An unmanned aerial vehicle traffic management system based on visual recognition

Abstract

The application discloses a kind of unmanned aerial vehicle traffic dispersal systems based on visual identification, and the present application relates to traffic dispersal technical field.The unmanned aerial vehicle traffic dispersal system based on visual identification, by navigation positioning traffic congestion road condition, and after unmanned aerial vehicle flight reaches, whether congestion is judged using visual target identification and clustering method to the density difference of area (full lane, or certain lane) vehicle before and after, and according to different types of congestion road condition, through airborne laser lamp projection and loudspeaker voice, dispersal intervention is carried out, and by grating, the jam point of queuing and lane is moved backward, lane projection, warning words and voice prompt are displayed in a larger range, and vehicle is guided in advance orderly shunt, it is favorable to limit uncivil queuing vehicle, warning accident, fault, dispersal traffic junction, and improve traffic efficiency.

Description

A visual recognition-based drone traffic management system Technical Field

[0001] This invention relates to the field of traffic management technology, specifically to a visual recognition-based unmanned aerial vehicle (UAV) traffic management system. Background Technology

[0002] Analysis of congestion points on roads without traffic lights, such as urban expressways and main roads in open areas, reveals that the main causes of congestion are: (1) individual vehicles traveling too slowly, reducing the speed of all vehicles in their lane, or occupying the overtaking lane for a long time; (2) vehicles at the intersections of expressways and highways do not pass slowly and alternately, but instead form extremely low-speed lane-changing congestion due to last-minute lane changes; and (3) road obstructions (such as construction or accidents) occupy lanes, reducing the number of lanes and causing congestion. For roads without traffic lights, there is no effective method other than relying on traffic police to direct traffic. However, traffic police directing traffic cannot quickly, timely, or conveniently reach all urban expressways and main roads in open areas, and the issued traffic guidance instructions cannot be disseminated to subsequent vehicles over a large area. Therefore, this invention proposes a visual recognition-based unmanned aerial vehicle (UAV) traffic guidance system. Summary of the Invention

[0003] (a) Technical problems to be solved

[0004] To address the shortcomings of existing technologies, this invention provides a visual recognition-based drone traffic management system, which solves the problems mentioned in the background section.

[0005] (II) Technical Solution

[0006] To achieve the above objectives, the present invention provides the following technical solution: a visual recognition-based UAV traffic management system, comprising a map import module, a flight positioning module, a panoramic camera module, a traffic information analysis module, a range flight module, a communication module, and a traffic management module. The output of the map import module is connected to the input of the flight positioning module, the output of the flight positioning module is connected to the input of the panoramic camera module, the output of the panoramic camera module is connected to the input of the traffic information analysis module, the output of the traffic information analysis module is connected to the inputs of the range flight module and the traffic management module, and the output of the range flight module is connected to the input of the communication module.

[0007] As an improved technical solution, the map import module is used to locate satellite information based on the real-time congestion start point in the map; the flight positioning module is used to generate a flight route and fly to the area above and in front of the road location; the panoramic camera module is used to take panoramic road condition photos of the flight arrival location; and the traffic information analysis module is used to determine whether there is a congestion start point and the type of congestion at the flight arrival location.

[0008] As an improved technical solution, the range flight module is used for the UAV to locate the starting point of congestion within the range of its arrival location; the communication module is used to transmit information about the inability to find the starting point of congestion back to the command center, or to receive manual intervention instructions from the command center; the traffic management module locates the congestion point and performs traffic management operations according to the congestion situation.

[0009] As an improved technical solution, the traffic information analysis module uses the following method: Through panoramic road condition photos, it identifies lane dividers and vehicles on one side of the road within the area; by abstracting vehicles as vehicle points, it performs density clustering on these vehicle points; and based on the clustering results, it comprehensively determines the overall road traffic status, abnormal vehicles or road surface conditions, and abnormal locations. The specific analysis method is as follows:

[0010] P1: The panoramic road condition photos are sent to the cloud storage. Using the built-in analysis processor, lane dividers and vehicles on designated sides of the road within the area are identified, separated by single solid lane lines. Vehicles are abstracted as points in the direction of travel. Density clustering is performed on these vehicle points to obtain the current partition density value P. 实 And the corresponding clusters C1, C2, ..., C n Any point outside the cluster is considered a noise point.

