Drone rescue methods, devices, equipment, storage media and software products

Abstract

This application discloses a drone rescue method, apparatus, device, storage medium, and program product, applied to the airborne system of a drone. The method includes: determining the target user's body movement information and facial information based on a video stream image of the target user; the facial information includes the target user's facial emotional state; determining the target user's action event based on the body movement information and the facial information; if the action event is a risk event, determining the distance between the drone and the target user: if the distance is less than or equal to a distance threshold, performing a rescue operation on the target user; or, if the distance is greater than the distance threshold, performing the rescue operation on the target user based on a preset strategy; the risk event includes a fall event.

Description

Technical Field

[0001] This application relates to the fields of public safety and artificial intelligence, and in particular to a drone rescue method, apparatus, equipment, storage medium, and program product. Background Technology

[0002] In recent years, there have been numerous incidents of falls from heights on school campuses, posing a serious challenge to campus safety. To address this issue, school management has increased protective facilities such as guardrails, anti-climb fences, fall protection nets, and fall-proof air cushions to prevent falls. However, these traditional protective methods still have limitations such as high cost and limited applicability, resulting in insufficient capacity to cope with safety risks. Summary of the Invention

[0003] To address the aforementioned technical problems, embodiments of the present invention provide a method, apparatus, device, storage medium, and program product for drone rescue.

[0004] The drone rescue method provided in this application embodiment is applied to the onboard system of a drone and includes:

[0005] Based on the video stream images of the target user, determine the target user's body movement information and facial information; the facial information includes the target user's facial emotional state.

[0006] Based on the body movement information and the facial information, the target user's action events are determined;

[0007] If the action event is a risk event, then determine the distance between the drone and the target user: if the distance is less than or equal to a distance threshold, then perform a rescue operation on the target user; or,

[0008] If the distance is greater than the distance threshold, the rescue operation is performed on the target user based on a preset strategy; the risk event includes a fall event.

[0009] The drone rescue device provided in this application embodiment is applied to the airborne system of a drone and includes:

[0010] A determining unit is configured to determine the target user's body movement information and facial information based on a video stream image of the target user; the facial information includes the target user's facial emotional state; and determine the target user's action events based on the body movement information and the facial information.

[0011] The processing unit is configured to determine the distance between the drone and the target user if the action event is a risk event: if the distance is less than or equal to a distance threshold, a rescue operation is performed on the target user; or, if the distance is greater than the distance threshold, the rescue operation is performed on the target user based on a preset strategy; the risk event includes a fall event.

[0012] The processing device provided in this application includes a processor and a memory. The memory is used to store computer programs, and the processor is used to call and run the computer programs stored in the memory to execute any of the above-described drone rescue methods.

[0013] The computer-readable storage medium provided in this application embodiment is used to store a computer program that causes a computer to execute any of the above-described drone rescue methods.

[0014] The computer program product provided in this application includes computer program instructions that cause a computer to execute any of the above-described drone rescue methods.

[0015] In the technical solution of this application embodiment, the onboard system of the UAV determines the target user's body movement information and facial information based on the target user's video stream image, and determines the target user's action events based on the body movement information and facial information. If the action event is a risk event, the distance between the UAV and the target user is determined: if the distance is less than or equal to a distance threshold, a rescue operation is performed on the target user; or, if the distance is greater than the distance threshold, a rescue operation is performed on the target user based on a preset strategy; wherein, risk events include fall events. In this way, by capturing the body movements of the person to be rescued in real time, estimating the body characteristic data of the person to be rescued, and generating the action events of the person to be rescued, a rapid safety incident emergency response can be provided to the person to be rescued according to the degree of danger of the action event, and a timely rescue decision can be made. This not only fills the problem of insufficient response capabilities of traditional protection methods, but also effectively improves the accuracy of preventive measures and enhances the timeliness of rescue operations. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating a drone rescue method for an airborne system of a drone, as provided in an embodiment of this application.

[0017] Figure 2 This is a schematic diagram illustrating the principle of a campus high-rise fall prevention method based on a rescue drone provided in this application embodiment;

[0018] Figure 3 This is a flowchart illustrating a campus high-rise fall prevention method based on a rescue drone provided in an embodiment of this application.

[0019] Figure 4 This is a schematic diagram of the structure of a drone rescue device for an airborne system of a drone, provided in an embodiment of this application;

[0020] Figure 5 This is a schematic diagram of the processing device provided in the embodiments of this application. Detailed Implementation

[0021] The technical solutions of the embodiments of this application will now be described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0022] It should be noted that, in the embodiments of this application, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, in the embodiments of this application, the character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0023] In the description of the embodiments of this application, the term "correspondence" may indicate that there is a direct or indirect correspondence between two things, or that there is an association between two things, or that there is a relationship of instruction and being instructed, configuration and being configured, etc.

[0024] To facilitate understanding of the technical solutions of the embodiments of this application, the relevant technologies of the embodiments of this application are described below. The following relevant technologies are optional solutions and can be combined with the technical solutions of the embodiments of this application in any way, and they all fall within the protection scope of the embodiments of this application.

[0025] In recent years, there have been numerous incidents of falls from heights on school campuses, posing a serious challenge to campus safety. To address this issue, school management has added railings and anti-climb fences to rooftop platforms and windows in high-rise buildings to prevent falls. Fall protection nets have also been installed on the exterior of windows and balcony railings in high-rise buildings to provide an additional layer of protection. Furthermore, to cope with emergencies, some schools have provided movable fall cushioning devices to provide a landing buffer in the event of a fall, maximizing the safety of those falling.

[0026] While the aforementioned rescue and protection measures can reduce the occurrence of falls from heights to some extent, they still have some shortcomings, including:

[0027] (1) High cost. The installation and maintenance of equipment such as rooftop platforms, guardrails at windows of high-rise buildings, anti-climb fences and fall protection nets require a lot of financial support. In addition, the purchase and maintenance costs of mobile fall protection air cushions are also high. Schools are limited by their budget and cannot implement them, which reduces the level of safety protection.

[0028] (2) Insufficient adaptability. Some campus buildings may be unable to install similar safety facilities due to structural limitations, or the installed protective measures may not provide sufficient protection. Therefore, the risk of falls from heights remains, and timely rescue may be impossible after a fall, leading to serious safety consequences. These technical measures may only be effective in certain high-rise areas, while other areas remain under safety threats.

[0029] (3) Insufficient response capabilities. Although these technical measures provide an additional layer of protection, they cannot completely eliminate the risk of falling from heights, especially when the accident is driven by subjective will, in which case these technical measures are basically unable to provide sufficient safety protection.

[0030] In other words, guardrails, anti-climb fences, and fall protection nets may be exploited by students to intentionally bypass them. Specifically, these systems may not fully cover the campus, leaving areas that students can easily find and circumvent. Furthermore, potential problems in design, construction, inspection, and maintenance could reduce their protective capabilities, rendering them ineffective. Fall protection air cushions, on the other hand, are typically only usable in limited locations and under specific circumstances, requiring professional operation. Their movement, deployment, and inflation all demand significant manpower and time, making them unsuitable for time-sensitive rescue operations.

