Sparse road network monitoring deployment and control method based on cooperation of ground equipment and unmanned aerial vehicle

Abstract

The invention discloses a sparse road network monitoring deployment and control method based on cooperation of ground equipment and an unmanned aerial vehicle. By integrating ground traffic detection equipment and an aerial unmanned aerial vehicle monitoring system, the traffic condition of a sparse road network in a complex environment is efficiently monitored. The ground sensing part uses a high-definition bayonet system and an intelligent video camera to detect traffic events, collects data in real time and optimizes equipment layout; the air maneuvering part carries high-resolution camera equipment through an unmanned aerial vehicle cluster, carries out scene feature analysis work, completes road scene three-dimensional reconstruction, optimizes a cruising path and divides a monitoring area. The invention further comprises a static and dynamic ground-air cooperation method and a TSP optimization algorithm, so that the coverage rate and accuracy of traffic monitoring are improved, and the cost is reduced. The method is suitable for traffic monitoring of complex terrains and remote areas, and is of great significance in improving the traffic safety management level.

Description

Technical Field

[0001] The present invention relates to the field of road traffic detection, and in particular to a three-dimensional monitoring and control method for a sparse road network based on the collaboration of ground equipment and a drone cluster. Background Art

[0002] At present, the road traffic monitoring of sparse road networks in my country mainly relies on traditional ground fixed traffic detection equipment, such as coil detectors, infrared detectors, radar detectors, ultrasonic detectors, video traffic detectors and other sensor detection, which provide point detection traffic information. Road traffic monitoring in central and eastern my country mainly realizes the collection of road traffic parameters and the monitoring of traffic events by deploying a large number of fixed ground detectors on the roads. The sparse road network area is vast, the highway network density is low, the road mileage is long, the traffic flow is small, and the node spacing is large. If the conventional traffic detection equipment deployment method is used, huge traffic detection equipment purchase, installation and maintenance costs will be incurred. For mobile traffic detection equipment, such as global positioning system, electronic tag technology, floating car technology, under the condition of expressway driving, at least 15% to 20% of the vehicles need to be equipped with GPS and RFID equipment to make the trip estimation error less than 5%. Limited by the economic development level of the region where the sparse road network is located, under the current socio-economic conditions, the high-density coverage of fixed and mobile traffic detection equipment is difficult to achieve on the roads of sparse road networks. However, the existing road traffic monitoring system with sparse road network finds it difficult to detect the traffic operation status information of the entire road, and cannot quickly transmit detailed information of traffic incidents to the traffic management command center. This greatly reduces the speed and efficiency of rescue of long-distance traffic incidents, especially traffic accidents, and it is difficult to reduce the loss of life and property caused by accidents.

[0003] In response to the above problems, the present invention proposes a sparse road network monitoring and control method based on the collaboration of ground equipment and drones. Under the condition that the sparse road traffic flow is basically not interrupted by traffic incidents, it can effectively improve the existing technology and detect the occurrence of traffic accidents more economically and effectively. Summary of the invention

[0004] In view of the shortcomings of the prior art, the present invention proposes a sparse road network monitoring and control method based on the collaboration of ground equipment and unmanned aerial vehicles to solve the above problems existing in the prior art. The specific scheme is as follows:

[0005] S1. Detect sparse road events through high-definition camera systems and intelligent video cameras, and optimize the layout of ground traffic detection equipment based on traffic accident prediction and road segmentation;

[0006] S2. Reconstruct the image information captured by the drone in three dimensions and arrange the air traffic detection equipment according to the cruise path collaborative optimization method and traffic monitoring area division;

[0007] S3. Optimization of the layout of ground-air coordinated traffic detection equipment, including static ground-air coordination and dynamic ground-air coordination.

[0008] Preferably, the sparse road network monitoring and control method based on the collaboration of ground equipment and unmanned aerial vehicles comprises:

[0009] High-definition camera systems and intelligent video cameras are introduced to track and identify individual vehicles. When a vehicle passes through the sensitive area of ​​the vehicle position sensor, the sensor sends a signal to the image acquisition control part, which controls the camera to collect a car image and send it to the image preprocessing module. The preprocessing module performs simple processing on the input image and then sends it to the PC. The software module in the PC finally reproduces the license plate characters through several links such as image preprocessing, license plate extraction, license plate image binarization, character segmentation, and character recognition, and gives the recognition result. The recognition result and image are stored in the database for future license plate query, traffic flow statistics and toll management. When a vehicle passes through the circular induction coil installed in the monitoring lane, it triggers the high-definition license plate camera to take a photo of the vehicle, and the photo is stored in the information platform.