[0011] P2: Calculate the distance between the front and rear of the vehicles in the photo, and then calculate the scale between the distance between the front and rear of the vehicles in the photo and the actual distance based on the drone's flight altitude. This allows you to calculate the actual distance between the front and rear of the vehicles, denoted as M. 实 ;

[0012] Based on lane classification standards and typical road conditions for this lane, a default reasonable distance between vehicles in front and behind is determined, denoted as M0 as the standard driving distance, M... 缓 For slow-moving spacing, M 堵 The congestion spacing is defined. Based on the vehicle distance, the preset cluster density is P0 for normal traffic and P0 for slow traffic. 缓 and congestion density P 堵 Preset road condition parameter thresholds W: normal traffic threshold W0 = 0.2, slow traffic threshold W... 缓 =1.0, congestion threshold W 堵 =3.0;

[0013] When P 实 <=P 缓 If so, the road condition is determined to be slow-moving, P 实 <=P 堵 If the signal is positive, the road condition is considered congested; otherwise, the road condition is considered normal.

[0014] P3: Based on the density clustering results, find the cluster C1 in the direction of travel, find the edge point AC1 on the side of the direction of travel in C1, and determine whether the car B outside the cluster of car AC1 belongs to a noise point or belongs to another cluster (assumed to be C2).

[0015] If car B belongs to cluster C2, the program skips cluster C1, starts from C2 to find the forward edge point of cluster C2, and repeats this process P3.

[0016] If car B is a noise point, calculate the distance D between cars A and B. AB If the distance If the distance between vehicles A and B exceeds twice the normal speed limit while moving slowly, then vehicle A is judged to be moving too slowly; otherwise, vehicle B is judged to be moving too slowly. For example, if the distance between vehicles A and B exceeds twice the normal speed limit while moving slowly, then vehicle A is judged to be moving too slowly.

[0017] Next, determine if the lane where car A / B is located is a solid line. If it is, directly alert car A / B in the current lane. If it is not a solid line, then identify vehicle C as the innermost lane's side edge in the direction of travel. 1内 A has been occupying the overtaking lane for an extended period, prompting a warning to vehicle C. 1内 A;

[0018] P4: Repeat the above operation after a time interval of T seconds (default is 5 seconds). If the results of the three judgments are consistent, the traffic management module will perform the corresponding traffic management operation.

[0019] As an improved technical solution, the specific traffic management method of the traffic management module is as follows:

[0020] Q1: When the spatial text information of the POI (Point of Interest) on the same side of the road on the map contains "road entrance / exit" or "road intersection" within the congestion point C threshold range (default 120 meters), it is determined to be near an intersection. When congestion occurs at the intersection, a drone flies over the intersection and uses laser lights to project more solid lines onto the lane dividers, simulating the extension of the single solid line length at the intersection in the form of a raster. Text and voice prompts are provided: "Attention, the lane solid line has been extended; cutting in across the solid line will be captured," and "Alternating traffic at the intersection." This shifts the potential congestion point for lane-jumping to the new solid line starting point, mitigating the reduced traffic efficiency caused by vehicles forcibly cutting in near the intersection. Simultaneously, information is fed back to the command center, and patrol officers are promptly dispatched to assist in traffic control.

[0021] Q2: When the vehicle in front is moving too slowly, while maintaining a relatively stable speed with the vehicle in front, use laser lights to project text and voice prompts on the road surface in front of the vehicle in that lane, displaying the maximum speed limit and the words "Do not drive slowly", or displaying various voice and text reminders such as "Do not occupy the overtaking lane for a long time", to guide vehicles to speed up and give way.

[0022] Q3: In case of temporary construction failure or accident failure, the projected text and voice will say "There is a failure in the left / middle / right lane ahead. Please move slowly to the right / both sides / left side in advance and pass quickly" to warn the following vehicles of the failure information and lane merging suggestions. At the same time, the single solid line of the intersection will be simulated in the form of a raster and the length of the single solid line will be gradually extended according to the lane merging situation of the vehicles to make room for accident handling.

[0023] As an improved technical solution, the drone also includes: prominent police markings, traffic guidance cameras, colored laser lights, laser light control gimbals, simple loudspeakers, and self-controlled flight cameras.