[0031] To address the aforementioned technical issues, this application proposes a drone rescue method. This method employs drone rescue technology, combined with real-time monitoring and artificial intelligence (AI) motion analysis, to capture the limb movements of the person to be rescued in real time, estimate body characteristic data, and generate motion events. This provides a rapid emergency response to safety incidents and enables timely rescue decisions. In the event of a crash, the drone will quickly calculate its trajectory, adjust its position to capture the person to be rescued, and control its descent speed by calculating lift and activating an auxiliary landing device after capture, ensuring a safe landing for the person to be rescued. This not only fills the gaps in traditional protective measures but also effectively improves the accuracy of preventative measures and greatly enhances the timeliness of rescue operations.

[0032] To facilitate understanding of the technical solutions of the embodiments of this application, the technical solutions of this application are described in detail below through specific embodiments. The above-mentioned related technologies are optional solutions and can be arbitrarily combined with the technical solutions of the embodiments of this application, all of which fall within the protection scope of the embodiments of this application. The embodiments of this application include at least some of the following contents.

[0033] This application proposes a drone rescue method applied to an airborne system of a drone. Figure 1 This is a flowchart illustrating a drone rescue method for an airborne system of a drone provided in an embodiment of this application. Figure 1 As shown, the method includes the following steps:

[0034] Step 101: Based on the video stream images of the target user, determine the target user's body movement information and facial information.

[0035] In this embodiment, after the UAV arrives at the location given in the alarm information, it will combine the target recognition technology of the command center and the airborne system to accurately identify the target user to be rescued. Once the target user is confirmed, the UAV's airborne system will immediately start a real-time data acquisition program to capture a video stream image of the target user. This video stream image includes multiple video frame images, each containing various key information of the target user, such as facial images, 3D facial models, and key point information of current limb movements. After capturing the video stream image of the target user, the UAV's airborne system determines the target user's current limb movement information and facial information based on the various key information contained in each video frame image of the target user's video stream image. Among them, the limb movement information includes changes in the target user's limb movements, and the facial information includes the target user's facial emotional state.

[0036] It should be noted that after the drone's onboard system captures a video stream image of the target user, key information from the video stream is transmitted to the command center in real time. This information assists the command center in remotely analyzing the drone's behavior and making rescue decisions. Simultaneously, the drone's onboard system fully records and saves the target user's key information for more accurate real-time target location.

[0037] In some implementations, the video stream image includes multiple video frame images; wherein, step 101 can be implemented through steps 1011-1012, specifically including:

[0038] Step 1011: Based on each video frame image in the multiple video frame images, determine the body part information of the target user in each video frame image.

[0039] Step 1012: Determine limb movement information and facial information based on the target user's body part information in each video frame image.

[0040] In this embodiment, since the video stream image includes multiple video frame images, and each video frame image contains various key information of the target user, computer vision and machine learning techniques such as Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) algorithms can be used to detect the key points of the target user in each video frame image. These key points are the joint information of the target user, which are usually easily identifiable and trackable feature points, including the head, shoulders, elbows, wrists, hips, knees, etc. After detecting the key points of the target user in each video frame image, features are extracted from the key points of the target user. These features can be position information, color information, or other information that helps to identify body parts. Based on the extracted features, the body part information of the target user in each video frame image can be obtained, thereby determining the target user's limb movement information and facial information based on the body part information of the target user in each video frame image.

[0041] In some implementations, the target user's body part information includes joint point information; wherein, step 1011 may specifically include:

[0042] For the keypoint information of the target user in each video frame, determine the changes in the keypoint information of the target user between each video frame and adjacent video frames;

[0043] The changes in the target user's joint information between each video frame and adjacent video frames are analyzed to obtain limb movement information and facial information.

[0044] Here, the keypoint information of the target user can provide important information including body movements and facial changes. For the keypoint information of the target user in each video frame, the displacement of the keypoint information between consecutive video frames can be calculated by optical flow technology to track the changes in the target user's body movements and facial expressions. That is, analyzing the changes in the target user's keypoint information between each video frame and adjacent video frames can help understand the dynamic changes in the user's movement patterns and facial expressions, thereby obtaining the target user's body movement information and facial information.

[0045] Specifically, by using optical flow algorithms such as Lucas-Kanade and Farneback, the motion vectors of each keypoint (limb keypoint and facial keypoint) of the target user across consecutive video frames are estimated, resulting in an optical flow field representing the motion vector of each keypoint across consecutive video frames. The motion vector of each keypoint across consecutive video frames represents the positional change of that keypoint across consecutive video frames. By concatenating the motion vectors of each keypoint across consecutive video frames, the changes in limb movements and facial expressions (such as speed and direction) of each keypoint in the video stream can be obtained. Based on the changes in limb movements of each keypoint in the video stream, the limb movement information of the target user can be obtained. For the facial changes of each keypoint in the video stream, facial expression recognition can be performed by combining the facial changes of each keypoint to obtain facial expression changes. For example, by observing the movement of keypoints at the corners of the eyes and mouth, expressions such as smiling or frowning can be identified, thereby obtaining the facial information of the target user based on these facial expression changes.

[0046] Step 102: Based on body movement information and facial information, determine the target user's action events.

[0047] In this embodiment, after determining the target user's body movement and facial information, a comprehensive analysis of the changes in the target user's body movements and facial expressions can be performed to predict the target user's movement characteristics. Then, specific action events can be generated based on these characteristics. For example, if it is predicted that the target user's movement speed at a certain joint point suddenly increases, and the target user's face changes to an expression containing negative emotions such as frowning or fright, this may indicate that the target user is about to jump or fall. In this case, the drone can perform emergency rescue operations on the target user based on the predicted action events. This method, which comprehensively considers both movement and emotion, can generate more accurate action event priorities, provide more comprehensive decision support for rescue, and optimize the execution plan of rescue operations.

[0048] In some implementations, step 102 can be achieved through steps 1021-1023, specifically including:

[0049] Step 1021: Determine the movement characteristics of the target user based on body movement information.

[0050] Step 1022: Determine the emotional characteristics of the target user based on facial information.

[0051] Step 1023: Identify the action event based on motion and emotional characteristics.

[0052] In this embodiment, the changes in limb movements in the target user's limb movement information can be analyzed using action recognition models such as 3D convolutional neural networks (3D-CNN) to obtain the target user's action features, such as movement direction, joint angles, and limb speed. Similarly, the changes in facial expressions in the target user's facial information can be analyzed using emotion recognition models such as those based on open-source computer vision libraries (OpenCV) and facial expression recognition (FER), or facial expression analysis systems, to obtain the target user's emotional features, such as smiling or frowning. These emotional features are then integrated into the target user's action features for comprehensive analysis, thereby further revealing the target user's action events. Both the action recognition model and the emotion recognition model are pre-trained models.

[0053] In some implementations, step 1023 may specifically include:

[0054] By fusing motion and emotional characteristics, a comprehensive feature profile of the target user is obtained.

[0055] Based on comprehensive features, the action tendency of the target user is predicted by an action classification model to obtain action events.