[0010] Among them, the high-definition toll station system mainly uses the vehicle license plate comparison method to detect traffic incidents in closed sections of road; the intelligent video camera is mainly used for manual or automatic video monitoring at intersections (or overpass areas), and the intelligent video camera can be used as a redundant supplementary device for the high-definition toll station in closed sections of road.

[0011] Preferably, the sparse road network monitoring and control method based on the collaboration of ground equipment and unmanned aerial vehicles predicts road section traffic accidents and determines the length of the road section according to the fixed-length method or the variable-length method; when laying out ground traffic detection equipment, it is necessary to base it on the safety analysis of the road. Intuitively, the layout density of traffic detection equipment is large in places with poor road safety levels, and the layout density of traffic detection equipment is small in places with good road safety levels.

[0012] The number of traffic accidents that occur on a certain road section each year can be predicted by the following method:

[0013] S31 Count the turning angles and number of horizontal curves in the research section. The sum of the turning angles of each horizontal curve divided by the number of horizontal curves is recorded as a.

[0014] S32 Statistical study of the number of vertical curves in the road section and the angle and slope length of each vertical curve. The sum of the ratio of the angle of each vertical curve to the slope length divided by the number of horizontal curves is recorded as b;

[0015] S33 calculates the percentage of the slope length of each vertical curve to the sum of the slope lengths of the total vertical curves and multiplies it by the angle of each vertical curve, and records it as c;

[0016] S34 counts the proportion of large vehicles passing through the studied road section and denotes it as d;

[0017] S35 integrates the above calculated values ​​using the following formula:

[0018] Q=e (-2676614+0.0071095a+0.2539616c+6.14963d) ·F

[0019] Where P is the predicted number of traffic accidents occurring on the study section each year, and F is the exposure variable.

[0020] Preferably, the sparse road network monitoring and control method based on the collaboration of ground equipment and unmanned aerial vehicles, wherein the road segment division method comprises: setting up high-definition checkpoint modeling, assuming that the road length is L, and dividing it into N sections according to the fixed-length division method, the number of high-definition checkpoints set up is m+2 in total, corresponding to m+1 high-definition checkpoint sections; according to the increasing mileage of the setting up, the set of high-definition checkpoints is recorded as R={x 0 ,x 1 ,x m ,…,x m+1}, where the position of the high-definition checkpoint at the starting point of the road is denoted by x 0 = 0, the position of the high-definition checkpoint at the end of the road is recorded as x m+1 = L. Assume that there are n intersections (or flyover areas) on the road, and the starting mileage of each intersection is recorded as S = {p 1 ,p 2 ,…,p n}, the road end mileage of each intersection (or overpass area) is E = {q 1 ,q 2 ,…,q n}, then the location set of the road intersection (or overpass area) can be recorded as Y = {y|p k ≤y≤q k ,k=1,2,…,n}.

[0021] The following assumptions were made during the modeling process:

[0022] (1) Based on the spatial distribution characteristics of road accidents under sparse road conditions

[0023] (2) Assume that the traffic accidents in each fixed-length sub-section are evenly distributed.

[0024] (3) Assume that the event detection environment of the high-definition camera is under normal weather conditions

[0025] (4) Assume that the high-definition camera records all vehicles entering and leaving the road

[0026] (5) Assume that the layout of the high-definition camera is located in a closed section

[0027] The optimization objective function is as follows:

[0028]

[0029] The constraints are as follows:

[0030] d·(m+2)≤D

[0031]

[0032] Where: Δ min Indicates the minimum spacing of the HD camera mount, Δ max Indicates the maximum spacing of the HD mount, Q i is the number of traffic accidents on the road section with high-definition checkpoints, f(x i -x i+1 ) is the relationship between the detection rate of traffic detection events on the i-th HD checkpoint interval section and the checkpoint interval, d is the cost of a single HD checkpoint, and D is the total investment amount.

[0033] Preferably, the sparse road network monitoring and control method based on the collaboration of ground equipment and drones, wherein the drone 3D reconstruction method comprises:

[0034] First, while the UAV is operating along the planned aerial photography flight path, it takes aerial photos of the accident scene from different heights and angles to generate a series of image sequences. Then, the aerial photos (including the corresponding latitude, longitude, and altitude) are imported into the 3D reconstruction software, and the OpenMP multi-threaded processing mechanism is used to first optimize the scale of the image, extract the GPS information in the image, and obtain the relative position information between adjacent image elements. The SIFT GPU feature extraction algorithm is then used to extract feature points in the image in parallel, and only feature matching is performed on adjacent images, thereby quickly calibrating the image. A sparse 3D point cloud model is then generated using the motion recovery algorithm, and a dense 3D point cloud model is generated using a patch-based multi-dimensional stereo vision algorithm. The 3D point cloud model is then meshed and texturized, and the digital surface model and digital elevation model of the accident scene are output after optimization.