[0024] As an improved technical solution, the map import module uses maps such as Gaode Map, Google Map, or Baidu Map that have real-time traffic conditions and whose traffic information can be imported and read, or maps specified by the command center, and connects to 4G or 5G networks.

[0025] As an improved technical solution, the flight range of the range flight module is set to a circular range with a radius of 2000m.

[0026] (III) Beneficial Effects

[0027] This invention provides a visual recognition-based drone traffic management system. Compared with existing technologies, it has the following advantages: This visual recognition-based drone traffic management system locates traffic congestion through navigation. After the drone arrives, it uses visual target recognition and density clustering analysis to calculate the density difference of vehicles in front and behind a region (the entire lane or a specific lane) to determine whether there is congestion. Based on different types of congestion, it intervenes through onboard laser projection and speakers to manage traffic flow. It uses raster projection to extend the solid lane lines to move the congestion points of queuing and lane changing back, and provides voice prompts and guidance to divert vehicles. This helps to increase the speed of vehicles queuing before and after lane changing, restricts uncivilized temporary queuing that causes vehicles to merge at extremely low speeds, and cultivates good habits of speeding up and merging in advance according to road conditions under regulations, thereby improving overall traffic efficiency. Attached Figure Description

[0028] Figure 1 is a block diagram of the UAV traffic management system based on visual recognition according to the present invention.

[0029] In the diagram, 1-Map import module, 2-Flight positioning module, 3-Panoramic camera module, 4-Traffic information analysis module, 5-Range flight module, 6-Communication module, and 7-Traffic management module. Detailed Implementation

[0030] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0031] Please refer to Figure 1. This invention provides two technical solutions:

[0032] Example 1: A visual recognition-based UAV traffic management system includes a map import module 10, a flight positioning module 20, a panoramic camera module 30, a traffic information analysis module 40, a range flight module 50, a communication module 60, and a traffic management module 70. The output of the map import module 10 is connected to the input of the flight positioning module 20, the output of the flight positioning module 20 is connected to the input of the panoramic camera module 30, the output of the panoramic camera module 30 is connected to the input of the traffic information analysis module 40, the output of the traffic information analysis module 40 is connected to the inputs of the range flight module 50 and the traffic management module 70, and the output of the range flight module 50 is connected to the input of the communication module 60.

[0033] In this embodiment of the invention, the map import module 10 is used to locate satellite information based on the real-time congestion start point in the map; the flight positioning module 20 is used to generate a flight route and fly to the area above and in front of the road location; the panoramic camera module 30 is used to take panoramic road condition photos of the location reached by the flight; and the traffic information analysis module 40 is used to determine whether there is a congestion start point at the location reached by the flight and to determine the type of congestion and possible causes.

[0034] In this embodiment of the invention, the range flight module 50 is used for the UAV to locate the starting point of congestion within the range of its arrival location; the communication module 60 is used to transmit information about the inability to find the starting point of congestion back to the command center; and the traffic management module 70 locates the congestion point and performs traffic management operations according to the congestion situation.

[0035] In this embodiment of the invention, the method of the traffic information analysis module 40 is as follows: Using panoramic road condition photos, lane dividers and vehicles on one side of the road within the area are identified. Vehicles are abstracted as vehicle points, and density clustering is performed on these vehicle points. Based on the clustering results, the overall road traffic status, abnormal vehicles or abnormal road conditions, and abnormal locations are comprehensively determined. The specific analysis method is as follows:

[0036] P1: The panoramic road condition photos are sent to the cloud storage. Using the built-in analysis processor, lane dividers and vehicles on designated sides of the road within the area are identified, separated by single solid lane lines. Vehicles are abstracted as points in the direction of travel. Density clustering is performed on these vehicle points to obtain the current partition density value P. 实 And the corresponding clusters C1, C2, ..., C n Any point outside the cluster is considered a noise point.

