Tunnel-based unmanned aerial vehicle control method and system

Abstract

The invention discloses a tunnel-based unmanned aerial vehicle control method and system, and relates to the technical field of unmanned aerial vehicle control methods, an unmanned aerial vehicle flies in a tunnel along an inspection route, and a traffic state diagram of the tunnel is determined according to a plurality of road surface images collected by the unmanned aerial vehicle in the flight process and a current image of an early warning event; dynamic inspection of the tunnel by the unmanned aerial vehicle is realized, overall consideration of a plurality of pavement images and the current image of the early warning event is compatible, and the accuracy of the traffic state diagram of the tunnel is ensured. Therefore, the accident level of the tunnel is determined based on the congestion degree of the traffic state diagram of the tunnel and the early warning degree of the early warning event. The cooperative control of the unmanned aerial vehicle and the fire-fighting equipment is triggered according to the accident level of the tunnel so as to control the fire-fighting management and control of the early warning event and the dispersion of the vehicle, so that the cooperative control of the unmanned aerial vehicle and the fire-fighting equipment is realized, and the fire-fighting management and control of the early warning event and the dispersion of the vehicle are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicle (UAV) control methods, and particularly to a UAV control method and system based on a tunnel. Background Art

[0002] With the development of technology, vehicles are allowed to pass through tunnels, generally with 3 to 6 lanes. Multiple vehicles are in the same tunnel at the same time and drive on the tunnel road surface. Some vehicle accidents also occur in tunnels. However, the lanes in the tunnel are limited and are single-lane, which affects the inspection of the tunnel and cannot ensure the accuracy of the traffic status map of the tunnel. Summary of the Invention

[0003] The purpose of the present invention is to overcome the deficiencies of the prior art, and the present invention provides a UAV control method and system based on a tunnel.

[0004] An embodiment of the present invention provides a UAV control method based on a tunnel, including: Determining an early warning event in the tunnel according to the real-time image of the tunnel and the early warning information of the tunnel; Determining the inspection route of the UAV relative to the tunnel according to the location of the early warning event and the distribution map of the tunnel, and the inspection route is from the current position of the UAV to the location of the early warning event; The UAV flies in the tunnel along the inspection route, and determines the traffic status map of the tunnel according to multiple road surface images collected during the flight of the UAV and the current image of the early warning event; Determining the accident level of the tunnel based on the congestion degree of the traffic status map of the tunnel and the early warning degree of the early warning event; Triggering the collaborative control of the UAV and fire-fighting equipment according to the accident level of the tunnel to control the fire control of the early warning event and the diversion of vehicles.

[0005] An embodiment of the present invention provides a UAV control system based on a tunnel. The UAV control system based on a tunnel is applied to the above-mentioned UAV control method based on a tunnel. The UAV control system based on a tunnel includes: An early warning event module, configured to determine an early warning event in the tunnel according to the real-time image of the tunnel and the early warning information of the tunnel; An inspection route module, configured to determine the inspection route of the UAV relative to the tunnel according to the location of the early warning event and the distribution map of the tunnel, and the inspection route is from the current position of the UAV to the location of the early warning event; A traffic status map module, configured to fly the UAV in the tunnel along the inspection route, and determine the traffic status map of the tunnel according to multiple road surface images collected during the flight of the UAV and the current image of the early warning event; An accident level module, which is used to determine the accident level of the tunnel based on the congestion degree of the traffic status map of the tunnel and the warning degree of the warning event; A collaborative control module, which is used to trigger the collaborative control of the drone and the fire-fighting equipment according to the accident level of the tunnel, so as to control the fire control and vehicle diversion of the warning event.

[0006] Compared with the prior art, the beneficial effects of the present invention are: In the embodiment of the present invention, by the method in the embodiment of the present invention, the warning event in the tunnel is determined according to the real-time image of the tunnel and the warning information of the tunnel; according to the location of the warning event and the distribution map of the tunnel, the inspection route of the drone relative to the tunnel is determined, and the inspection route is from the current position of the drone to the location of the warning event; the drone flies in the tunnel along the inspection route, and determines the traffic status map of the tunnel according to a plurality of road surface images collected during the flight of the drone and the current image of the warning event, realizing the dynamic inspection of the tunnel by the drone, taking into account the overall consideration of a plurality of road surface images and the current image of the warning event, and ensuring the accuracy of the traffic status map of the tunnel.

[0007] Therefore, the accident level of the tunnel is determined based on the congestion degree of the traffic status map of the tunnel and the warning degree of the warning event; the collaborative control of the drone and the fire-fighting equipment is triggered according to the accident level of the tunnel, so as to control the fire control and vehicle diversion of the warning event, realizing the collaborative control of the drone and the fire-fighting equipment, ensuring the collaborative control of the drone and the fire-fighting equipment in different dimensions, and realizing the fire control and vehicle diversion of the warning event. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Figure 1 is a schematic flowchart of the method for controlling a drone based on a tunnel in an embodiment of the present invention; Figure 2 is a schematic flowchart of step S11 in the method for controlling a drone based on a tunnel in an embodiment of the present invention; Figure 3 is a schematic flowchart of step S12 in the method for controlling a drone based on a tunnel in an embodiment of the present invention; Figure 4 is a schematic flowchart of step S13 in the method for controlling a drone based on a tunnel in an embodiment of the present invention; Figure 5 is a schematic flowchart of step S14 in the method for controlling a drone based on a tunnel in an embodiment of the present invention; Figure 6 is a schematic flowchart of step S15 in the method for controlling a drone based on a tunnel in an embodiment of the present invention; Figure 7It is a schematic diagram of the structural composition of the tunnel-based UAV control system in the embodiments of the present invention. Detailed implementation manners

[0009] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention.

[0010] Please refer to Figures 1 to 7 , a tunnel-based UAV control method, which is applied to a tunnel-based UAV control scenario; the tunnel-based UAV control method includes: Step S11: Determine the warning events in the tunnel according to the real-time image of the tunnel and the warning information of the tunnel; Step S12: Determine the inspection route of the UAV relative to the tunnel according to the location of the warning event and the distribution map of the tunnel, and the inspection route is from the current position of the UAV to the location of the warning event; Step S13: The UAV flies in the tunnel along the inspection route, and determines the traffic status map of the tunnel according to multiple road surface images collected during the flight of the UAV and the current image of the warning event; Step S14: Determine the accident level of the tunnel based on the congestion degree of the traffic status map of the tunnel and the warning degree of the warning event; Step S15: Trigger the collaborative control of the UAV and the fire-fighting equipment according to the accident level of the tunnel to control the fire control of the warning event and the diversion of vehicles; Refer to Figure 2 , in step S11, determine the warning events in the tunnel according to the real-time image of the tunnel and the warning information of the tunnel; In the specific implementation process of the present invention, the specific steps are as follows: S111: Collect multiple images at different positions according to the shooting of multiple cameras in the tunnel, determine the real-time image of the tunnel according to the synthesis of the multiple images, and determine multiple warning areas by identifying the real-time image of the tunnel; S112: Determine the corresponding warning targets and corresponding warning positions according to the recheck of the multiple warning areas, collect the warning information of the tunnel, and determine the warning events in the tunnel according to the matching of the warning information of the tunnel, the multiple warning areas and the corresponding warning positions.