[0035] Preferably, the sparse road network monitoring and control method based on the collaboration of ground equipment and unmanned aerial vehicles, wherein the flight path optimization method of the unmanned aerial vehicle comprises:

[0036] S61. Generate a number of drone cruise path populations, a real weight vector for each cruise path, and a corresponding neighborhood cruise path; set an objective function reference point for the cruise path population;

[0037] S62. Based on the cruise path of the UAV, various constraints are considered to divide the sub-paths and generate the objective function value corresponding to each cruise path; the cruise path with the smallest objective function value is taken as the elite path, and the objective function reference point is updated;

[0038] S63. Calculate the Chebyshev value of each cruise path in the neighborhood where the elite path is located. Each cruise path corresponds to a real weight value and an objective function value, and its corresponding Chebyshev value = max{real weight value x(objective function value-value of the objective function reference point)}; when the Chebyshev value of the elite path ≤ the Chebyshev value of the neighborhood cruise path, the elite path is used to replace other cruise paths in the neighborhood, and the drone cruise path population is updated to implement the elite strategy;

[0039] S64. Set the crossover and mutation probabilities, perform crossover and mutation operations on all cruise paths, and improve the diversity of flight paths;

[0040] S65. Calculate the objective function value of each cruise path of the cruise path population after crossover and mutation, find the cruise path with the smallest objective function value as the elite path, and update the objective function reference point;

[0041] S66. Calculate the Chebyshev value of each cruise path in the neighborhood where the elite path obtained in step S65 is located. When the Chebyshev value of the elite path is ≤ the Chebyshev value of the neighborhood cruise path, replace the neighborhood cruise path with the elite path, update the drone cruise path population, and implement the elite strategy;

[0042] S67. Perform loop iterations to calculate the optimal UAV cruising path and obtain the corresponding objective function value.

[0043] Preferably, the sparse road network monitoring and control method based on the collaboration of ground equipment and drones is as follows: Figure 3 As shown, the method for dividing the monitoring area includes:

[0044] For the selection of monitoring objects, based on the traffic safety level evaluation of sparse roads, road sections and traffic nodes (such as overpass areas, service areas, toll stations, etc.) with low road safety levels are selected as traffic monitoring objects for drones.

[0045] Monitoring area division, according to the spatial distribution of drone monitoring objects, the monitoring objects are clustered, and the number of monitoring areas is continuously increased to ensure that there is one drone in each monitoring area and the maximum flight distance constraint of the drone can be met.

[0046] Aircraft path planning within the area: The drone starts from the base in the monitoring area, inspects all monitored objects, and then returns to the base. The drone's cruising path is required to be the shortest.

[0047] Preferably, the sparse road network monitoring and control method based on the collaboration of ground equipment and unmanned aerial vehicles, the static ground-air collaboration method includes:

[0048] According to the importance of intersections and road sections (such as intersections and road sections with poor traffic safety levels), and taking into account the constraints of transportation investment, traffic detection equipment is installed at relevant intersections and road sections to monitor road traffic. For intersections and road sections without traffic detection equipment, drones are assigned to monitor them, thereby effectively supplementing traditional ground traffic monitoring.

[0049] Preferably, the sparse road network monitoring and control method based on the collaboration of ground equipment and drones is as follows: Figure 4 As shown, the dynamic ground-air coordination method includes:

[0050] The vehicle carries multiple drones, which take off and land from the vehicle; the ground control station located in the vehicle allocates reconnaissance targets for the vehicle and drones, and gives inspection / cruise plans for the vehicles and drones; when there is a sudden demand for traffic monitoring, the vehicle (ground control station) dynamically plans the cruise path of the drone (cluster), and then the drone (cluster) performs cruise reconnaissance according to the new planned path; the drone monitors traffic in the air, and after detecting an accident, the drone obtains images and point cloud information of the accident scene through the onboard camera or onboard radar, and returns it to the vehicle (ground control station); finally, based on the image and point cloud information, a reconstruction study of the accident scene is carried out.

