Unmanned aerial vehicle pursuit-based escorting method, device and system

Abstract

The invention provides an escorting method, device and system based on unmanned aerial vehicle pursuit, and relates to the technical field of unmanned aerial vehicle application, and the method comprises the steps: obtaining real-time positioning information from an escorted target; when it is determined that the real-time positioning information deviates from the escorting route corresponding to the escorted target, a current positioning scene is determined according to the real-time positioning information, and a pursuit instruction is sent to the unmanned aerial vehicle; inputting a target image to a target recognition model corresponding to the current positioning scene to obtain all to-be-selected targets; a following instruction is sent to the unmanned aerial vehicle, the following instruction is used for indicating the unmanned aerial vehicle to follow the escorted target, and the user input instruction is used for selecting the escorted target from all the to-be-selected targets. According to the invention, target persons in different scenes can be accurately identified, misjudgment and missed judgment risks are reduced, rapid deployment and flexible flight route adjustment are realized by using the unmanned aerial vehicle, a target position is rapidly reached, real-time monitoring and tracking are carried out, and the monitoring and tracking efficiency is significantly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of drone application, and in particular to a method, device and system for escorting fugitives based on drone pursuit. Background Art

[0002] In the current practical application of the intelligent escort system, despite the continuous advancement of technology, its monitoring and tracking capabilities still face many challenges and limitations. First, as the core of traditional monitoring methods, the layout of ground monitoring equipment is often limited by complex and changeable terrain conditions. Whether it is rugged mountains, dense forests, or narrow alleys between high-rise buildings in cities, they may form insurmountable monitoring blind spots, resulting in omissions in the monitoring of key areas and increasing security risks. In addition, obstacles such as buildings and trees may also cause signal obstruction or reflection, further reducing the clarity and accuracy of the monitoring image and affecting the continuous tracking of the target; secondly, in the face of emergencies such as pursuit, ground monitoring equipment is often difficult to respond quickly and adjust the monitoring range due to its fixed or inconvenient characteristics. This not only delays the speed of obtaining information, but also limits the timeliness of decision-making, which brings considerable pressure to emergency response work.

[0003] With its unique aerial perspective and high flexibility, drones can be quickly deployed to the scene of the incident, regardless of terrain and building obstacles, directly fly over the monitoring blind spots, and provide all-round, multi-angle real-time images. This immediacy not only ensures the accuracy of information, but also greatly shortens the decision-making cycle, enabling it to more comprehensively grasp the on-site situation and formulate more accurate and effective response strategies. At the same time, the powerful maneuverability and coverage of drones enable them to flexibly adjust the flight altitude and route according to mission requirements. This high degree of flexibility and adaptability is difficult for ground monitoring equipment to achieve.

[0004] At present, there is no technical solution for using drone linkage to track and pursue the target when the target escapes during the escort process, nor is there a technical solution for using a command and control center to realize the linkage between the drone platform and the escort vehicle communication, and there is no escort method, device and system based on drone pursuit. Summary of the invention

[0005] The present invention provides an escort method, device and system based on drone pursuit. The application of drone technology in the intelligent escort process can not only effectively make up for the shortcomings of ground monitoring equipment, but also improve the efficiency and accuracy of monitoring and tracking.

[0006] In a first aspect, the present invention provides an escort method based on drone pursuit, comprising:

[0007] Obtaining real-time positioning information from the escorted target, wherein the real-time positioning information is determined after the escorted target's foot ring reports the real-time position to the command and control center;

[0008] In the case where it is determined that the real-time positioning information deviates from the escort route corresponding to the escorted target, the current positioning scene is determined according to the real-time positioning information, and a pursuit instruction is sent to the drone, wherein the pursuit instruction is used to instruct the drone to obtain a target image corresponding to the target positioning after arriving at the target positioning according to the target planning path, and transmit the target image back to the command and control center, wherein the target planning path is determined after path planning from the starting point of the drone to the real-time position;

[0009] Inputting the target image to a target recognition model corresponding to the current positioning scene to obtain all targets to be selected output by the target recognition model, wherein the current positioning scene is a desert scene, a forest scene, a road scene, a grassland scene or a river scene;

[0010] A user input instruction is received, and in response to the user input instruction, a follow instruction is sent to the drone, wherein the follow instruction is used to instruct the drone to follow the escorted target, and the user input instruction is used to select the escorted target from all targets to be selected.

[0011] According to the escort method based on drone pursuit provided by the present invention, before determining the current positioning scene according to the real-time positioning information, the method also includes:

[0012] When the straight-line distance between the real-time positioning information and the escort route is greater than a first preset distance, determining that the real-time positioning information deviates from the escort route corresponding to the escorted target;

[0013] Alternatively, when the distance between the foot ring and the escort vehicle escorting the escorted target is greater than a second preset distance, it is determined that the real-time positioning information deviates from the escort route corresponding to the escorted target.

[0014] According to the escort method based on drone pursuit provided by the present invention, the drone is located on a drone platform of an escort vehicle escorting the escorted target. Before sending a pursuit instruction to the drone, the method further includes:

[0015] Determine the starting point of the drone according to the real-time position of the escort vehicle;

[0016] Utilize the starting point of the drone and the real-time position reported by the foot ring of the escorted target to perform path planning and determine the target planning path;

[0017] The real-time position of the escort vehicle is determined by using the on-board positioning of the escort vehicle when it is determined that the real-time positioning information deviates from the escort route corresponding to the escorted target.