[0037] P2: Calculate the distance between the front and rear of the vehicles in the photo, and then calculate the scale between the distance between the front and rear of the vehicles in the photo and the actual distance based on the drone's flight altitude. This allows you to calculate the actual distance between the front and rear of the vehicles, denoted as M. 实 ;

[0038] Based on lane classification standards and typical road conditions for this lane, a default reasonable distance between vehicles in front and behind is proposed, denoted as M0, which is the median of the upper and lower speed limits specified for this lane classification. 缓 For slow-moving spacing, M 堵 The congestion spacing is defined. Based on the vehicle distance, the preset cluster density is P0 for normal traffic and P0 for slow traffic. 缓 and congestion density P 堵 Preset road condition parameter thresholds W: normal traffic threshold W0 = 0.2, slow traffic threshold W... 缓 =1.0, congestion threshold W 堵 =3.0;

[0039] When P 实 <=P 缓 If so, the road condition is determined to be slow-moving, P 实 <=P 堵 If the signal is positive, the road condition is considered congested; otherwise, the road condition is considered normal.

[0040] P3: Based on the density clustering results, find the cluster C1 in the direction of movement, find the edge point on the side of the direction of movement in C1, i.e. the foremost car AC1, and determine whether the car B outside the cluster of car AC1 belongs to a noise point or belongs to another cluster (assumed to be C2).

[0041] If car B belongs to cluster C2, the program skips cluster C1, starts from C2 to find the forward edge point of cluster C2, and repeats this process P3.

[0042] If car B is a noise point, calculate the distance D between cars A and B. AB If the distance Among them, W 实 The current cluster density P M实If the applicable road condition parameter threshold is applied, then vehicle A's speed is determined to be too low; otherwise, vehicle B's speed is determined to be too low. For example, in slow-moving conditions, if the distance between vehicles A and B exceeds twice the normal slow-moving distance, then vehicle A's speed is determined to be too low.

[0043] Next, it determines whether the lane where car A / B is located is a solid line. If it is a solid line, it directly alerts car A / B in the current lane. If it is not a solid line, it identifies car C as the innermost lane's side edge point in the direction of travel. 1内 A has been occupying the overtaking lane for an extended period, prompting a warning to vehicle C. 1内 A;

[0044] P4: After a default 5-second time interval, repeat the above operation. If the results of the three judgments are consistent, the traffic management module (70) will perform the corresponding traffic management operation.

[0045] In this embodiment of the invention, the specific traffic management method of the traffic management module (70) is as follows:

[0046] Q1: When the POI spatial text information on the same side of the road map within the default 120-meter threshold of congestion point C contains "road entrance / exit" or "road intersection," it is determined to be near an intersection. When congestion occurs at the intersection, a drone flies over the intersection and uses laser lights to project more solid lines onto the lane dividers, simulating the extension of the single solid line length at the intersection in a raster form. Text and voice prompts are provided: "Attention, lane solid lines have been extended; cutting in across solid lines will be captured," and "Alternating traffic at the intersection." This shifts the potential congestion point for lane-jumping to the new solid line starting point, mitigating the reduced traffic efficiency caused by vehicles forcibly cutting in near the intersection. Simultaneously, information is fed back to the command center, and patrol officers are promptly dispatched to assist in traffic control.

[0047] Q2: When the vehicle in front is moving too slowly, while maintaining a relatively stable speed with the vehicle in front, use laser lights to project text and voice prompts on the road surface in front of the vehicle in that lane, displaying the maximum speed limit and the words "Do not drive slowly", or displaying various voice and text reminders such as "Do not occupy the overtaking lane for a long time", to guide vehicles to speed up and give way.

[0048] Q3: In case of temporary construction failure or accident failure, the projected text and voice will say "There is a failure in the left / middle / right lane ahead. Please move slowly to the right / both sides / left side in advance and pass quickly" to warn the following vehicles of the failure information and lane merging suggestions. At the same time, the single solid line of the intersection is simulated in the form of a raster and the length of the single solid line is gradually extended according to the lane merging situation of the vehicles to make room for accident handling.

[0049] In this embodiment of the invention, the drone also includes: prominent police markings, a traffic guidance camera, a colored laser light, a laser light control gimbal, a simple megaphone, and a self-controlled flight camera.

[0050] In this embodiment of the invention, the map import module 10 uses maps such as Gaode Map, Google Map, or Baidu Map that have real-time traffic conditions and whose traffic information can be imported and read, or maps specified by the command center, and connects to 4G or 5G networks.

[0051] In this embodiment of the invention, the flight range is set within the range flight module 50 as a circular range with a radius of 2000m.