[0051] The present invention provides a sparse road network monitoring and control method based on the collaboration of ground equipment and unmanned aerial vehicles, which has the beneficial effects of improving the monitoring coverage of sparse road networks, optimizing information transmission and response efficiency, and reducing operation and maintenance costs:

[0052] 1. Based on the ground-air collaborative monitoring mechanism, through the fusion of heterogeneous sensor networks of ground high-definition card-entry systems, intelligent video cameras and drone clusters, three-dimensional monitoring is established to achieve comprehensive perception of sparse road networks and improve the coverage of traffic event detection;

[0053] 2. Through ground-air collaboration, traffic accident information is quickly transmitted to the command center. Combining high-definition images taken by ground equipment with 3D reconstruction by drones, comprehensive on-site information is provided to assist the command center in making accurate decisions and shorten rescue response time.

[0054] 3. The optimization of ground equipment layout reduces the deployment of redundant equipment. At the same time, drones are used flexibly to fill in the gaps, reducing hardware investment costs. The license plate recognition process is fully automated, reducing the need for manual intervention and reducing the labor costs of long-term operation and maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and the embodiments in the drawings do not constitute any limitation to the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0056] Figure 1 It is a schematic diagram of the layout of the ground-to-air coordinated traffic monitoring equipment of the present invention.

[0057] Figure 2 It is a schematic diagram of the logical relationship of the ground-to-air collaborative monitoring of the present invention.

[0058] Figure 3 It is a schematic diagram of the working logic of the aerial drone equipment of the present invention.

[0059] Figure 4 It is a schematic diagram of the workflow of ground-air coordinated traffic accident monitoring of the present invention. DETAILED DESCRIPTION

[0060] The technical solution of the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments, which are preferred embodiments of the present invention. It should be understood that the described embodiments are only part of the embodiments of the present invention, not all of the embodiments; it should be noted that the embodiments of the present invention and the features in the embodiments can be combined with each other without conflict. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0061] The main idea of ​​the technical solution of the embodiment of the present invention: The present invention discloses a sparse road network monitoring and control method based on the collaboration of ground equipment and drones. By integrating ground traffic detection equipment with aerial drone monitoring systems, efficient monitoring of traffic conditions in sparse road networks in complex environments is achieved. The ground perception part uses a high-definition camera system and an intelligent video camera to detect traffic events, collect data in real time and optimize equipment layout; the aerial mobility part uses a drone cluster equipped with high-resolution camera equipment to conduct scene feature analysis to complete three-dimensional reconstruction of road scenes, optimize cruise paths and divide monitoring areas. The present invention also includes static and dynamic ground-to-air collaboration methods, as well as TSP optimization algorithms, to improve the coverage and accuracy of traffic monitoring and reduce costs.

[0062] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0063] Embodiment 1

[0064] Embodiment 1 of the present invention provides a method for ground detection equipment layout, such as Figure 1 As shown, the detection method may specifically include the following steps:

[0065] S1. Detect sparse road events through high-definition camera systems and intelligent video cameras, and optimize the layout of ground traffic detection equipment based on traffic accident prediction and road segmentation;

[0066] S2. Reconstruct the image information captured by the drone in three dimensions and arrange the air traffic detection equipment according to the cruise path collaborative optimization method and traffic monitoring area division;

[0067] S3. Optimization of the layout of ground-air coordinated traffic detection equipment, including static ground-air coordination and dynamic ground-air coordination.

[0068] Furthermore, the sparse road network monitoring and control method based on the collaboration of ground equipment and unmanned aerial vehicles comprises:

[0069] High-definition camera systems and intelligent video cameras are introduced to track and identify individual vehicles and jointly collect signal image information in multiple scenes. When a vehicle passes through the sensitive area of ​​the vehicle position sensor, the sensor sends a signal to the image acquisition control part, which controls the camera to collect a car image and send it to the image preprocessing module. The preprocessing module performs simple processing on the input image and then sends it to the PC. The software module in the PC completes the reproduction of the license plate characters through several links such as image preprocessing, license plate extraction, license plate image binarization, character segmentation, and character recognition, and finally generates recognition results. The recognition results and images are stored in the database for subsequent license plate queries, traffic flow statistics, and toll management. When a vehicle passes through the circular induction coil installed in the monitoring lane, it triggers the high-definition license plate camera to take a photo of the vehicle, and the photo is stored in the information platform.

[0070] Among them, the high-definition checkpoint system mainly uses the vehicle license plate comparison method to detect traffic incidents in closed sections of road; the intelligent video camera is mainly used for manual or automatic video monitoring at intersections (or overpass areas), and the intelligent video camera can be used as a redundant supplementary device for the high-definition checkpoint camera in closed sections of road.