[0018] According to the escort method based on drone pursuit provided by the present invention, the drone is located on a drone platform on the escort route. Before sending a pursuit instruction to the drone, the method further includes:

[0019] Determine the platform position corresponding to each of the drone platforms on the escort route, calculate the distance difference between the real-time position of the escort vehicle and each platform position, and determine the drone platform corresponding to the platform position with the smallest distance difference as the target platform;

[0020] The starting point of the UAV is determined according to the platform position corresponding to the target platform, and the path planning is performed using the starting point of the UAV and the real-time position reported by the foot ring of the escorted target to determine the target planning path.

[0021] According to the escort method based on drone pursuit provided by the present invention, before inputting the target image into the target recognition model corresponding to the current positioning scene, the method also includes:

[0022] Get the corresponding sample images under different scenes;

[0023] For each scene, the initial model is trained according to the sample images and the character head portrait annotations corresponding to each sample image to obtain a recognition model corresponding to the scene;

[0024] Traverse all scenes and obtain each recognition model corresponding to all scenes.

[0025] According to the escort method based on drone pursuit provided by the present invention, the current positioning scene is determined according to the real-time positioning information, and a pursuit instruction is sent to the drone, including:

[0026] When it is determined that the current positioning scene is a desert scene or a road scene, determining to use a first number of drones for pursuit;

[0027] When it is determined that the current positioning scene is a grassland scene or a river scene, determining to use a second number of drones for pursuit;

[0028] When it is determined that the current positioning scene is a forest scene, it is determined to use a third number of drones for pursuit;

[0029] The first number is smaller than the second number, and the second number is smaller than the third number.

[0030] According to the escort method based on drone pursuit provided by the present invention, when there is more than one escorted target, sending a pursuit instruction to the drone includes:

[0031] Send pursuit instructions to a preset number of drones;

[0032] The number of the escorted targets is less than or equal to the preset number.

[0033] According to the escort method based on drone pursuit provided by the present invention, after sending pursuit instructions to a preset number of drones, the method further includes:

[0034] For a target image corresponding to any drone, all targets to be selected in the target image are obtained, and each target image corresponding to all drones is traversed to obtain a set of all targets to be selected;

[0035] After removing duplicate avatars from all sets of targets to be selected, the avatars of the targets to be selected are displayed on a preset display screen.

[0036] In the second aspect, a drone-based escort device is provided, comprising:

[0037] An acquisition unit, the acquisition unit is used to acquire real-time positioning information from the escorted target, the real-time positioning information is determined after the escorted target's foot ring reports the real-time position to the command and control center;

[0038] A sending unit, wherein the sending unit is used to determine the current positioning scene according to the real-time positioning information when it is determined that the real-time positioning information deviates from the escort route corresponding to the escorted target, and send a pursuit instruction to the drone, wherein the pursuit instruction is used to instruct the drone to obtain a target image corresponding to the target positioning after arriving at the target positioning according to the target planning path, and transmit the target image back to the command and control center, wherein the target planning path is determined after path planning from the starting point of the drone to the real-time position;

[0039] An input unit, the input unit is used to input the target image to a target recognition model corresponding to the current positioning scene, and obtain all targets to be selected output by the target recognition model, wherein the current positioning scene is a desert scene, a forest scene, a road scene, a grassland scene or a river scene;

[0040] A receiving unit, the receiving unit is used to receive a user input instruction, and in response to the user input instruction, sends a follow instruction to the drone, the follow instruction is used to instruct the drone to follow the escorted target, and the user input instruction is used to select the escorted target from all targets to be selected.

[0041] In a third aspect, an escort system based on drone pursuit is provided, including an escort device based on drone pursuit, wherein the escort device based on drone pursuit is a command and control center, and also includes an escort vehicle and a drone platform;

[0042] The command and control center is communicatively connected with the UAV platform and the escort vehicle respectively.

[0043] The present invention aims to use drones to achieve rapid deployment and flexible adjustment of flight routes in response to the escape of an escorted target, quickly reach the target location and conduct real-time monitoring and tracking, and significantly improve the efficiency of monitoring and tracking; drones can cover a wider area and are not restricted by obstacles such as terrain and buildings, effectively making up for the shortcomings of ground monitoring equipment. When an escorted target escapes, the drone can respond quickly and provide real-time intelligence and support, providing strong support for commanders to formulate response strategies. The present invention also combines different recognition scenarios, and by training target recognition models suitable for each scenario, it can accurately identify and distinguish target persons in different scenarios, reducing the risk of misjudgment and missed judgment; the present invention can also support flexible adjustment of the number of drones and flight routes according to real-time positioning information and characteristics of different scenarios, and has strong flexibility and scalability. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0045] Figure 1 It is a flow chart of the escort method based on drone pursuit provided by the present invention;

[0046] Figure 2 It is a structural schematic diagram of an escort device based on drone pursuit provided by the present invention;

[0047] Figure 3 It is a structural schematic diagram of the escort system based on drone pursuit provided by the present invention;

[0048] Figure 4 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0049] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. 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.

[0050] Figure 1 : is a flow chart of the escort method based on drone pursuit provided by the present invention, and the escort method based on drone pursuit includes:

[0051] Step 101: obtaining real-time positioning information from the escorted target, wherein the real-time positioning information is determined after the escorted target's foot ring reports the real-time position to the command and control center;

[0052] Step 102: when it is determined that the real-time positioning information deviates from the escort route corresponding to the escorted target, the current positioning scene is determined according to the real-time positioning information, and a pursuit instruction is sent to the drone. The pursuit instruction is used to instruct the drone to obtain a target image corresponding to the target positioning after arriving at the target positioning according to the target planning path, and transmit the target image back to the command and control center. The target planning path is determined after path planning from the starting point of the drone to the real-time position;

[0053] Step 103: input the target image to the target recognition model corresponding to the current positioning scene, and obtain all the targets to be selected output by the target recognition model, wherein the current positioning scene is a desert scene, a forest scene, a road scene, a grassland scene or a river scene;

[0054] Step 104: receiving a user input instruction, and in response to the user input instruction, sending a follow instruction to the drone, wherein the follow instruction is used to instruct the drone to follow the escorted target, and the user input instruction is used to select the escorted target from all targets to be selected.