[0011] In the embodiments of the present application, multiple images at different positions are collected according to the shooting of multiple cameras in the tunnel, the real-time image of the tunnel is determined according to the synthesis of the multiple images, and multiple warning areas are determined by identifying the real-time image of the tunnel; At this time, multiple cameras installed in the tunnel are used to capture different positions inside the tunnel; these cameras are usually installed at the entrance, exit, turning points, key nodes, and different positions along the line of the tunnel to ensure comprehensive coverage of the tunnel interior; each camera will capture images within its field of view in real time and transmit this image data to the central processing system.

[0012] After receiving images from multiple cameras, the central processing system will perform composite processing on these images and effectively splice these images to ensure that the synthesized image can accurately reflect the actual situation inside the tunnel; image stitching technology can seamlessly stitch images from different cameras together to form a comprehensive real-time image of the tunnel.

[0013] After obtaining the real-time image of the tunnel, the central processing system will analyze the image using image recognition methods to identify possible warning areas; these warning areas may be the locations where potential dangerous behaviors such as vehicle congregation, abnormal driving, and pedestrian intrusion occur; image recognition methods can be implemented based on technologies such as preset deep learning. Optionally, the central processing system uses image recognition methods to analyze the real-time image of the tunnel; assume that in the field of view of a camera in the middle section of the tunnel, several vehicles are congregating and their driving speed has significantly slowed down; according to the preset rules, this situation may be regarded as a potential traffic accident warning area; therefore, the system will demarcate a warning area around this area and mark it on the real-time image of the tunnel.

[0014] Specifically, assume that a tunnel is 2 kilometers long and a high-definition camera is installed every 200 meters, for a total of 11 cameras installed; these cameras are all connected to an image acquisition system in the central control room; when there are vehicles driving or pedestrians passing through the tunnel, each camera will capture images within its field of view in real time and transmit these images to the central processing system at a speed of 30 frames per second.

[0015] The central processing system has received image data from 11 cameras; the system will first preprocess these images, including removing noise, adjusting brightness and contrast, etc.; then, using image stitching technology, stitch these images according to their actual positions in the tunnel to form a complete real-time image of the tunnel; this image can be displayed in real time on the screen in the central control room for operators to monitor.

[0016] The central processing system analyzes the real-time images of the tunnel using image recognition. Suppose it is detected by the method that within the field of view of a camera in the middle section of the tunnel, several vehicles are gathered together and their driving speeds are significantly reduced. According to the preset rules, this situation may be regarded as a potential traffic accident warning area. Therefore, the system will delimit a warning area around this area and mark it on the real-time tunnel image.

[0017] Furthermore, based on the review of multiple warning areas, the corresponding warning targets and corresponding warning positions are determined, the warning information of the tunnel is collected, and the warning events in the tunnel are determined according to the matching of the warning information of the tunnel, multiple warning areas, and the corresponding warning positions. Considering the overall matching of the warning information of the tunnel, multiple warning areas, and the corresponding warning positions, the accuracy of the warning events in the tunnel is ensured.

[0018] At this time, multiple warning areas are reviewed, and the corresponding warning targets and corresponding warning positions are determined according to the review of multiple warning areas. At the same time, target marking is carried out for the warning areas, so as to determine the warning targets for the detection of some targets and mark the warning positions of the warning targets.

[0019] The warning information of the tunnel is collected, the warning information of the tunnel, multiple warning areas, and the corresponding warning positions are introduced, and multiple-dimensional marking is carried out on the warning information of the tunnel, multiple warning areas, and the corresponding warning positions. Thus, warning events in the tunnel are formed according to the matching of the warning information of the tunnel, multiple warning areas, and the corresponding warning positions.

[0020] In an embodiment of the present application, the warning areas are specific areas delimited in the tunnel based on historical data and risk assessment. These areas may have relatively high safety risks. For example, positions such as the tunnel entrance, exit, bend, long downhill, etc. are usually set as warning areas. The warning position refers to the specific position where the warning event occurs, which is usually determined by combining image recognition algorithms with GIS map data. For example, when the camera captures a vehicle breakdown, the algorithm can identify the specific position of the faulty vehicle and map it to the GIS map of the tunnel.

[0021] The time dimension, space dimension, and attribute dimension are introduced. At this time, the time when the warning information occurs is recorded to analyze the development trend and periodicity of the event; combining the warning position and warning area, the spatial distribution characteristics of the event are analyzed; according to the type of warning information (such as environmental monitoring data, traffic flow data, video image data) and specific values, the event is marked with attributes. After marking in multiple dimensions, the system can associate the warning information with the warning area and warning position through a matching algorithm, thus forming a warning event.

[0022] Specifically, assume that three warning areas are set in a certain tunnel: Area A (tunnel entrance), Area B (long downhill section), and Area C (curve); one day, the system collects the following warning information: environmental monitoring data: the smoke concentration in Area B suddenly increases and reaches the alarm threshold; video image data: the camera in Area C captures a truck with a breakdown parked on the lane, and the vehicles behind are queuing; traffic flow data: the traffic flow in Area A suddenly increases and the vehicle speed slows down.

[0023] The system makes multi-dimensional markings based on this information: Time: all events occur within the same time period; Space: the smoke concentration in Area B increases, there is a truck breakdown in Area C, and there is traffic congestion in Area A; Attribute: smoke alarm in Area B, vehicle breakdown in Area C, and abnormal traffic flow in Area A.

[0024] Then, the system associates these warning information with warning areas and warning positions through a matching algorithm, forming the following warning events: Event 1: The smoke concentration in Area B exceeds the standard, and there may be a fire risk. It is necessary to immediately activate the fire emergency plan; Event 2: The truck breakdown in Area C causes traffic congestion, and it is necessary to dispatch rescue vehicles to deal with it; Event 3: The traffic flow in Area A increases and the vehicle speed slows down, and there may be a risk of traffic accidents. It is necessary to strengthen traffic guidance; Based on these warning events, the tunnel management department can quickly take actions to ensure traffic safety in the tunnel.

[0025] Reference Figure 3 , in step S12, according to the location of the warning event and the distribution map of the tunnel, determine the inspection route of the drone relative to the tunnel, and this inspection route is from the current position of the drone to the location of the warning event; In the specific implementation process of the present invention, the specific steps are as follows: S121: Determine the distribution map of the tunnel according to the name of the tunnel and the tunnel database, mark the location of the warning event and the current position of the drone based on the distribution map of the tunnel. At this time, the drone stays in the helipad of the tunnel; S122: Determine the flight route according to the current position of the drone, the location of the warning event, and the traffic direction of the tunnel, and determine the corresponding flight space by traversing along this first flight video. Determine the inspection route of the drone relative to the tunnel according to the flight route and the corresponding flight space. At this time, the drone adjusts its attitude according to the corresponding flight space to smoothly pass through the flight space.