[0052] Example 2: A visual recognition-based UAV traffic management system includes a map import module 10, a flight positioning module 20, a panoramic camera module 30, a traffic information analysis module 40, a range flight module 50, a communication module 60, and a traffic management module 70. The output of the map import module 10 is connected to the input of the flight positioning module 20, the output of the flight positioning module 20 is connected to the input of the panoramic camera module 30, the output of the panoramic camera module 30 is connected to the input of the traffic information analysis module 40, the output of the traffic information analysis module 40 is connected to the inputs of the range flight module 50 and the traffic management module 70, and the output of the range flight module 50 is connected to the input of the communication module 60.

[0053] In this embodiment of the invention, the map import module 10 is used to locate satellite information based on the real-time congestion start point in the map; the flight positioning module 20 is used to generate a flight route and fly to the area above and in front of the road location; the panoramic camera module 30 is used to take panoramic road condition photos of the flight arrival location; and the traffic information analysis module 40 is used to determine whether there is a congestion start point at the flight arrival location.

[0054] In this embodiment of the invention, the range flight module 50 is used for the UAV to locate the starting point of congestion within the range of its arrival location; the communication module 60 is used to transmit information about the inability to find the starting point of congestion back to the command center; and the traffic management module 70 is used to locate the congestion point and manage traffic according to the congestion situation.

[0055] In this embodiment of the invention, the method of the traffic information analysis module 40 is as follows: Using panoramic road condition photos, lane dividers and vehicles on one side of the road within the area are identified; vehicles are abstracted as vehicle points, and density clustering is performed on these vehicle points; based on the clustering results, the overall road traffic status, abnormal vehicles or abnormal road conditions, and abnormal locations are comprehensively determined; the specific analysis method is as follows:

[0056] P1: The panoramic road condition photos are sent to the cloud storage. Using the built-in analysis processor, lane dividers and vehicles on designated sides of the road within the area are identified, separated by single solid lane lines. Vehicles are abstracted as points in the direction of travel. Density clustering is performed on these vehicle points to obtain the current partition density value P. 实 And the corresponding clusters C1, C2, ..., Cn Any point outside the cluster is considered a noise point.

[0057] P2: Calculate the distance between the front and rear of the vehicles in the photo, and then calculate the scale between the distance between the front and rear of the vehicles in the photo and the actual distance based on the drone's flight altitude. This allows you to calculate the actual distance between the front and rear of the vehicles, denoted as M. 实 ;

[0058] Based on lane classification standards and typical road conditions for this lane, a default reasonable distance between vehicles in front and behind is proposed, denoted as M0, which is the median of the upper and lower speed limits specified for this lane classification. 缓 For slow-moving spacing, M 堵 The congestion spacing is defined. Based on the vehicle distance, the preset cluster density is P0 for normal traffic and P0 for slow traffic. 缓 and congestion density P 堵 Preset road condition parameter thresholds W: normal traffic threshold W0 = 0.2, slow traffic threshold W... 缓 =1, congestion threshold W 堵 =3.0;

[0059] When P 实 <=P 缓 If so, the road condition is determined to be slow-moving, P 实 <=P 堵 If the signal is positive, the road condition is considered congested; otherwise, the road condition is considered normal.

[0060] P3: Based on the density clustering results, find the cluster C1 in the direction of movement, find the edge point on the side of the direction of movement in C1, i.e. the foremost car AC1, and determine whether the car B outside the cluster of car AC1 belongs to a noise point or belongs to another cluster (assumed to be C2).

[0061] If car B belongs to cluster C2, the program skips cluster C1, starts from C2 to find the forward edge point of cluster C2, and repeats this process;

[0062] If car B is a noise point, calculate the distance D between cars A and B. AB If the distance Among them, W 实 The current cluster density P M实 If the applicable road condition parameter threshold is applied, then vehicle A's speed is determined to be too low; otherwise, vehicle B's speed is determined to be too low. For example, in slow-moving conditions, if the distance between vehicles A and B exceeds twice the normal slow-moving distance, then vehicle A's speed is determined to be too low.