[0055] In step 101, the system obtains the location information of the escorted target in real time through the anklet device worn by the escorted target. The location information is sent to the command and control center periodically or in real time. The anklet usually integrates GPS or other positioning technologies, such as Beidou, GLONASS, etc., to ensure that accurate positioning data can be provided in any environment. Once the location changes, the system can immediately detect and respond.

[0056] In step 102, when the command and control center detects that the real-time position of the escorted target deviates from the predetermined escort route, the system will plan a path for the drone from the current position of the drone to the target position based on the current real-time position information, and then send a pursuit command to the drone, instructing it to fly to the target position according to the planned path and take a target image and transmit it back to the control center.

[0057] Optionally, before determining the current positioning scene according to the real-time positioning information, the method further includes:

[0058] When the straight-line distance between the real-time positioning information and the escort route is greater than a first preset distance, determining that the real-time positioning information deviates from the escort route corresponding to the escorted target;

[0059] Alternatively, when the distance between the foot ring and the escort vehicle escorting the escorted target is greater than a second preset distance, it is determined that the real-time positioning information deviates from the escort route corresponding to the escorted target.

[0060] Optionally, in the escort method based on drone pursuit, in order to more accurately determine whether the escorted target has deviated from the escort route, the method introduces two preset distances as a basis for judgment, a first preset distance and a second preset distance. For the first solution, the first preset distance is used to determine whether the straight-line distance between the real-time position of the escorted target and the escort route exceeds the allowed range. When the straight-line distance between the real-time positioning information reported by the foot ring and the preset escort route is greater than the first preset distance, the system believes that the escorted target may have deviated from the escort route, thereby triggering further scene judgment and pursuit instruction sending process; for the second solution, the second preset distance also takes into account the actual distance between the foot ring and the escort vehicle. In some cases, even if the straight-line distance between the real-time position of the escorted target and the escort route does not exceed the range, if the distance between the foot ring and the escort vehicle exceeds the second preset distance, it may also indicate that the escorted target is trying to escape.

[0061] Optionally, the present invention can adopt the above two solutions at the same time, or can select one of the solutions to determine whether the real-time positioning information deviates from the escort route corresponding to the escorted target. By introducing two preset distances as the basis for judgment, the present invention can more accurately judge whether the escorted target has actually deviated from the escort route, avoiding misjudgment or missed judgment caused by a single judgment criterion. Once it is found that the escorted target may escape, the system can quickly trigger subsequent processes, including determining the current positioning scene and sending a pursuit command to the drone, so as to quickly respond to and deal with potential escape risks.

[0062] Optionally, the drone is located on a drone platform of an escort vehicle escorting the escorted target, and before sending a pursuit instruction to the drone, the method further includes:

[0063] Determine the starting point of the drone according to the real-time position of the escort vehicle;

[0064] Utilize the starting point of the drone and the real-time position reported by the foot ring of the escorted target to perform path planning and determine the target planning path;

[0065] The real-time position of the escort vehicle is determined by using the on-board positioning of the escort vehicle when it is determined that the real-time positioning information deviates from the escort route corresponding to the escorted target.

[0066] Optionally, the drone can be set at multiple locations. In an optional embodiment shown in the present invention, the drone is located on a drone platform of an escort vehicle that escorts the escorted target. Before sending a pursuit instruction to the drone, the system will determine the take-off point of the drone, that is, the starting point, according to the real-time position of the escort vehicle. This can be achieved by obtaining the vehicle's geographic location information in real time through the on-board positioning device on the escort vehicle. When it is detected that the real-time positioning information of the escorted target deviates from the predetermined escort route, the drone take-off preparation process will be immediately triggered, and the current vehicle position will be determined as the drone's take-off starting point. Furthermore, the system uses the drone starting point and the real-time position information reported by the escorted target's foot ring to perform path planning. The purpose of the path planning is to generate an optimal or feasible flight path for the drone from the starting point to the target position. In one embodiment, the present invention can also take into account various factors in the process of designing this planned path, such as terrain, obstacles, flight restricted areas, etc., to ensure that the drone can reach the target position safely and efficiently.

[0067] In the process of determining the target planning path, the present invention will dynamically adjust the path according to the actual situation. For example, if the escort vehicle is moving, the starting point of the drone will be updated in real time and the path will be flexibly planned. In addition, if the command center has a priority selection for the pursuit mission or needs to adjust the tracking target, the flight mission and path of the drone will be adjusted accordingly. The present invention deploys the drone on the escort vehicle and takes off immediately when needed, which significantly shortens the conversion time of the drone from the standby state to the mission execution state, and improves the response speed to emergencies. The drone can dynamically plan the path according to the real-time position of the escort vehicle and the real-time position of the target. This flexibility enables the drone to quickly adapt to different environments and mission requirements. At the same time, the drone's high-altitude flight capability and all-round monitoring capability can reduce the risk of ground contact and conflict. At the same time, the security equipment on board can also provide additional security when necessary.