[0026] In the embodiment of the present application, determine the distribution map of the tunnel according to the name of the tunnel and the tunnel database, mark the location of the warning event and the current position of the drone based on the distribution map of the tunnel. At this time, the drone stays in the helipad of the tunnel; At this time, the name of the tunnel and the tunnel database are introduced, and the name of the tunnel and the tunnel database are controlled to facilitate finding the distribution map of the tunnel corresponding to the name of the tunnel in traversing the tunnel database, and further controlling the distribution map of the tunnel, so as to mark the location of the warning event and the current location of the drone based on the distribution map of the tunnel.

[0027] Specifically, by inputting the name of the tunnel as the unique identifier for querying the database and obtaining relevant information, an input box is set in the user interface or the background management interface of the system for the operator to input the name of the tunnel. This name is usually part of the unique key or primary key when the tunnel is stored in the database.

[0028] The tunnel database should contain key information such as the name of the tunnel, geographical location, structural layout, location of key facilities, historical warning event records, etc. This information should be stored in a structured manner for easy querying and updating. Regularly maintain and update the database to ensure the accuracy and timeliness of the information. This includes adding information about new tunnels, updating data of existing tunnels, deleting outdated or invalid records, etc. To prevent data loss or damage, regularly back up the database. The backup can be stored on a local or remote server to ensure that the data can be quickly restored when needed.

[0029] Find the distribution map of the tunnel corresponding to the input tunnel name in the database. At this time, according to the input tunnel name, construct an SQL query statement (or other database query language) to retrieve relevant information in the database. Submit the query statement to the database server for execution and wait for the return result. Process the query result and extract the distribution map information of the tunnel that matches the tunnel name. This usually includes the file path of the distribution map, image data, or the URL linking to the online map service, etc.

[0030] Ensure the accuracy and availability of the distribution map to provide a basis for subsequent location marking and flight route planning. At the same time, mark the location of the warning event and the current location of the drone on the distribution map. At this time, obtain the specific longitude and latitude coordinates, mileage stake numbers, or other location identifiers from the warning event report or the drone status information. Find the point corresponding to the warning event location on the distribution map and use different colors, icons, or text markings to distinguish. Similarly, find the current location of the drone (such as the apron) on the distribution map and use the corresponding markings to represent.

[0031] Furthermore, determine the flight route based on the current position of the drone, the location of the warning event, and the traffic direction of the tunnel. Determine the corresponding flight space according to the traversal of the first flight video. Determine the inspection route of the drone relative to the tunnel based on the flight route and the corresponding flight space. At this time, the drone adjusts its attitude according to the corresponding flight space to smoothly pass through the flight space, taking into account the overall consideration of the flight route and the corresponding flight space, and ensuring the accuracy of the inspection route of the drone relative to the tunnel.

[0032] At this time, the current position of the drone, the location of the warning event, and the traffic direction of the tunnel are introduced, and the current position of the drone, the location of the warning event, and the traffic direction of the tunnel are synthesized to output the corresponding flight route, so as to guide the flight of the drone based on the flight route. Traverse the first flight route, and determine the corresponding flight space during the traversal of the first flight route, realizing the combination of the flight route and the corresponding flight space, and outputting the inspection route of the drone relative to the tunnel, introducing the inspection route. At this time, the drone adjusts its attitude according to the corresponding flight space to smoothly pass through the flight space, taking into account the overall consideration of the flight route and the corresponding flight space, and ensuring the accuracy of the inspection route of the drone relative to the tunnel.

[0033] Specifically, synthesize the current position of the drone, the location of the warning event, and the traffic direction of the tunnel to output the corresponding flight route. At the same time, first, ensure that the current position of the drone and the location of the warning event have been accurately obtained, and the traffic direction of the tunnel is known. This information can be obtained through GPS positioning, tunnel management systems, or warning event reports. Based on this information, use path planning algorithms (such as A* algorithm, Dijkstra algorithm, or heuristic search algorithm) to preliminarily plan a flight route from the current position of the drone to the location of the warning event. This route should avoid obstacles as much as possible while considering the traffic direction of the tunnel and the flight performance of the drone.

[0034] During the traversal of the first flight route, determine the three-dimensional space (i.e., flight space) occupied by the drone during flight. At this time, establish a three-dimensional space model according to the structural layout of the tunnel and the size of the drone. This model should include key information such as the width, height, bend, and slope of the tunnel, as well as the additional space (such as safety margin) required by the drone during flight. Along the preliminarily planned flight route, gradually determine the flight space occupied by the drone at each position. This usually involves the analysis of the tunnel cross-section, the prediction of the drone's attitude, and possible collision detection.

[0035] Combine the flight route and the corresponding flight space, output the inspection route of the drone relative to the tunnel, and ensure that the drone can smoothly pass through the flight space. At this time, based on the flight space and the flight performance of the drone, optimize the initially planned flight route; this may include adjusting parameters such as flight altitude, speed, direction, etc., to ensure that the drone does not collide with the tunnel wall or other obstacles during flight; generate attitude control instructions for the drone according to the optimized inspection route and flight space; these instructions should be able to guide the drone to maintain the correct attitude during flight to adapt to the changes in the tunnel and possible flight challenges; send the generated attitude control instructions to the drone and monitor its flight status and position in real time; during flight, adjust the instructions as needed to ensure that the drone can complete the inspection task safely and accurately.

[0036] Therefore, ensure that the determination processes of the flight route and the flight space are mutually compatible to ensure the accuracy of the inspection route of the drone relative to the tunnel; in the process of planning the flight route and determining the flight space, adopt an iterative optimization method; this means that after each determination of the flight space, evaluate and adjust the flight route to ensure that it conforms to the flight performance of the drone and the structural characteristics of the tunnel; always consider a certain safety margin when planning the flight route and determining the flight space; this includes leaving a certain margin in terms of flight altitude, speed, direction, etc. to cope with possible flight errors or emergencies.

[0037] Reference Figure 4 , in step S13, the drone flies in the tunnel along the inspection route, and determines the traffic state map of the tunnel according to multiple road surface images collected by the drone during flight and the current image of the warning event; In the specific implementation process of the present invention, the specific steps are as follows: S131: Trigger the flight of the drone based on the inspection route. The drone is compatible with attitude control and dynamic shooting during flight. In the dynamic shooting of the drone, the drone synchronously shoots in each lane and collects multiple road surface images, and determines the traffic distribution map of the tunnel according to the stitching of the multiple road surface images; S132: Collect the current image of the warning event, determine the corresponding traffic accident type according to the recognition of the current image of the warning event, match the corresponding traffic impact range based on the traffic accident type, and determine the traffic state map of the tunnel according to the traffic distribution map of the tunnel and the corresponding traffic impact range.

[0038] In an embodiment of the present application, the flight of the drone is triggered based on the inspection route. During the flight, the drone is compatible with attitude regulation and dynamic shooting. During the dynamic shooting of the drone, the drone synchronously shoots in each lane and collects multiple road surface images. The traffic distribution map of the tunnel is determined based on the stitching of the multiple road surface images, ensuring the accuracy of the traffic distribution map of the tunnel.