[0071] Furthermore, the sparse road network monitoring and control method based on the collaboration of ground equipment and drones is characterized in that the traffic accidents on the road section are predicted, and the length of the road section is determined by the fixed length method or the variable length method. When laying out the ground traffic detection equipment, it is necessary to base it on the safety analysis of the road. Intuitively, the density of traffic detection equipment is high in places with poor road safety levels, and low in places with good road safety levels. Generally speaking, the more corners there are, the more likely a road section is to have an accident.

[0072] The number of traffic accidents that occur on a certain road section each year can be predicted by the following method:

[0073] S31 Count the turning angles and number of horizontal curves in the research section. The sum of the turning angles of each horizontal curve divided by the number of horizontal curves is recorded as a.

[0074] S32 Statistical study of the number of vertical curves in the road section and the angle and slope length of each vertical curve. The sum of the ratio of the angle of each vertical curve to the slope length divided by the number of horizontal curves is recorded as b;

[0075] S33 calculates the percentage of the slope length of each vertical curve to the sum of the slope lengths of the total vertical curves and multiplies it by the angle of each vertical curve, and records it as c;

[0076] S34 counts the proportion of large vehicles passing through the studied road section and denotes it as d;

[0077] S35 integrates the above calculated values ​​using the following formula:

[0078] Q=e (-2676614+0.0071095a+0.2539616c+6.14963d) ·F

[0079] Where P is the predicted number of traffic accidents occurring on the study section each year, and F is the exposure variable.

[0080] Furthermore, the method for monitoring and controlling a sparse road network based on the collaboration of ground equipment and drones includes: setting up high-definition checkpoints for modeling, the road length is 21 kilometers, and it is divided into 21 sections according to the fixed-length division method. The number of high-definition checkpoints set up is 23 in total, corresponding to 22 high-definition checkpoint sections. In increasing order of the set up mileage, the set up location of the high-definition checkpoints is recorded as R = {x 0 ,x 1 ,x m ,…,x m+1}, where the position of the high-definition checkpoint at the starting point of the road is denoted by x 0 = 0, the position of the high-definition checkpoint at the end of the road is recorded as x 23 = 21. The road has 32 intersections (or flyover areas), and the starting mileage of each intersection is recorded as S = {p 1 ,p 2 ,…,p n}, the road end mileage of each intersection (or overpass area) is E = {q 1 ,q 2 ,…,q n}, then the location set of the road intersection (or overpass area) can be recorded as Y = {y|p k ≤y≤q k ,k=1,2,…,n}.

[0081] The following assumptions were made during the modeling process:

[0082] (1) Based on the spatial distribution characteristics of road accidents under sparse road conditions

[0083] (2) Assume that the traffic accidents in each fixed-length sub-section are evenly distributed.

[0084] (3) Assume that the event detection environment of the high-definition camera is under normal weather conditions

[0085] (4) Assume that the high-definition camera records all vehicles entering and leaving the road

[0086] (5) Assume that the layout of the high-definition camera is located in a closed section

[0087] The optimization objective function is as follows:

[0088]

[0089] The constraints are as follows:

[0090] d(m+2)≤D

[0091]

[0092] Where: Δ min Indicates the minimum spacing of the HD camera mount, Δ max Indicates the maximum spacing of the HD mount, Q i is the number of traffic accidents on the road section with high-definition checkpoints, f(x i -x i+1 ) is the relationship between the detection rate of traffic detection events on the i-th HD checkpoint interval section and the checkpoint interval, d is the cost of a single HD checkpoint, and D is the total investment amount.

[0093] Embodiment 2

[0094] Embodiment 2 of the present invention provides a sparse road network monitoring and control method based on the collaboration of ground equipment and unmanned aerial vehicles, such as Figure 2 As shown, the ground-to-air coordination module in the detection and control method mainly includes:

[0095] First, the drone is operated according to the planned aerial flight path, and the accident scene is photographed from different heights and angles to generate a series of image sequences. Then, the aerial photos (including the corresponding latitude, longitude, and altitude) are imported into the 3D reconstruction software, and the OpenMP multi-threaded processing mechanism is used to first optimize the scale of the image, extract the GPS information in the image, and obtain the relative position information between adjacent image elements; then the SIFTGPU feature extraction algorithm is used to extract feature points in the image in parallel, and only feature matching is performed on adjacent pictures, so as to quickly perform image calibration. Then, a sparse 3D point cloud model is generated using the recovery from motion algorithm, and a dense 3D point cloud model is generated using a multi-dimensional stereo vision algorithm based on patches. The 3D point cloud model is meshed and textured, and the digital surface model and digital elevation model of the accident scene are output after optimization.