[0063] Next, it determines whether the lane is a solid line. If it is, it directly alerts vehicle A in the current lane; if it is not a solid line, it identifies vehicle C as the innermost lane's side edge in the direction of travel. 1内 A. Occupying the overtaking lane for an extended period of time;

[0064] P4: After a time interval of T seconds, repeat the above operation. If the results of the three judgments are consistent, the traffic management module will then perform the traffic management operation.

[0065] In this embodiment of the invention, the specific traffic management method of the traffic management module 70 is as follows:

[0066] Q1: When the spatial text information of the POI (Point of Interest) on the same side of the road as the congestion point C within 120 meters contains "road entrance / exit" or "road intersection," it is determined to be near an intersection. When congestion occurs at the intersection, a drone flies over the intersection and uses laser lights to project more solid lines onto the lane dividers, simulating the extension of the single solid line length at the intersection in a raster form. Text and voice prompts are provided: "Attention, lane solid lines have been extended; cutting in across solid lines will be captured," and "Alternating traffic at the intersection," mitigating the reduced traffic efficiency caused by vehicles forcibly cutting in near the intersection. Simultaneously, information is fed back to the command center, and patrol officers are promptly dispatched to assist in traffic control.

[0067] Q2: When the vehicle in front is too slow, while maintaining a relatively stable speed with the vehicle in front, use laser lights to project text and voice prompts on the road surface in front of the vehicle in that lane, displaying the maximum speed limit and the words "Do not drive slowly", or displaying various voice and text reminders such as "Do not occupy the overtaking lane for a long time", to guide the vehicle to speed up and give way.

[0068] Q3: In case of temporary construction failure or accident failure, the projected text and voice will say "Left / middle / right lane failure, merge in advance and proceed slowly" to warn following vehicles of the failure information and lane merging suggestions. At the same time, the single solid line at the intersection will be simulated in the form of a raster and the length of the single solid line will be gradually extended according to the lane merging situation of vehicles to make room for accident handling.

[0069] Furthermore, any content not described in detail in this specification is existing technology known to those skilled in the art.

[0070] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A visual recognition-based unmanned aerial vehicle (UAV) traffic management system, characterized in that: The system includes a map import module (10), a flight positioning module (20), a panoramic camera module (30), a traffic information analysis module (40), a range flight module (50), a communication module (60), and a traffic management module (70). The output of the map import module (10) is connected to the input of the flight positioning module (20), the output of the flight positioning module (20) is connected to the input of the panoramic camera module (30), the output of the panoramic camera module (30) is connected to the input of the traffic information analysis module (40), the output of the traffic information analysis module (40) is connected to the inputs of the range flight module (50) and the traffic management module (70), and the range flight module (50)... The output of the 50) is connected to the input of the communication module (60); the method of the traffic information analysis module (40) is as follows: by using panoramic road condition photos, identify the lane dividers and vehicles in the area, by abstracting the vehicles into vehicle points, and performing density clustering on the vehicle points, and comprehensively judging the overall road traffic status, abnormal vehicles, abnormal road conditions and abnormal locations based on the clustering results. The specific analysis method is as follows: P1: send the panoramic road condition photos to the storage cloud, use the built-in analysis processor to identify the lane dividers and vehicles on the designated side of the road in the area, and divide them with a single solid line lane as the boundary, and abstract the vehicles into vehicle points in the forward direction; by performing density clustering on the above vehicle points, obtain the current division density value P. 实 And the corresponding clusters C1, C2, ..., C n Any point outside the cluster is considered a noise point. P2: Calculate the distance between the front and rear of the vehicles in the photo, and then calculate the scale between the distance between the front and rear of the vehicles in the photo and the actual distance based on the drone's flight altitude. This allows you to calculate the actual distance between the front and rear of the vehicles, denoted as M. 实 ; Based on lane classification standards and typical road conditions, a reasonable distance between vehicles in front and behind each lane is determined, denoted as M0 as the standard driving distance, and M... 缓 For slow-moving spacing, M 堵 The congestion spacing is defined as follows: Normal traffic spacing is preset to a cluster density P0, and slow traffic spacing is preset to a cluster density P0. 缓 and congestion density P 堵 ; Preset Road condition parameter thresholds W: Normal traffic threshold W0 = 0.2, Slow traffic threshold W 缓 =1.0, congestion threshold W 堵 =3.0; P 实 ≥P 堵 If so, the road condition is determined to be congested; P3: Based on the density clustering results, find cluster C1 in the direction of travel. From C1, find the edge point on the direction of travel, i.e., the foremost car A. Determine whether the car B outside the cluster of car A belongs to a noise point or to another cluster C2. If car B belongs to cluster C2, the program skips cluster C1 and starts searching for the edge point on the direction of travel of cluster C2 from C2, repeating this process P3. If car B is a noise point, calculate the distance D between cars A and B. AB If the distance to D AB >=P M实 *(1+W) 实 ), where W 实 The current cluster density P M实 If the applicable road condition parameter threshold is used, it is determined that vehicle A's speed is too low; otherwise, it is determined that vehicle B's speed is too low. Then, it is determined whether the lanes where vehicles A and B are located are solid lines. If they are solid lines, vehicles A and B in the current lane are directly reminded. If they are not solid lines, it is determined that vehicle A, which is at the innermost lane's forward direction side edge, has been occupying the overtaking lane for a long time, and vehicle A is reminded. P4: After a time interval of T seconds, repeat the above operation. If the results of the three judgments are consistent, the traffic management module (70) will perform the corresponding traffic management operation.