[0039] At this time, the system triggers the takeoff and flight of the drone according to the preset inspection route; the inspection route is usually planned during the tunnel design stage, considering factors such as the tunnel structure, key nodes, potential risk areas, etc.; the system generates flight instructions for the drone according to this route, including parameters such as the takeoff point, flight altitude, speed, and heading. Optionally, the system sends a takeoff command to the drone through the drone control software or API; this command contains detailed information about the inspection route, such as the starting point coordinates, ending point coordinates, coordinates of key points passed through, flight altitude limits, etc.; after receiving the command, the drone performs a self-check before takeoff, including battery power, sensor status, communication link, etc., and starts takeoff and flies according to the command after ensuring everything is normal.

[0040] During the flight, the drone needs to continuously perform attitude regulation to ensure that it can fly stably and move forward along the preset route; attitude regulation usually involves adjusting the drone in three directions: pitch, yaw, and roll; sensors on the drone (such as gyroscopes, accelerometers, magnetometers, etc.) will detect the attitude information of the drone in real time and adjust it through the flight control system to maintain the stability and accuracy of the drone. Optionally, during the flight of the drone, its built-in flight control system continuously receives attitude information from the sensors and calculates the deviation between the current actual attitude and the desired attitude of the drone through algorithms; then, the control system sends corresponding control commands to the motors of the drone to adjust their rotation speed and direction, thereby changing the attitude of the drone and making it gradually approach the desired attitude; this process is carried out in real time, ensuring the stability and accuracy of the drone during flight.

[0041] During the flight, the drone needs to perform dynamic shooting, that is, capture images of the tunnel road surface in real time; to achieve this, the drone is usually equipped with a high-resolution camera and a stable gimbal system; the camera is responsible for capturing images, and the gimbal system is responsible for maintaining the stability and accuracy of the camera, ensuring image quality even when encountering interference factors such as wind shear or bumps during flight. Optionally, during the flight of the drone, its camera captures images in real time according to the preset shooting parameters (such as exposure time, shutter speed, ISO, etc.); at the same time, the gimbal system adjusts according to the attitude information and flight parameters of the drone to maintain the stability and accuracy of the camera; the captured images are transmitted to the ground control station or cloud server in real time for storage and processing.

[0042] To achieve comprehensive traffic monitoring, drones need to take synchronized pictures on each lane; this usually means that drones need to be equipped with wide-angle lenses or multi-camera systems to ensure that they can capture the road surface conditions of all lanes simultaneously; multiple road surface images obtained from the shooting are then stitched together to form a complete traffic distribution map of the tunnel; this distribution map can display key traffic information such as the distribution of vehicles in the tunnel, driving speeds, and lane occupancy; optionally, during flight, the camera system on the drone takes synchronized pictures on each lane according to preset shooting parameters and lane information; the images obtained from the shooting are transmitted in real time to a ground control station or a cloud server for processing; the image processing algorithm identifies the vehicles and road surface features on each lane and stitches multiple road surface images together to form a complete traffic distribution map of the tunnel; this distribution map can be displayed and analyzed through visualization software or an API.

[0043] Specifically, assume there is a tunnel named "YY Tunnel", and the system has planned a detailed inspection route for it; the drone takes off from a helipad near the tunnel entrance and starts flying according to the preset route; during flight, the drone continuously adjusts its attitude to maintain its stability and accuracy; at the same time, the camera system on the drone takes synchronized pictures on each lane to capture real-time images of the tunnel road surface; the images obtained from the shooting are transmitted in real time to a ground control station for processing; the image processing algorithm identifies the vehicles and road surface features on each lane and stitches multiple road surface images together to form a complete traffic distribution map of the "YY Tunnel"; through this distribution map, the operator can clearly see the distribution of vehicles in the tunnel, driving speeds, and any signs of possible traffic congestion or accidents.

[0044] Furthermore, collect the current image of the warning event, determine the corresponding traffic accident type based on the recognition of the current image of the warning event, match the corresponding traffic impact range based on the traffic accident type, and determine the traffic status map of the tunnel according to the traffic distribution map of the tunnel and the corresponding traffic impact range, taking into account the overall consideration of the traffic distribution map of the tunnel and the corresponding traffic impact range, and ensuring the accuracy of the traffic status map of the tunnel.

[0045] At this time, image acquisition is carried out on the scene where the warning event occurs; this is usually done through a high-definition camera carried on a drone; when the drone receives an instruction to fly to the location of the warning event, it will adjust its flight attitude and altitude to ensure that it can clearly capture the current image of the warning event; the acquired image needs to contain sufficient information for subsequent image recognition and accident type judgment. Optionally, after receiving the instruction, the drone will automatically adjust its flight path and quickly reach the location where the warning event occurs; after arriving, the drone will hover at an appropriate altitude and use a high-definition camera to capture the image of the scene; these images will be transmitted in real time to the ground control station or cloud server for storage and processing.

[0046] After acquiring the current image of the warning event, the next step is to use an image recognition algorithm to process and analyze the image to determine the type of traffic accident that occurred; this usually involves using a deep learning model (such as a convolutional neural network CNN) to extract features and classify the image; through the trained model, elements such as vehicles, pedestrians, and road obstacles in the image can be accurately identified, and the type of traffic accident can be judged based on the spatial relationship and features of these elements. Optionally, on the ground control station or cloud server, the image recognition algorithm will process the acquired warning event image; the algorithm will first preprocess the image, such as denoising and enhancing contrast, to improve the image quality; then, the algorithm will use the trained deep learning model to extract features and classify the image; by comparing the features in the image with the preset traffic accident type feature library, the algorithm can determine the most matching traffic accident type.

[0047] After determining the type of traffic accident, the next step is to match the corresponding traffic impact range according to the accident type; this usually involves using a preset traffic accident impact range database, which stores information on the traffic impact ranges that different traffic accident types may cause; by matching the accident type and the records in the database, the traffic congestion, road closure, etc. impact ranges that the accident may cause can be determined; optionally, after determining the type of traffic accident, the system will query the corresponding impact range information in the traffic accident impact range database; the database may contain impact range data for different types of accidents under different conditions, such as accident duration, number of affected lanes, estimated recovery time, etc.; the system will determine the specific traffic impact range based on these data and the current actual situation (such as road conditions, traffic flow, etc.).

[0048] It is necessary to generate a traffic status map of the tunnel based on the traffic distribution map of the tunnel and the determined traffic impact range; the traffic distribution map is usually a two-dimensional or three-dimensional map that shows information such as the distribution of vehicles in the tunnel and their driving speeds; while the traffic status map is based on the traffic distribution map and superimposes information about traffic accidents and their impact ranges to visually display the traffic conditions in the tunnel; optionally, after determining the traffic impact range, the system will superimpose this information on the traffic distribution map of the tunnel; this usually involves using geographic information system (GIS) technology or map APIs to achieve; through GIS technology or map APIs, the system can display information such as the location and impact range of traffic accidents graphically on the traffic distribution map, thereby generating the traffic status map of the tunnel; this status map can be updated in real time to reflect changes in the traffic conditions in the tunnel.