[0068] Optionally, the drone is located on a drone platform on the escort route, and before sending a pursuit instruction to the drone, the method further includes:

[0069] Determine the platform position corresponding to each of the drone platforms on the escort route, calculate the distance difference between the real-time position of the escort vehicle and each platform position, and determine the drone platform corresponding to the platform position with the smallest distance difference as the target platform;

[0070] The starting point of the UAV is determined according to the platform position corresponding to the target platform, and the path planning is performed using the starting point of the UAV and the real-time position reported by the foot ring of the escorted target to determine the target planning path.

[0071] Optionally, in another optional embodiment shown in the present invention, drones are deployed on multiple drone platforms on the escort route. In this case, before sending a pursuit instruction to the drone, the precise positions of all preset drone platforms on the escort route are determined. These platform positions are usually marked and recorded in the escort route planning stage. During the execution of the escort mission, once it is determined that the escorted target has escaped and it is confirmed that the drone needs to be called for pursuit or monitoring, the system will calculate the distance difference between the escort vehicle and each drone platform on the escort route based on the real-time position of the escort vehicle. By comparing these distance differences, the drone platform with the smallest distance difference will be selected as the target platform, that is, the drone platform closest to the current position of the escort vehicle. Once the target platform is determined, the present invention will use the precise position of the target platform as the starting point of the drone, and perform real-time offline path planning based on the real-time position reported by the foot ring of the escorted target and the starting point of the drone to determine the optimal flight path required for the drone to track the target, that is, the target planning path.

[0072] The present invention can significantly shorten the preparation time before the drone takes off by setting up multiple drone platforms on the escort route and selecting the nearest platform as the drone starting point when necessary, thereby improving the response speed of the overall system. Compared with the method of setting up a drone platform on the escort vehicle, the advantage of this embodiment is that after completing the target tracking of the escorted target, it is easier to return to the drone platform, and it is easier to realize return path planning and charging. At the same time, the number of deployable drones will also be more than the method of deploying on the escort vehicle.

[0073] Optionally, determining the current positioning scene according to the real-time positioning information and sending a pursuit instruction to the drone includes:

[0074] When it is determined that the current positioning scene is a desert scene or a road scene, determining to use a first number of drones for pursuit;

[0075] When it is determined that the current positioning scene is a grassland scene or a river scene, determining to use a second number of drones for pursuit;

[0076] When it is determined that the current positioning scene is a forest scene, it is determined to use a third number of drones for pursuit;

[0077] The first number is smaller than the second number, and the second number is smaller than the third number.

[0078] Optionally, after the real-time positioning information of the escorted target is uploaded, the present invention first determines the scene type of the escorted target according to its current location information. Possible scenes include but are not limited to deserts, roads, grasslands, rivers and forests. The number of required drones is dynamically determined according to different scene types. Generally, different scenes have different requirements for the number of drones. For desert scenes or road scenes, since these scenes are relatively open, there are fewer obstacles and the drone field of view is clearer, the first number of drones can be selected for pursuit. This number can be relatively small to save resources and costs. For grassland scenes or river scenes, These scenarios may require more drones to work together to cover more ground areas and possible hiding spots. Therefore, it is determined that the second number of drones will be used for pursuit, and the number of drones used is more than that in desert and road scenarios. Forest scenes are complex and changeable, with dense trees and obstructed vision. Drones require more intensive deployment and coordination to ensure blind-spot monitoring. Therefore, the third number of drones is used for pursuit, and the third number is the largest among the three. Finally, according to the determined number of drones, pursuit instructions are sent to the corresponding number of drones, including target location, flight route, mission requirements, etc. The drones take off, track and monitor according to the instructions.

[0079] The present invention dynamically adjusts the number of drones according to the real-time positioning scene, making the drone pursuit system more flexible and intelligent, and able to better adapt to various complex environments. Specifically, the number of drones is reduced in open and simple scenes to save resources and costs; the number of drones is increased in complex and hidden scenes to ensure the monitoring effect. Such a dynamic adjustment strategy helps to achieve the optimal allocation of resources. Different numbers of drones are used for collaborative operations in different scenes, which can more effectively cover the target area, reduce the possibility of missed detection and misjudgment, and improve tracking efficiency.

[0080] Optionally, when there is more than one escorted target, sending a pursuit instruction to the drone includes:

[0081] Send pursuit instructions to a preset number of drones;

[0082] The number of the escorted targets is less than or equal to the preset number.

[0083] Optionally, in the case where there is more than one escorted target, in order to improve the efficiency and flexibility of pursuit, the present invention proposes a strategy of sending pursuit instructions to a preset number of drones. The preset number is a value determined in advance based on factors such as mission requirements, drone performance, and available resources. It aims to ensure that under any circumstances, there are a sufficient number of drones to perform tasks simultaneously without causing excessive waste of resources.

[0084] Specifically, when multiple escorted targets are identified, drones not exceeding a preset number will be automatically allocated to track these targets according to the preset number rules. If the number of escorted targets is less than or equal to the preset number, each target will be assigned at least one drone for tracking; if the number of targets exceeds the preset number, it may be necessary to decide which targets are tracked first based on priority or other evaluation criteria, or take other measures such as requesting reinforcements, temporarily giving up tracking certain targets, etc. The present invention allocates the pursuit task by presetting the number of drones, which can maximize the use of limited drone resources and avoid waste of resources while ensuring the smooth completion of the task.