[0096] Furthermore, in the sparse road network monitoring and control method based on the collaboration of ground equipment and unmanned aerial vehicles, the flight path optimization method of the unmanned aerial vehicle includes:

[0097] S61. Generate a number of UAV cruise path populations, a real weight vector for each cruise path, and a corresponding neighborhood cruise path; the cruise path population sets an objective function reference point.

[0098] S62. Based on the cruise path of the UAV, various constraints are considered, sub-paths are divided, and the objective function value corresponding to each cruise path is generated; the cruise path with the smallest objective function value is taken as the elite path, and the objective function reference point is updated.

[0099] S63. Calculate the Chebyshev value of each cruise path in the neighborhood where the elite path is located. Each cruise path corresponds to a real weight value and an objective function value, and its corresponding Chebyshev value = max{real weight value x(objective function value-value of the objective function reference point)}; when the Chebyshev value of the elite path ≤ the Chebyshev value of the neighborhood cruise path, the elite path is used to replace other cruise paths in the neighborhood, and the drone cruise path population is updated to implement the elite strategy.

[0100] S64. Set the crossover and mutation probabilities, perform crossover and mutation operations on all cruise paths, and improve the diversity of flight paths.

[0101] S65. Calculate the objective function value of each cruise path of the cruise path population after crossover and mutation, find the cruise path with the smallest objective function value as the elite path, and update the objective function reference point.

[0102] S66. Calculate the Chebyshev value of each cruise path in the neighborhood where the elite path obtained in step S65 is located. When the Chebyshev value of the elite path is ≤ the Chebyshev value of the neighborhood cruise path, the elite path is used to replace the neighborhood cruise path, and the drone cruise path population is updated to implement the elite strategy.

[0103] S67. Perform loop iterations to calculate the optimal UAV cruising path and obtain the corresponding objective function value.

[0104] Furthermore, the sparse road network monitoring and control method based on the collaboration of ground equipment and unmanned aerial vehicles is as follows: Figure 3 As shown, the method for dividing the monitoring area includes:

[0105] Selection of monitoring objects: Based on the traffic safety level evaluation of sparse roads, road sections and traffic nodes (such as overpass areas, service areas, toll stations, etc.) with low road safety levels are selected as traffic monitoring objects for drones.

[0106] Division of monitoring areas. According to the spatial distribution of the drone monitoring objects, the monitoring objects are clustered, and the number of monitoring areas is continuously increased to ensure that there is one drone in each monitoring area and the maximum flight distance constraint of the drone is met.

[0107] Aircraft path planning in the area. The drone starts from the base in the monitoring area, inspects all monitored objects, and then returns to the base. The drone's cruising path is required to be the shortest.

[0108] Furthermore, the sparse road network monitoring and control method based on the collaboration of ground equipment and unmanned aerial vehicles is as follows: Figure 4 As shown, the static ground-air coordination method includes:

[0109] According to the importance of intersections and road sections (such as intersections and road sections with poor traffic safety levels), and taking into account the constraints of transportation investment, traffic detection equipment is installed at relevant intersections and road sections to monitor road traffic. For intersections and road sections without traffic detection equipment, drones are assigned to monitor them, thereby effectively supplementing traditional ground traffic monitoring.

[0110] Furthermore, the sparse road network monitoring and control method based on the collaboration of ground equipment and unmanned aerial vehicles, the dynamic ground-air collaboration method includes:

[0111] The vehicle carries multiple drones, which take off and land from the vehicle; the ground control station located in the vehicle allocates reconnaissance targets for the vehicle and drones, and gives inspection / cruise plans for the vehicles and drones; when there is a sudden demand for traffic monitoring, the vehicle (ground control station) dynamically plans the cruise path of the drone (cluster), and then the drone (cluster) performs cruise reconnaissance according to the new planned path; the drone monitors traffic in the air, and after detecting an accident, the drone obtains images and point cloud information of the accident scene through the onboard camera or onboard radar, and returns it to the vehicle (ground control station); finally, based on the image and point cloud information, a reconstruction study of the accident scene is carried out.

Claims

1. A sparse road network monitoring and control method based on the collaboration of ground equipment and unmanned aerial vehicles, characterized in that: The steps include: S1. Detect sparse road traffic events through high-definition camera systems and intelligent video cameras, and optimize the layout of ground traffic detection equipment based on traffic accident prediction and road segmentation; S2. Reconstruct the image information captured by the drone in three dimensions and arrange the air traffic detection equipment according to the cruise path collaborative optimization method and traffic monitoring area division; S3. Optimization of the layout of ground-air coordinated traffic detection equipment, including static ground-air coordinated and dynamic ground-air coordinated modes.

