compensation correction algorithm, the initial queue length is dynamically corrected to obtain the real-time queue length of each lane in the congested road segment, specifically including: In the cruise video data, keyframe sequences are identified for the appearance of the lead vehicle, the lead vehicle's imminent departure, and the appearance of the tail vehicle; and in the keyframe sequences, the operating parameters of the lead vehicle and the tail vehicle are extracted; wherein, the operating parameters of the lead vehicle include at least the time of its appearance, the time of its imminent departure, and the instantaneous speed and acceleration of the lead vehicle in the keyframe image; the operating parameters of the tail vehicle include at least the time of its appearance. The estimated speed of the lead vehicle is calculated based on the instantaneous speed and acceleration of the lead vehicle in the keyframe image and the time difference between the sampling times of adjacent keyframes. Based on the estimated speed and the second preset speed threshold, it is determined whether the lead vehicle has left the queue of vehicles. A dynamic correction model for real-time queue length is constructed based on the moving distance between the first and last vehicles. The running parameters of the lead vehicle and the running parameters of the tail vehicle are input into the dynamic correction model to obtain the real-time queue length of each lane in the congested road segment.

2. The method for dynamically estimating vehicle queue length based on UAV patrol according to claim 1, characterized in that, By using inspection drones, patrol video data of congested road sections is collected, specifically including: Real-time monitoring of road traffic operation status information based on electronic maps; When a traffic jam is detected on a certain road section and the duration of the traffic jam reaches the preset judgment condition, the drone patrol mission is automatically triggered. The system locates the head of the traffic jam and controls the inspection drone to reach above the head of the traffic jam. Then, it activates a reverse cruise mode, flying in the opposite direction of the traffic flow to collect cruise video data and calibrate the flight trajectory in real time. The flight parameters of the inspection drone are set to a vertical ground-view perspective.

3. The method for dynamically estimating vehicle queue length based on UAV patrol according to claim 1, characterized in that, Based on the aforementioned vehicle characteristic parameters, a lane-level queuing vehicle discrimination model is constructed based on two indicators: vehicle speed and straight-line vehicle spacing. Specifically, this includes: Based on the vehicle speed and a first preset speed threshold, a preliminary judgment model for queuing vehicles based on vehicle speed index is constructed for preliminary judgment of queuing vehicles; wherein, the first preset speed threshold is related to the lane free flow speed; Based on the initial determination of the queued vehicles, using the rear coordinates of each queued vehicle as a reference, a neighborhood search algorithm is used to search for the set of upstream vehicles located within the same lane neighborhood; wherein, when the front coordinates of an upstream vehicle are within the neighborhood, the upstream vehicle is included in the set of upstream vehicles. A re-discrimination model for queuing vehicles based on the straight-line distance between the queuing vehicles and each upstream vehicle in the upstream vehicle set, and a preset distance threshold, is constructed; wherein, the preset distance threshold is related to the vehicle speed and deceleration of the upstream vehicles. The preliminary queuing vehicle discrimination model and the re-queuing vehicle discrimination model are combined to form the lane-level queuing vehicle discrimination model.

4. The method for dynamically estimating vehicle queue length based on UAV patrol according to claim 3, characterized in that, The lane-level queuing vehicle discrimination model is used to obtain an ordered set of queuing vehicles in the congested road section, specifically including: The vehicle speed of each vehicle target in the same frame image is input into the preliminary queuing vehicle determination model to obtain the preliminary queuing vehicle sequence in each lane; Based on the initial queue of vehicles, obtain the corresponding upstream vehicle set; The straight-line distance between each queued vehicle and its corresponding upstream vehicle in the initial queued vehicle sequence, as well as the vehicle speed and deceleration of the upstream vehicle, are input into the queued vehicle re-discrimination model to obtain the complete queued vehicle sequence for each lane. Combine the complete queuing vehicle sequences of all lanes into a queuing vehicle set; The collection of all queuing vehicles in each frame of the cruise video data is integrated into an ordered collection of queuing vehicles for the congested road section.

5. A vehicle queue length dynamic estimation system based on UAV patrol, employing the vehicle queue length dynamic estimation method based on UAV patrol as described in any one of claims 1-4, characterized in that, The system includes: The parameter extraction module is used to collect patrol video data of congested road sections through patrol drones; perform image processing on the patrol video data to extract vehicle feature parameters of the congested road sections; wherein, the vehicle feature parameters include at least one or more of the following: the front coordinates, rear coordinates, vehicle type, vehicle speed, acceleration, vehicle color, and straight-line distance between adjacent vehicles in the same lane for each vehicle target. The queuing vehicle identification module is used to construct a lane-level queuing vehicle discrimination model based on two indicators: vehicle speed and straight-line vehicle spacing, based on the vehicle feature parameters; and to obtain an ordered set of queuing vehicles in the congested road section through the lane-level queuing vehicle discrimination model. The queue length estimation module is used to determine the initial queue length of each lane in the congested road segment based on the ordered set of queued vehicles; and to dynamically correct the initial queue length based on the spatiotemporal compensation correction algorithm to obtain the real-time queue length of each lane in the congested road segment.

Images