[0085] Optionally, after sending the pursuit instruction to a preset number of drones, the method further includes:

[0086] For a target image corresponding to any drone, all targets to be selected in the target image are obtained, and each target image corresponding to all drones is traversed to obtain a set of all targets to be selected;

[0087] After removing duplicate avatars from all sets of targets to be selected, the avatars of the targets to be selected are displayed on a preset display screen.

[0088] Optionally, for each drone, when they perform pursuit missions, they will capture and transmit back target images through the high-definition cameras they are equipped with. The system will collect these target images from different drones, and identify the person portraits in each image. By traversing the target images returned by all drones, the system will generate a set of targets to be selected that includes the person portraits in all images. There may be duplicate target portraits in the set because they may be taken by different drones at different angles or time periods. The present invention can use image processing techniques such as feature extraction or comparison algorithms to detect and remove duplicate portraits in the set, ensuring that each target is listed only once in the final displayed list of targets to be selected. After removing duplicate portraits, the present invention will organize all unique target portrait sets into an ordered list or graphical interface, and display them on a preset display screen, so that the command center or relevant personnel can quickly view and identify the escorted targets.

[0089] In step 103, the target image transmitted back by the drone is input into a target recognition model corresponding to the current positioning scene. The current positioning scene is a desert scene, a forest scene, a road scene, a grassland scene or a river scene. The target recognition model can be a desert scene recognition model, a forest scene recognition model, a road scene recognition model, a grassland scene recognition model or a river scene recognition model. These models are trained to recognize the portraits of people in the scene and output all possible targets to be selected.

[0090] Optionally, before inputting the target image into the target recognition model corresponding to the current positioning scene, the method further includes:

[0091] Get the corresponding sample images under different scenes;

[0092] For each scene, the initial model is trained according to the sample images and the character head portrait annotations corresponding to each sample image to obtain a recognition model corresponding to the scene;

[0093] Traverse all scenes and obtain each recognition model corresponding to all scenes.

[0094] Optionally, before inputting the target image into the target recognition model corresponding to the current positioning scene, collect sample images in different scenes, such as deserts, forests, roads, grasslands, rivers, etc. These sample images need to cover a wide range of natural environments and possible human concealment situations, collect human images and annotation information in various scenes, and then make a human sample set, compress it into a ZIP package, and the annotated data set covers various possible scenes and angles as much as possible to improve the generalization ability of the model, so as to facilitate the identification of small human targets in the video in various scenes. Each sample image needs to have a corresponding human portrait annotation to ensure accuracy during the training process. Upload the generated sample set ZIP package to the target recognition model training platform, where all uploaded sample set information can be seen. Create a new training task on the target recognition model training platform, and select the training name and model training type as YOLOV8. This usually involves using the training script provided by YOLOV8 for training, and adjusting the training parameters such as learning rate, batch size, etc. as needed to improve the model's recognition ability in this specific scenario. After the training is completed, the character model can be downloaded, and the model can be evaluated using the validation set or test set to understand the accuracy and performance of the model. This can be done by calculating the model's accuracy, recall rate, F1 score and other indicators. For each scenario, the initial target recognition model is trained using the corresponding sample images and character head portrait annotation information. Each training will generate a specific recognition model for the scenario. Repeat the above steps and train each preset scenario until the recognition model corresponding to all scenarios is obtained.

[0095] In step 104, at the command and control center, the operator displays all targets to be selected on a preset display screen according to the targets to be selected output by the target recognition model. The targets to be selected may be human portraits. The escorted targets are selected through the user interface. Once the selection is completed, the system sends a follow command to the drone, instructing it to lock on to and follow the escorted target.

[0096] In combination with step 101 to step 104, in an optional embodiment, after the escorted target escapes, real-time offline path planning is performed according to the real-time positioning information reported by the foot ring carried by the escorted target, and a preliminary tracking plan for the ground target is formulated, and a tracking instruction matching the preliminary tracking plan for the ground target is issued to the drone. The escaped escorted target is prioritized according to the command center, such as selecting to track escorted target A or escorted target B, and the ground target person tracking plan can be dynamically adjusted. According to the adjusted ground target tracking plan, a tracking instruction matching the tracking plan is issued to the drone, such as selecting target A or target B, and the drone flies to the tracking location. According to the change in the position of the target person, the drone plans the route in real time to ensure that the drone is where the escorted target is.

[0097] Optionally, the present invention uses a drone to perform coverage aerial photography of the suspected area of ​​the target object, obtains image data and flight record data, and uniformly scales the collected drone aerial photography image data into a 304*304*32 feature map. At the same time, the scaled feature map and the original image are marked and associated. The original image collected by the drone is large in size and high in precision, so the original image is uniformly scaled to a standard 304*304*32 feature map and then sent to the detection network. This can greatly improve the matching calculation speed of the target object without affecting the accuracy.

[0098] The present invention adopts the YOLOV8 algorithm to identify the pictures related to the target in the model training in the aerial picture data; according to the resolution matching degree of the pictures related to the target object, the YOLOV8 algorithm is used to deeply calculate and identify the original picture of the corresponding target object, and the YOLOV8 algorithm is used to perform Mosaic data enhancement on the target object in the picture related to the target object, and perform anchor frame processing. The input end of YOLOV8 adopts the method of Mosaic data enhancement to perform Mosaic data enhancement on the target object in the picture related to the target object, and performs anchor frame processing, which can greatly enhance the calculation speed of the target area, eliminate the interference items around the target object, and improve the recognition accuracy. Through Mosaic data enhancement, the model can be exposed to more target variants during the training process, thereby accelerating the recognition speed of the target object. At the same time, Mosaic data enhancement reduces the interference of the surrounding environment of the target object, so that the model is more focused on the target object itself, thereby improving the recognition accuracy.