2. The sparse road network monitoring and control method based on the collaboration of ground equipment and unmanned aerial vehicles according to claim 1 is characterized in that: High-definition camera systems and intelligent video cameras are introduced to track and identify individual vehicles and jointly collect signal image information in multiple scenes. When a vehicle passes through the sensitive area of ​​the vehicle position sensor, the sensor sends a signal to the image acquisition control part. The acquisition control part controls the camera to collect a car image and send it to the image preprocessing module. The preprocessing module performs simple processing on the input image and then sends it to the PC. The software module in the PC completes the reproduction of the license plate characters through image preprocessing, license plate extraction, license plate image binarization, character segmentation, and character recognition. The recognition results and images are stored in the database for subsequent license plate queries, traffic flow statistics, and toll management. When a vehicle passes through the circular induction coil installed in the monitoring lane, the high-definition license plate camera is triggered to take a photo of the vehicle, and the photo is stored in the information platform. Among them, the high-definition checkpoint system mainly uses the vehicle license plate comparison method to detect traffic incidents in closed sections of road; the intelligent video camera is mainly used for manual or automatic video monitoring at intersections or overpasses, and the intelligent video camera can be used as a redundant supplementary device for the high-definition checkpoint camera in closed sections of road.

3. The sparse road network monitoring and control method based on the collaboration of ground equipment and unmanned aerial vehicles according to claim 2 is characterized in that: Predict road section traffic accidents and determine the length of the road section by fixed length method or variable length method; when laying out ground traffic detection equipment, the safety analysis of the road should be the basis. Intuitively, the density of traffic detection equipment is high in places with poor road safety levels, and low in places with good road safety levels. Generally speaking, the more corners a road section has, the more likely it is to have an accident. The number of traffic accidents that occur on a certain road section each year can be predicted by the following method: S31. Count the turning angles and number of the horizontal curves in the research section, and the sum of the turning angles of the horizontal curves divided by the number of horizontal curves is recorded as a; S32. The number of vertical curves in the road section under study, the angle and slope length of each vertical curve, and the sum of the ratios of the angle and slope length of each vertical curve divided by the number of horizontal curves are recorded as b; S33. Calculate the percentage of the slope length of each vertical curve to the sum of the slope lengths of the total vertical curves and multiply it by the angle of each vertical curve, denoted as c; S34. Count the proportion of large vehicles passing through the study section and denote it as d; S35. Integrate the above calculated values ​​using the following formula: Q=e (-2676614+0.0071095a+0.2539616c+6.14963d) ·F Where P is the predicted number of traffic accidents occurring on the study section each year, and F is the exposure variable.

4. The sparse road network monitoring and control method based on the collaboration of ground equipment and unmanned aerial vehicles according to claim 3 is characterized in that: The road segment division method includes: setting up high-definition checkpoint modeling, assuming that the road length is L, and dividing it into N sections according to the fixed-length division method, the number of high-definition checkpoints set up is m+2 in total, corresponding to m+1 high-definition checkpoint sections; according to the increasing mileage of the setting, the setting position set of the high-definition checkpoint is recorded as R={x0,x1,x m ,…,x m+1 }, where the position of the high-definition checkpoint at the starting point of the road is recorded as x0=0, and the position of the high-definition checkpoint at the end of the road is recorded as x m+1 = L; Assume that the road has n intersections or flyovers, and the starting mileage of each intersection or flyover is recorded as S = {p1, p2, ..., p n }, the road end mileage of each intersection or overpass area is E = {q1, q2, ..., q n }, then the location set of the road intersection or overpass area can be recorded as Y = {y|p k ≤y≤q k ,k=1,2,…,n}; The following assumptions were made during the modeling process: (1) Based on the spatial distribution characteristics of road accidents under sparse road conditions (2) Assume that the traffic accidents in each fixed-length sub-section are evenly distributed. (3) Assume that the event detection environment of the high-definition camera is under normal weather conditions (4) Assume that the high-definition camera records all vehicles entering and leaving the road (5) Assume that the layout of the high-definition camera is located in a closed section The optimization objective function is as follows: The constraints are as follows: d·(m+2)≤D Where: Δ min Indicates the minimum spacing of the HD camera mount, Δ max Indicates the maximum spacing of the HD mount, Q i is the number of traffic accidents on the road section with high-definition checkpoints, f(x i -x i+1 ) is the relationship between the detection rate of traffic detection events on the i-th HD checkpoint interval section and the checkpoint interval, d is the cost of a single HD checkpoint, and D is the total investment amount.