[0049] Specifically, assume that a vehicle rear-end collision occurred in the "ZZ Tunnel"; after receiving the instruction, the drone quickly arrived at the accident scene and captured the current image; through the processing and analysis of the image recognition algorithm, the system determined that this was a vehicle rear-end collision; then, the system queried the corresponding impact range information in the traffic accident impact range database according to the accident type and determined the traffic congestion and road closure situations that the accident might cause; finally, the system superimposed this information on the traffic distribution map of the "ZZ Tunnel" to generate the traffic status map of the tunnel; through this status map, the operator can clearly see the location of the accident scene, the impact range, and the traffic conditions in other areas of the tunnel, so as to take corresponding countermeasures in a timely manner.

[0050] In an embodiment of the present application, in order to quickly identify the type of traffic accident, we can adopt a preset traffic accident type matching table. This traffic accident type matching table associates image features with traffic accident types. The traffic accident type matching table is shown in Table 1: Table 1 Traffic Accident Type Matching Table

[0051] After the image recognition method extracts the features of the warning event image, it will compare them with the features in the traffic accident type matching table to find the most matching accident type. Different accident types have different impact ranges. At this time, a vehicle rear-end collision affects the accident lane and may cause short-term congestion; a vehicle rollover / overturn accident affects the accident lane and the adjacent lane and may cause long-term congestion; a vehicle-pedestrian collision accident affects the accident lane and may cause traffic interruption; a vehicle fire accident affects the entire tunnel and requires emergency evacuation and tunnel closure.

[0052] Reference Figure 5, in step S14, determine the accident level of the tunnel based on the congestion level of the traffic status map of the tunnel and the warning level of the warning event; In the specific implementation process of the present invention, the specific steps are as follows: S141: Generate traffic images of each lane based on the division of the traffic status map of the tunnel, determine the corresponding traffic congestion coefficient according to the recognition of the traffic images of each lane, and determine the congestion level of the traffic status map of the tunnel according to multiple lanes and the corresponding traffic congestion coefficients; S142: Collect the types of traffic accidents corresponding to the warning event, determine the warning level of the warning event according to the type of traffic accident and the real-time image of the warning event, and determine the accident level of the tunnel according to the warning level of the warning event, the congestion level of the traffic status map of the tunnel, and the accident level mapping relationship.

[0053] In the embodiment of the present application, traffic images of each lane are generated based on the division of the traffic status map of the tunnel, the corresponding traffic congestion coefficient is determined according to the recognition of the traffic images of each lane, and the congestion level of the traffic status map of the tunnel is determined according to multiple lanes and the corresponding traffic congestion coefficients, which takes into account the overall situation of multiple lanes and the corresponding traffic congestion coefficients, and ensures the accuracy of the congestion level of the traffic status map of the tunnel.

[0054] At this time, the system first needs to accurately divide the traffic status map of the tunnel in order to separate the traffic conditions of each lane from the entire traffic status map; this usually involves image processing techniques, such as image segmentation or edge detection, to identify and extract the boundaries of each lane in the tunnel; once the boundaries of the lanes are determined, the system can crop or segment the traffic images of each lane; these images should clearly show the vehicle distribution, driving speed, and any factors that may affect the traffic flow (such as accidents, construction, or road obstacles) on each lane. Optionally, assume that the "CC Tunnel" has three lanes and the system has generated an image containing the entire traffic status of the tunnel; by using the Canny edge detection operator, the system can identify the edges of the tunnel walls and lane dividers; then, through morphological operations, the system cleans the noise in the edge image and determines the exact boundaries of each lane; finally, the system crops the traffic images of each lane, which can now be used for calculating the traffic congestion coefficient in the subsequent steps.

[0055] The system needs to further analyze the traffic images of each lane to determine the traffic congestion level of that lane; this typically involves counting the vehicles in the image, measuring the distances between vehicles, calculating the average speed, or identifying any abnormal traffic behaviors (such as sudden braking, frequent lane changes, etc.); based on these analyses, the system can calculate a traffic congestion coefficient, which is a quantitative indicator used to reflect the congestion level of the lane; this coefficient can be based on vehicle density (such as the number of vehicles per kilometer), speed (such as the percentage reduction in average speed), or other traffic flow parameters. Optionally, for the traffic images of each lane in the "CC Tunnel", the system uses background subtraction to detect vehicles and calculates the vehicle density on each lane; at the same time, the system also calculates the average speed by tracking the vehicle positions in consecutive frames; based on these data, the system calculates a traffic congestion coefficient for each lane; for example, the congestion coefficient of Lane 1 is 0.6 (indicating moderate congestion), the congestion coefficient of Lane 2 is 0.4 (indicating mild congestion), and for Lane 3 which is closed due to construction, its congestion coefficient is set to 1.0 (indicating extreme congestion, even though there are actually no vehicles traveling in that lane).

[0056] The system needs to synthesize the traffic congestion coefficients of each lane to determine the traffic congestion level of the entire tunnel; this typically involves performing a weighted average, summing, or applying other aggregation functions to the congestion coefficients to obtain a quantitative indicator reflecting the overall traffic condition of the tunnel; in addition, the system can also consider the importance of the lanes (such as the difference between main lanes and auxiliary lanes), the imbalance of traffic flow, and any other factors that may affect traffic flow (such as weather conditions, special events, etc.); optionally, for the "CC Tunnel", the system considers the importance of the lanes and the imbalance of traffic flow; Lane 1 is the main lane and bears most of the traffic flow, so its congestion coefficient is given a higher weight during synthesis; Lane 2 is an auxiliary lane with less traffic flow, so the weight of its congestion coefficient is lower; for Lane 3 which is closed due to construction, its congestion coefficient is ignored during synthesis (or considered as a fixed high value to reflect the impact of construction on traffic); based on these weights and congestion coefficients, the system calculates the overall congestion level of the tunnel; for example, the calculated congestion level through weighted average is 0.55 (indicating between moderate and high congestion); this quantitative indicator can now be used to guide traffic management decisions, emergency response plans, or provide real-time traffic information to the public.

[0057] Furthermore, collect the types of traffic accidents corresponding to the warning event, determine the warning level of the warning event based on the type of traffic accident and the real-time image of the warning event, and determine the accident level of the tunnel according to the warning level of the warning event, the congestion level of the traffic status map of the tunnel, and the accident level mapping relationship. It takes into account the warning level of the warning event, the congestion level of the traffic status map of the tunnel, and the accident level mapping relationship as a whole, ensuring the accuracy of the accident level of the tunnel.