2. The UAV traffic management system based on visual recognition according to claim 1, characterized in that: The map import module (10) is used to locate satellite information based on the real-time congestion start point in the map; the flight positioning module (20) is used to generate a flight route and fly to the area above the road location; the panoramic camera module (30) is used to take panoramic road condition photos of the flight location; the traffic information analysis module (40) is used to determine whether there is a congestion start point at the flight location and to determine the type and cause of congestion.

3. The UAV traffic management system based on visual recognition according to claim 2, characterized in that: The range flight module (50) is used for the UAV to find the starting point of the congestion within the range of the arrival location; the communication module (60) is used to send back the information that the starting point of the congestion cannot be found to the command center, and to receive the manual intervention instructions from the command center; the traffic management module (70) finds the congestion point and performs traffic management operations according to the congestion situation.

4. The UAV traffic management system based on visual recognition according to claim 1, characterized in that: The specific traffic management method of the traffic management module (70) is as follows: Q1: When the spatial text information of the POI location on the same side of the road contains "road entrance / exit" or "road intersection" within the threshold range of the congestion point C, it is determined to be near the intersection. When congestion occurs at the intersection, the drone flies to the top of the intersection and uses laser lights to project more solid lines on the lane dividing line. It simulates the extension of the single solid line length at the intersection in the form of raster, and provides text and voice prompts: "Attention: the lane solid line has been extended. Cutting in across the solid line will be caught." The text and voice prompts for "Alternating traffic at intersections" will shift the congestion point of lane-jumping and lane-changing to the new starting point of the solid line; at the same time, the information will be fed back to the command center, and patrol officers will be dispatched in a timely manner to participate in the command; Q2: When the vehicle in front is moving too slowly, while maintaining a relatively stable speed with the vehicle in front, the laser lights will project text and voice prompts on the road surface in front of the vehicle in that lane, displaying the maximum speed limit and the voice and text prompts for "Do not occupy the overtaking lane for a long time", guiding vehicles to speed up and give way; Q3: In case of temporary construction failure or accident failure, the projected text and voice will say "There is a failure in the left, middle or right lane ahead. Please move slowly to the right or left lane in advance and pass quickly" to give the following vehicles an advance warning of the failure information and lane changing suggestions. At the same time, the single solid line of the intersection is simulated in the form of a raster and the length of the single solid line is gradually extended according to the lane changing situation of the vehicles to make room for accident handling.

5. The UAV traffic management system based on visual recognition according to claim 1, characterized in that: The drone also includes: prominent police markings, traffic guidance cameras, colored laser lights, laser light control gimbals, simple megaphones, and self-controlled flight cameras.

6. The UAV traffic management system based on visual recognition according to claim 1, characterized in that: The map import module (10) uses Gaode Map, Google Map or Baidu Map with real-time traffic conditions and traffic information that can be imported and read, or the map specified by the command center, and connects to 4G or 5G network.

7. The UAV traffic management system based on visual recognition according to claim 1, characterized in that: The flight range is set within the range flight module (50) as a circular range with a radius of 2000m.

Images