5. The sparse road network monitoring and control method based on the collaboration of ground equipment and unmanned aerial vehicles according to claim 1 is characterized in that: In step S2, the UAV 3D reconstruction method includes: First, while the UAV is operating according to the planned aerial photography flight path, it takes aerial photos of the accident scene from different heights and angles to generate a series of image sequences; then, the aerial photos, including the corresponding longitude, latitude and altitude, are imported into the 3D reconstruction software, and the OpenMP multi-threaded processing mechanism is used to first optimize the scale of the image, extract the GPS information in the image, and obtain the relative position information between adjacent image elements; then the SIFTGPU feature extraction algorithm is used to extract feature points in the image in parallel, and only feature matching is performed on adjacent images, so as to quickly perform image calibration; then a sparse 3D point cloud model is generated using the motion recovery algorithm, and a dense 3D point cloud model is generated using a patch-based multi-dimensional stereo vision algorithm, and the 3D point cloud model is gridded and texturized, and the digital surface model and digital elevation model of the accident scene are output after optimization.

6. The sparse road network monitoring and control method based on the collaboration of ground equipment and unmanned aerial vehicles according to claim 5 is characterized in that: The flight path optimization method of the UAV comprises: S61. Generate a number of drone cruise path populations, a real weight vector for each cruise path, and a corresponding neighborhood cruise path; set an objective function reference point for the cruise path population; S62. Based on the cruise path of the UAV, various constraints are considered to divide the sub-paths and generate the objective function value corresponding to each cruise path; the cruise path with the smallest objective function value is taken as the elite path, and the objective function reference point is updated; S63. Calculate the Chebyshev value of each cruise path in the neighborhood where the elite path is located. Each cruise path corresponds to a real weight value and an objective function value, and its corresponding Chebyshev value = max{real weight value x(objective function value-value of the objective function reference point)}; when the Chebyshev value of the elite path ≤ the Chebyshev value of the neighborhood cruise path, the elite path is used to replace other cruise paths in the neighborhood, and the drone cruise path population is updated to implement the elite strategy; S64. Set the crossover and mutation probabilities, perform crossover and mutation operations on all cruise paths, and improve the diversity of flight paths; S65. Calculate the objective function value of each cruise path of the cruise path population after crossover and mutation, find the cruise path with the smallest objective function value as the elite path, and update the objective function reference point; S66. Calculate the Chebyshev value of each cruise path in the neighborhood where the elite path obtained in step S65 is located. When the Chebyshev value of the elite path is ≤ the Chebyshev value of the neighborhood cruise path, replace the neighborhood cruise path with the elite path, update the drone cruise path population, and implement the elite strategy; S67. Perform loop iterations to calculate the optimal UAV cruising path and obtain the corresponding objective function value.

7. The sparse road network monitoring and control method based on the collaboration of ground equipment and unmanned aerial vehicles according to claim 6 is characterized in that: The method for dividing the monitoring area includes: Selection of monitoring objects: Based on the traffic safety level evaluation of sparse roads, sections and traffic nodes with low road safety levels are selected as traffic monitoring objects for drones; Monitoring area division: according to the spatial distribution of the drone monitoring objects, the monitoring objects are clustered, and the number of monitoring areas is continuously increased to ensure that there is one drone in each monitoring area and the maximum flight distance constraint of the drone can be met; Aircraft path planning within the area: The drone starts from the base in the monitoring area, inspects all monitored objects, and then returns to the base. The drone's cruising path is required to be the shortest.

8. The sparse road network monitoring and control method based on the collaboration of ground equipment and unmanned aerial vehicles according to claim 1 is characterized in that: The static ground-air coordination method comprises: According to the importance of intersections and road sections, traffic detection equipment will be installed at intersections and road sections with poor traffic safety levels, taking into account the constraints of traffic investment, and road traffic monitoring will be carried out; for intersections and road sections without traffic detection equipment, drones will be assigned to monitor them, thereby effectively supplementing traditional ground traffic monitoring.

9. The sparse road network monitoring and control method based on the collaboration of ground equipment and unmanned aerial vehicles according to claim 1 is characterized in that: The dynamic ground-air coordination method comprises: The vehicle carries multiple drones, which take off and land from the vehicle; the ground control station located in the vehicle allocates reconnaissance targets for the vehicle and drones, and gives inspection / cruise plans for the vehicles and drones; when there is a sudden demand for traffic monitoring, the vehicle or ground control station dynamically plans the cruise path of the drone cluster, and then the drone cluster conducts cruise reconnaissance along the new planned path; the drones conduct traffic monitoring in the air, and after detecting an accident, the drones obtain images and point cloud information of the accident scene through onboard cameras or onboard radars, and return them to the vehicle or ground control station; finally, based on the images and point cloud information, reconstruction research of the accident scene is carried out.

Images