[0058] At this time, the system needs to determine the type of traffic accident involved in the warning event; this is usually achieved by identifying relevant data of the warning event, which may come from traffic monitoring cameras, vehicle sensors, traffic accident reports, or emergency service calls; the system needs to be able to parse this data to identify the specific type of traffic accident, such as vehicle rear-end collision, rollover, fire, pedestrian collision, etc.; this may require the use of pattern matching, natural language processing, or machine learning techniques to accurately classify the accident type. Optionally, assume that a traffic accident occurred in the "DD Tunnel", and the system captured a video of the accident scene through a monitoring camera; through video analysis, the system identified that the accident involved a rear-end collision of two vehicles; therefore, the system marked the accident type of this warning event as "vehicle rear-end collision".

[0059] The system needs to analyze the real-time image of the warning event and combine the known accident type to determine the severity or urgency of the warning event; this usually involves the assessment of factors such as the degree of vehicle damage, the number of injured people, the degree of road damage, and the scale of the fire in the image; the system may need to use image processing techniques (such as edge detection, feature extraction) and machine learning models (such as classifiers or regressors) to automatically analyze the image and determine the warning level; the warning level can be a numerical score, level, or category, used to represent the urgency and potential impact of the accident. Optionally, in the vehicle rear-end collision accident in the "DD Tunnel", the system analyzed the real-time image and identified that both vehicles had obvious damage, but no injured people or fire were seen; based on these observations, the system used a machine learning model to predict a warning level score, such as level 3 (in a scoring system from 1 to 5, level 3 indicates moderate urgency).

[0060] The system needs to determine the accident level of the tunnel by combining the early warning level, the traffic congestion level of the tunnel, and the predefined accident level mapping relationship; the accident level mapping relationship may be a table, a function, or a decision tree, which establishes a connection between the early warning level, the congestion level, and the accident level; the system needs to be able to query or calculate these mapping relationships to determine the final accident level; the accident level can be used to guide the formulation of emergency response measures, traffic management strategies, or to provide real-time traffic information to the public; optionally, in the accident of the "DD Tunnel", the system knows that the early warning level is level 3, and the traffic congestion level of the tunnel is moderate (for example, the congestion coefficient is 0.6); according to the predefined accident level mapping relationship, the system determines an accident level, such as level 2 (in the accident level system from 1 to 4, level 2 means that urgent but not the highest-level response measures need to be taken); this accident level can now be used to guide the emergency response plan of the tunnel management department, including dispatching rescue vehicles, adjusting traffic lights, or issuing traffic warnings to the public.

[0061] In an embodiment of the present application, the system maintains an accident level mapping table, which determines the accident level according to the early warning level, the tunnel congestion level, and other possible factors (such as weather, special events); the accident level mapping table is shown in Table 2: Table 2 Accident Level Mapping Table

[0062] Suppose in the accident of the "EE Tunnel", the early warning level determined by the system is "medium to high" (which can be simplified to "high"), and the tunnel congestion level is "medium"; according to the accident level mapping table, the system determines that the accident level is level 3; therefore, in the accident of the "EE Tunnel", the traffic accident type determined by the system is "truck rear-end collision", the early warning level is "high", and the tunnel congestion level is "medium"; according to the accident level mapping relationship, the system finally determines that the accident level of the tunnel is level 3; this information will be used to guide the formulation of emergency response measures, traffic management strategies, or to provide real-time traffic information to the public.

[0063] Reference Figure 6 , in step S15, trigger the collaborative control of the drone and the fire-fighting equipment according to the accident level of the tunnel to control the fire control of the early warning event and the traffic diversion of the vehicle; In the specific implementation process of the present invention, the specific steps are as follows: S151: Determine the corresponding optimization event based on the accident level of the tunnel and the preset accident optimization matching table, and determine multiple sub-optimization items according to the traversal of the optimization event, and match the corresponding drone and fire-fighting equipment based on the multiple sub-optimization items; S152: Determine multiple synchronization events based on the comparison of the execution order of multiple sub-optimization items, and build a collaborative control relationship between the drone and the fire-fighting equipment in this synchronization event; determine the vehicle evacuation by the drone and the fire control of the early warning event by the fire-fighting equipment according to this collaborative control relationship and the synchronization event. At this time, as the control effect of the early warning event improves, the vehicle evacuation effect by the drone also improves.

[0064] In the embodiment of the present application, based on the accident level of the tunnel and a preset accident optimization matching table, determine the corresponding optimization event, and determine multiple sub-optimization items according to the traversal of the optimization event. Match the corresponding drone and fire-fighting equipment based on the multiple sub-optimization items, and introduce the drone and the fire-fighting equipment.

[0065] At this time, the system first receives the accident level from step S142. This accident level describes the severity of the accident in the current tunnel; the accident level is usually a numerical value or a classification label, which is used to represent the urgency and potential impact of the accident; the system maintains a preset accident optimization matching table, which associates different accident levels with optimization events; the optimization event refers to a series of operations or strategies that need to be executed to deal with the accident; the system looks up the corresponding optimization event in the matching table according to the accident level.

[0066] The system decomposes the found optimization event into multiple specific sub-optimization items; the sub-optimization item is a finer-grained component of the optimization event, and each sub-item represents a specific task or operation; at this time, the system traverses the optimization event and generates a list containing all sub-optimization items; this list will be used for subsequent device matching and scheduling.

[0067] Furthermore, the system analyzes the required device types and quantities for each sub-optimization item; this may include drones (for monitoring, communication, guidance, etc.), fire trucks (for fire extinguishing, rescue, etc.), ambulances (for treating the wounded, etc.), etc.; according to the results of the device requirement analysis, the system matches the corresponding drones and fire-fighting equipment from the available device resources; the matching process may involve checking the status of the devices (such as whether they are idle, available, remaining battery power, etc.), location information (such as the distance between the current location of the device and the accident site), and the priority of the devices (such as some devices may be more suitable for performing specific tasks); once the matching is completed, the system generates a device scheduling plan, including the dispatch order, arrival time, and task assignment of the devices.

[0068] Specifically, assume that a level 4 accident has occurred in the "GG Tunnel", and the system has determined the accident level to be level 4 according to step S142; the system queries the preset accident optimization matching table and finds that the optimization events corresponding to the level 4 accident are "urgently evacuating vehicles and personnel" and "rapid fire extinguishing and rescue".

[0069] The system decomposes "Emergency evacuation of vehicles and personnel" into the following sub-optimization matters: using drones to monitor traffic conditions in the tunnel and guide vehicle evacuation; using drones or a broadcast system to notify people in the tunnel to evacuate; setting up temporary traffic control to prevent new vehicles from entering the tunnel.

[0070] The system decomposes "Quick fire extinguishing and rescue" into the following sub-optimization matters: dispatching fire trucks to the accident scene; using water guns or foam fire extinguishing devices on the fire trucks to extinguish the fire; dispatching a rescue team into the tunnel to search for trapped people.

[0071] The system analyzes the types and quantities of equipment required for each sub-optimization matter and matches the corresponding drones and fire-fighting equipment; for "using drones to monitor traffic conditions in the tunnel and guide vehicle evacuation", the system matches 2 drones equipped with high-definition cameras and traffic indication functions; for "using drones or a broadcast system to notify people in the tunnel to evacuate", the system selects 1 drone with a broadcast function; for "setting up temporary traffic control", the system dispatches 1 police car to set up roadblocks at the tunnel entrance.