[0056] Here, after obtaining the target user's motion and emotional features, an emotion classification model is used to categorize the emotional features, with each category corresponding to a score. The emotional score is then obtained based on the category of the emotional feature. Next, the target user's emotional score is integrated into their action features. This involves representing the emotional score and action features as vectors and concatenating these vectors to fuse the emotional score and action features into a comprehensive feature vector. This comprehensive feature vector represents the target user's overall features, containing combined information about both action and emotion. Finally, the comprehensive feature is input into the action classification model. The action classification model categorizes the target user's behavior based on the comprehensive feature to predict the target user's action tendencies and outputs the action event corresponding to the category result. Both the emotion classification model and the action classification model are pre-trained models.

[0057] For example, a dual-stream architecture can be adopted, where one stream processes the target user's action features and the other stream processes the target user's emotion rating. A temporal convolutional module (TCM) captures the dependency between the action and emotion streams, and average pooling reduces the size of the action features and emotion ratings. The two are then concatenated to form a comprehensive feature vector containing integrated information about the target user's actions and emotions. This comprehensive feature vector is then input into a classification module of a multilayer perceptron, which classifies the target user's behavior, thereby predicting the target user's action events.

[0058] Step 103: If the action event is a risk event, determine the distance between the drone and the target user: if the distance is less than or equal to the distance threshold, perform a rescue operation on the target user; or, if the distance is greater than the distance threshold, perform a rescue operation on the target user based on a preset strategy.

[0059] In this embodiment, after determining the target user's action event, if the action event is a risky event such as a fall or a fall from a building, the distance between the drone and the target user is first determined. This distance can be Euclidean distance, Manhattan distance, etc. If the distance is less than or equal to a distance threshold, the drone immediately performs a rescue operation on the target user, such as firing a net gun at the target user to quickly restrict their risky activity. If the distance is greater than the distance threshold, the drone needs to perform a rescue operation on the target user according to a preset strategy. The distance threshold is a preset value, and its specific size can be adjusted according to different scenarios.

[0060] In some implementations, before a risk event occurs and before rescue operations are carried out on the target user, reassuring messages can be sent to the target user to help reduce panic and anxiety during the tense rescue process. These messages include:

[0061] Based on facial information, determine the identity information of the target user;

[0062] Based on identity information, determine the associated information of the target user within a preset time period; the associated information includes mental health and / or emotional assessment information of the target user;

[0063] Based on relevant information, determine persuasive messages for the target users;

[0064] Send reassuring messages to the target users.

[0065] Here, the target user's identity information can be determined through an identity database system based on their facial information. Based on this identity information, their associated information within a preset time period can be determined through an association system. For example, if the target user is a student, their identity information can be obtained through the school's student registration system based on their facial information. Based on this identity information, their campus experiences within a preset time period, such as grades, school performance, teacher evaluations, and reward / punishment records, can be obtained through the same system. By analyzing these campus experiences, their mental health level and current emotional fluctuations within the preset time period can be determined, thus obtaining their associated information. Through comprehensive analysis of this associated information, their mental health and / or emotional assessment information can be obtained. Based on this assessment, targeted reassurance messages can be developed and delivered to the target user via drone to reduce their panic and anxiety.

[0066] In some implementations, step 103, "performing rescue operations on the target user based on a preset strategy," may specifically include:

[0067] Determine the environmental information of the target user;

[0068] Based on body movement information and environmental information, the movement trajectory of the target user is determined;

[0069] Based on the motion trajectory, determine the target flight path of the UAV;

[0070] The drone tracks the target user based on the target flight path and determines the distance between the drone and the target user. If the distance is less than or equal to a distance threshold, a rescue operation is carried out on the target user.

[0071] Here, once a risk event occurs, the drone's onboard system can quickly and accurately predict the target user's fall trajectory (movement trajectory) based on the target user's body movements and environmental information such as wind speed and direction. Based on this fall trajectory, it automatically plans the optimal flight path for the drone towards the target user (target flight path). The drone then quickly adjusts its position and tracks the target user according to the optimal flight path. At the same time, the drone's onboard system judges the distance between the drone and the target user in real time. When the distance is less than or equal to a distance threshold, the drone's onboard system can calculate the shortest path between the drone and the target user in real time. Based on this shortest path, it can quickly calculate the best spatial position and direction for launching a rescue operation and carry out a rescue operation on the target user.

[0072] Here, for predicting the fall trajectory of a target user, a dynamic landing point calculation model can be used. The target user's initial position, limb movement information (velocity, direction), and environmental information (wind speed, wind direction) are used as inputs. The dynamic motion differential equations in the model are solved by numerical integration, and the predicted landing point position of the target user is calculated. Thus, the fall trajectory of the target user is calculated based on the target user's initial position and the predicted landing point position.

[0073] Here, regarding the optimal flight path planning for drones, specifically, the drone's onboard system will perceive the environmental conditions around the drone in real time, including the location of the target user and the location of obstacles. Secondly, a target reward value and a reward function are designed. This function considers distance rewards, first-time reconnaissance rewards, repeated reconnaissance rewards, drone collision rewards, etc., to incentivize the drone to approach the target user and avoid repeated reconnaissance and collisions with obstacles. Then, a path planning algorithm such as a genetic algorithm is used to optimize the path of the drone to the target user in real time while calculating the value of the reward function. When the value of the reward function reaches the target reward value, the optimal flight path of the drone can be planned.

[0074] It's important to note that before determining the target user's trajectory, the drone's onboard system first collects environmental information about the target user's location to prepare for rescue operations. Environmental information includes meteorological and site information. Meteorological data includes wind speed, wind direction, cloud height, precipitation, and atmospheric pressure, which can affect the drone's flight safety and performance optimization. Site information includes marking flat areas for safe landing and identifying obstacles to avoid. Site information helps plan the drone's safe landing path, ensuring the drone is not unexpectedly hindered during and after landing, thus guaranteeing flight stability and rescue safety. Specifically, the drone's onboard system can acquire meteorological information such as weather warnings and alerts in real time via high-speed communication networks. Through real-time perception and analysis of the surrounding environment, it identifies suitable open areas for flight and the location of obstacles. After obtaining meteorological and site information, the drone's onboard system quickly transmits it to the command center, providing real-time and accurate information support to assist the command center in conducting more precise rescue operations.

[0075] In some implementations, after a risk event occurs and rescue operations are performed on the target user, the process may further include:

[0076] Based on the UAV's spatial location information and motion status, determine the UAV's target landing path;

[0077] The drone is controlled to descend based on the target landing path, and the lift margin of the drone is calculated. If it is determined that the lift margin is less than or equal to the lift threshold and the descent speed of the drone is greater than the speed threshold, the descent speed of the drone is controlled by auxiliary equipment.