[0099] Optionally, screening can be performed based on the results of drone target identification. Targets that are difficult to identify with the naked eye can be identified, which is convenient for personnel on site or in the command center to manually determine whether the target is being escorted. If the target is being escorted, the drone camera angle will be adjusted in real time to track the target person in real time. After the target tracking is completed, the remote end of the mission is indicated, and the drone automatically returns and charges. Drones can monitor and track in the air, reducing the risk of ground contact and conflict. At the same time, drones can carry necessary security equipment, such as anti-riot spray, electric shock devices, etc., to enhance the safety of the escort process. In addition, drones can also be used for escort tasks in special environments, such as mountainous areas, waters and other difficult-to-reach areas.

[0100] The present invention pre-produces a sample set of target identification for people, vehicles and objects, and performs corresponding model training. The location information of the foot ring device carried by the escorted target is reported to the server background in real time. When the escorted target escapes, the drone performs offline path planning in real time based on the uploaded real-time location of the escorted target. The drone takes off to the location of the escorted target, and enables target recognition to identify small targets on the ground, assists in reconnaissance, and manually judges whether they are escorted targets based on the identified people. The present invention utilizes high-definition cameras and other sensor devices carried by drones, combined with advanced technologies such as deep learning, computer vision, and AI recognition, to achieve rapid and accurate recognition of escorted targets. By transmitting and processing image data in real time, the system can automatically identify and track targets, provide key target information to the command center, and assist it in making correct decisions.

[0101] The present invention aims to use drones to achieve rapid deployment and flexible adjustment of flight routes in response to the escape of an escorted target, quickly reach the target location and conduct real-time monitoring and tracking, and significantly improve the efficiency of monitoring and tracking; drones can cover a wider area and are not restricted by obstacles such as terrain and buildings, effectively making up for the shortcomings of ground monitoring equipment. When an escorted target escapes, the drone can respond quickly and provide real-time intelligence and support, providing strong support for commanders to formulate response strategies. The present invention also combines different recognition scenarios, and by training target recognition models suitable for each scenario, it can accurately identify and distinguish target persons in different scenarios, reducing the risk of misjudgment and missed judgment; the present invention can also support flexible adjustment of the number of drones and flight routes according to real-time positioning information and characteristics of different scenarios, and has strong flexibility and scalability.

[0102] The present invention can quickly respond to escort tasks and flexibly adjust flight routes to adapt to different terrains and environments. This high efficiency and flexibility make the drone escort platform have significant advantages in emergency situations or complex environments; the drone escort solution can reduce safety risks caused by human factors, such as fatigue driving, human negligence, etc. At the same time, the drone can be equipped with advanced sensors and monitoring equipment to achieve all-round and no-dead-angle monitoring to ensure the safety of the escort process; the drone escort platform can partially replace manpower, reduce personnel input, and reduce labor costs; the drone can transmit video and audio information of the escort process in real time, so that the command center can grasp the escort situation in real time and make quick decisions and dispatch; the drone's high-altitude flight capability enables it to cover a wider area, achieve comprehensive monitoring of the escort route, and effectively prevent the occurrence of incidents such as escape and hijacking; compared with the traditional vehicle escort method, the drone escort method has lower energy consumption and emissions, which is in line with the green and environmentally friendly development concept.

[0103] Figure 2 It is a structural schematic diagram of the escort device based on drone pursuit provided by the present invention, and the escort device based on drone pursuit includes an acquisition unit 1, and the acquisition unit is used to obtain real-time positioning information from the escorted target. The real-time positioning information is determined after the real-time position of the escorted target is reported to the command and control center by the foot ring of the escorted target. The working principle of the acquisition unit 1 can refer to the aforementioned step 101 and will not be repeated here.

[0104] The escort device based on drone pursuit also includes a sending unit 2, which is used to determine the current positioning scene according to the real-time positioning information when it is determined that the real-time positioning information deviates from the escort route corresponding to the escorted target, and send a pursuit instruction to the drone. The pursuit instruction is used to instruct the drone to arrive at the target positioning according to the target planning path, obtain the target image corresponding to the target positioning, and transmit the target image back to the command and control center. The target planning path is determined after path planning from the starting point of the drone to the real-time position. The working principle of the sending unit 2 can refer to the aforementioned step 102 and will not be repeated here.

[0105] The escort device based on drone pursuit also includes an input unit 3, which is used to input the target image to the target recognition model corresponding to the current positioning scene to obtain all targets to be selected output by the target recognition model. The current positioning scene is a desert scene, a forest scene, a road scene, a grassland scene or a river scene. The working principle of the input unit 3 can refer to the aforementioned step 103 and will not be repeated here.

[0106] The escort device based on drone pursuit also includes a receiving unit 4, which is used to receive user input instructions and send a follow instruction to the drone in response to the user input instruction, wherein the follow instruction is used to instruct the drone to follow the escorted target, and the user input instruction is used to select the escorted target from all targets to be selected. The working principle of the receiving unit 4 can refer to the aforementioned step 104 and will not be repeated here.

[0107] The present invention aims to use drones to achieve rapid deployment and flexible adjustment of flight routes in response to the escape of an escorted target, quickly reach the target location and conduct real-time monitoring and tracking, and significantly improve the efficiency of monitoring and tracking; drones can cover a wider area and are not restricted by obstacles such as terrain and buildings, effectively making up for the shortcomings of ground monitoring equipment. When an escorted target escapes, the drone can respond quickly and provide real-time intelligence and support, providing strong support for commanders to formulate response strategies. The present invention also combines different recognition scenarios, and by training target recognition models suitable for each scenario, it can accurately identify and distinguish target persons in different scenarios, reducing the risk of misjudgment and missed judgment; the present invention can also support flexible adjustment of the number of drones and flight routes according to real-time positioning information and characteristics of different scenarios, and has strong flexibility and scalability.