[0072] For "dispatching fire trucks to the accident scene", the system matches 2 fire trucks, each equipped with a water gun and a foam fire extinguishing device; for "using water guns or foam fire extinguishing devices on the fire trucks to extinguish the fire", the system designates 1 fire truck to be responsible for extinguishing the fire and the other as a backup; for "dispatching a rescue team into the tunnel to search for trapped people", the system dispatches 1 rescue team composed of firefighters and medical staff; through the above steps, the system has successfully matched and dispatched the corresponding drones and fire-fighting equipment for a Level 4 accident in the "GG Tunnel", providing strong support for subsequent emergency response and accident handling.

[0073] Furthermore, multiple synchronous events are determined based on the comparison of the execution sequences of multiple sub-optimization matters, and a cooperative control relationship between the drones and the fire-fighting equipment is constructed in the synchronous events; based on the cooperative control relationship and the synchronous events, the guidance of the drones to the vehicles and the fire control of the fire-fighting equipment for early warning events are determined. At this time, as the control effect of the early warning events improves, the guidance effect of the drones to the vehicles also improves.

[0074] At this point, the system first analyzes the execution order and interdependencies of the multiple sub-optimization items determined in step S151; the execution order may be determined based on factors such as the urgency of the accident site, task priority, and resource availability. At the same time, by comparing the execution order and interdependencies of the sub-optimization items, the system identifies task groups that can be executed in parallel, namely synchronous events; synchronous events refer to a collection of sub-optimization items that can be started at the same time and will not interfere with each other. Optionally, a level 3 accident occurred in the "HH Tunnel", and the system has determined multiple sub-optimization items according to step S151, and matched the corresponding drones and fire-fighting equipment; the system analyzes the execution order and interdependencies of the sub-optimization items, and identifies the following synchronous events: Synchronous event A: drones guide vehicle evacuation and fire trucks to the accident site; synchronous event B: drones monitor the accident site and rescue teams prepare to enter the tunnel.

[0075] For each synchronization event, the system analyzes the collaborative control requirements between the UAV and the fire-fighting equipment; this may include requirements for information sharing, task coordination, resource allocation, etc.; based on the results of the collaborative control requirements analysis, the system builds a collaborative control relationship between the UAV and the fire-fighting equipment; the system ensures that the UAV and the fire-fighting equipment can share information and coordinate actions in real time to achieve efficient emergency response.

[0076] Optionally, for synchronous event A, the system builds a collaborative control relationship between the drone and the fire truck; the drone is responsible for guiding the orderly evacuation of vehicles in the tunnel, while providing real-time traffic information to the fire truck to ensure that the fire truck can reach the accident scene safely and quickly.

[0077] For synchronous event B, the system established a collaborative control relationship between the drone and the rescue team; the drone was responsible for monitoring the accident scene and transmitting image data to the rescue team in real time, helping the rescue team understand the situation on the scene and prepare to enter the tunnel.

[0078] The system assigns specific tasks to drones and firefighting equipment based on collaborative control relationships and synchronization events. Drones may be assigned to guide vehicle evacuation, monitor accident sites, and other tasks; firefighting equipment may be assigned to firefighting, rescue, and other tasks. During task execution, the system dynamically adjusts the operating strategies of drones and firefighting equipment based on the management and control effects of early warning events (such as whether the flames are effectively controlled, the progress of vehicle evacuation, etc.). If the management and control effects are good, the system may reduce resource input or adjust task priorities. If the effects are not good, resource input will be increased or strategies will be changed. At the same time, as the management and control effects of early warning events improve, the system will correspondingly improve the vehicle diversion effects of drones. This may involve measures such as increasing the number of drones, increasing the diversion frequency, and optimizing diversion paths.

[0079] Optionally, the system assigns specific tasks to drones and firefighting equipment: Two drones are assigned the task of guiding vehicle evacuation, and they are respectively responsible for traffic control near the tunnel entrance and the accident site; One fire truck is assigned the task of going to the accident site to extinguish the fire; One rescue team is assigned the task of preparing to enter the tunnel for search and rescue; During the task execution, the system dynamically adjusts the strategy according to the control effect of the early warning event; as the fire is effectively controlled, the system increases the number of drones, improving the frequency and efficiency of vehicle evacuation; At the same time, the system monitors the accident site through drones, and evaluates the evacuation progress and fire control situation in real time to ensure that all tasks can be completed efficiently and orderly; through the above steps, the system successfully constructs a collaborative control relationship between drones and fire-fighting equipment in a Level 3 accident in the "HH Tunnel", and realizes efficient vehicle guidance and fire control; this provides strong support for subsequent emergency response and accident handling, ensuring the safety of personnel and the rapid restoration of tunnel traffic.

[0080] In an embodiment of the present application, a Level 4 accident occurred in the "II Tunnel", and the system has determined multiple sub-optimization matters according to step S151 and matched the corresponding drones and fire-fighting equipment; the system identifies the following synchronization events: Synchronization Event 1: Drone guiding vehicle evacuation and fire truck going to the accident site; Synchronization Event 2: Drone monitoring the accident site and rescue team preparing to enter the tunnel; Synchronization Event 3: Fire truck starting to extinguish the fire (drones continue to monitor).

[0081] The system constructs the following collaborative control relationships: Synchronization Event 1: The drone provides traffic information to guide vehicle evacuation; the fire truck follows the drone's instructions to go to the accident site; Synchronization Event 2: The drone transmits images of the accident site for the rescue team to refer to; the rescue team prepares to enter the tunnel for search and rescue; Synchronization Event 3: The drone continuously monitors the fire situation and provides real-time updates; the fire truck adjusts the fire extinguishing strategy according to the fire situation.

[0082] The system calculates and evaluates the control effect of the early warning event according to the weight and score: Initial fire size: 6 (score 4) Smoke concentration: 7 (score 3) Personnel evacuation progress: 25% (score 25) Control effect score: 4 + 3 + 25 * 0.3 = 12.5 (full score 20) As the fire extinguishing progresses and personnel are evacuated, the control effect score is increased to 16 (the flame is reduced to 3, the smoke concentration is reduced to 5, and the evacuation progress is increased to 40%); accordingly, the system increases the number of drones to 3, improves the guidance frequency, and speeds up the vehicle evacuation speed; through the above steps, the system successfully constructs a collaborative control relationship between drones and fire-fighting equipment in a Level 4 accident in the "II Tunnel" and dynamically adjusts the guidance strategy of drones for vehicles according to the control effect of the early warning event.