[0078] Here, after a risk event occurs and the drone tracks the target user to the rescue airspace and captures the target user along the target flight path, the drone's onboard system immediately acquires its own motion status and performs environmental perception to obtain spatial position information such as altitude, speed, and direction. Based on this spatial position information, it performs 3D reconstruction to generate 3D scene information of the drone's location. Based on this 3D scene information, the initial starting point and initial ending point in space can be determined. Based on these initial starting and ending points, the drone's initial landing path can be generated, and the initial landing path is optimized in real time using a local path optimization algorithm, ultimately planning the drone's optimal landing path (target landing path). During the drone's descent along the optimal landing path, the onboard system calculates the drone's lift margin in real time. If it detects that at a certain moment the drone's lift margin is less than or equal to a lift threshold, and the drone's descent speed is greater than a speed threshold, the onboard system will release auxiliary equipment such as a parachute to control the drone's descent speed, ensuring a safe landing. The lift threshold and speed threshold are preset values, and their specific values ​​can be adjusted according to different scenarios.

[0079] It should be noted that, as the spatial position information of the drone changes during its descent, and the local path optimization algorithm optimizes the drone's landing path in real time, the optimal landing path will also change when the drone's spatial position information is different.

[0080] In some implementations, if an obstacle is detected on the target landing path during the descent of the UAV based on the target landing path, the target landing path is updated to obtain an updated target landing path; the UAV is then controlled to descend based on the updated target landing path.

[0081] Here, as the drone descends along the target landing path, its onboard system will also detect whether there are any obstacles on the target landing path in real time. If an obstacle is detected, the target landing path will be replanned and updated, thereby controlling the drone to descend according to the updated target landing path to adapt to the dynamic environment. Through continuous dynamic path planning and real-time obstacle perception, the drone will eventually land safely, which not only takes into account the safety of the drone, but also ensures the best adaptation to the flight environment.

[0082] In some implementations, it can also coordinate with one or more other drones to carry out rescue operations on the target user, specifically including:

[0083] The drone communicates with one or more other drones for coordinated rescue operations against a target user. Each drone in the coordinated rescue operation is configured with a corresponding mission and / or execution strategy, and the mission and / or execution strategy are different for each drone.

[0084] Here, when drones face difficulties performing rescue missions alone, they can collaborate with other drones to conduct rescue operations on target users, ensuring that the rescue operation is not hindered. Specifically, in a multi-drone collaborative rescue process, each drone first checks its own status and capabilities, such as its maximum rescue payload weight, spatial location, and battery endurance, to ensure it is in optimal operating condition. Simultaneously, each drone receives key information about the target user, such as estimated weight, location, and the urgency of the rescue. Based on this information, each drone is assigned different tasks (different task content and execution strategies) and begins the rescue operation. During the rescue, each drone monitors its own status and surrounding environment in real time to ensure the safety and efficiency of its flight path. Furthermore, each drone shares information with other drones through a high-speed communication network to achieve information synchronization and transparency, thereby enabling multi-drone collaborative rescue of the target user.

[0085] It should be noted that during the coordinated rescue operation of multiple drones targeting a user, each drone can adjust its rescue priority and action plan in real time based on its current task status. For example, if a drone finds that its battery or payload capacity is insufficient to complete the current rescue mission, it can quickly transfer the task to another drone in better condition, while it switches to monitoring or auxiliary tasks. This allows for flexible task switching and resource optimization, improving the responsiveness and adaptability of the rescue mission. For instance, for a lighter target user, a drone can choose to quickly approach and use a net gun to capture the target user; for a heavier target user or a more complex situation, the drone can adopt a more cautious rescue strategy, gradually approaching the target user and carrying out the rescue operation.

[0086] In the technical solution of this application embodiment, the onboard system of the UAV determines the target user's body movement information and facial information based on the target user's video stream image, and determines the target user's action events based on the body movement information and facial information. If the action event is a risk event, the distance between the UAV and the target user is determined: if the distance is less than or equal to a distance threshold, a rescue operation is performed on the target user; or, if the distance is greater than the distance threshold, a rescue operation is performed on the target user based on a preset strategy; wherein, risk events include fall events. In this way, by capturing the body movements of the person to be rescued in real time, estimating the body characteristic data of the person to be rescued, and generating the action events of the person to be rescued, a rapid safety incident emergency response can be provided to the person to be rescued according to the degree of danger of the action event, and a timely rescue decision can be made. This not only fills the problem of insufficient response capabilities of traditional protection methods, but also effectively improves the accuracy of preventive measures and enhances the timeliness of rescue operations.

[0087] This application also proposes a method for preventing falls from high-rise buildings on campus based on rescue drones. Figure 2 This is a schematic diagram illustrating the principle of a campus high-rise fall prevention method based on a rescue drone provided in this application embodiment. The onboard system of the rescue drone consists of the following basic components:

[0088] (1) UAV Basic Platform. The UAV basic platform includes the basic frame and the flight control system. The flight control system is responsible for flight control and ensures that the UAV can fly stably; the basic frame is the supporting structure of the entire system, providing stability and load-bearing capacity. It is made of lightweight and strong materials to ensure that the UAV can maintain structural integrity during flight.

[0089] (2) Environmental Detection and Obstacle Avoidance System. The environmental detection and obstacle avoidance system can be implemented by modules such as phased array radar, lidar, and infrared sensors. It is a key sensor for sensing and detecting the surrounding environment, and can provide high-resolution, wide-area coverage target detection and tracking capabilities. Among them, phased array radar is suitable for long-distance target detection; lidar is suitable for accurate ranging and 3D environment modeling; and infrared modules can provide environmental and obstacle detection capabilities with a certain level of accuracy and cost-effectiveness.

[0090] (3) Camera. Cameras are used to monitor and acquire visible light or infrared image data in real time for target identification, navigation, and surveillance. Cameras can also work in conjunction with other sensors to provide comprehensive environmental awareness.

[0091] (4) Net gun. A net gun can fire a net at a certain distance. This net can be a specially designed mesh structure to ensure rapid and secure capture of the target and restrict the target's activities. It can be used for security and interception needs in various application scenarios.

[0092] (5) Parachute. The parachute is used in emergency situations, such as when the drone has insufficient lift or other emergencies. The parachute can be deployed to slow the drone's descent and ensure the safety of the drone and the surrounding environment.

[0093] (6) Calculation Unit. The calculation unit is used to identify the person to be rescued from the video stream acquired by the camera and to analyze the person's limb movements and movement trends.

[0094] (7) Decision-Making Unit. The decision-making unit is the intelligent core of the UAV system, responsible for processing sensor data, executing tasks, and making flight decisions. It can analyze sensor data through algorithms to determine whether the person to be rescued is falling and decide whether to activate net gun interception, release parachutes, or take other countermeasures. The decision-making unit can also coordinate the work of different parts, enabling the entire airborne system to respond quickly and accurately based on real-time conditions. The introduction of the decision-making unit allows the UAV airborne system to cope more flexibly with complex environments and diverse tasks. Through continuous learning and adaptation, the decision-making unit can improve the system's intelligence level, enabling it to more effectively cope with various challenges, including interception of different types of targets and emergency handling. The intelligence of the decision-making unit enhances the autonomy and adaptability of the entire UAV airborne system, making the UAV more practical and reliable.

[0095] Figure 3 This is a flowchart illustrating a campus high-rise fall prevention method based on a rescue drone, as provided in an embodiment of this application. Figure 3 As shown, the method consists of four stages, specifically including:

[0096] Step 301: Gather initial weather and site information to prepare for the rescue mission.