[0108] Figure 3 It is a structural schematic diagram of the escort system based on drone pursuit provided by the present invention, wherein the escort system based on drone pursuit includes the escort device based on drone pursuit, wherein the escort device based on drone pursuit is a command and control center, and also includes an escort vehicle and a drone platform; the command and control center is respectively communicatively connected with the drone platform and the escort vehicle.

[0109] The present invention aims to use drones to achieve rapid deployment and flexible adjustment of flight routes in response to the escape of an escorted target, quickly reach the target location and conduct real-time monitoring and tracking, and significantly improve the efficiency of monitoring and tracking; drones can cover a wider area and are not restricted by obstacles such as terrain and buildings, effectively making up for the shortcomings of ground monitoring equipment. When an escorted target escapes, the drone can respond quickly and provide real-time intelligence and support, providing strong support for commanders to formulate response strategies. The present invention also combines different recognition scenarios, and by training target recognition models suitable for each scenario, it can accurately identify and distinguish target persons in different scenarios, reducing the risk of misjudgment and missed judgment; the present invention can also support flexible adjustment of the number of drones and flight routes according to real-time positioning information and characteristics of different scenarios, and has strong flexibility and scalability.

[0110] Figure 4 Schematic diagram of the structure of the electronic device provided by the present invention. Figure 4 As shown, the electronic device may include: a processor 410, a communication interface 420, a memory 430 and a communication bus 440, wherein the processor 410, the communication interface 420 and the memory 430 communicate with each other through the communication bus 440. The processor 410 may call the logic instructions in the memory 430 to execute the escort method based on the pursuit of the drone, the method comprising: obtaining the real-time positioning information from the escorted target, the real-time positioning information is determined after the foot ring of the escorted target reports the real-time position to the command and control center; when it is determined that the real-time positioning information deviates from the escort route corresponding to the escorted target, the current positioning scene is determined according to the real-time positioning information, and a pursuit instruction is sent to the drone, the pursuit instruction is used to instruct the drone to arrive at the target location according to the target planning path, obtain the target image corresponding to the target location, and transmit the target image back to the command and control center, the target planning path is determined after path planning from the starting point of the drone to the real-time position; input the target image to the target recognition model corresponding to the current positioning scene, and obtain all the targets to be selected output by the target recognition model, and the current positioning scene is a desert scene, a forest scene, a road scene, a grassland scene or a river scene; receive a user input instruction, and in response to the user input instruction, send a follow instruction to the drone, the follow instruction is used to instruct the drone to follow the escorted target, and the user input instruction is used to select the escorted target from all targets to be selected.

[0111] In addition, the logic instructions in the above-mentioned memory 430 can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on such an understanding, the technical solution of the present invention can be essentially or partly embodied in the form of a software product that contributes to the prior art, and the computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a disk or an optical disk.

[0112] On the other hand, the present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute a drone-based escort method provided by the above methods, the method including: obtaining real-time positioning information from the escorted target, the real-time positioning information is determined after the real-time position is reported to the command and control center by the foot ring of the escorted target; when it is determined that the real-time positioning information deviates from the escort route corresponding to the escorted target, determining the current positioning scene according to the real-time positioning information, and sending a pursuit instruction to the drone, the pursuit instruction is used to instruct the drone to plan a path according to the target After arriving at the target location, a target image corresponding to the target location is obtained, and the target image is transmitted back to the command and control center, and the target planning path is determined after path planning from the starting point of the UAV to the real-time position; the target image is input to the target recognition model corresponding to the current positioning scene, and all targets to be selected output by the target recognition model are obtained, and the current positioning scene is a desert scene, a forest scene, a road scene, a grassland scene or a river scene; a user input instruction is received, and in response to the user input instruction, a follow instruction is sent to the UAV, and the follow instruction is used to instruct the UAV to follow the escorted target, and the user input instruction is used to select the escorted target from all targets to be selected.

[0113] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the escort method based on drone pursuit provided by the above-mentioned methods, the method comprising: obtaining real-time positioning information from the escorted target, the real-time positioning information is determined after the real-time position is reported to the command and control center by the foot ring of the escorted target; when it is determined that the real-time positioning information deviates from the escort route corresponding to the escorted target, determining the current positioning scene according to the real-time positioning information, sending a pursuit instruction to the drone, the pursuit instruction is used to instruct the drone to obtain the target positioning after arriving at the target positioning according to the target planning path. The command and control center calculates the target image corresponding to the target, and transmits the target image back to the command and control center, the target planning path is determined after path planning from the starting point of the drone to the real-time position; inputs the target image to the target recognition model corresponding to the current positioning scene, and obtains all the targets to be selected output by the target recognition model, and the current positioning scene is a desert scene, a forest scene, a road scene, a grassland scene or a river scene; receives a user input instruction, and in response to the user input instruction, sends a follow instruction to the drone, the follow instruction is used to instruct the drone to follow the escorted target, and the user input instruction is used to select the escorted target from all the targets to be selected.