[0083] Please refer to Figure 7 , Figure 7 which is a schematic structural diagram of the tunnel-based drone control system in an embodiment of the present invention; the tunnel-based drone control system includes: An early warning event module 21, configured to determine an early warning event in the tunnel according to the real-time image of the tunnel and the early warning information of the tunnel; An inspection route module 22, configured to determine the inspection route of the drone relative to the tunnel according to the location of the early warning event and the distribution map of the tunnel, and the inspection route is from the current position of the drone to the location of the early warning event; A traffic status map module 23, configured to fly the drone along the inspection route in the tunnel and determine the traffic status map of the tunnel according to a plurality of road surface images collected during the flight of the drone and the current image of the early warning event; An accident level module 24, configured to determine the accident level of the tunnel based on the congestion degree of the traffic status map of the tunnel and the warning degree of the early warning event; A collaborative control module 25, configured to trigger the collaborative control of the drone and the fire-fighting equipment according to the accident level of the tunnel to control the fire control of the early warning event and the guidance of the vehicle.

[0084] For any combination of the technical features of the above embodiments, for the sake of brevity of description, not all combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope recorded in this specification.

Claims

1. A tunnel-based drone control method, characterized in that: include: Determine the warning events in the tunnel based on the real-time images of the tunnel and the warning information of the tunnel; Determine the inspection route of the drone relative to the tunnel according to the location of the warning event and the distribution map of the tunnel, where the inspection route is from the current location of the drone to the location of the warning event; The UAV flies in the tunnel along the inspection route and determines the traffic status map of the tunnel based on multiple road surface images collected by the UAV during the flight and the current image of the warning event; Determine the accident level of the tunnel based on the congestion level of the tunnel traffic state diagram and the warning level of the warning event; The coordinated control of drones and fire-fighting equipment is triggered according to the accident level in the tunnel to control the fire control of warning events and the evacuation of vehicles.

2. The tunnel-based drone control method according to claim 1, characterized in that: Determining the warning event in the tunnel according to the real-time image of the tunnel and the warning information of the tunnel includes: Collect multiple images at different positions according to the shooting of multiple cameras in the tunnel, determine the real-time image of the tunnel according to the synthesis of the multiple images, and determine multiple warning areas by identifying the real-time image of the tunnel; Based on the review of multiple warning areas, the corresponding warning target and the corresponding warning position are determined, the warning information of the tunnel is collected, and the warning event in the tunnel is determined based on the matching of the tunnel warning information, multiple warning areas and the corresponding warning positions.

3. The tunnel-based drone control method according to claim 1, characterized in that: The method of determining the inspection route of the drone relative to the tunnel according to the location of the warning event and the distribution map of the tunnel, wherein the inspection route is from the current location of the drone to the location of the warning event, includes: The distribution map of the tunnel is determined according to the name of the tunnel and the tunnel database, and the location of the warning event and the current location of the drone are marked based on the distribution map of the tunnel. At this time, the drone stays in the apron of the tunnel; The flight route is determined according to the current position of the UAV, the location of the warning event and the direction of passage of the tunnel, and the corresponding flight space is determined along the traversal of the first flight video. The inspection route of the UAV relative to the tunnel is determined according to the flight route and the corresponding flight space. At this time, the UAV performs attitude control according to the corresponding flight space to smoothly pass through the flight space.

4. The tunnel-based drone control method according to claim 1, characterized in that: The drone flies in the tunnel along the inspection route, and determines the traffic status map of the tunnel according to a plurality of road surface images collected by the drone during the flight and the current image of the warning event, including: The flight of the drone is triggered based on the inspection route. During the flight, the drone is compatible with attitude control and dynamic shooting. In the dynamic shooting of the drone, the drone takes synchronous shots in each lane and collects multiple road surface images. The traffic distribution map of the tunnel is determined based on the stitching of multiple road surface images.

5. The tunnel-based drone control method according to claim 4, characterized in that: The UAV flies in the tunnel along the inspection route, and determines the traffic status map of the tunnel according to multiple road surface images collected by the UAV during the flight and the current image of the warning event, and also includes: The current image of the warning event is collected, and the corresponding traffic accident type is determined based on the recognition of the current image of the warning event. The corresponding traffic impact range is matched based on the traffic accident type, and the traffic status map of the tunnel is determined according to the traffic distribution map of the tunnel and the corresponding traffic impact range.

6. The tunnel-based drone control method according to claim 1, characterized in that: The step of determining the accident level of the tunnel based on the congestion level of the tunnel traffic state diagram and the warning level of the warning event includes: Based on the division of the traffic state diagram of the tunnel, traffic images of each lane are generated, and the corresponding traffic congestion coefficient is determined according to the recognition of the traffic images of each lane. The congestion degree of the traffic state diagram of the tunnel is determined according to the multiple lanes and the corresponding traffic congestion coefficients.

7. The tunnel-based drone control method according to claim 6, characterized in that: The step of determining the accident level of the tunnel based on the congestion level of the tunnel traffic state diagram and the warning level of the warning event further includes: The traffic accident type corresponding to the warning event is collected, the warning degree of the warning event is determined according to the traffic accident type and the real-time image of the warning event, and the accident level of the tunnel is determined according to the warning level of the warning event, the congestion level of the traffic status diagram of the tunnel, and the accident level mapping relationship.

8. The tunnel-based drone control method according to claim 1, characterized in that: The coordinated control of the drone and the fire-fighting equipment is triggered according to the accident level of the tunnel to control the fire control of the warning event and the evacuation of vehicles, including: The corresponding optimization event is determined based on the accident level of the tunnel and the preset accident optimization matching table, and multiple sub-optimization items are determined according to the traversal of the optimization event. The corresponding drones and fire-fighting equipment are matched based on the multiple sub-optimization items.

9. The tunnel-based drone control method according to claim 8, characterized in that: The coordinated control of the drone and the fire-fighting equipment is triggered according to the accident level of the tunnel to control the fire control of the warning event and the evacuation of vehicles, and also includes: Based on the comparison of the execution order of multiple sub-optimization items, multiple synchronization events are determined, and a collaborative control relationship between the drone and the fire-fighting equipment is established in the synchronization event; based on the collaborative control relationship and the synchronization event, the drone's guidance of vehicles and the fire-fighting management and control of the early warning event by the fire-fighting equipment are determined. At this time, as the management and control effect of the early warning event is improved, the drone's guidance of vehicles is improved.

10. A tunnel-based drone control system, characterized in that: The tunnel-based UAV control system is applied to the tunnel-based UAV control method as described in any one of claims 1 to 9, and the tunnel-based UAV control system includes: A warning event module, used to determine the warning events in the tunnel according to the real-time image of the tunnel and the warning information of the tunnel; An inspection route module is used to determine the inspection route of the drone relative to the tunnel according to the location of the warning event and the distribution map of the tunnel, and the inspection route is from the current location of the drone to the location of the warning event; The traffic status diagram module is used for the drone to fly in the tunnel along the inspection route and determine the traffic status diagram of the tunnel based on multiple road surface images collected by the drone during the flight and the current image of the warning event; An accident level module, used to determine the accident level of the tunnel based on the congestion level of the tunnel traffic status diagram and the warning level of the warning event; The collaborative control module is used to trigger the collaborative control of drones and fire-fighting equipment according to the accident level in the tunnel, so as to control the fire control of warning events and the evacuation of vehicles.

Images