[0097] Meteorological data is crucial for the safe and effective operation of rescue drones. Meteorological data includes wind speed, wind direction, cloud height, precipitation, and atmospheric pressure, which can affect the flight safety and performance optimization of drones. It also helps drone onboard systems plan obstacle avoidance strategies, ensures good visibility for drone operators, improves drone stability, and triggers timely emergency responses in emergencies. In this embodiment, real-time weather warnings and alerts can be obtained via a 5G network, enabling the drone onboard system to respond more sensitively to sudden weather conditions and ensure the safe and efficient execution of flight missions. The collected meteorological data specifically includes:

[0098] (1) Wind speed and direction. Wind speed and direction information obtained through meteorological sensors is crucial for UAV flight. This data can be used to calculate flight trajectories, adjust flight paths, and ensure the stability of UAVs under different wind conditions.

[0099] (2) Temperature and humidity. Temperature and humidity data are crucial for flight performance and the stable operation of electronic equipment. High temperatures, low temperatures, or extreme humidity may negatively impact the operation and equipment performance of the UAV.

[0100] (3) Cloud height and cloud cover. Understanding cloud height and cloud cover helps in planning safe flight altitudes. Cloud data can also be used to plan obstacle avoidance strategies, ensuring that drones can navigate safely in complex weather conditions.

[0101] (4) Precipitation. Precipitation data is a key meteorological parameter. Accurate precipitation information can be used to formulate flight plans, select appropriate flight altitudes, and adjust flight speeds.

[0102] (5) Atmospheric pressure. Atmospheric pressure data is crucial for calculating and adjusting flight altitude. Unmanned aerial vehicle (UAV) systems need to acquire atmospheric pressure information in real time to ensure the accuracy of altitude measurements.

[0103] (6) Thunderstorm activity. Real-time monitoring of thunderstorm activity can help UAV systems plan lightning avoidance strategies and reduce the risk of lightning strikes. Early warning information about thunderstorm activity will trigger corresponding safety measures, such as reducing flight altitude or returning to base quickly.

[0104] (7) Visibility. Visibility data is crucial to ensuring that drone operators can clearly monitor the flight area. Low visibility can lead to unsafe flight, so flight plans need to be adjusted according to the actual situation.

[0105] (8) Early warning and alerts for the external environment. Real-time acquisition of external early warning and alert information, including meteorological and electromagnetic environment data, through 5G networks enables UAV systems to respond more promptly to sudden weather conditions, such as storms, typhoons, or other extreme weather events. Such real-time data helps the decision-making unit in the airborne system to adjust flight plans more accurately, ensuring the safety and effectiveness of rescue flights.

[0106] Site information is crucial for the safe landing of rescue drones. This includes identifying flat areas for safe landing, avoiding the location of obstacles, and understanding interactions with the surrounding environment. Site information helps plan the landing path of the rescue drone, ensuring it is not unexpectedly hindered during landing, thus guaranteeing the stability of the rescue flight and the safety of the rescue operation.

[0107] Information on open areas, such as campus airspace (specific areas designated in campus planning) and uncontrolled areas (G-level areas), can be pre-entered into the system. Alternatively, after the rescue drone takes off, its onboard detection equipment (such as airborne infrared / visible light cameras and airborne phased array radar) combined with the onboard artificial intelligence system can actively explore and discover these areas. These open areas can be used as obstacle avoidance paths or emergency landing points during flight, ensuring that the rescue drone has sufficient space to maneuver and land safely in potentially risky situations. The drone's onboard environmental detection and obstacle avoidance system can identify suitable open areas for flight by sensing and analyzing the surrounding environment in real time, improving its ability to respond to potential risks. Furthermore, the onboard system can quickly transmit the acquired data to the command center, providing decision-makers with real-time and accurate information support. Simultaneously, this data also serves as an important reference for drone operators, helping them to control the drone more precisely and ensure the smooth execution of flight missions.

[0108] If a rescue air cushion has been deployed, its location information can be sent to the rescue drone in advance or during the rescue mission. Knowing the exact location of the rescue air cushion after a fall from a height will help the rescue drone quickly plan an evacuation route after detecting the person being rescued, and guide the person to the air cushion. Utilizing the cushion's cushioning and safety protection, the injury from the fall can be minimized.

[0109] Step 302: Identify the person to be rescued and estimate their physical characteristics.

[0110] After takeoff, the rescue drone will immediately fly to the location specified in the high-altitude fall warning message. Once in position, under the control of the operator, the drone, combined with the target recognition technology of the command center, operators, and onboard systems, will accurately identify the person to be rescued. Once the person is identified, the rescue drone immediately initiates a real-time data acquisition program, capturing video stream images of the person to be rescued. This video stream includes various key information such as facial images, 3D facial models, the person's body temperature, and current limb movements. This key information is transmitted in real time to the rescue command center to assist in remote behavioral analysis and decision-making. Simultaneously, the rescue drone will ensure the complete recording and preservation of the identifiable features of the person to be rescued, such as physical appearance, clothing, and any identifiable personal markings, in order to more accurately locate the person to be rescued in real time.

[0111] After acquiring 3D image information of the person to be rescued, the UAV's onboard system automatically analyzes and estimates their body characteristics. Through pre-trained model algorithms, the onboard system can determine the height and weight of the person to be rescued with acceptable accuracy. These automated estimations are crucial to the rescue process, providing a basis for subsequent rescue decisions. It should be noted that operators and commanders have the authority to modify the automatically estimated height, weight, and other predicted data from the rescue UAV. This key feature allows the rescue team to adjust according to the actual situation, ensuring that the final body characteristic data is more accurate and consistent with reality. This real-time data modification mechanism helps the rescue team respond more flexibly to complex rescue scenarios during the rescue process, ensuring optimal rescue decisions.

[0112] Step 303: Capture the body movements of the person to be rescued in real time and generate motion events.

[0113] The rescue drone initiates a limb motion capture process. The core of this process lies in the drone's high-definition camera and cutting-edge motion sensing technology. These technologies accurately capture key body parts of the person being rescued in the video stream, such as joint information of the head, arms, legs, and hands and feet. After capturing this joint information, the onboard system uses optical flow technology to further analyze it. Optical flow technology calculates the motion trajectory of these joints in the image sequence by comparing the differences in joint information between consecutive video frames, thereby outputting the limb motion information of the person being rescued, including the speed, direction, and possible movement patterns of each joint.

[0114] After obtaining information about the limb movements of the person awaiting rescue, pre-trained deep learning models such as CNN and LSTM can be used to predict their movement events. For example, if a sudden increase in the movement speed of a certain joint is detected, it may indicate that the person is about to jump. By analyzing the motion characteristics of these joints, the deep learning model can predict the details of the movement, such as the direction and force of the jump, and thus generate specific movement events.

[0115] Meanwhile, the rescue drone system can utilize high-definition facial images or data acquired through optical flow technology to recognize the facial expressions of those to be rescued and assess their emotional state. Facial expression recognition technology identifies different emotional states by analyzing subtle changes in facial features. For example, it employs the TV-L1 optical flow method and an improved ShuffleNet model, extracting optical flow features from keyframes and pre-training the model, then using a Support Vector Machine (SVM) classifier to recognize the micro-expressions of those to be rescued. Furthermore, publicly available facial expression recognition datasets can be used to train the model to improve the accuracy of expression recognition.