[0114] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0115] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0116] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for escorting fugitives based on drone pursuit, characterized in that: include: Obtaining real-time positioning information from the escorted target, wherein the real-time positioning information is determined after the escorted target's foot ring reports the real-time position to the command and control center; In the case where it is determined that the real-time positioning information deviates from the escort route corresponding to the escorted target, the current positioning scene is determined according to the real-time positioning information, and a pursuit instruction is sent to the drone, wherein the pursuit instruction is used to instruct the drone to obtain a target image corresponding to the target positioning after arriving at the target positioning according to the target planning path, and transmit the target image back to the command and control center, wherein the target planning path is determined after path planning from the starting point of the drone to the real-time position; Inputting the target image to a target recognition model corresponding to the current positioning scene to obtain all targets to be selected output by the target recognition model, wherein the current positioning scene is a desert scene, a forest scene, a road scene, a grassland scene or a river scene; A user input instruction is received, and in response to the user input instruction, a follow instruction is sent to the drone, wherein the follow instruction is used to instruct the drone to follow the escorted target, and the user input instruction is used to select the escorted target from all targets to be selected.

2. The escort method based on drone pursuit according to claim 1 is characterized in that: Before determining the current positioning scene according to the real-time positioning information, the method further includes: When the straight-line distance between the real-time positioning information and the escort route is greater than a first preset distance, determining that the real-time positioning information deviates from the escort route corresponding to the escorted target; Alternatively, when the distance between the foot ring and the escort vehicle escorting the escorted target is greater than a second preset distance, it is determined that the real-time positioning information deviates from the escort route corresponding to the escorted target.

3. The escort method based on drone pursuit according to claim 1 is characterized in that: The drone is located on a drone platform of an escort vehicle escorting the escorted target. Before sending a pursuit instruction to the drone, the method further includes: Determine the starting point of the drone according to the real-time position of the escort vehicle; Path planning is performed using the starting point of the drone and the real-time position reported by the foot ring of the escorted target to determine the planned path of the target; The real-time position of the escort vehicle is determined by using the on-board positioning of the escort vehicle when it is determined that the real-time positioning information deviates from the escort route corresponding to the escorted target.

4. The escort method based on drone pursuit according to claim 1 is characterized in that: The drone is located on a drone platform on the escort route, and before sending a pursuit instruction to the drone, the method further includes: Determine the platform position of each drone platform corresponding to all drone platforms on the escort route, calculate the distance difference between the real-time position of the escort vehicle and the position of each platform, and determine the drone platform corresponding to the platform position with the smallest distance difference as the target platform; The starting point of the UAV is determined according to the platform position corresponding to the target platform, and the path planning is performed using the starting point of the UAV and the real-time position reported by the foot ring of the escorted target to determine the target planning path.

5. The escort method based on drone pursuit according to claim 1 is characterized in that: Before inputting the target image into the target recognition model corresponding to the current positioning scene, the method further includes: Get the corresponding sample images under different scenes; For each scene, the initial model is trained according to the sample images and the character head portrait annotations corresponding to each sample image to obtain a recognition model corresponding to the scene; Traverse all scenes and obtain each recognition model corresponding to all scenes.

6. The escort method based on drone pursuit according to claim 1 is characterized in that: The determining the current positioning scene according to the real-time positioning information and sending a pursuit instruction to the drone includes: When it is determined that the current positioning scene is a desert scene or a road scene, determining to use a first number of drones for pursuit; When it is determined that the current positioning scene is a grassland scene or a river scene, determining to use a second number of drones for pursuit; When it is determined that the current positioning scene is a forest scene, it is determined to use a third number of drones for pursuit; The first number is smaller than the second number, and the second number is smaller than the third number.

7. The method for escorting fugitives by drone according to claim 1 is characterized in that: In the case where there is more than one escorted target, sending a pursuit instruction to the drone includes: Send pursuit instructions to a preset number of drones; The number of the escorted targets is less than or equal to the preset number.

8. The escort method based on drone pursuit according to claim 7 is characterized in that: After sending the pursuit instruction to a preset number of drones, the method further includes: For a target image corresponding to any drone, all targets to be selected in the target image are obtained, and each target image corresponding to all drones is traversed to obtain a set of all targets to be selected; After removing duplicate avatars from all sets of targets to be selected, the avatars of the targets to be selected are displayed on a preset display screen.

9. An escort device based on drone pursuit, characterized in that: include: An acquisition unit, the acquisition unit is used to acquire real-time positioning information from the escorted target, the real-time positioning information is determined after the escorted target's foot ring reports the real-time position to the command and control center; A sending unit, wherein the sending unit is used to determine the current positioning scene according to the real-time positioning information when it is determined that the real-time positioning information deviates from the escort route corresponding to the escorted target, and send a pursuit instruction to the drone, wherein the pursuit instruction is used to instruct the drone to obtain a target image corresponding to the target positioning after arriving at the target positioning according to the target planning path, and transmit the target image back to the command and control center, wherein the target planning path is determined after path planning from the starting point of the drone to the real-time position; An input unit, the input unit is used to input the target image to a target recognition model corresponding to the current positioning scene, and obtain all targets to be selected output by the target recognition model, wherein the current positioning scene is a desert scene, a forest scene, a road scene, a grassland scene or a river scene; A receiving unit, the receiving unit is used to receive a user input instruction, and in response to the user input instruction, sends a follow instruction to the drone, the follow instruction is used to instruct the drone to follow the escorted target, and the user input instruction is used to select the escorted target from all targets to be selected.

10. A drone-based escort system, characterized in that: The device comprises the escort device based on drone pursuit as claimed in claim 9, wherein the escort device based on drone pursuit is a command and control center, and further comprises an escort vehicle and a drone platform; The command and control center is communicatively connected with the UAV platform and the escort vehicle respectively.

Images