[0116] By combining facial expression information of those awaiting rescue, the rescue drone system can score their emotional state and integrate these scores into previously predicted action events. This comprehensive approach, which considers both emotions and actions, can generate more accurate action event priorities, providing more comprehensive decision support for rescue efforts and optimizing the execution plan of rescue operations.

[0117] Once the drone anticipates a potentially dangerous maneuver, it will immediately activate its emergency plan, adjusting its flight path to quickly approach the person awaiting rescue. Throughout the process, the drone's obstacle avoidance system will ensure that no collision occurs during the approach, guaranteeing the safety of the rescue operation. Simultaneously, the drone will communicate with the ground control center in real time, transmitting critical information to enable the rescue team to react rapidly.

[0118] In the intelligent upgrade of campus safety management, the multi-directional deployment of high-definition surveillance cameras constitutes the first line of defense. These cameras can capture every corner of the campus in real time, ensuring a rapid response to potential dangers such as students engaging in activities at heights. Combined with AI technology, these surveillance devices not only achieve basic visual monitoring but also, by analyzing data collected from existing security facilities such as cameras, infrared sensors, and ultrasonic sensors, enable automatic identification and real-time early warning of abnormal events. Furthermore, the advanced analytical capabilities of AI technology enable it to identify abnormal behaviors and risks, generating alarms in real time, providing more acute insights for campus safety management. Based on this, the linkage mechanism between the campus system and rescue drones is activated. Once the AI ​​function in the campus system detects an emergency, it immediately issues instructions to the rescue drone, which then quickly arrives at the scene, providing a first-hand, real-time perspective to help rescue personnel accurately assess the situation and immediately participate in rescue operations when necessary.

[0119] Once the drone approaches the person to be rescued, it can remotely read the ID information from the person's electronic student ID. If the person does not have their electronic student ID, the drone can quickly obtain their identity information by capturing a facial image and utilizing the school's student registration system. This identity information allows the team to determine the person's school-related information over a period of time, such as grades, academic performance, teacher evaluations, and recent awards and punishments. By analyzing the person's past records, the rescue team can objectively assess their mental health and the reasons for their current emotional fluctuations, and then develop targeted reassurance messages. These messages are then transmitted to the person through the drone's onboard loudspeaker to help them remain calm and reduce panic and anxiety during the tense rescue process.

[0120] Step 304: Conduct rescue operations on rescue personnel through proactive intervention and descent tracking.

[0121] Once the drone identifies a person at risk of falling, it will immediately initiate proactive intervention. Specifically, the drone will fire a net gun to quickly restrict the person's movement and prevent them from falling. If, during proactive intervention, extreme situations such as obstructed visibility or obstacles prevent successful intervention, the drone will quickly change its rescue strategy and conduct the rescue operation as the person begins to fall. By combining proactive intervention with fall rescue, effective control can be achieved in the early stages of risk, significantly improving the success rate of rescue operations and ensuring the safety of the person being rescued.

[0122] For rescue operations involving falling individuals, rescue drones utilize onboard high-definition cameras and multiple sensors to track the movements of those being rescued in real time, capturing key motion parameters. Once the drone detects signs of movement and potential jumping or falling, it quickly activates its emergency plan and calculates the individual's fall trajectory. Specifically, based on the individual's initial velocity, fall speed, and environmental information such as wind speed and direction, the rescue drone rapidly and accurately predicts the fall trajectory. It then automatically plans its optimal flight path and adjusts its position to track the individual along this path. Simultaneously, the drone autonomously calculates the best spatial location for deployment and the precise direction for firing the net gun. This data and calculation results are rapidly synchronized with the rescue team and other drones participating in the rescue mission.

[0123] Once the rescue drone has tracked the person to be rescued along the optimal flight path to the rescue airspace, it will fire a net to capture the falling person. After capture, the rescue drone will immediately increase lift to reduce the person's descent speed. Simultaneously, the drone will begin to tow the person towards a safe airspace, thus preventing injury from collisions with building structures during the descent.

[0124] When the rescue drone detects the person in need and begins its descent, its onboard system plans a safe landing path. During the initialization phase, the onboard system first acquires its own spatial position information and motion status, such as altitude, speed, and orientation, and plans the drone's landing path based on this information. This path is designed to avoid any potential obstacles as much as possible and ensure a safe landing within the designated landing area. As the rescue drone begins its descent, it continuously detects obstacles along the path. When a new obstacle is detected, the landing path is replanned to allow the drone to bypass it. This obstacle detection process is real-time, with the onboard system continuously updating the landing path to adapt to the dynamic environment. Through continuous dynamic path planning and real-time obstacle detection, the rescue drone ultimately lands safely, ensuring not only its own safety but also optimal adaptation to the flight environment.

[0125] Once the descent of the rescue drone towing the person awaiting rescue begins, the drone's onboard system immediately calculates the drone's real-time lift margin. When there is no lift margin remaining, and the drone's overall descent rate exceeds a set threshold (e.g., 6 meters per second, which can be preset by the operator), the onboard system will control the drone's descent speed by deploying an auxiliary parachute. Among these, F... d C represents air resistance (unit: N). d Let ρ represent the drag coefficient, which is the effect of an object's shape on air resistance; ρ represent air density (in kilograms per cubic meter); A represent the cross-sectional area of ​​the object in the direction of motion (in square meters); and V represent the object's velocity relative to the air (in meters per second). The formula for calculating air resistance is:

[0126]

[0127] When the weight of the person to be rescued is 100kg and the maximum rescue payload of the drone (excluding the drone itself, fuel, battery, and rescue equipment) is 75kg, in order to ensure that the descent speed of the drone dragging the person to be rescued does not exceed 6m / s, the effective area of ​​the parachute required by the drone can be calculated to be 5.67m² according to the above formula (1). 2 That is, a parachute with an effective diameter of 2m can meet the requirements.

[0128] It should be noted that, considering the possibility of a single rescue drone mission failing, this embodiment of the application introduces multiple drones to jointly execute rescue missions, thereby optimizing the emergency rescue mechanism and making the rescue mission more fault-tolerant to reduce the possibility of rescue failure. If a drone encounters problems during mission execution, other drones can coordinate in real time and take over the mission, ensuring that the rescue operation is not hindered. This collaborative operation method enhances the robustness of the entire rescue system, effectively responds to unexpected situations that may occur during mission execution, and maximizes the success rate of the mission. It not only possesses efficient rescue planning and dynamic adjustment capabilities but also further ensures the reliability and safety of mission execution.

[0129] Specifically, in scenarios involving multiple drones collaborating on rescue missions, each drone, upon receiving a mission, first performs a self-check of its own status and capabilities to ensure it is in optimal operating condition. This includes checking key parameters such as its maximum rescue payload weight, spatial location, and battery endurance. Simultaneously, the drone receives information about the person to be rescued, such as estimated weight, location, and potential urgency level. Based on this information, the drone performs initial mission planning and dynamically identifies the target person. During the rescue process, the drones utilize advanced sensors and algorithms to monitor their own status and the surrounding environment in real time, ensuring the safety and efficiency of their flight paths. At the same time, the drones share this information with other drones through high-speed communication networks to achieve transparency and synchronization. This close collaborative mechanism allows the drone swarm to dynamically adjust mission allocation and execution strategies based on real-time conditions. For example, if a drone finds its battery or payload capacity insufficient to complete the current rescue mission, it can quickly transfer the mission to another drone in better condition, while it may switch to monitoring or auxiliary tasks. This flexible mission switching and resource optimization improves the responsiveness and adaptability of the entire rescue mission.

[0130] Furthermore, the collaborative work of drone swarms is also reflected in handling complex situations. When faced with multiple individuals requiring rescue, drones can adjust their rescue priorities and action plans in real time based on their respective mission progress. For example, for lighter individuals, drones can choose to quickly approach and use net guns to capture them; while for heavier individuals or those in more complex situations, drones may need to adopt a more cautious rescue strategy, approaching and rescuing them gradually.

[0131] Through this dynamic task allocation and close collaboration based on real-time data, drone swarms not only improve rescue efficiency but also ensure that everyone in need of rescue receives timely and effective support. This intelligent and automated rescue method provides strong technical support for rapid response in emergency situations, greatly enhancing the success rate and safety of rescue operations.

[0132] The technical solution of this application combines advanced drone technology, real-time monitoring, artificial intelligence analysis, and proactive intervention measures to form a comprehensive campus security protection solution. Its specific technical advantages include:

[0133] (1) By analyzing and identifying potential risks of falling from heights in real time, we can not only improve the accuracy of preventive measures, but also greatly enhance the timeliness of rescue operations.

[0134] (2) By using the active intervention mechanism of drones, including the ability to fire net guns to restrict the movement of people to be rescued, a new, more direct and effective method is provided to prevent falls from heights.

[0135] (3) The ability to calculate the fall trajectory in real time and accurately track and capture it during the fall ensures that rescue missions can be carried out effectively even in complex and ever-changing campus environments.

[0136] (4) Through the ability of multi-machine collaborative operation and seamless integration with the campus monitoring system, it is possible to further realize and enhance the intelligence and automation of campus security protection.

[0137] This application also proposes a drone rescue device for use in an airborne system of a drone. Figure 4 This is a schematic diagram of the structure of a drone rescue device for an airborne system of a drone provided in an embodiment of this application, as shown below. Figure 4 As shown, the device includes:

[0138] The determining unit 401 is used to determine the target user's body movement information and facial information based on the video stream image of the target user; the facial information includes the target user's facial emotional state; and to determine the target user's action events based on the body movement information and facial information.

[0139] The processing unit 402 is used to determine the distance between the drone and the target user if the action event is a risk event: if the distance is less than or equal to a distance threshold, a rescue operation is performed on the target user; or, if the distance is greater than the distance threshold, a rescue operation is performed on the target user based on a preset strategy; risk events include crash events.

[0140] In some embodiments, the determining unit 401 includes a first sub-determining unit and a second sub-determining unit; wherein,

[0141] The first sub-determination unit is used to determine the body part information of the target user in each video frame image based on each of the multiple video frame images; and to determine limb movement information and facial information based on the body part information of the target user in each video frame image.

[0142] The second sub-determination unit is used to determine the target user's motion characteristics based on body movement information; determine the target user's emotional characteristics based on facial information; and determine the action event based on both motion and emotional characteristics.

[0143] In some implementations, the first sub-determining unit is specifically used to determine the changes in the joint point information of the target user between each video frame image and adjacent video frames, based on the joint point information of the target user in each video frame image; and to analyze the changes in the joint point information of the target user between each video frame image and adjacent video frames to obtain limb movement information and facial information; wherein, the target user's body part information includes joint point information.

[0144] In some implementations, the second sub-determining unit is specifically used to fuse motion features and emotional features to obtain the comprehensive features of the target user; based on the comprehensive features, the action tendency of the target user is predicted by an action classification model to obtain action events.

[0145] In some implementations, the processing unit 402 is specifically used to determine the environmental information of the target user; determine the movement trajectory of the target user based on the limb movement information and the environmental information; determine the target flight path of the drone based on the movement trajectory; track the target user based on the target flight path and determine the distance between the drone and the target user; if the distance is less than or equal to a distance threshold, then perform a rescue operation on the target user.

[0146] In some embodiments, the device further includes a control unit; wherein,

[0147] The control unit is used to determine the target landing path of the UAV based on its spatial location information and motion state after a risk event occurs and a rescue operation is carried out on the target user; control the UAV to descend based on the target landing path; calculate the lift margin of the UAV; if it is determined that the lift margin is less than or equal to the lift threshold and the descent speed of the UAV is greater than the speed threshold, then the descent speed of the UAV is controlled by auxiliary equipment.

[0148] In some embodiments, the device further includes an updating unit; wherein,

[0149] The update unit is used to update the target landing path if an obstacle is detected in the target landing path during the descent of the UAV based on the target landing path, and then control the UAV to descend based on the updated target landing path.

[0150] In some embodiments, the device further includes an association unit; wherein,

[0151] The association unit is used to determine the target user's identity information based on facial information; determine the target user's association information within a preset time period based on the identity information; the association information includes mental health and / or emotional assessment information for the target user; determine reassurance information for the target user based on the association information; and send the reassurance information to the target user.

[0152] In some embodiments, the device further includes a coordination unit; wherein,

[0153] The coordination unit is used to communicate with one or more other drones for coordinated rescue of target users. Each drone in the coordinated rescue is configured with corresponding task content and / or execution strategy, and the task content and / or execution strategy of each drone are different.

[0154] In the technical solution of this application embodiment, the onboard system of the UAV determines the target user's body movement information and facial information based on the target user's video stream image, and determines the target user's action events based on the body movement information and facial information. If the action event is a risk event, the distance between the UAV and the target user is determined: if the distance is less than or equal to a distance threshold, a rescue operation is performed on the target user; or, if the distance is greater than the distance threshold, a rescue operation is performed on the target user based on a preset strategy; wherein, risk events include fall events. In this way, by capturing the body movements of the person to be rescued in real time, estimating the body characteristic data of the person to be rescued, and generating the action events of the person to be rescued, a rapid safety incident emergency response can be provided to the person to be rescued according to the degree of danger of the action event, and a timely rescue decision can be made. This not only fills the problem of insufficient response capabilities of traditional protection methods, but also effectively improves the accuracy of preventive measures and enhances the timeliness of rescue operations.

[0155] Those skilled in the art should understand that Figure 4 The functions of each unit in the drone rescue device shown can be understood by referring to the relevant descriptions of the aforementioned methods. Figure 4 The functions of each unit in the drone rescue device shown can be implemented through a program running on a processor or through specific logic circuits.

[0156] Figure 5This is a schematic diagram of the processing device provided in the embodiments of this application. Figure 5 The processing device shown includes a processor 501, which can call and run computer programs from memory to implement the methods in the embodiments of this application.

[0157] Optionally, such as Figure 5 As shown, the processing device may further include a memory 502. The processor 501 can retrieve and run computer programs from the memory 502 to implement the methods described in this embodiment.

[0158] The memory 502 can be a separate device independent of the processor 501, or it can be integrated into the processor 501.

[0159] Optionally, such as Figure 5 As shown, the processing device may also include a transceiver 503, which the processor 501 can control to communicate with other devices. Specifically, it can send information or data to other devices or receive information or data sent by other devices.

[0160] The transceiver 503 may include a transmitter and a receiver. The transceiver 503 may further include an antenna, which may be one or more.

[0161] The processing device may specifically be the drone rescue device of this application embodiment, and the processing device can implement the corresponding processes of the various methods implemented in this application embodiment. For the sake of brevity, it will not be described in detail here.

[0162] It should be understood that the processor in the embodiments of this application may be an integrated circuit chip with signal processing capabilities. In implementation, the steps of the above method embodiments can be completed by integrated logic circuits in the processor's hardware or by instructions in software form. The processor described above can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules can be located in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The storage medium is located in memory, and the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method.

[0163] It is understood that the memory in the embodiments of this application can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDR SDRAM), Enhanced Synchronous DRAM (ESDRAM), Synchlink DRAM (SLDRAM), and Direct Rambus RAM (DR RAM). It should be noted that the memory used in the systems and methods described herein is intended to include, but is not limited to, these and any other suitable types of memory.

[0164] It should be understood that the above-described memory is exemplary and not a limiting description. For example, the memory in the embodiments of this application may also be static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct memory bus RAM (DR RAM), etc. That is to say, the memory in the embodiments of this application is intended to include, but is not limited to, these and any other suitable types of memory.

[0165] This application also provides a computer-readable storage medium for storing a computer program. This computer-readable storage medium can be applied to the processing device in this application embodiment, and the computer program causes the computer to execute the corresponding processes implemented by the various methods in this application embodiment; for brevity, further details are omitted here.

[0166] This application also provides a computer program product, including computer program instructions. This computer program product can be applied to the processing device in this application embodiment, and the computer program instructions cause the computer to execute the corresponding processes implemented by the various methods in this application embodiment; for brevity, further details are omitted here.

[0167] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0168] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0169] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0170] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0171] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0172] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0173] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A drone rescue method, characterized by, The method is applied to an airborne system of a UAV, and comprises the following steps: determining body movement information and face information of a target user based on a video stream image of the target user; the face information comprises a facial emotional state of the target user; determining a movement event of the target user based on the body movement information and the face information; if the movement event is a risk event, determining a distance between the UAV and the target user; if the distance is less than or equal to a distance threshold, implementing a rescue operation on the target user; or, if the distance is greater than the distance threshold, implementing the rescue operation on the target user based on a preset strategy; the risk event comprises a falling event.

2. The drone rescue method of claim 1, wherein, The video stream image comprises a plurality of video frame images; the step of determining the body movement information and the face information of the target user based on the video stream image of the target user comprises the following steps: determining body part information of the target user in each video frame image in the plurality of video frame images; determining the body movement information and the face information based on the body part information of the target user in each video frame image. 3.The UAV rescue method of claim 2, wherein, The body part information of the target user comprises joint information; the step of determining the body movement information and the face information based on the body part information of the target user in each video frame image comprises the following steps: determining a change in joint information of the target user between each video frame image and an adjacent video frame image based on the joint information of the target user in each video frame image; analyzing the change in joint information of the target user between each video frame image and an adjacent video frame image to obtain the body movement information and the face information. 4.The UAV rescue method of claim 1, wherein, The step of determining the movement event of the target user based on the body movement information and the face information comprises the following steps: determining a movement feature of the target user based on the body movement information; determining an emotional feature of the target user based on the face information; determining the movement event based on the movement feature and the emotional feature.

5. The drone rescue method of claim 4, wherein, The step of determining the movement event based on the movement feature and the emotional feature comprises the following steps: fusing the movement feature and the emotional feature to obtain a comprehensive feature of the target user; predicting a movement tendency of the target user through a movement classification model based on the comprehensive feature to obtain the movement event. 6.The UAV rescue method of claim 1, wherein, The step of implementing the rescue operation on the target user based on the preset strategy comprises the following steps: determining environmental information in which the target user is located; determining a movement trajectory of the target user based on the body movement information and the environmental information; determining a target flight path of the UAV based on the movement trajectory; tracking the target user based on the target flight path and determining a distance between the UAV and the target user; if the distance is less than or equal to the distance threshold, implementing the rescue operation on the target user.

7. The drone rescue method of any one of claims 1 to 6, wherein, The method further comprises the following steps: After the risk event occurs and the rescue operation is implemented on the target user, a target landing path of the UAV is determined based on spatial position information and a motion state of the UAV; The UAV is controlled to descend based on the target landing path, and a lift margin of the UAV is calculated, and if it is determined that the lift margin is less than or equal to a lift threshold value and a descending speed of the UAV is greater than a speed threshold value, the descending speed of the UAV is controlled by an auxiliary device.

8. The drone rescue method of claim 7, wherein, The method further includes: During the process of controlling the UAV to descend based on the target landing path, if it is detected that there is an obstacle on the target landing path, the target landing path is updated to obtain an updated target landing path; The UAV is controlled to descend based on the updated target landing path. 9.The UAV rescue method of claim 7, wherein, The method further includes: Identity information of the target user is determined based on the facial information; Correlation information of the target user within a preset time is determined based on the identity information; the correlation information includes psychological health and / or emotional assessment information for the target user; Comfort information for the target user is determined based on the correlation information; The comfort information is sent to the target user. 10.The UAV rescue method of claim 7, wherein, The method further includes: Communication is performed with one or more other UAVs, and the communication is used for the UAV and the one or more other UAVs to perform cooperative rescue on the target user; wherein each UAV in the cooperative rescue is configured with corresponding task content and / or execution strategy, and the corresponding task content and / or execution strategy of each UAV are different.

11. A drone rescue device, characterized in that, An on-board system applied to a UAV, the device includes: A determination unit configured to determine body movement information and facial information of a target user based on a video stream image of the target user; the facial information includes a facial emotional state of the target user; and determine a movement event of the target user based on the body movement information and the facial information; A processing unit configured to, if the movement event is a risk event, determine a distance between the UAV and the target user; if the distance is less than or equal to a distance threshold value, implement a rescue operation on the target user; or, if the distance is greater than the distance threshold value, implement the rescue operation on the target user based on a preset strategy; the risk event includes a falling event.

12. A processing device, characterized by A processor and a memory for storing a computer program, the processor being configured to invoke and run the computer program stored in the memory to perform the UAV rescue method according to any one of claims 1 to 10. A computer program for storing, the computer program causing a computer to perform the UAV rescue method according to any one of claims 1 to 10.

13. A computer-readable storage medium, characterized in that, Computer program instructions for causing a computer to perform the UAV rescue method according to any one of claims 1 to 10.

14. A computer program product, characterised in that, ​

Images